Myzus persicae (green peach aphid) feeding on Arabidopsis thaliana induces a defence response, quantified as reduced aphid progeny production, in infested leaves but not in other parts of the plant. Similarly, infiltration of aphid saliva into Arabidopsis leaves causes only a local increase in aphid resistance. Further characterization of the defence-eliciting salivary components indicates that Arabidopsis recognizes a proteinaceous elicitor with a size between 3 and 10 kD. Genetic analysis using well-characterized Arabidopsis mutants shows that saliva-induced resistance against M. persicae is independent of the known defence signalling pathways involving salicylic acid, jasmonate and ethylene. Among 78 Arabidopsis genes that were induced by aphid saliva infiltration, 52 had been identified previously as aphid-induced, but few are responsive to the well-known plant defence signalling molecules salicylic acid and jasmonate. Quantitative PCR analyses confirm expression of saliva-induced genes. In particular, expression of a set of O-methyltransferases, which may be involved in the synthesis of aphid-repellent glucosinolates, was significantly up-regulated by both M. persicae feeding and treatment with aphid saliva. However, this did not correlate with increased production of 4-methoxyindol-3-ylmethylglucosinolate, suggesting that aphid salivary components trigger an Arabidopsis defence response that is independent of this aphid-deterrent glucosinolate.
In their natural environment, plants are continuously challenged by unfavourable conditions, including attack by pathogenic microorganisms and insect herbivores. Pre-existing physical barriers, such as hairs, thorns and spines, are an essential first line of defence against herbivory. In addition, plants rely heavily on chemical defences to fend off herbivores (Karban & Baldwin 1997). The evolutionary chemical arms race between host and attacker has rendered plants resistant to many attackers, but also has led to specialization by insect herbivores (Schoonhoven, Van Loon & Dicke 2005).
Plants need to recognize attacker-specific factors to induce a timely response that will minimize fitness losses. Previously identified insect-derived elicitors of plant defenses include fatty acid amino acid conjugates of oral secretions from Spodoptera exigua (Alborn et al. 1997) and Manduca sexta (Halitschke et al. 2001), fragments of the plant's chloroplastic ATP synthase produced by Spodoptera frugiperda (Schmelz et al. 2006, 2007), and disulfooxy fatty acids in the regurgitant of the grasshopper Schistocerca americana (Alborn et al. 2007). Genetic evidence suggests that classic gene-for-gene interactions, similar to those involved in pathogen resistance, confer resistance to some insect herbivores, in particular phloem-feeding insects (Klingler et al. 2001, 2005; Liu et al. 2001; Liu, Smith & Gill 2002; Gao et al. 2008). Tomato Mi1-2, an NB-LRR gene, is required to trigger resistance against phloem feeders, including aphids (Rossi et al. 1998), white flies (Nombela, Williamson & Muniz 2003) and nematodes (Milligan et al. 1998). However, to date, specific elicitors have not been identified from any phloem-feeding insects or nematodes.
The generalist insect herbivore, Myzus persicae (green peach aphid), has a very broad host range (Blackman & Eastop 2000). By feeding from a single-cell type, the phloem sieve element, the aphid's feeding strategy clearly differs from that utilized by chewing insects and those such as thrips that feed from epidermal cell contents. Aphids use their slender stylets to manoeuvre between cells, until they reach a sieve element. Once contact with the phloem is initiated, aphids can feed for a prolonged period of time from a single sieve element. Salivation is crucial for successful colonization by aphids, and saliva is thought to play a major role in the aphid virulence (Tjallingii 2006). A gelling, or sheath, saliva physically protects the stylet, and its enzymes might also play a role in degradation of the middle lamella by aphids. Once prolonged feeding has been established, aphids repeatedly inject watery saliva into the sieve tubes. In addition to having a protective function in the form of polyphenol oxidases and other detoxifying enzymes, aphid saliva likely contains factors that facilitate uptake of phloem sap (Miles 1999; Cherqui & Tjallingii 2000; Tjallingii 2006; Will & van Bel 2006). Although phloem sieve elements clog quickly after mechanical wounding, this does not happen when aphids puncture the sieve element, suggesting that, similar to the anti-clotting saliva of blood-feeding insects, aphid saliva possesses anti-clogging properties.
Although aphid feeding has been linked to increased levels of salicylic acid (SA) and expression of SA-responsive genes (Moran & Thompson 2001; Moran et al. 2002; De Vos et al. 2005), jasmonate (JA)-induced defences have been implicated in plant defence responses and control of aphid population growth. Exogenously applied JA significantly reduced M. persicae fecundity on Arabidopsis (Ellis, Karafyllidis & Turner 2002). Conversely, mutant coi1-16 plants, which have a reduced JA response but also harbour a pen2 mutation (Westphal, Scheel & Rosahl 2008), allowed increased aphid reproduction compared with wild-type plants (Ellis et al. 2002). It has been hypothesized that aphids and Bemisia tabaci (white flies) actively suppress effective JA defences by induction of the antagonistic SA pathways (De Vos, Kim & Jander, 2007; Goggin 2007; Zarate, Kempema & Walling 2007). Several microarray studies have been undertaken to study gene expression changes upon aphid infestation. A large number of expression changes were observed after 48 and 72 h of M. persicae feeding (De Vos et al. 2005). Similarly, infestation of three Arabidopsis accessions for 75 h with M. persicae and Brevicoryne brassicae (cabbage aphid) induced a multitude of transcriptional changes, including the up-regulation many general stress-related genes (Kusnierczyk et al. 2007). However, very few transcriptional changes were detected after 2 and 36 h of M. persicae feeding in another study (Couldridge et al. 2007). Kusnierczyk et al. (2008) monitored the transcriptional changes induced by B. brassicae feeding in a time course experiment, which suggests the involvement of reactive oxygen species, SA and JA in the regulation of defence responses. The large differences observed between the transcriptional analyses mentioned earlier are likely caused by a combination of environmental factors, aphid dosage and timing of tissue harvest (De Vos et al. 2007).
Recent advances in genomics and proteomics have resulted into a renewed interest in aphid saliva. Large-scale cDNA sequencing has identified genes expressed specifically in the salivary glands of aphids (Ramsey et al. 2007). Moreover, protein sequencing has identified numerous proteins of unknown function in M. persicae and Acyrthosiphon pisum (pea aphid) salivary secretions (Harmel et al. 2008; Carolan et al. 2009). However, to date, there has been relatively little functional analysis of aphid salivary proteins. RNAi-mediated knock-down of C002, one of the most abundant A. pisum salivary proteins, reduced aphid reproduction on Vicia faba (broad bean) but not artificial diet, suggesting that this protein is essential for aphid feeding on host plants (Mutti et al. 2008). Will et al. (2007) found calcium-binding proteins in the saliva of Megoura viciae (vetch aphid) on V. faba. Legumes contain forisomes, which are involved in occlusion of the sieve element upon injury. Calcium has been proposed as an important signalling component in this phloem-clogging response, and in vitro calcium capture by salivary proteins prevents the forisomes conformational changes that result in phloem clogging (Will et al. 2007).
Here, we used alternative strategies for the functional analysis of aphid salivary components. Whereas other studies have focused on the functionality of individual salivary proteins through RNAi knockdown (Mutti et al. 2006, 2008), we study the in vivo effect of the entire composition of aphid saliva on Arabidopsis gene expression and resistance to M. persicae, followed by the initial steps toward identifying M. persicae salivary factors that trigger plant defence.
MATERIALS AND METHODS
Plant growth and insect rearing
Seeds of wild-type Arabidopsis (A. thaliana) land race Columbia-0 (Col-0) were obtained from the Arabidopsis Biological Resource Center (http://www.arabidopsis.org). Mutants were kindly supplied by C.M.J. Pieterse (Utrecht University, the Netherlands; jar1-1, sid2-1, NahG transgenic and ein2-1), G. Howe (Michigan State University, MI, USA; acx1 acx5), J. Glazebrook (University of Minnesota, MN, USA; pad4-1) and S. Sommerville (University of California, Berkeley, CA, USA; pen2-1). Seeds were kept in 0.1% Phytagar (Invitrogen, Carlsbad, CA, USA) for 24 h at 4 °C prior to planting on Cornell mix (Landry, Chapple & Last 1995) with Osmocote fertilizer (Scotts). Plants were grown in Conviron growth chambers in 20 × 40 cm nursery flats at a photosynthetic photon flux density of 200 µmol m−2 s−1 and a 16 h photoperiod. The temperature in the chambers was 23 °C and the relative humidity was 50%. Unless stated differently, plants were grown for 3 weeks and used in experiments before flowering.
All experiments were conducted with a tobacco-adapted red lineage of M. persicae that was obtained from S. Gray (USDA Plant Soil and Nutrition Laboratory, Ithaca, NY, USA). Aphids were raised on cabbage (Brassica oleracea var. Wisconsin Golden Acre; Seedway) with a 16 h day (150 µmol m−2 s−1 at 24 °C) and an 8 h night (19 °C) at 50% relative humidity.
Individual wild-type plants were infested with 50 aphids or kept free of aphids for 48 h. After removal of the aphids, one adult aphid was placed on each plant and its offspring number was determined after 7 d. To determine if aphid-induced resistance is effective in local or distal leaf tissue, movement of 50 aphids was restricted onto one leaf (local tissue) with a cage, whereas a distal (systemic tissue) leaf was caged without any aphids as a within-plant comparison. Control plants also had two cages, both without aphids (Supporting Information Fig. S1a). After 48 h of feeding, all aphids were removed and replaced by a single adult aphid per cage, and the number of offspring was counted after 7 d.
Fifty aphids were allowed to feed from 50 µL artificial diet, containing sucrose and amino acids (Kim & Jander 2007; Supporting Information Fig. S1b) between two layers of Parafilm (Pechiney Plastic Packaging, Chicago, IL, USA). After 24 h, artificial diet from aphid and control (0 aphid) treatments was collected and infiltrated into leaves of intact Arabidopsis plants using a 1 mL syringe without the needle (Supporting Information Fig. S1c). Another 24 h later, infiltrated leaves were infested with a single adult aphid and the number of offspring was determined 10 d later. A similar experimental set-up was used upon fractionation of the saliva-containing diet. All saliva-containing diets were pooled upon collection of the diet from 30 individual diet cups. Subsequently, the diet was fractionated into separate fractions using an Amicon Microcon spin-column filter according to the manufacturer's guidelines (YM-10 and YM-3; Millipore, Bedford, MA, USA). The number of offspring from a single aphid was assessed in leaves infiltrated with total diet, <10, >10, <3, or 3< × >10 kD fractions, as well as the corresponding controls from diet that had not been fed upon by aphids. In another experiment, aphid fecundity was determined after 10 d on leaves infiltrated with untreated, boiled (30 min at 95 °C) and protease K (100 µg; 30 min at 37 °C) treated <10 kD fractions.
In order to get insight into the number of potential protein elicitors in M. persicae saliva, we collected saliva from ∼5000 aphids feeding on artificial diet. Diet with salivary components was collected after 24 h and fractionated as described earlier. M. persicae salivary proteins were coupled with a fluorescent Cy2 dye (CyDye flour) according to the manufacturer's instructions (GE Healthcare Bio-Sciences Corp, Piscataway, NJ, USA). A 10% NuPAGE Novex Bis-Tris Mini gel was run on an XCell SureLock Mini Cell system (Invitrogen, Carlsbad, CA, USA) at 200 V for 35 min. The NuPAGE MES SDS Running Buffer contained 200 mL 1× NuPAGE MES SDS with 500 µL of the NuPAGE antioxidant (upper buffer). The lower buffer existed of 600 mL of 1× NuPAGE MES SDS Running Buffer. Scanning of the 1D protein gel was done at 570–650 V using a Typhoon 9400 fluorescent scanner (GE Healthcare Bio-Sciences Corp).
To test for direct effects of saliva on aphid performance, diet was collected from control and aphid-fed cups and transferred to a new sachet of Parafilm. One adult aphid was allowed to feed from these cups and the number of offspring was assessed after 3 days (n = 25).
Sample preparation for affymetrix ATH1 genechip arrays
Fifty aphids were allowed to feed from 50 µL artificial diet, containing sucrose and amino acids (Kim & Jander 2007) between two layers of Parafilm (Pechiney Plastic Packaging). After 24 h, artificial diet from 20 aphids and control (0 aphid) diet cups was collected and infiltrated into leaves of intact Arabidopsis plants using a 1 mL syringe without the needle. Plants for control diet and aphid saliva containing diet were grown in the same pot to allow for a paired comparison. Eighteen leaves (three leaves from six plants) treated with control and aphid saliva containing diet were harvested and immediately frozen in liquid nitrogen. This experiment was repeated three times to function as independent biological replicates. RNA was extracted using the Qiagen Plant RNeasy kit (Qiagen, Valencia, CA, USA). RNA quality and quantity was assessed with an Agilent BioAnalyser 2100 (Agilent, Santa Clara, CA, USA), before the samples were processed by the Cornell Microarray facility. Whole genome gene expression profiling was done using Affymetrix ATH1 GeneChips. Briefly, plant mRNA was converted into cRNA, labelled and hybridized to the ATH1 GeneChips.
Microarray data analysis
Raw data from the microarrays was normalized at probe-level using gcRMA algorithm (Irizarry et al. 2003; Wu et al. 2004). The detection calls (present, marginal, absent) for each probe set was obtained using the Affymetrix GeneChip® Operating Software (GCOS) system (http://www.affymetrix.com/support/technical/datasheets/gcos_datasheet.pdf). Only genes with at least one present call across all the compared samples were used to identify differentially expressed genes. Significance of gene expression was determined using the LIMMA (Smyth 2004) program and raw P values of multiple tests were corrected using false discovery rate (FDR, Benjamini et al. 2001). Genes differentially expressed between aphid saliva treated and control plants were defined by the following criteria: FDR less than 0.05 from LIMMA analysis and changes greater than 1.5-fold. Genes showing significantly altered gene expression between leaves treated with control and aphid saliva containing diet were subjected to additional criteria to obtain robust sets of genes that are regulated by aphid saliva. We selected those genes that are significant different in expression between control and saliva treated leaves and that showed at least 1.5-fold change in expression (up or down; Supporting Information Tables S1 & S2, respectively). Micro-array data, including an overall project description, a matrix table and raw ATH1 CEL-files, have been submitted to the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE16497) and to the Nottingham Arabidopsis Stock Centre (NASC; http://affymetrix.arabidopsis.info/narrays/experimentpage.pl?experimentid=504).
RNA extraction and quantitative RT-PCR (Q-RT-PCR) analysis
Q-RT-PCR analysis was performed similarly to what has been described previously (De Vos et al. 2005). To remove genomic DNA, total RNA (2.5 µg) from independent experiments (biological replicates) was digested with Turbo DNaseTM (Ambion, Austin, TX, USA) according to the manufacturer's instructions. To check for genomic DNA contamination, a PCR reaction with primers designed based on intron sequences of TUB4 (At5g44340; Supporting Information Table S3) was carried out. Subsequently, DNA-free total RNA was converted into cDNA using oligo-dT20 primers, 10 mM dNTPs, and Clontech PowerScriptTM Reverse Transcriptase (Clontech, Mountain View, CA, USA) according to the manufacturer's instructions. Efficiency of cDNA synthesis was assessed by Q-RT-PCR using primers of the constitutively expressed gene UBI10 (At4g05320; see Supporting Information Table S3). Gene-specific primers were designed using Primer3 v. 0.4.0. (Rozen & Skaletsky 2000) for all Arabidopsis genes tested (see Supporting Information Table S3). Q-RT-PCR analysis was done in optical 384 well clear optical reaction plates and optical adhesive covers (Applied Biosystems, Forest City, CA, USA) with an ABI 7900HT Fast Real-Time PCR System (Applied Biosystems), using SYBR Green to monitor double-stranded (ds) DNA synthesis. Each reaction contained 1 µL of cDNA, 0.5 µL of each of the two gene-specific primers (10 pmol µL−1), and 7.5 µL of 2× SYBR Green PCR mix reagent (Applied Biosystems) in a final volume of 15 µL. The following PCR program was used for all PCR reactions: 95 °C for 2 min, followed by 40 cycles of 95 °C for 30 s, 59.5 °C for 30 s, and 72 °C for 30 s. When primers were used for the first time, the program was followed by a dissociation curve to determine if only one product was formed. Threshold cycle (Ct) values were calculated using Applied Biosystems Software (SDS version 2.3, for Windows XP). Subsequently, Ct values were normalized for differences in dsDNA synthesis using the UBI10 Ct values. Normalized transcript levels of all genes were compared with those of the mock-treated controls and the fold change in expression level was calculated based on the ΔΔCt.
Glucosinolate-containing extracts of plant tissue were prepared as described by Barth & Jander (2006). Desulfoglucosinolates were separated using a Waters 2695 high-performance liquid chromatography (HPLC) and detected using a Waters 2996 photodiode array detector and a Micromass Platform LC mass spectrometer (Waters, Milford, MA; http://www.waters.com/), with minor modifications of a previously described protocol (Kim et al. 2004). For HPLC separation, the mobile phases were A = water and B = 90% acetonitrile, at a flow rate of 1.2 mL min−1) at 23°C. Column linear gradients for samples were: 0–1 min, 98% A; 1–6 min 94% A; 6–8 min, 92% A; 8–16 min, 77% A; 16–20 min, 60% A; 20–25 min, 0% A; 25–27 min hold 0% A; 27–28 min, 98% A; 28–37 min, 98% A.
All results were tested for statistical significance using a paired t-test (α = 0.05), an unpaired t-test (α = 0.05) or a χ2-test, using JMP (2007) for Windows (SAS Institute Inc., Cary, NC, USA).
Aphid-induced resistance is local
To determine whether prior aphid feeding leads to an induced resistance response in Arabidopsis, 3-week-old plants were either mock-infested or infested for 48 h with 50 aphids. After the removal of the initial infestation, we assessed the fecundity of individual adult aphids on these plants by counting their offspring after 7 d. Figure 1a shows no differences in aphid fecundity on control and aphid-induced plants, suggesting that the plants are unable to adequately mount a defence response or that the aphids are able to move away from parts of the plant in which a defence response has been induced. A different result was obtained by caging aphids on individual leaves, thereby focusing on the site of interaction, where defence responses are more likely to occur (De Vos et al. 2005). Aphid fecundity is significantly reduced in leaves that were previously fed upon by 50 aphids, whereas the number of offspring from M. persicae on uninfested, systemic leaves of the aphid-infested plant was not significantly different from mock-treated plants (Fig. 1b). Together, these results show that prior feeding by aphids reduces subsequent aphid fecundity in local tissue, but that this resistance does not spread throughout the plant.
Aphid saliva has a role in induced defence against M. persicae
To investigate the role of salivary components in aphid-induced local resistance, plants were infiltrated with saliva samples collected by allowing aphids to feed from an artificial diet between two layers of Parafilm, a setup similar to those that have been used to collect aphid saliva for proteomic analyses (Harmel et al. 2008; Carolan et al. 2009). Figure 2a demonstrates that, compared with control infiltrations, aphid fecundity is significantly reduced (P < 0.05) upon infiltration of diet containing aphid salivary components into wild-type Columbia-0 (Col-0) Arabidopsis leaves. Aphid progeny production was unaffected in systemic tissues of saliva-infiltrated plants (14.5 ± 1.1 progeny on control treated plants and 15.4 ± 0.9 on saliva-treated plants; mean ± SE of n = 30; P = 0.39, Student's t-test). This indicates that, similar to aphid feeding, aphid saliva can induce a local resistance response in Arabidopsis leaves. As demonstrated in Fig. 2b no significant difference in aphid survival and reproduction was observed when aphids were fed on artificial diet containing saliva in comparison with a control diet. Therefore, the saliva itself is not toxic or repellent to M. persicae, and saliva-induced resistance depends on some plant component.
To determine whether known defence signalling pathways are required for saliva-induced resistance, we undertook an analysis using well-characterized mutants in JA (jar1 and acx1 acx5), SA (sid1, sid2 and NahG transgenic) and ethylene (ET; ein2) signalling. Compared with control leaves, aphid fecundity was significantly reduced on saliva-treated leaves in all tested mutants (Fig. 2a). This indicates that saliva-induced resistance against M. persicae is independent of these major defence-signalling pathways. In addition, we tested the phytoalexin deficient 4 (pad4) mutant for its ability to recognize salivary factors and mount an efficient defence response against M. persicae. PAD4 activity was recently shown to regulate resistance against aphids through activation of early senescence (Pegadaraju et al. 2005, 2007). However, similar to the other mutants impaired in defence signalling, the number of aphid progeny on pad4-1 plants was reduced by prior treatment with saliva (Fig. 2a).
Salivary peptides induce M. persicae resistance in Arabidopsis
To gain a better insight into the origin of salivary components that induce resistance in A. thaliana leaves, total aphid-fed diet was separated to yield <10 and >10 kD fractions. Subsequent infiltration of these salivary fractions and infestation by one adult aphid indicated that leaves trigger defence responses only upon recognition of salivary components in <10 kD fraction, whereas the >10 kD fraction was an ineffective inducer of aphid resistance (Fig. 3a). Additional size fractionation of the active fraction into a <3 kD and a 3–10 kD fraction showed that the latter fraction contains the active salivary component (Fig. 3b). Boiling or protease K treatment of the active <10 kD fraction prevented induction of resistance (Fig. 3c), suggesting that a proteinaceous salivary component functions as an elicitor. Lack of activity in the <3 kD fraction suggests that changes in the amino acid content of the diet caused by aphid feeding are not the cause of the elicited resistance. Moreover, analysis of the aphid-fed diet showed that, with the exception of tryptophan, amino acid levels were not significantly altered by 24 h of aphid feeding (Supporting Information Fig. S2). Together, these results indicate that plants can trigger local aphid resistance upon detection of a peptide or protein with a mass of 3–10 kD in aphid saliva.
In order to identify the potential eliciting factors, saliva collected from ∼5000 aphids was separated by sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS–PAGE). However, although the >10 kD fraction contained proteins, we were unable to detect any protein bands in the 3–10 kD fraction.
Aphid saliva changes gene expression in Arabidopsis
Recent advances in microarray technology have enabled whole-genome transcript profiling of Arabidopsis subjected to M. persicae feeding (De Vos et al. 2005; Couldridge et al. 2007; Kusnierczyk et al. 2007). To identify what part of these transcriptional changes is caused by M. persicae salivary components, we used Affymetrix whole-genome microarrays to profile the transcriptional regulation in Arabidopsis leaves 24 h after infiltration of artificial diet containing aphid salivary components. Compared with control diet infiltration, treatment with aphid saliva caused differential expression of many genes. A total of 125 transcripts were differentially regulated by aphid saliva treatment, of which 78 are significantly up-regulated (Supporting Information Table S1) and 47 are suppressed by components of M. persicae saliva (Supporting Information Table S2). A high percentage (67%; significantly more than the 23% expected by random chance; P = 0.002; χ2 test) of the genes up-regulated by aphid saliva was also induced by aphid feeding under similar experimental conditions (De Vos et al. 2005; Table 1). A comparison of the down-regulated genes showed a smaller overlap (41%; P = 0.008; χ2-test) with those suppressed by aphid feeding (Table 2).
Table 1. Arabidopsis genes induced by both M. persicae feeding (De Vos et al. 2005) and saliva infiltration
A biological function for the proteins encoded by these genes was determined using the Gene Ontology tool at The Arabidopsis Information Resource (Berardini et al. 2004).
Response to stress
Receptor-like protein kinase
Leucine-rich repeat family protein
Putative mitogen-activated protein kinase
Myb family transcription factor
Myb family transcription factor
Invertase/pectin methylesterase inhibitor protein family
Proteins encoded by aphid saliva responsive genes have relevant functions
Q-RT-PCR analysis confirmed the expression patterns of 7 out of 10 of the genes identified to be up- or down-regulated in the microarray experiment (Supporting Information Fig. S3). In order to understand the processes underlying aphid saliva-induced transcriptional changes, we determined the biological function and the cellular components using the Gene Ontology tool in The Arabidopsis Information Resource (TAIR) (http://www.arabidopsis.org; Berardini et al. 2004). Table 1 shows all genes that are significantly induced upon saliva treatment and aphid infestation (De Vos et al. 2005). There is a definite enrichment of genes involved in the response to stress conditions (e.g. 3.5-fold enrichment for response to stress, P < 0.001, χ2-test; and 2.7-fold enrichment for response to abiotic or biotic stimuli, P = 0.015, χ2-test). For example, the expression of pathogenesis-related protein (PR2), which encodes a β-1,3-glucanase has been shown to be responsive to SA and is induced upon infection by a variety of pathogens. Expression of PR-2 is highly induced in aphid-infested leaves (Moran & Thompson 2001; Moran et al. 2002; De Vos et al. 2005). In addition, it is induced upon feeding by the whitefly Bemicia tabacci (Kempema et al. 2007). Here we show that infiltration of aphid saliva factors can induce PR-2 expression. Several other genes characterized as ‘responsive to stress’ relate to induction of early senescence. These include, a ribonuclease, several senescence associated genes (SAG13; SAG21) and senescence associated protein 1 (SEN1). It has been hypothesized that a controlled early senescence response allows plants to move nutrients away from the M. persicae-infested tissues and can thereby reduce aphid population growth (Pegadaraju et al. 2005). Interestingly, recent transcriptional analysis of Arabidopsis upon feeding by Brevicoryne brassicae (cabbage aphid) also indicates an early senescence response (Kusnierczyk et al. 2008). Altogether, the induction of similar senescence-associated genes by aphids and their salivary component suggest that plants are able to recognize saliva-derived factors and mount an appropriate defence response. Aphid probing and injection of aphid salivary components during feeding are likely to trigger cell wall changes in their host plants. A phloem-specific xyloglucan endotransglycosylase/hydrolase cDNA was highly induced by M. persicae feeding in celery (Apium graveolens), and its homolog in Arabidopsis (XTH33, At1g10550) is also induced in response to aphid feeding (De Vos et al. 2005). Here, we demonstrate that saliva factors that are infiltrated into leaves of Arabidopsis also show increase expression of XTH33 (Table 1). Furthermore, Divol et al. (2007) show that Arabidopsis xth33 mutant plants are a preferred host for M. persicae compared with Col-0 wild-type plants, indicating that changes in cell walls can contribute to aphid defence.
Indole glucosinolates play an in vivo role in defence against M. persicae
Glucosinolates have been implicated as an important crucifer defence against insect herbivores (Halkier & Gershenzon 2006). Aphid feeding on Arabidopsis induces the conversion of indol-3-ylmethylglucosinolate (I3M) into 4-methoxyindol-3-ylmethylglucosinolate (4MI3M), and the latter is sixfold more deterrent to M. persicae feeding on artificial diets (Kim & Jander 2007). Arabidopsis cyp81F2 mutants, which are unable to convert I3M to 4-hydroxyindol-3-ylmethylglucosinolate (4OHI3M) as an intermediate in 4MI3M biosynthesis, are deficient in both 4OHI3M and 4MI3M (Bednarek et al. 2009; Clay et al. 2009; Fig. 4a). As has been reported previously with somewhat different experimental conditions (Pfalz, Vogel & Kroymann 2009), the number of offspring from a single adult female was found to be significantly higher on cyp81F2 plants compared with wildtype (Fig. 4b), indicating that 4MI3M and/or 4OHI3M play an in vivo role in Arabidopsis defence against aphids (Fig. 4b).
O-methyltransferases are induced by aphid feeding and saliva
To date, no O-methyltransferases that covert 4OHI3M into 4MI3M have been confirmed, but it has been speculated that a set of four highly homologous aphid-inducible O-methyltransferases (At1g21100, At1g21110, At1g21120 and At1g21130; >94% identical at the amino acid level) on chromosome 1 could play a role in the biosynthesis of repellent 4MI3M (Kim & Jander 2007). Affymetrix microarrays cannot differentiate the four homologous O-methyltransferase genes, but transcript profiling data shows that feeding by M. persicae significantly induces the expression of the gene family as a whole (De Vos et al. 2005), a result that we confirmed by Q-RT-PCR using primers complementary to sequences conserved in all four genes (Fig. 5a). In a time-course experiment, expression was found to be significantly increased at all time points except 12 h after infestation, with the highest induction after 24 h of M. persicae feeding (Fig. 5b). In order to differentiate between the four homologous genes, we designed gene-specific Q-RT-PCR primers to determine their individual contributions to the increased mRNA levels. As indicated in Fig. 5c two of the O-methyltransferases were significantly induced by 24 h of aphid feeding, suggesting that they could have redundant functions.
Figure 5d shows that aphid saliva also induces the expression of the O-methyltransferases 24 h after the start of the treatment. In order to correlate induction of O-methyltransferases, increased synthesis of 4MI3M, and saliva-induced resistance in Arabidopsis leaves, we assessed the expression of O-methyltransferases in response to infiltration of size-fractionated saliva samples. As observed for saliva-induced resistance, O-methyltransferase expression was induced by the <10 kD but not the >10 kD saliva fraction (Fig. 5d).
T-DNA insertions were identified in the promoters or coding regions of these O-methyltransferases to determine whether they affect 4MI3M accumulation, but no indole glucosinolate phenotypes were observed in single-gene T-DNA knockout mutants (Supporting Information Fig. S4). This lack of a glucosinolate phenotype does not confirm a lack of involvement in this pathway, but rather may reflect the high sequence similarity and overlapping expression patterns (Fig. 5c) of these genes.
4MI3M production is not induced by aphid saliva
Myzus persicae feeding has been shown to increase 4MI3M concentrations locally at the site of interaction (Kim & Jander 2007). However, leaves infiltrated with aphid saliva did not have increased 4MI3M in local tissue relative to control plants (Fig. 6a). The PEN2 β-thioglucosidase catalyzes 4MI3M breakdown (Bednarek et al. 2009), and 4MI3M accumulates to a higher level after flagellin (Flg22) treatment of pen2 plants than wild-type plants (Clay et al. 2009). Therefore, indole glucosinolate accumulation was also measured after aphid saliva infiltration of pen2 and wild-type plants, but no significant differences in 4MI3M accumulation were observed (Fig. 6b). Compared with wild-type Arabidopsis, pen2-1 mutant plants also do not display altered resistance to M. persicae. One adult aphid produced 27.9 ± 2.2 progeny on Col-0 and 25.3 ± 3.1 on pen2, respectively, over 10 d (mean ± SE of n = 16; P < 0.05, Student's t-test).
To effectively combat microbial pathogens and insect herbivores, plants have evolved sophisticated mechanisms that ensure early detection and induction of appropriate defence responses to the encountered invader. Ideally, signals indicating that attack is ongoing or imminent need to be both reliable and specific to the invader being encountered. In bacterial pathogens these elicitors often are essential for life or virulence on the host plant, and thus function as highly reliable cues. Despite their economic importance as agricultural pests, no elicitors have been identified from phloem feeding-insects. Here, we present molecular evidence that plants trigger an appropriate defence response upon detection of M. persicae salivary components. Although syringe infiltration of aphid-fed diet is clearly different from actual aphid feeding from the phloem, the large overlap in gene expression patterns (Table 1), as well as the induction of known defence-related genes, indicates a similar pathway to elicitation of plant responses. Further characterization rules out a role for aphid-derived small molecules in this particular defence response and indicates an elicitor or elicitors of proteinaceous origin and between 3 and 10 kD in size (Fig. 3). Based on a rough estimate of average amino acid size, the presumed peptide or small protein that triggers Arabidopsis defences is between 25 and 100 amino acids long. Although several recent publications have identified specific aphid salivary proteins (Will et al. 2007; Harmel et al. 2008; Carolan et al. 2009), none of these are in the 3–10 kD size range, primarily because these analyses did not allow detection of proteins and peptides of this size. Our own attempt to use protein gels approach to identify potential elicitors in the 3–10 kD size range was unsuccessful. Therefore, it appears that Arabidopsis is responsive to as yet unknown, low abundance proteinaceous elicitors from aphid saliva. As we saw no protein bands at all in gels of the 3–10 kD fraction, it is also unlikely that salivary proteins that appear larger than 10 kD on SDS-PAGE gels somehow pass through the 10 kD cutoff Amicon filters that were used in our experiments. Furthermore, although the data presented here do not rule out the possibility that a peptide elicitor is formed by the action of aphid saliva on amino acids contained in the artificial diet, this seems unlikely given what is currently known about the function of aphid saliva.
Will et al. (2007) have characterized two proteins from M. viciae with Ca2+-binding activity that neutralize increases in free Ca2+ that under normal conditions would induce forisome plugging of the sieve elements. Furthermore, C002, a highly abundant pea aphid (Acrythosiphon pisum) salivary protein, is required for the successful establishment of an aphid feeding site (Mutti et al. 2006, 2008). Although all of these proteins are >10 kD in size, we cannot rule out the possibility that Arabidopsis recognizes proteolytic breakdown products. Future peptide analysis of the active fraction (3–10 kD) will contribute to the identification of the elicitors responsible for induction of defences in Arabidopsis.
Both aphid feeding and saliva infiltration induce a local, rather than a systemic response in Arabidopsis (Figs 1 & 2), which could indicate that metabolite transport has a role in resistance, either through the removal of nutrients or caused by the movement of deterrents to the site of infestation. In other plant species systemic resistance to aphids has been observed (e.g. wild potato, F. Tjallingii; Medicago truncatula, Edwards, personal communications). Analysis of well-studied mutants in SA, JA and ET biosynthesis and signalling shows that aphid saliva-induced local resistance is not dependent on these pathways (Fig. 2), though the action of some related plant signalling molecules, for example, 12-oxo-phytodienoic acid (OPDA), cannot be ruled out at this point. Our analysis of the transcriptional reprogramming of Arabidopsis upon treatment with M. persicae saliva also suggests that transcriptional changes occur independently of the major defence signalling pathways (Table 1; Supporting Information Tables S1 & S2). Comparison with Affymetrix microarray data of plants treated with SA, JA (Koornneef 2008) shows that most genes differentially expressed upon treatment with aphid saliva (61% of up-regulated and 82% of down-regulated genes, respectively) have not been identified as responsive to JA or SA. Therefore, it is likely that as yet unknown plant signalling pathways contribute to the initiation of defence against aphids.
Recent publications have characterized CYP81F2, a cytochrome P450 required for in planta production of 4MI3M (Bednarek et al. 2009; Clay et al. 2009). Here, we show through analysis of cyp81F2 Arabidopsis plants that 4MI3M (or 4OH-I3M) also has in vivo aphid-deterrent properties (Fig. 4). Whereas cyp81F2 mutant plants are more susceptible to aphid feeding, CYP81F2 does not influence the interaction between lepidopteran larvae and Arabidopsis (Pfalz et al. 2009).
Altered transcript levels of a cluster of predicted O-methyltransferase genes on chromosome 1 indicate that these genes are induced upon aphid feeding and infiltration of aphid saliva (Fig. 5). Interestingly, gene expression of the O-methyltransferases correlates with the Arabidopsis defences triggered by saliva infiltration. The <10 kD saliva fraction highly increases O-methyltransferase expression, whereas the >10 kD fraction does not (Fig. 5d). Although these O-methyltransferase genes co-localize with an indole glucosinolate quantitative trait locus (QTL; Pfalz et al. 2007), we were not able to identify significant glucosinolate changes caused individual T-DNA insertions in this four-gene cluster (Supporting Information Fig. S4). Although methods exist for making knockout mutations in tandemly duplicated genes (Barth & Jander 2006 and references therein), these are currently difficult and time-consuming to implement.
Clay et al. (2009) showed that PEN2 is required for 4MI3M breakdown in response to treatment with the Flg22 flagellin fragment, and that 4MI3M-derived breakdown products are likely to be signals for callose formation during and plant defence against pathogens. Therefore, it is tempting to speculate that accumulation of callose in sieve elements in response to 4MI3M breakdown would contribute to aphid resistance. However, neither wild-type plants, nor pen2-1 mutants, show increased levels of 4MI3M upon treatment of Arabidopsis with aphid saliva, and the pen2 mutant does not exhibit reduced aphid resistance. Therefore, although the coi1-16 mutant was recently shown to also contain a pen2 mutation (Westphal et al. 2008), our results suggest that the decreased aphid resistance observed in this line (Ellis et al. 2002) comes entirely from the coi1-16 rather than the pen2 mutation. However, at this point, we cannot rule out the possibility that another of the numerous thioglucosidases encoded in Arabidopsis (Xu et al. 2004) has a function similar to PEN2 and thereby contributes to 4MI3M breakdown during aphid feeding and perhaps defence-related signalling.
Our results show that both M. persicae feeding and infiltration of salivary components lead to locally increased aphid resistance in Arabidopsis leaves. In particular, a 3–10 kD aphid saliva fraction reliably induces defence responses. Further characterization of this saliva fraction will likely lead to the identification of a previously unknown elicitor of plant defences against a phloem-feeding insect.
We would like to thank Wei Wang and Jie Zhao at the Cornell Microarray facility, Zhangjun Fei for help with the statistical analysis of the Affymetrix data, and Theodore Thannhauser and Tara Fish for their insights and technical assistance with detecting low-abundance proteins. This work was funded by NSF grant IOS-0718733 to G.J. and a postdoctoral fellowship from the Netherlands Organisation for Scientific Research (NWO) to M.D.V.