Towards global understanding of plant defence against aphids – timing and dynamics of early Arabidopsis defence responses to cabbage aphid (Brevicoryne brassicae) attack


A. M. Bones. Fax: +47 596100; e-mail:


Insect feeding on plants causes a complex series of coordinated defence responses. Little is known, however, about the time-dependent aspect of induced changes. Here we present a time series-based investigation of Arabidopsis thaliana Ler subjected to attack by a specialist pest of Brassicaceae species, Brevicoryne brassicae. Transcriptome and metabolome changes were studied at 6, 12, 24 and 48 h after infestation to monitor the progress of early induced responses. The use of full-genome oligonucleotide microarrays revealed the initiation of extensive gene expression changes already during the first 6 h of infestation. Data indicated the involvement of reactive oxygen species (ROS) and calcium in early signalling, and salicylic acid (SA) and jasmonic acid (JA) in the regulation of defence responses. Transcripts related to senescence, biosynthesis of anti-insect proteins, indolyl glucosinolates (GS) and camalexin, as well as several uncharacterized to date WRKY transcription factors, were induced. Follow-up studies of defence-involved secondary metabolites revealed depositions of callose at the insects’ feeding sites, a decrease in the total level of aliphatic GS, particularly 3-hydroxypropyl glucosinolate, and accumulation of 4-methoxyindol-3-ylmethyl glucosinolate 48 h after the attack. The novel role of camalexin, induced as a part of defence against aphids, was verified in fitness experiments. Fecundity of B. brassicae was reduced on camalexin-accumulating wild-type (WT) plants as compared with camalexin-deficient pad3-1 mutants. Based on experimental data, a model of plant–aphid interactions at the early phase of infestation was proposed.


Aphids are an example of a ‘stealthy’ pest. In contrast to chewing herbivores, which macerate plant tissue, they are adapted to feed on phloem sap (Goggin 2007). The amount of phloem sap an individual aphid is able to consume during its relatively short lifetime is small compared to the plant's size (Will & van Bel 2006). However, short generation times and an extremely high asexual fecundity can result in a rapid increase in aphid population density and subsequent elevated consumption levels of phloem sap. Initiating and sustaining plant defence responses additionally detracts resources otherwise used for plant growth and development. Thus, the depletion of nutrients can become a serious problem and may have a severe impact on host plants. Crop losses resulting from colonization by fast-reproducing aphids show how vulnerable to infestation monocultures of cultivated plants are. Some aphid species are oligophagous and feed on a narrow range of host plants. The cabbage aphid, Brevicoryne brassicae, only thrives on cruciferous plants. It is so specialized that, in addition to exploiting nutrients from phloem sap, it also takes advantage of the host defence system utilizing plant-derived glucosinolates (GS) and an endogenous myrosinase for its own protection against predators (Jones et al. 2001, 2002; Bridges et al. 2002; Kazana et al. 2007).

Plant survival upon aphid attack depends on a multicomponent protection strategy, involving constitutive and inducible defences. Triggering the latter requires attacker recognition, cell signalling, and subsequent transcriptional and metabolic reprogramming leading to the production of toxic or deterrent compounds. Reactive oxygen species (ROS) and calcium signalling are both induced as a general response to wounding of plant cell membranes and may prove to be important in plant–insect interactions (Kehr 2006). Defence responses of an aphid-infested plant are modulated by the activation of three signal transduction pathways, each based on a different phytohormone: salicylic acid (SA), jasmonic acid (JA) and ethylene (ET) (Thompson & Goggin 2006). Further, the activation of other important regulators (often transcription factors) enabling a cross-talk mediates antagonistic or synergistic relationships between the pathways, and defines gene expression profiles of induced resistance (Li et al. 2006a). Previous studies have shown the induction of JA- and SA-dependent signalling after aphid attack (Moran & Thompson 2001; Moran et al. 2002; De Vos et al. 2005; Kuśnierczyk et al. 2007), and their important roles in plant defence against aphids (Mewis et al. 2005; Pegadaraju et al. 2005; Gao et al. 2007).

Cruciferous plants are able to produce a range of glucose-derived, sulphur-containing secondary metabolites, commonly named glucosinolates (GS); for review see Grubb & Abel (2006) and Halkier & Gershenzon (2006). Upon tissue damage, encountered during herbivore attack or mechanical wounding, these components are hydrolysed by myrosinases into a range of toxic or deterrent products. The final composition of the generated mixture of compounds is determined by the pool of GS available for hydrolysis, chemical conditions and the presence of various myrosinase-interacting proteins (Bones & Rossiter 1996, 2006). Because of the spatial separation of GS and myrosinases in plant tissue (Kelly, Bones & Rossiter 1998; Koroleva et al. 2000; Husebye et al. 2002; Thangstad et al. 2004), only the breakdown of cellular integrity is believed to initiate a massive GS hydrolysis. Aphids introduce minimal tissue damage and are therefore more exposed to intact GS rather than the hydrolysis products. This may explain why knocking out foliar myrosinase genes (TGG1 and TGG2) did not have any effect on aphids’ (Myzus persicae and B. brassicae) performance (Barth & Jander 2006), while changes in GS profiles influenced aphids’ fecundity and choice of host plants (Levy et al. 2005; Mewis et al. 2006). Moreover, a tryptophan-derived GS, 4-methoxyindol-3-ylmethyl (4MI3M), was shown to be a strong feeding deterrent for M. persicae (Kim & Jander 2007). Recent research has provided evidence for changes in GS content and expression profiles of genes from the GS biosynthesis pathway upon aphid infestation (Mewis et al. 2005, 2006; Kim & Jander 2007; Kuśnierczyk et al. 2007).

Phytoalexins belong to a different class of secondary metabolites involved in plant defence. In Arabidopsis, the biosynthesis pathway of camalexin (3-thiazol-2’-yl-indole), one of the most studied phytoalexins, shares the first steps with the synthesis of indolyl GS (for review see Glawischnig 2007). Camalexin appears to accumulate at infection sites and has a crucial role in defence against fungal and bacterial pathogens (Rogers, Glazebrook & Ausubel 1996; Ferrari et al. 2007; Sellam et al. 2007). To our knowledge, there are no reports about the possible induction of camalexin biosynthesis in plants infested with aphids, nor any evidence for toxicity of camalexin to aphids.

Several studies have been performed to elucidate plant responses to aphid infestation including gene expression as well as metabolite analysis. Transcriptional changes were investigated both in Arabidopsis (Moran & Thompson 2001; Moran et al. 2002; De Vos et al. 2005; Pegadaraju et al. 2005; Couldridge et al. 2007; Kuśnierczyk et al. 2007) and other plant species including Nicotiana attenuata (Heidel & Baldwin 2004; Voelckel, Weisser & Baldwin 2004), Medicago truncatula (Gao et al. 2007), Sorgum bicolor (Zhu-Salzman et al. 2004; Park, Huang & Ayoubi 2006) and Apium graveolens (Divol et al. 2005). Unravelling ‘genes that matter’ or ‘key proteins’ in plant–aphid interactions has greatly contributed to our understanding of plant perception and response to attack. The vast majority of studies, however, describe transcriptional responses after 48 or 72 h of sustained infestation, thus missing the temporal aspect of plant–insect interactions. Moreover, transcript levels of a limited number of genes were usually investigated, and only expression changes of particularly interesting transcripts were monitored by northern blots or quantitative RT-PCR methods in additional time points (Zhu-Salzman et al. 2004; De Vos et al. 2005; Gao et al. 2007). Only one study has attempted to use a full-genome microarray to capture Arabidopsis genome responses at two time points after M. persicae infestation, 2 and 36 h (Couldridge et al. 2007). However, in contrast to previous reports, surprisingly fewer responding genes were identified. Secondary metabolite profiling, aimed at investigating changes in GS levels, was carried out at 72 h or 1 week after aphid attack (Mewis et al. 2005, 2006; Kim & Jander 2007). Thus, very little is known about early transcriptional responses of plants to aphid infestation and the accompanying changes in secondary metabolite profiles, including GS levels, in the first 48 h of infestation.

In this study, with the advantage of time course experiment, we aim to investigate the complexity and dynamics of Arabidopsis defensive responses during the early phase of infestation with the cruciferous specialist aphid B. brassicae. Arabidopsis plants are considered to be moderately resistant to B. brassicae infestation when compared with other Brassicaceae species (Singh et al. 1994). Landsberg erecta (Ler) ecotype was chosen for analysis because of the previously observed high induction of genes involved in the indolyl GS pathway after 72 h of aphid infestation (Kuśnierczyk et al. 2007). Transcriptional events were captured with the use of full-genome oligoarrays at four time points after infestation: 6, 12, 24 and 48 h. The microarray findings were complemented with the follow-up analysis aiming to assess the accumulation of the induced genes’ products. The generation of hydrogen peroxide and callose depositions at the aphids’ feeding places were examined histochemically. Furthermore, the concentrations of the defence-related secondary metabolites, GS and camalexin, were measured. The importance of camalexin in defence against B. brassicae was evaluated by carrying out fitness experiments. The study investigates the local but not the constitutive defence responses. By using an integrated approach, combining data acquisition on different levels, we give a broad illustration of the time-dependent development of early Arabidopsis defensive responses after infestation with B. brassicae, and develop a model of plant–aphid interactions.


Plant material

All time series experiments were performed on a single seed line derived from Arabidopsis thaliana ecotype Landsberg erecta (Ler) seeds produced by Lehle Seeds (Round Rock, TX, USA; Catalog No.: WT-04-14, Seed Lot No.: GH198-01). Seeds were sown into 6-cm-diameter pots filled with a sterile soil mix (1.0 part soil and 0.5 part horticultural perlite). Plants were kept in growth chambers Vötsch VB 1514 (Vötsch Industrietechnik GmbH, Balingen-Frommern, Germany) with a 16/8 h (light/dark) photoperiod at 22/18 °C, 40/70% relative humidity, and 70/0 µmol m−2 s−1 light intensity.


B. brassicae clones were maintained on Brassica napus or cabbage plants in a growth chamber with a 16/8 h (light/dark) photoperiod at 22/18 °C, 40/70% relative humidity, and 70/0 µmol m−2 s−1 light intensity.

Infestation experiments

Plants at the growth stage 1.08 (Boyes et al. 2001) having eight developed rosette leaves, 21–25 d old, were infested with 32 apterous aphids (4 aphids per leaf) transferred to the leaves with a fine paint brush. Insect bioassays were conducted in plexiglass cylinders as described before (Kuśnierczyk et al. 2007). Each pot contained three plants. Infested plants and corresponding aphid-free controls were harvested at 6, 12, 24 and 48 h after infestation (hpi). To minimize diurnal differences in glucosinolate levels between different time points (de Felice et al. 2006), plants at 6, 24 and 48 hpi were always harvested between 16:00 and 18:00 h. Plants at 12 hpi were harvested between 21:00 and 23:00 h, as harvesting at the same day time as the other time points would require an interruption of dark period during infestation, which might influence gene expression. The experiments were repeated five times for each of the time points, and plant material from the five repetitions was divided into four biological replicates to balance the experimental effect. For each of the time points, each biological replica consisted of approximately 24 plants (Jørstad, Langaas & Bones 2007). Whole rosettes were cut at the hypocotyls, harvested, and aphids were removed by washing with Milli-Q-filtered water (Millipore Corporation, Bedford, MA, USA). Any bolts appearing during the experiments were removed before harvesting. The harvested material was immediately frozen in liquid nitrogen.

RNA isolation, cDNA synthesis and microarray experiments

RNA was isolated with the Spectrum Plant Total RNA Kit (Sigma, St. Louis, MO, USA) and was quantified with Nanodrop ND 1000 (Nanodrop Technologies, Wilmington, DE, USA). The quality of RNA was assessed by formaldehyde agarose gel electrophoresis. Total RNA (15 µg) and SuperScript III reverse transcriptase (Invitrogen, Carlsbad, CA, USA) were used in a reverse transcription reaction. Microarray slides were printed by the Norwegian Microarray Consortium (Trondheim, Norway). A 3DNA Array 350 kit with Cy3- and Cy5-labelled dendrimers (Genisphere, Inc., Hatfield, PA, USA) was used for labelling. The microarray slides contained 31 811 unique 70 mer oligos with C6-amino linker, a total of 33 696 spots, covering 26 604 genes. Of these oligos, 29 110 were from the Qiagen-Operon Arabidopsis Genome Array Ready Oligo Set (AROS) Version 3.0, while the others were custom-made and produced by Operon (Alameda, CA, USA) or MWG (Ebersberg, Germany). The oligonucleotides were dissolved in Milli-Q-filtered grade water and 50% dimethyl sulfoxide (DMSO) (20 pmol/µL), and were spotted on aminosilane-coated UltraGaps slides (Corning, New York, NY, USA) using a BioRobotics MicroGrid II robot (Genomic Solutions, Ann Arbor, MI, USA). Printing of the microarray slides was carried out at the Norwegian Microarray Consortium. Hybridizations were performed in a Slide Booster Hybridization Station (Advalytix, Brunnthal, Germany), and the slides were washed according to the manufacturers’ descriptions (Genisphere and Advalytix). The slides were scanned at 10 mm resolution on a G2505B Agilent DNA microarray scanner (Agilent Technologies, Palo Alto, CA, USA). The resulting images were processed using GenePix 5.1 software (Axon Instruments, Union City, CA, USA).

Quantitative RT-PCR

Quantitative RT-PCR analysis was performed with the use of two out of four biological replicas. The total RNA was DNAse treated (Invitrogen), and SuperScript III reverse transcriptase (Invitrogen) was used in cDNA synthesis. Sequences for gene-specific primers used in PCR are given in Supplementary Table S1. TIP41-like gene (At4g34270) was used as a normalizer (Czechowski et al. 2005). LightCycler 480 System and corresponding SYBR Green I Master (Roche, Indianapolis, IN, USA) were used in a three-step programme including (1) preincubation at 95 °C for 5 min; (2) 40 cycles of amplification consisting of 95 °C for 10 s, 55 °C for 7 s and 72 °C for 7 s; and (3) melting curve analysis 95 °C for 5 s, 65 °C for 1 min, 97 °C. Each 20 µL reaction contained 0.2 µm of each forward and reverse primer, and cDNA quantity corresponding to 0.0125 µg of RNA (for HIG1, CYP79B2, PR1) or 0.125 µg of RNA (for PAD3 and SAG21). Relative expression values were calculated using 2−ΔΔCt method.

Statistical analysis

The GenePix processed data were filtered to remove the spots whose exact position could not be determined by the software because of a low signal intensity (GenePix flag -50), the empty spots for which no probes were printed (GenePix flag -75) and the spots that were recognized as being of low quality by manual inspection (GenePix flag -100). The spots were also filtered out if they had more than 50% saturated pixels, or had a median foreground intensity less than the local median background intensity. No background subtraction was performed. The data from each array were log-transformed and normalized using the printtip-loess approach (Yang et al. 2001). Within-array replicated measurements for the same gene were merged by taking the average over the replicates. The data were then scaled so that all array data sets had the same median absolute deviation. In the data processing that followed, we focused on genes for which at least three out of four biological replicates for the time points had passed quality control.

For making statistical inference about differentially regulated genes, the limma package (Smyth 2004) was used. The limma approach is based on fitting a linear model to the expression data of each probe on a microarray. For our experiment, a model was fitted with a coefficient for each comparison of infested and control plants at the four different time points. Comparisons (contrasts) were also made to evaluate changes between time points. To identify differentially regulated genes for any particular comparison, the associated P-values were used. From the P-values, a set of corresponding q-values (Storey 2002) could be calculated. The q-value for a gene is the expected proportion of false positives one will get when calling that gene significant. For a gene to be considered statistically significantly differentially expressed, its q-value was required to be lower than 0.05, in effect controlling the false discovery rate (FDR) (Benjamini & Hochberg 1995) of the comparison at a 0.05 level. All genes discussed in this paper were found to be significantly differentially expressed.

To compare measured glucosinolate levels of infested plants with those of control plants and to compare camalexin levels, a two-tailed Welch's t-test was used. To compare aphid counts of camalexin-deficient plants with those of control plants, a two-tailed Wilcoxon rank sum test was used. For both tests, the significance level used for each individual comparison was 0.05.

Extraction of GS and high-performance liquid chromatography–mass spectrometry (HPLC-MS) analysis

GS were extracted from freeze-dried material with 3 × 3.5 mL of 70% MeOH (v/v) at 70 °C. After heating at 70 °C for 10 min and subsequent cooling, homogenates were centrifuged at 4000 g for 10 min at 4 °C. Combined extracts were filtered through 0.45 µm filters (Millipore Corporation), loaded onto an anion exchange column (Sephadex DEAE A25; Amersham Bioscences, Uppsala, Sweden) and treated for 18 h with a solution of aryl sulphatase (H-1 type, from Helix pomatia; Sigma-Aldrich), prepared as described by Graser et al. (2001). The desulfoglucosinolates were eluted with Milli-Q-filtered water, and the eluate was frozen at –20 °C and then freeze-dried for 72 h. The samples were redissolved in water and analysed by HPLC-diode array detector (DAD) coupled to LC/MSD trap mass spectrometer (Agilent HP 1100 Series; Agilent Technologies) equipped with an APCI interface. The separation was carried out in a Supelcosil LC 18 column (250 × 2.1 mm, 5 µm) (Supelco, Bellefonte, PA, USA) at a flow rate of 0.3 mL min−1, at room temperature. The mobile phase consisted of solvent A (Milli-Q-filtered water) and solvent B (acetonitrile; Merck, Darmstadt, Germany), and the following gradient procedure was used: 0–2 min, 3% (v/v) B; 2–17 min, 3–40% (v/v) B; 17–22 min, 40% (v/v) B; 22–22.10 min, 40–100% (v/v) B; 22.10–32 min, 100% (v/v) B; 32–32.10 min, 3% (v/v) B; and 32.10–60 min, 3% (v/v) B. The mass spectrometer was configured in a positive ion chemical ionization. The APCI settings were as follows: nebulizer pressure, 60 psi; drying gas flow, 5 L min−1; drying gas temperature, 350 °C; APCI vap, 400 °C; and Corona current, 4000 nA. The individual GS were identified by a conjunct analysis of diode array ultraviolet (UV) spectra at 229 nm and mass spectra. The identities of GS were confirmed by a comparison with retention times and UV spectra of standard desulfoglucosinolates (mixture of isolated B. napus L. desulfoglucosinolates, Bioraf Denmark Foundation, Aakirkeby, Denmark). Relative quantification was based on sinigrin (Sigma-Aldrich) added as an internal standard at the start of the extraction procedure. Values of relative response factors of individual desulfoglucosinolates used in the calculation were taken from previous publications (Daun & McGregor 1991; Brown et al. 2003). The response factor of desulfo-7-methylsulfinylheptyl glucosinolate was assumed to be 1.

Determination of camalexin

Camalexin was extracted from 200 mg of frozen leaves by incubation for 20 min with 800 µL of 80% methanol at 65 °C. For 72 hpi, the plant material from previously described experiment was used (Kuśnierczyk et al. 2007). After the removal of the tissue, the methanol was evaporated under vacuum, and camalexin was extracted from the aqueous phase with 150 µL of chloroform, which was then evaporated to dryness. The residue was dissolved in 5 µL of chloroform, spotted on a thin-layer silica gel-coated chromatography plates (silica gel 60, 20 × 20; Merck KGaA) and developed in ethyl acetate hexane (8:1, v/v). The probes were compared with a camalexin standard (kindly provided by Professor Jane Glazebrook, University of Minnesota) under UV illumination (305 nm). The silica-containing camalexin was scraped, and camalexin was extracted into 1 mL of methanol. Fluorolog ISA HORIBA JOBIN YVON-SPEX (Jobin Yvon Ltd., Stanmore, UK) was used to measure the emission after excitation at 315 nm. The camalexin concentration was estimated at 400 nm emission by comparison with the standard curve of pure camalexin. For quantification of camalexin after UV-B treatment, only the leaves, which have been exposed to the treatment, were used. Newly developing leaves, which have not received the treatment, were discarded during harvesting.

Aphid fitness experiments

To assess the susceptibility of B. brassicae to camalexin, we used pad3-1 mutant plants defected in camalexin accumulation (kindly donated by Professor Jane Glazebrook, University of Minnesota), and as a control, wild-type (WT) Col-0 plants. Thirty-day-old plants, grown in pots, were first exposed for 5 min to UV-B light (302 nm) to stimulate camalexin accumulation. The plants were illuminated with the use of Transluminator (Model LM-26E; UVP, Upland, CA, USA) from a 21 cm distance. Immediately after the exposure, two first instar nymphs were placed on each plant, and the plants were placed in plexiglass cylinders (3 pad3-1 plants or 3 WT plants per cylinder). A total of 24 pad3-1 plants (eight cylinders) and 24 WT plants (eight cylinders) were used in the experimental setup. After 13 d, the numbers of progeny on the three plants in each cylinder were counted. During the experiment, the newly developing leaves and bolting stems were being removed to restrict aphid feeding to the leaves that have been previously exposed to UV-B light.


Visualization of H2O2 accumulation and callose depositions was carried out in plants infested with B. brassicae and in wounded plants, 6, 12, 24 and 48 h after treatments, and additionally 1 h after wounding (for H2O2 accumulation) and immediately after wounding (for callose depositions). Infestation was carried out as for time series experiments. The leaves were wounded by puncturing with a fine glass needle, produced by pulling a glass pipette over a flame. H2O2 accumulation was detected with 3,3’-diaminobenzidine (DAB)-HCl, pH 3.8 (Sigma) as described by (Thordal-Christensen et al. 1997). To detect callose, the leaves were rinsed with 96% ethanol, rinsed with water and stained for 2 h in 150 mm K2HPO4 (pH 9.5) containing 0.01% aniline blue. Samples were examined by epifluorescent illumination. Images were captured with a SPOT CCD camera (Diagnostic Instruments, Sterling Heights, MI, USA) coupled to a Nikon S800 microscope equipped with a filter with the following parameters: EX 465-495, DM 505, BA 515-555, or alternatively with a Nikon COOLPIX camera coupled to a Nikon (C-DSD230) stereo microscope (Tokyo, Japan).


Intensity and timing of defensive responses

Most of the transcriptional changes observed during B. brassicae attack involved an induction rather than a suppression of genes. At 6 h post-infestation (hpi), 736 genes were significantly up-regulated with a log2 ratio > 0.5, and 165 genes with a log2 ratio > 1.0. There were considerably fewer down-regulated genes (397 having a log2 ratio < −0.5, and 31 having a log2 ratio < −1.0). The corresponding numbers of induced and repressed genes at 48 hpi were 2–2.5 times greater, suggesting a dynamic progress in the development of plant responses. Genes coding for calcium-binding proteins, proteins involved in the detoxification of ROS, and a group of WRKY transcription factors were among the highest up-regulated from 6 hpi, indicating that the regulation of transcriptional reprogramming is most probably initiated by the early activation of ROS, calcium and WRKY signalling. Induced genes, as revealed by our microarray data, were grouped in seven different categories for simplicity reasons, although in reality, the relationships between the presented genes and the processes they are involved in are more complex. Some genes were connected to more than one category (for details, see Supplementary Table S2). A number of responding genes in each of the categories, as well as an average expression change in response to B. brassicae attack, were visualized on the basis of the experiment's time frame (Fig. 1). Gene responses in each of the shown categories were switched on by 6 hpi, with most of them showing progressively stronger responses with the later time points. The number of genes with altered expression also increased over time, especially at 24 and 48 hpi. In particular, after 24 h of infestation, there was a sudden rise in the number and intensity of the induction of genes attributed to the ROS, cell wall and SA groups. At the same time, the quantity of genes classified to indolyl GS biosynthesis, senescence and insecticides increased proportionally with the time of B. brassicae attack. The highest induction through all the time points was observed in genes connected to camalexin synthesis. Notably, the number of genes and the intensity of transcriptional changes of the wound-inducible JA responses considerably dropped at 12 hpi, indicating the presence of a tuning mechanism regulating transcriptional changes after the initiation of JA-connected responses. It should be noted that the total numbers of known genes attributed to date to each of the presented categories are different, and thus the relative contribution of the categories to B. brassicae-induced resistance should be interpreted with caution. For example, only four genes in our study were associated with camalexin biosynthesis, but all were highly induced at the first time point, indicating a strong responsiveness of this pathway to B. brassicae attack.

Figure 1.

Quantitative and qualitative development of Brevicoryne brassicae-induced responses within the eight different categories of Arabidopsis defence-related genes. The number of up-regulated genes and changes in their expression level indicate a progress in responsiveness within each of the categories over the first 48 h of infestation. Bars represent the number of genes significantly up-regulated with log2 ratio > 0.5, and colour code indicates the average expression of genes contributing to each of the categories after 6, 12, 24 and 48 h of infestation, respectively. It should be noted that the total numbers of known genes, which have been attributed to each of the categories, differ, and thus the relative contribution of each of the categories to B. brassicae-induced defence should be interpreted with caution. Designated gene categories include the following classes of genes: reactive oxygen species (ROS), up-regulated genes associated with the generation and detoxification of ROS; cell wall, up-regulated genes connected with cell wall metabolism; insecticides, up-regulated protease inhibitors and lectins; senescence, up-regulated genes associated with senescence; salicylic acid (SA), up-regulated genes involved in the SA biosynthesis pathway and connected to SA signalling; jasmonic acid (JA), up-regulated genes involved in the JA biosynthesis pathway and signalling; indolyl glucosinolates (IGS), up-regulated genes involved in the IGS biosynthesis pathway; and camalexin, genes connected to camalexin biosynthesis. It should also be noted that genes involved in the common branch of IGS and camalexin biosynthesis (upstream of indole-3-acetaldoxime) are presented under the IGS category. Some genes are attributed to more than one category because of overlapping functions. For detailed information about genes in each of the categories, see Table 1 and Supplementary Table S2.

The largest differences in transcript abundance between the first and the last time point studied were found in the following groups: pathogenesis related genes (PR1, PR2, PR4 and PR5) and genes coding for protease inhibitors (pEARLI 1-like and containing leucine-rich-repeat [LRR] domains). These were late-responding genes and generally not regulated at 6 hpi, but highly induced at the last time point of infestation. The most striking regulation was observed in the case of genes coding for homeobox proteins: knotted 1-like (KNAT1) and SHOOT MERISTEMLESS (STM), where a substantial down-regulation was observed at 6 hpi and not detectable at the later time points (Supplementary Table S3).

Oxidative stress- and calcium signalling

ROS are both directly and indirectly involved in plant defence. They are produced as the first reaction to injury and are toxic to pathogens including insects (Bi & Felton 1995; Kehr 2006). Moreover, they trigger a hypersensitive response (HR) and contribute to stress signalling, causing an induction of defence-related genes not only locally but also in remote organs (Divol et al. 2005). The expression levels of several genes associated with oxidative stress, especially ROS scavengers, were changed in response to B. brassicae attack, suggesting that the redox status in the attacked plants had been disturbed (Fig. 1 & Supplementary Table S2). Transcripts levels of many genes coding for proteins contributing to ROS detoxification accumulated already at 6 hpi, and reached maximum expression at 24 hpi. Among the up-regulated genes, the following groups were found: ascorbate reductases (MDAR 1 and MDAR 4); L-ascorbate oxidase; RESPONSIVE TO HIGH LIGHT 41 (RHL41), which is required for the expression of cytosolic ascorbate peroxidase 1 during oxidative stress; copper homeostasis factor (CCH); blue copper protein precursor (AtBCB); four glutaredoxin family proteins; thioredoxin (TRX5); glutathione S-transferases (induced already at 6 hpi: ATGST6, ATGST7 and ATGST10, and induced strongly at 48 hpi: ATGSTU3, ATGSTU10, ATGSTU11 and At5g02780); glutathione S-conjugate transporters (MRP3 and MRP4); and peroxidase precursors (Atperox P34, P37, P50, P52 and P71). Genes coding for hydrogen peroxide (H2O2) generating copper amine oxidases (At4g12290 and At1g62810) and NADPH oxidase (RBOHD) were moderately induced at 24 and 48 hpi. The latter has been shown to be required for H2O2 generation during bacterial and fungal challenge (Torres, Dangl & Jones 2002). In contrast, at the same two time points, genes coding for superoxide dismutases (FSD2 and FSD3) and a group of six polygalacturonases, the products of which lead to H2O2 generation, were moderately down-regulated. The expression of four genes coding for germin-like protein precursors was strongly induced at 48 hpi. Germin-like proteins of barley were recently reported to be induced by an exogenous hydrogen peroxide application and give a basal pathogen resistance, possibly by possessing a superoxide dismutase activity (Zimmermann et al. 2006). At the same time, the genes encoding inhibitors of H2O2 generated cell death: BAP1, BAP2 and their interactive partner BON were induced upon B. brassicae infestation (BAP1 and BON at all time points, and BAP2 at 48 hpi). A recent study has shown that BAP1 and BAP2 have overlapping functions and are required for negative regulation of programmed cell death (Yang et al. 2007).

To determine if the accumulation of H2O2 was detectable after aphid attack, infested and control (wounded) leaves were stained with DAB at 6, 12, 24 and 48 h after treatment, and in addition, the control leaves at 1 h after wounding. In the control plants, staining was detectable locally in cells at the edge of wounding, already after 1 h (Fig. 2a) and visible at all time points. In contrast, DAB staining of B. brassicae-infested leaves showed neither local nor systemic H2O2 accumulation in any of the studied time points (Fig. 2b, c).

Figure 2.

Histochemical staining of H2O2 and callose depositions in the control and Brevicoryne brassicae-infested leaves. Visualization of H2O2: (a) 1 h after wounding; (b, c) after 12 h of infestation with B. brassicae; (c) a closer view at the stylet's insertion site. Detection of callose depositions in the control plants: (d) immediately after wounding; (e) 48 h after wounding; (f) in the infested plants after 12 h of B. brassicae attack. Callose sealed holes in the cell walls probed by the aphids are seen as lighter fluorescent points. Although the results for detection of H2O2 and callose depositions are presented only for 12 h after infestation, they are representative for all the other time points studied (6, 24 and 48 h) as no differences in results between these time points were observed. (a, b) bars = 1mm; (c) bars = 0.5 mm; (d, e) bars = 200 µm; (f) bars = 100 µm.

Hydrogen peroxide activates the protein phosphorylation cascade, which modulates gene expression in response to external stimuli. This cascade involves subsequent phosphorylation events of MAPK, where the last of them is finally translocated to the nucleus and activates transcription factors (Neill, Desikan & Hancock 2002a). The expression of five genes coding for MAPKs (MKK1, MKK2, MKK4, MKK9 and MKK11) was positively regulated in our experiment, indicating a possible activation of this signal transduction cascade in response to B. brassicae attack.

Mechanical damage of cell membranes leads not only to the formation of ROS, but also to an elevation in cytosolic Ca2+ levels. Phloem calcium level is normally very low, and a sudden increase upon injury is believed to initiate a systemic signalling cascade (van Bel & Gaupels 2004). Genes coding for numerous calcium-binding proteins were up-regulated upon B. brassicae infestation as early as at 6 hpi (Supplementary Table S2), indicating a massive gene response to elevated calcium levels. Transcripts of several calmodulins, calmodulin-binding proteins (MLO2, MLO3, MLO12 and At2g26190), calmodulin-regulated nucleotide-gated ion channels (CNGC3, CNGC10, CNGC11 and CNGC20), calcium-dependent protein kinases (CDPK1, CPK6, CPK29, CPK32 and At4g23150), calcium-binding calretikulin 3 (CRT3) and pinoid-binding protein 1 (PBP1) were induced. Transcripts of calcium transporting ATPases (ACA1, ACA2, ACA10, ACA 11, ACA12, ACA 13, CAX3 and CAX5) were up-regulated. Interestingly, some of these genes' products are located in the plasma membrane (ACA2), in the vacuole membrane (ACA11) and in the chloroplast membrane (ACA1). In contrast, a gene coding for IQ calmodulin-binding protein and CIPK20 were both down-regulated.

Activation of WRKY transcription factors

Transcription factors are the important messengers between a perceived stimuli and the induced response. They control the expression of selected groups of genes by binding to the regulatory regions in the promoters of genes, and thus activate the transcription. Infestation with B. brassicae resulted in a coordinated, early induction of 16 genes coding for transcription factors belonging to the WRKY family (Supplementary Table S2). The WRKY transcription factors contain at least one conserved WRKY domain, about 60 amino acid residues with the WRKYGQK sequence followed by a C2H2 or C2HC zinc finger motif (Wu et al. 2005). Some of B. brassicae responsive WRKY proteins were previously associated with pathogen resistance (WRKY18, WRKY25, WRKY33 and WRKY40), senescence (WRKY6) or both processes (WRKY70 and WRKY53) (Robatzek & Somssich 2002; Miao et al. 2004; Xu et al. 2006; Ulker, Mukhtar & Somssich 2007; Zheng et al. 2007). WRKY75, which was the strongest induced of all WRKY transcription factors at 48 hpi, was reported to be up-regulated by hydrogen peroxide-induced cell death (Gechev, Minkov & Hille 2005) and during phosphate deprivation (Devaiah, Karthikeyan & Raghothama 2007). Its role in plant defence, however, has so far not been assessed. We have observed the induction of the following transcription factors involved in plant resistance, for which the specific regulatory functions have yet to be described: WRKY15, WRKY26, WRKY30, WRKY38, WRKY46, WRKY50, WRKY51 and WRKY54.

B. brassicae induces changes in cell wall metabolism

When phloem elements are punctured, the concentration of cytosolic Ca2+ rises and initiates defensive mechanisms, such as protein plugging and callose sealing, to prevent the loss of phloem sap. Up-regulation of callose synthase gene (CALS1) at 24 and 48 hpi, and changes in expression patterns of many genes related to cell wall metabolism and remodelling (Supplementary Table S2) indicate the involvement of cell wall modifications in defence response to B. brassicae attack. Four genes encoding cell wall-associated protein kinases were induced already at 6 hpi (WAK1, WAK2, At1g21245 and At1g22720), and two (RFO1 and At1g79680) at 48 hpi. Transcripts of glycine-rich protein (GRP-3) interacting with WAK1 were also induced at 24 and 48 hpi. Several genes coding for pectin esterases were up-regulated, and a transcript of a pectin esterase inhibitor was found to be down-regulated at 6 and 48 hpi. The expression profiles of several xyloglucan endotransglycosylases/hydrolases showed a differential expression where XTH22 and XTH25 were induced, and XTH4, XTH8, XTH9, XTH16, XTH31 and XTH32 were down-regulated. A fucosyltransferase gene, FUT4, was induced during the whole infestation period, and another, FUT7, and a galactoside 2-alpha-L-fucosyltransferase (At2g15360) at 24 and 48 hpi. Genes coding for mannan synthase (CSLA09), ARABINOGALACTAN PROTEIN 17 (AGP17), polygalacturonase inhibitor (FLR1), HOTHEAD (HTH), CWLP-related protein (At1g62500) and an unknown cell wall protein precursor (At2g20870) were found to be down-regulated.

Deposition of callose has been previously reported after attack by phloem feeding aphids (Botha & Matsiliza 2004) and silverleaf whitefly nymphs (Kempema et al. 2007). To investigate if B. brassicae feeding causes callose depositions, the infested and control (wounded) leaves were stained with aniline blue. Immediately after wounding, no callose depositions were detectable (Fig. 2d), but already at 6 hpi, callose was visible at mechanically damaged wound edges (Fig. 2e) and at the stylets insertion sites in the infested leaves (Fig. 2f).

Contributions of SA-, JA- and ET-signalling pathways to resistance against B. brassicae

A wide range of defence responses depend on SA-signalling. Genes involved in SA synthesis (ICS1 and ICS2), the induction of the SA-signalling pathway (EDS1, EDS5 and PAD4) and stress-responsive SA-dependent genes (PR1, PR2, PR4, PR5, NIMIN-1, NIMIN-2 and SABP2-like) were up-regulated (Supplementary Table S2). Some of these genes were induced already at 6 hpi, indicating the early recognition of B. brassicae attack. Genes coding for PR proteins were, however, induced significantly higher at the later time points suggesting that a time period between 12 and 24 h is required to develop the SA-related responses. An adequate microarray-based estimation of PR1 expression ratio was unsuccessful for all but the 24 h time point. This was mainly because of a very high abundance of PR1 transcripts in Arabidopsis tissue that caused a saturation of spots corresponding to this gene. An alternative quantitative RT-PCR-based method was used to estimate the regulation of PR1 and revealed its induction at 12, 24 and 48 hpi (Fig. 3). Unlike the SA-signalling-related genes, the expression profiles of genes associated with ET-signalling were not changed except for ET-responsive transcription factor 11 (ERF11), which was highly up-regulated at all time points, and ERF1-like, which was moderately induced at 48 hpi (Supplementary Table S2). The wound-inducible JA synthesis pathway was slightly less responsive to B. brassicae feeding than the SA-signalling pathway (Fig. 1 & Supplementary Table S2). Genes coding for the important enzymes of JA synthesis (PLD gamma 1, PLD gamma 2, and PLD gamma 3, LOX3-like, AOC2, OPR3) and coronatine-inducible genes (CORI3, PDF1.2, PDF1.2b, PDF1.2c and PDF1.3) were, however, induced generally already at 6 hpi. Genes coding for ZIM-domain proteins (JAZ1, JAZ5, JAZ9 and JAZ10), which are negative feedback regulators of jasmonate signalling (Chini et al. 2007; Thines et al. 2007), were also induced from the first time point. In contrast, JA-inducible genes coding for vegetative storage proteins VSP1 and VSP2, which has been shown to be an acid phosphatase with anti-insect properties (Liu et al. 2005), were found to be repressed at 6 and 48 hpi.

Figure 3.

Quantitative RT-PCR results for five selected genes. Bars represent averages of log2-transformed expression values for two biological replicates of Brevicoryne brassicae-infested plants relative to aphid-free control plants. The expression level of PR1 gene at 6 h after infestation was equal in the infested and control plants. HIG1, high indole glucosinolate 1 (MYB51 transcription factor); CYP79B2, cytochrome P450 79B2; PAD3, phytoalexin deficient 3 (cytochrome P450 71B15); PR1, pathogenesis-related 1; SAG21, senescence-associated gene 21.

Genes coding for proteins with anti-insect activities

In order to take up sufficient amounts of amino acids, phloem-feeding insects ingest large volumes of phloem sap. Some of the phloem sap proteins, by having an anti-nutrition effect or being toxic for insects, are potentially a part of the plant direct defence system, especially against phloem-sucking insects (Murdock & Shade 2002). A number of protease inhibitor transcripts were induced with different timing (Fig. 1 & Supplementary Table S2). Among these were genes coding for four pEARLI 1-like proteins and a putative serpin (At2g26390), which were up-regulated at 12 hpi and their induction increased gradually until 48 hpi. The induction of another putative serpin gene (At2g25240) was only pronounced at 48 hpi. Potato type II protease inhibitor gene was strongly up-regulated from the first time point, transcripts of trypsin inhibitors (At1g73260 and At2g43510) were induced at 24 and at 48 hpi, while the induction of protease inhibitors transcripts (At5g55460, At5g55450 and At5g55420) had a maximum differential expression at 24 hpi. Lectins, another class of phloem proteins with anti-insect activities, are difficult to digest and can bind to specific mono- or oligosaccharides in chitin. Their toxicity to both chewing and phloem-sucking insects is explained by binding to midgut tissue, and thus disturbing digestion and nutrient uptake (Murdock & Shade 2002). It has been shown that the addition of lectins originated from resistant brassicae species to artificial diet significantly reduced B. brassicae survival (Cole 1994). In response to B. brassicae, four genes coding for lectins had expression patterns revealing the induction during the whole infestation period (At5g18470, At3g16530, At3g15356 and At5g03350). A phloem-specific lectin transcript (At1g56240) was up-regulated at 24 and 48 hpi.

Glucosinolate profile changes upon B. brassicae infestation

Recent reports point out that aphid attack can quantitatively and qualitatively change the composition of GS in infested Arabidopsis plants (Mewis et al. 2005, 2006; Kim & Jander 2007). The transcript levels of genes associated with aliphatic GS biosynthesis pathway were either not changed or slightly down-regulated upon B. brassicae infestation. Conversely, several genes associated with indolyl GS biosynthesis and HIGH INDOLE GLUCOSINOLATE (HIG1) transcription factor controlling indolyl GS synthesis (Gigolashvili et al. 2007) were induced (Table 1). Genes early in the biosynthesis pathway, such as tyrosine aminotransferase (TAT3), anthranilate synthases (ASA1, ASB1 and ASB beta) and indole-3-glycerol phosphate synthase (IGPS), were induced at 6 or 12 hpi, while genes coding for cytochrome P450 enzymes (CYP79B2, CYP79B3 and CYP83B1) were up-regulated at the later time points. The induction of HIG1 and CYP79B2 was confirmed by qRT-PCR (Fig. 3). The tryptophane-derived indolyl GS and camalexin share the very first part of the biosynthesis pathway with the synthesis of phenylalanine-derived flavonoids and lignins (Fig. 4). The common steps lead from tyrosine to prephenate, which is a branching point for the biosynthesis of tryptophane and phenylalanine. Remarkably, in contrast to indolyl GS and genes coding for enzymes involved in camalexin biosynthesis, several genes placed downstream from prephenate and dedicated either to flavonoids or lignin biosynthesis, such as 4-coumarate-CoA ligases (4CL3 and 4CL10), chalcone synthases (CHSs), chalcone-flavanone isomerases (At1g53520 and At5g05270), flavonol synthase (FLS), probable caffeoyl-CoA O-methyltransferase (CCoAMT), cinnamyl-alcohol dehydrogenase (CAD1) and isoflavone reductase related (PCB2), were down-regulated, especially at 48 hpi (Table 1). The repression of the phenylalanine-connected branch presumably spares tyrosine-derived resources, which can instead be directed into indolyl GS and camalexin biosynthesis.

Table 1.  Expression profiles of genes involved in the biosynthesis or regulation of biosynthesis of indolyl and aliphatic glucosinolates (GS), camalexin, flavonoids and lignin Thumbnail image of
Figure 4.

A simplified representation of transcriptional changes induced by Brevicoryne brassicae in the tyrosine, phenylalanine and tryptophan pathways. Arrow width indicates the predicted rate of flow of tyrosine-derived resources, as implied by the differential regulation of key genes involved in the flavonoid and lignin biosynthetic pathways, compared with genes involved in indolyl glucosinolates and camalexin biosynthesis. Names of genes that were found to be up-regulated in at least one of the time points of B. brassicae infestation are indicated in bold red, while down-regulated genes are indicated in green. For simplification, details of the biosynthetic steps are omitted. TAT3, tyrosine aminotransferase 3; 4CL3, 4-coumarate-CoA ligase 3; 4CL10, 4-coumarate-CoA ligase 10; CHSs, chalcone synthases; CFI, chalcone-flavanone isomerase; FLS, flavonol synthase; CCoAMT, caffeoyl-CoA O-methyltransferase; CAD1, cinnamyl-alcohol dehydrogenase; ASA1, anthranilate synthase component 1; ASB1, anthranilate synthase beta subunit 1; ASB beta, anthranilate synthase beta subunit; TSA1, tryptophan synthase beta chain 1; IGPS, indole-3-glycerol phosphate synthase; CYP79B2, cytochrome P450 79B2; CYP79B3, cytochrome P450 79B3; CYP83B1, cytochrome P450 83B1; CYP71A13, cytochrome P450 71 A13; CYP71A12, cytochrome P450 71 A12; PAD3, phytoalexin deficient 3 (CYP71B15).

To determine how the GS content was changing over the time of infestation, the levels of individual GS in the infested and control plants were measured (Fig. 5). For metabolic profiling, the same plant material as for transcriptional analysis was used. Over the 48 h infestation period analysed, we noticed a trend of gradually declining quantities of total aliphatic GS in the infested plants. After 48 h, the total aliphatic GS level was significantly lower in the B. brassicae-attacked plants than in the aphid-free controls (Fig. 5f). This reduction was mainly because of a decline in the level of 3-hydroxypropyl GS (3HP) and its precursor, 3-methylsulfinylpropyl GS (3MSP) (Fig. 5d, e). The observed greater decrease in 3HP than in 3MSP levels is consistent with the down-regulation of the AOP3 gene coding for a 2-oxoglutarate-dependent dioxygenase, which catalyses the inversion of 3MSP to 3HP. In contrast, 7-methylsulfinylheptyl GS (7MSH) accumulated at 48 hpi (Fig. 5d). Conversely to aliphatic GS, we did not observe any trend of declining of the total indolyl GS level. At 6 hpi, the total indolyl GS level was significantly higher in the infested plants, but the only individual GS whose level statistically significantly differed between the control and the attacked plants was 1-methoxyindol-3-ylmethyl GS (1MI3M) (Fig. 5a). At 12 and 24 hpi, the total indolyl GS levels were not different from the untreated plants, whereas at 48 hpi, the level was slightly but not significantly increased (Fig. 5f). Despite the absence of a significant change in the total amount of indolyl GS, infestation with B. brassicae resulted in the significant accumulation of indol-3-ylmethyl (I3M) at 12 hpi (Fig. 5b), and 4-methoxyindol-3-ylmethyl (4MI3M) at 48 hpi (Fig. 5d). Noteworthy, there was a clear shift from an elevated level of I3M to an accumulation of 4MI3M between the 12 and 48 h of B. brassicae attack, suggesting the conversion of the I3M excess to 4MI3M. Such a change could possibly be the result of an increase in enzymatic activity leading to the addition of a methoxy group to I3M (Kim & Jander 2007).

Figure 5.

Glucosinolate content in rosette leaves of Brevicoryne brassicae-infested Ler plants in comparison with aphid-free controls at 6 h (a), 12 h (b), 24 h (c) and 48 h (d) after infestation. (e) amount of 3-hydroxypropyl glucosinolate in infested versus control plants 6, 12, 24 and 48 h after infestation. (f) total amount of methionine-derived (aliphatic) and tryptophan-derived (indolyl) glucosinolates expressed as a percentage of the levels in control plants. All bars represent averages (+SD) for at least three biological replicates, *P < 0.05, **P < 0.01 as revealed by a two-tailed Welch's t-test. Abbreviations of glucosinolate side chains: 3MSP, 3-methylsulfinylpropyl; 7MSH, 7-methylsulfinylheptyl; 8MSO, 8-methylsulfinyloctyl; I3M, indol-3-ylmethyl; 4MI3M, 4-methoxyindol-3-ylmethyl; 1MI3M, 1-methoxyindol-3-ylmethyl; and 8MTO, 8-methylthiooctyl.

Infestation induces accumulation of camalexin

Interestingly, three genes encoding enzymes involved in camalexin biosynthesis or its regulation [two cytochrome P450 enzymes (PAD3/CYP71B15 and CYP71A13) and triacylglycerol lipase (PAD4)] were strongly induced from the first time point of infestation, and their induction gradually increased during the assay period (Table 1). The changes in the transcript level of PAD3 were confirmed with qRT-PCR (Fig. 3). It should also be noted that the gene encoding CYP71A12, which is structurally similar to CYP71A13, had a similar expression pattern as PAD3/CYP71B15 and CYP71A13.

Because all known genes implicated in the camalexin biosynthetic pathway were highly induced as revealed by the microarray and qRT-PCR data, we decided to measure the actual levels of camalexin in the plant material previously used for transcriptional profiling. At 6 and 12 hpi, camalexin levels in the leaves attacked by aphids were unchanged. A substantial accumulation was first noticeable at 48 hpi (Fig. 6). To estimate the possibility of further increase in camalexin synthesis during prolonged infestation period, we also assessed the level at 72 hpi. On average, the accumulation at 72 hpi was 48% higher than at 48 hpi.

Figure 6.

Accumulation of camalexin in leaves of Ler after 6, 12, 24, 48 and 72 h lasting infestation with Brevicoryne brassicae in comparison with aphid-free controls. Bars represent averages (+SD) of four biological replicates; stars indicate P < 0.01 as revealed by a two-tailed Welch's t-test.

Camalexin influences aphid fitness

The biological relevance of camalexin-connected defences in ArabidopsisB. brassicae interactions was evaluated in aphid fitness experiments. We assessed the asexual fecundity of aphids on camalexin-deficient pad3-1 mutants in comparison with Col-0 WT plants (Fig. 7). Because only two first instar nymphs were placed on each plant in the beginning of the experiment and the aphids were allowed to move freely between different rosette leaves, we could not expect any significant accumulation of camalexin in connection to aphid feeding during the early phase of the experiment. To ensure a considerable difference in camalexin levels during the fitness experiment and its uniform accumulation in all WT plants’ leaves, both WT and pad3-1 plants were exposed for 5 min to UV-B light prior to infestation. UV-B treatment was previously reported to induce the accumulation of camalexin in Arabidopsis WT plants (Mert-Turk et al. 2003). Although this kind of treatment can induce a wide range of defensive processes, the only predicted difference in reaction to UV-B light between WT and pad3-1 plants was the ability to accumulate camalexin, as PAD3 encoding an enzyme catalysing the last step of camalexin synthesis is knocked out in pad3-1 mutant plants. The accumulation of camalexin after UV-B treatment in WT versus pad3-1 plants during the period of aphid fitness experiment is shown in Supplementary Fig. S1. The aphid fecundity on camalexin-accumulating WT plants was significantly lower than on pad3-1 mutants (Fig. 7), suggesting that camalexin does have an impact on aphid fitness.

Figure 7.

Asexual fecundity of Brevicoryne brassicae on wild-type (WT) and pad3-1 plants. WT Col-0 control and pad3-1 plants were pretreated with UV-B light prior to the experiment to induce camalexin accumulation. The two first instar nymphs were introduced to each of the plants, and three plants were kept in the same cage. Twenty-four WT plants (eight cages) and 24 pad3-1 plants (eight cages) were used in the experiment. Bars represent an average number (+SD) of aphid progeny per cage after 13 d. A star indicates P < 0.01 as revealed by a two-tailed Wilcoxon rank sum test.

Senescence-associated genes (SAGs) and induction of HR

When the density of an aphid colony established on an individual plant reaches a certain level, the shortage of metabolites consumed by the aphids can become a serious problem, impeding plant development or even leading to plant death. Premature leaf senescence was proposed as being a plant defence strategy to prevent this problem (Pegadaraju et al. 2005). By redirecting the flow of nutrients from heavily attacked leaves to other aphid-free organs, infested plants possibly rescue resources that otherwise would be consumed by aphids. Although no visual symptoms of senescence were observed up to 48 h of B. brassicae attack, transcriptional profiles of several SAGs were changed, suggesting that the induction of senescence starts already during the first 48 h of infestation (Fig. 1 & Supplementary Table S2). PHYTOALEXIN DEFICIENT 4 gene (PAD4), which is crucial for premature senescence (Pegadaraju et al. 2005), and several genes connected to senescence (SAG21, SAG13, SAG101, SEN1, SRG1, YLS5, YLS9, YSL3, SIRK, TRX-H-5, DIN6, ACD11-like and AOX1D) were induced. Some transcripts revealed an up-regulation already at 6 hpi, but all were strongly induced at 48 hpi. This substantial increase in intensity of changes indicates a progressive development of premature senescence with the time of infestation. The changes in the expression profile of SAG21 were additionally confirmed with qRT-PCR (Fig. 3).

Four genes coding for cysteine-rich receptor-like protein kinases (CRK5, CRK6, CRK10, CRK11 and DUF26-family protein) and a number of transcripts of their possible interaction partners, kinase-associated type 2C protein phosphatases, were up-regulated at 24 and 48 hpi (Supplementary Table S2). Interestingly, the same CRKs were identified as targets of WRKY transcription factors (Du & Chen 2000). The overexpression of CRK5 increased the plant resistance to the bacterial pathogen Pseudomonas syringae by enhancing the response of defence-related genes, especially PR1, and triggered HR-like cell death, probably in an SA-independent manner (Chen, Du & Chen 2003). The observed induction of CRKs during B. brassicae attack indicates their involvement in resistance towards aphids.


Model of plant–aphid interactions

Based on the results from this and other published studies, we propose a general model that describes some of the main aspects of plant–aphid interactions (Fig. 8). In our model, both the attacker and the host are locked together in a biological arms race where the actions of one trigger counteraction of the other. Plant responses are largely triggered by two distinct processes: (1) the perception of B. brassicae attack and (2) cell signalling giving rise to defence responses. The content of oral secretions may play a role in the recognition of the attacker by plant (De Vos, Kim & Jander 2007); however, the disruption of cell wall and membrane integrity, either by enzymes introduced through the penetrating stylet or simply by mechanical damage following the incursion, is likely to be the first factor that triggers the plant response. Calcium and ROS signalling are initiated rapidly following attack, and together with WRKY, transcription factors play a key role in signal transduction and activation of defensive processes. Signalling molecules, as well as some of defence-related products, can be transported from companion cells to SE, where they can fulfil their function. At the same time, aphids try to prevent the plant defensive responses by (1) specialized feeding strategy avoiding a membrane damage-provoked stimulation and (2) injecting Ca2+-binding proteins into the punctured cells, thus hindering calcium-signalling-induced sealing of the sieve pores and protein clogging inside the stylet. Despite all the aphids’ effort, the plant is able to perceive the attack and responds with a broad range of gene expression that is reflected in metabolic changes.

Figure 8.

Model of plant–aphid interactions at cellular level. An oxidative burst at the place of injury generates oxygen species (ROS), which can lead to induced cell death. Synthesis of H2O2 activates a protein MAPK phoshorylation cascade, which triggers the transcription of plant defence genes. As neither transcription nor translation takes place in sieve element (SE), products are transported into SE from adjacent companion cells via proteins with chaperone activity. Puncturing of a cell membrane causes a leakage of the cell content into the intercellular space and relaxation of the cell turgor pressure, resulting in the opening of mechanosensitive Ca2+ membrane channels. Rising concentration of Ca2+ in SE initializes protein plugging and callose sealing of the sieve pores. Plugged pores would prevent the connection of pierced SE with adjacent SEs, blocking the phloem flow. The pressure drop and clogging of proteins inside the aphid's food canal would make feeding impossible. Active components of injected saliva bind Ca2+ thereby preventing Ca2+-signalling and protein clogging. In addition, salivary sheets, enveloping the penetrating stylet, seal the holes, minimizing the leakage into intercellular spaces and reducing the flux of Ca2+ and hypersensitive responses.

The role of changes in redox conditions and calcium-signalling as the very first factors triggering plant response

The coordinated regulation of ROS-induced genes observed at the first time point of infestation indicates that the generation of ROS is one of the first events upon B. brassicae attack. Degradation of linolenic acid induced by oligogalacturonides, which are released from cell walls by polygalacturonases present in aphid saliva (Miles 1999), may lead to the accumulation of JA but also activates the synthesis of ROS (Gatehouse 2002). In addition to having a direct adverse effect on insects (Bi & Felton 1995), ROS are known elicitors of a multitude of induced defences (Orozco-Cardenas, Narvaez-Vasquez & Ryan 2001). Hydrogen peroxide is one of the most important radical messengers. Mutant plants, which are not able to generate H2O2 during pathogen challenge, are compromised in induced resistance (Torres et al. 2002). Through the activation of the MAPK cascade, the rapidly synthesized H2O2 initiates global expression changes causing the accumulation of several gene transcripts (Neill et al. 2002b; Li et al. 2006b). Furthermore, some genes can be directly regulated by H2O2 concentration via H2O2 responsive promoters (Desikan et al. 2001). Because different types of stimuli can activate specific ROS-generating enzymes (Neill et al. 2002a), it is not surprising that not all genes involved in ROS synthesis were consistently co-regulated upon aphid attack in our study and the previously published studies (Moran et al. 2002). Interestingly, up-regulated genes that could possibly be involved in H2O2 synthesis upon B. brassicae attack (copper amine oxidases, germines and RBOHD) were not found to be induced before 24 hpi. Challenging with B. brassicae, however, caused the induction of several genes encoding proteins involved in the detoxification of ROS-generated products already at 6 hpi, indicating that changes in redox conditions happened during the first hours of infestation. These changes could result from rapidly occurring, transcription-independent activation of ROS-generating enzymes. Taken together, our data suggest that Arabidopsis attacked by B. brassicae tends to maintain a fine balance between generation and detoxification of ROS. Genes coding for enzymes involved in H2O2 synthesis (copper amine oxidases, germins and RBOHD) are up-regulated, and at the same time, the induction of endogenous radical scavenger transcripts keeps the ROS levels under control, preventing a massive accumulation of radicals. In addition, high levels of transcription of calcium-dependent phospholipid-binding proteins BAP1, BAP2 and their interacting partner BON repress H2O2-induced cell death. No accumulation of H2O2 could be detected at aphid feeding sites with DAB staining, but we cannot eliminate a possibility that the H2O2 accumulation was only transient. In fact, very small changes in H2O2 levels maybe enough to activate the transcription of a wide range of defence-related genes, as Arabidopsis cells are sensitive to millimolar changes in H2O2 concentration (Neill et al. 2002b).

When damage occurs in a sieve element (SE) membrane, it results in a leakage of phloem sap. As a consequence, mechanosensitive membrane Ca2+ channels open, and subsequent changes in intracellular Ca2+ concentration initiate protein plugging and callose sealing (Will & van Bel 2006) (Fig. 8). As a result, the pores in sieve plates connecting the punctured cell with the adjacent SEs are blocked, and the damaged SE no longer supports phloem transport. This mechanism, preventing a possible loss of phloem sap upon SE injury, is beneficial for plants, but highly unfavourable for phloem feeders. The normally high pressure in the phloem is critical for the passive acquisition of phloem sap by aphids. Moreover, the clogging of proteins inside the stylet's food canal would make feeding impossible. Aphids seem to actively overcome this problem by regular injections of salivary secretions containing Ca2+-binding proteins into the SE they feed from (Will et al. 2007) (Fig. 8). By doing so, they are able to prevent increases in Ca2+ concentration to the level that would trigger protein clogging and occlusion of SE. However, this strategy is not sufficient to keep intracellular Ca2+ at low levels. The mechanical damage of cell membranes during stylet penetration apparently resulted in Ca2+ influx that overwhelmed the Ca2+-binding capacity of salivary components, as reflected in our experiment by the early induction of many calcium-binding proteins. Calcium signalling seems thus to be very important in early signal transduction in response to B. brassicae attack. Remarkably, calcium and ROS signalling are linked by the calcium-dependent activation of NADPH oxidases, which are involved in H2O2 synthesis (Sagi & Fluhr 2001).

A multitude of responses are initiated from the first hours of B. brassicae attack and build-up with time

Despite the limited tissue damage caused by aphids and a feeding strategy aimed to prevent the activation of plant defence responses, Arabidopsis is able to perceive B. brassicae attack and responds with a broad range of transcriptional changes, activated already at 6 hpi. Investigation of molecular responses to M. persicae conducted by Couldridge et al. (2007) resulted in finding only two genes whose expression was changed at 2 hpi. Neither of these two genes had similar regulation at 6 hpi with B. brassicae. In contrast, our study revealed 196 genes induced or repressed twofold or more at 6 hpi. It seems rather unlikely that the main reason for such dissimilarity could be the difference in aphid species used in infestation experiments. Either the larger-scale gene expression changes are initiated some time after 2 hpi, or for some experimental or technical reasons, Couldridge and co-workers were not able to identify most of the genes responding to aphid attack. Similarly, they found only 25 transcripts responding at 36 hpi, which is again quite a small number in the light of other research (De Vos et al. 2005; Kuśnierczyk et al. 2007). However, 20 of those genes had also changed expression upon 48 h of infestation with B. brassicae, indicating that the Arabidopsis responses (regardless of the ecotype) to the two aphid species are similar, as previously reported (Kuśnierczyk et al. 2007).

The majority of plant defence responses is presumably located close to the aphid feeding sites. Induction of genes, such as PR-1 or CYP79B2, was previously shown to be strongest at stylets’ insertion places (De Vos et al. 2005; Kuśnierczyk et al. 2007), and this may be the case for many other genes. Consequently, it can be anticipated that also secondary metabolites accumulate locally, and that their concentration diffuses gradually towards the uninfected tissue, as observed in ArabidopsisBotrytis interactions (Kliebenstein, Rowe & Denby 2005). In our experiment, the intensity of transcriptional changes, especially of genes connected to senescence, synthesis of proteins with anti-insect activities, and biosynthesis of indolyl GS and camalexin, showed gradual increases with the time of B. brassicae infestation (Fig. 1). Because whole leaves were harvested for transcriptional and metabolic profiling, the resulting data presumably represent an average of ‘induced’ and ‘non-induced’ areas. When aphids move to the new feeding places, they induce defence responses at previously untouched leaf parts. The proportion of ‘induced’ leaf areas will thereby increase with time and the number of aphid probes, influencing the overall magnitude of detected changes.

Senescence, also referred to as programmed cell death, is a common protective strategy against bacterial and fungal pathogens. Recent research elucidated the importance of premature senescence, controlled by the PAD4 gene, in resistance to M. persicae (Pegadaraju et al. 2005, 2007). Synchronized up-regulation of PAD4 together with a number of other genes associated with senescence and the dynamic progress of senescence-connected responses confirm that the acceleration of senescence as an ecologically favourable form of defence is also relevant in ArabidopsisB. brassicae interactions. The slight suppression of genes involved in chlorophyll biosynthesis, cell cycle and cell division, especially at 48 hpi (data not shown), is an additional indication of initiated leaf senescence. PAD4, however, is also important for the synthesis of camalexin as the pad4 mutant has a reduced ability to accumulate camalexin upon pathogen attack. The function of PAD4 in limiting phloem sap ingestion and the aphids’ population growth rates, reported by Pegadaraju et al. (2007), could conceivably be connected to its role in camalexin induction (in addition to its involvement in senescence-related defences). The up-regulation of many genes involved in senescence is possibly governed by SA- or JA-signalling as it was shown for the SEN1 gene (Schenk et al. 2005) or WRKY53, which is a regulatory factor initiating senescence events by influencing the expression of SAGs (Hinderhofer & Zentgraf 2001; Miao et al. 2004).

The complex regulation of JA- and SA-signalling

The activation of SA-associated defence in response to aphids (Moran et al. 2002; De Vos et al. 2005) and silver leaf whitefly (Zarate, Kempema & Walling 2007) in Arabidopsis has been previously reported. Bioassays using mutant plants have shown that the SA-signalling pathway does not contribute to basal resistance to silver leaf whitefly. Two independent experiments assessing the fitness of M. persicae on NahG and npr1 mutants (which have reduced SA content or are impaired in SA signalling) resulted in somewhat contradictory results (Mewis et al. 2005; Pegadaraju et al. 2005). At the same time, accumulating evidence underlines the importance of an alternative JA-mediated defence against phloem feeders not only in Arabidopsis (Ellis, Karafyllidis & Turner 2002; Kuśnierczyk et al. 2007; Zarate et al. 2007) but also in other plant species (Kessler, Halitschke & Baldwin 2004; Gao et al. 2007). In the light of an antagonistic crosstalk between the SA- and JA-signalling pathways, Zarate et al. (2007) hypothesize that whiteflies can manipulate the plant-induced resistance and are able to suppress the effective JA defences by induction of the less efficient SA-signalling-based responses. A comparable model has been proposed for plant–aphid interactions (Zhu-Salzman, Bi & Liu 2005; De Vos et al. 2007), but lacks experimental support and remains to be proven.

Resent research has demonstrated the existence of a negative regulatory system in JA-signalling pathway, allowing for desensitization of the cell to rising levels of JA. In the proposed model, JAZ family proteins have a key function in targeting the JA-signal transduction messenger: COI1 protein for proteasome degradation (Chini et al. 2007; Thines et al. 2007). This regulatory mechanism seems to be activated at 6 h of B. brassicae attack. The early induction of genes coding for JAZ proteins during B. brassicae attack could explain the negative feedback loop causing a slight regression of JA-connected responses at 12 hpi, and the observed down-regulation of the JA-inducible VSP1 and VSP2 transcripts together with a slight decrease in the induction of PDF1 gene over time.

Changes in glucosinolate levels induced by B. brassicae attack are similar to those upon M. persicae infestation

The glucosinolate-myrosinase system of Brassicaceae is an important part of plant induced resistance (Mewis et al. 2006). Recently, the 4-methoxyindol-3-ylmethyl (4MI3M) glucosinolate was identified as an important component in Arabidopsis–aphid interactions revealing that not only the GS hydrolysis products, as previously anticipated, but also the intact GS can affect pest performance. Not only does 4MI3M accumulate in the aphid-infested tissue, but it also deters M. persicae feeding (Kim & Jander 2007). We observe a similar increase in the formation of 4MI3M after B. brassicae infestation. Interestingly, an SA treatment of Ler has been reported to specifically induce 4MI3M levels, while a simultaneous treatment with methyl jasmonate hinders the SA-mediated induction at 24 h, but not at 48 h after treatment (Kliebenstein, Figuth & Mitchell-Olds 2002). The up-regulation of SA-related genes indicates a possible elevation of SA levels in response to B. brassicae infestation, which could in turn influence the accumulation of 4MI3M at 48 h after attack. Furthermore, we notice a decrease in the total level of aliphatic GS, in particular 3HP, which is comparable to the results reported from M. persicae-infested plants (Kim & Jander 2007). In contrast to our findings, Mewis et al. (2006) observed an increase in short-chain aliphatic GS and did not see any change in indolyl GS levels in Col-0 plants attacked by B. brassicae. The dissimilarity between these and our results can be a consequence of the differences between the two experimental settings, for example, the Arabidopsis ecotype used, the age of studied plants and how long aphids had been allowed to feed before GS were measured. For example, the 3HP glucosinolate, which was mainly responsible for the drop in aliphatic GS level observed in Ler after B. brassicae attack, is not detectable in Col-0 ecotype at all (Kliebenstein et al. 2001). The changes we observed in the GS profiles correlated well with the changes in the expression of genes encoding proteins involved in GS biosynthesis or its regulation. Together, these and previously published results (Kim & Jander 2007; Kuśnierczyk et al. 2007) demonstrate that the transcriptional and metabolic reprogramming of the GS system is very similar in response to the generalist and specialist aphid attack. Whether these responses have any adverse effect on B. brassicae, which because of its specialization should be adapted to GS-related defences, needs to be further investigated.

A novel role of camalexin in response to B. brassicae attack

The role of camalexin in plant responses to aphid attack has not been reported previously. We found that several genes, known to be involved in camalexin synthesis, were highly up-regulated in a concerted fashion already at 6 hpi. Particularly strong induction of CYP71A13 and PAD3 (CYP71B15), which are placed downstream of the branching point of camalexin and indolyl GS biosynthetic pathways (Fig. 4), could suggest an enhanced synthesis of camalexin by channelling a great proportion of indole acetaldoxime into the camalexin-producing branch. The co-expression of CYP71A12 with CYP 71A13 and PAD indicates a possible involvement of this gene in camalexin biosynthesis. Interestingly, camalexin did not accumulate during the first phase of B. brassicae infestation despite the high induction of PAD3 and CYP71A13, and the major changes in camalexin levels were first detected at 48 hpi. Remarkably, despite a high up-regulation of genes involved in its biosynthesis, the accumulation of camalexin detected after induction by B. brassicae attack was much lower than previously reported during infection with pathogenic bacteria or fungi (Kliebenstein et al. 2005; Nafisi et al. 2007; Schuhegger, Rauhut & Glawischnig 2007). There are several possible explanations for this observation. One is a simple assumption that the regulation of camalexin synthesis happens at the post-transcriptional level. This hypothesis is, however, not supported by any earlier reports attempting to correlate the accumulation of camalexin with the expression levels of genes involved in the camalexin production. In particular, CYP71A13 and PAD3 seem to be induced with similar kinetics as camalexin accumulation (Nafisi et al. 2007; Schuhegger et al. 2007). Another possibility could be a partial biotransformation of camalexin by the enzyme(s) injected by the feeding aphids with watery saliva. It has been previously demonstrated that some phytopathogenic fungi are capable of detoxifying camalexin (Pedras & Khan 2000), and B. brassicae-induced modification of camalexin is an open possibility. Given that B. brassicae is sensitive to high camalexin concentrations, as revealed by the fitness experiments, such ability would be beneficial for aphid fitness. Alternatively, camalexin may not be an end product and could be further metabolized to other metabolites by endogenous enzyme(s), induced upon infestation with aphids, and not induced with bacterial or fungal pathogens. We cannot rule out the possibility that the metabolite(s) formed downstream of camalexin may be a part of defence arsenal effecting aphid performance. Interestingly, Pegadaraju et al. (2005) did not see any effect of pad3-1 mutation on generalist aphid, M. persicae, fitness and thus concluded that this species is not affected by camalexin. Further research is needed to unravel the details of the defensive mechanism involving camalexin or related compounds in the context of aphid attack.


A multitude of transcriptional responses is activated at 6 h after B. brassicae attack. ROS and calcium signalling are initiated rapidly after infestation, and together with WRKY transcription factors play a key role in signal transduction and induction of plant defences. Extended periods of aphid probing cause a stronger regulation of transcriptional changes and activate more genes, whose number is finally doubled at 48 hpi. Detailed studies of selected genes, including analysis of mutant plants and assessment of their susceptibility to infestation, are needed to estimate the relative importance of particular genes contribution to plant defence against aphids. The changes in transcript levels correspond well to the altered profiles of secondary metabolites including aliphatic and indolyl GS and camalexin. The substantial change in camalexin levels, first detected at 48 hpi, is, however, smaller than previously reported upon bacterial and fungal infections. These observations suggest the existence of either unknown factors regulating camalexin synthesis, unknown metabolic or catabolic steps in the camalexin turnover or aphid-induced modifications of the product. Future research will unravel which of the proposed scenarios is applicable. B. brassicae fitness is impaired by camalexin accumulation as revealed by assays comparing aphid fecundity on WT and camalexin-deficient pad3 mutants.

The microarray data described in this study are available in ArrayExpress Database located at European Bioinformatics Institute (


The authors would like to thank Professor Jane Glazebrook (University of Minnesota) for the pad3-1 seeds, the camalexin standard, and the protocol for camalexin isolation and quantification; Professor Thor Bernt Melø for the introduction to fluorometry; and Hong Tran for the introduction to HPLC analysis. Torfinn Sparstad and Marianne Nymark are gratefully acknowledged for their excellent technical assistance with microarray hybridizations, Jenny Bytingsvik for the extraction of GS, and Wacek Kuśnierczyk for helping with the editing of the figures. We also thank Ralph Kissen and other members of the Cell Molecular Biology Group for the insightful comments on the manuscript. This work was supported by the Biotechnology and Functional genomics (FUGE) programmes of the Norwegian Research Council through grants NFR 159959, 164583 and 151991.