Introduction
The interrelationship between plants and herbivorous organisms, particularly insects, is one of the most important factors determining the stability of ecosystems, both natural and man-managed. One important guild of herbivores includes sap-sucking insects such as aphids, which are pests of many temperate and glasshouse crops (Slosser, Reference Slosser1993; Abate et al., Reference Abate, van Huis and Ampofo2000; Douglas, Reference Douglas2003). These insects cause direct feeding damage by the removal of large quantities of plant sap and, in some species, also transmit plant virus diseases. Physiological studies have revealed important information that describes the plant-aphid interaction (Tjallingii, Reference Tjallingii2006); but, currently, little is known about the underlying biochemical and molecular mechanisms involved.
Optimal growth and reproduction of aphids is facilitated by sustained sap ingestion from the phloem sieve element. There are many potential barriers that aphids must overcome before the sieve element is successfully accessed. The electrical penetration graph system (EPG) and microscopy studies have revealed that the pathway to the sieve element is predominantly intercellular, passing between cells; however, individual cells may be punctured en route (Tjallingii & Hogen Esch, Reference Tjallingii and Hogen Esch1993). Passage of the stylets between and within cells may be facilitated by the secretion of a range of cell wall degrading enzymes (Cherqui & Tjallingii, Reference Cherqui and Tjallingii2000). During the pathway phase, aphids produce a gelling saliva composed of lipoprotein, which forms a sheath and supports and protects the stylets as they penetrate the plant tissues (Douglas, Reference Douglas2003). A variety of plant chemical cues and physical barriers facilitate or restrict location of the phloem tissue (Pritchard et al., Reference Pritchard, Griffiths and Hunt2007). Once the sieve element has been accessed, a watery saliva is injected (Tjallingii, Reference Tjallingii2006), which may overcome phloem-localised defence proteins mediated by calcium (Will & van Bel, Reference Will and van Bel2006). Both watery and gelling saliva are potential elicitors of plant responses to insect feeding (Moran et al., Reference Moran, Cheng, Cassell and Thompson2002).
Phloem sap consists of water, carbohydrate (often as sucrose), amino acids, inorganic ions, proteins, hormones, mRNA, lipids and various secondary compounds (van Bel, Reference Van Bel2003). Considerable amounts of phloem sap can be extracted from plants by aphids; for example, the willow aphid (Tuberolachnus salignus, Gmelin) produces around 2 μl of honeydew per hour (Mittler, Reference Mittler1958). Aphids are, therefore, a considerable sink on the plant and additionally can alter phloem sap composition to increase the quality of their diet (Sandström et al., Reference Sandström, Telang and Moran2000).
The EPG technique has provided considerable detail about the location and acceptance of sieve elements by aphids and modification of the process by altered environments (Ponder et al., Reference Ponder, Pritchard, Harrington and Bale2000, Reference Ponder, Pritchard, Harrington and Bale2001; Hale et al., Reference Hale, Bale, Pritchard, Masters and Brown2003). This information is now being supplemented by genetic techniques to reveal the molecular mechanisms that underlie the plant-insect interaction (Qubbaj et al., Reference Qubbaj, Reineke and Zebitz2005; Hunt et al., Reference Hunt, Pritchard, Bennett, Zhu, Barrett, Allen, Bale and Newbury2006). Recently, this approach has been further developed by a number of studies using gene expression array technology to examine the response of a subset of plant defence genes to sap-feeding pests (Moran et al., Reference Moran, Cheng, Cassell and Thompson2002; Voelckel et al., Reference Voelckel, Weisser and Baldwin2004; Zhu-Salzman et al., Reference Zhu-Salzman, Salzman, Ahn and Koiwa2004). However, these methods reveal the response of only those genes that are already implicated in the interaction. Alternative strategies have used subtraction cDNA libraries to identify aphid responsive genes in phloem-enriched tissue of celery (Divol et al., Reference Divol, Vilaine, Thibivilliers, Amselem, Palauqui, Kusiak and Dinant2005). Importantly, the availability of whole genome arrays for A. thaliana provides an opportunity to examine the response of the whole plant genome to pathogen attack. Currently, studies on nematodes (Jammes et al., Reference Jammes, Lecompte, de Almida-Engler, Bitton, Martin-Magniette, Renou, Abad and Favery2005), whitefly (Kempema et al., Reference Kempema, Cui, Holzer and Walling2007) and aphids (De Vos et al., Reference De Vos, van Oosten, van Poecke, van Pelt, Pozo, Mueller, Buchala, Metraux, van Loon, Dicke and Pieterse2005) have suggested that very large numbers of Arabidopsis genes (i.e. ranging from 832 to 1349 in the De Vos study) are up- or down-regulated in response to pest infestation. While these studies have produced tantalising clues as to the complexity of the plant response to biotic stress, they are limited by having used only one or two biological replicates. The paucity of replication reduces considerably the statistical confidence that can be placed on changes in gene expression, and, therefore, the models of the molecular regulation of the interaction that can be constructed from the data.
The aim of the present study was to identify, with rigorous analysis, Arabidopsis genes that were up- or down-regulated in leaf tissue during the first 2 h (when EPG experiments indicate that aphids predominantly exhibit the pathway phase of stylet penetration) and the following 34 h of aphid infestation (when the predominant activity is sustained phloem sap ingestion). Disseminating information on such a list of responsive genes will be important for stimulating further studies by other researchers in this area. To achieve this objective, six independent biological replicates were utilised for infested and uninfested tissue at each time point. Each leaf was infested with ten fourth instar aphids to ensure that infestation levels were constant throughout the experimental period (i.e. there was no increase in aphid numbers through reproduction).
Materials and methods
Plant and insect cultivation
Arabidopsis thaliana (Col-0) were grown individually in soil (four parts loam-based John Innes Compost Number 2, four parts peat-based compost, one part Silvaperl, Osmocote) and maintained at 22°C in an 18:6 L:D cycle with unregulated relative humidity (≈70%). Plants were used when rosette growth was complete, just prior to inflorescence bolt production at principal growth stage 3.90, which was usually 3–4 weeks after sowing (Boyes et al., Reference Boyes, Zayed, Ascenzi, McCaskill, Hoffman, Davis and Gorlach2001).
Myzus persicae (Sulzer) were cultured on mature rosette stage or older A. thaliana (Col-0) plants grown under the conditions described above. Insects were obtained from a long-term clonal culture maintained at the University of Birmingham. Only apterous aphids were used for experimental purposes.
Insect bioassays
For both the 2 and 36 h time points, six independent biological replicates were performed; control and infested treatments at each time point, therefore, required a total of 24 chips. For each replicate, 20 plants were selected and paired according to rosette size and leaf number. Each pair of plants was assigned to a control or aphid infestation treatment, making ten plants per treatment or control. Two fully-expanded rosette leaves were chosen; however, the size of the clip ages prevented use of adjacent leaves. Each of the chosen leaves was infested on the upper surface with ten, fourth instar nymphs; this ensured that no reproduction would take place during the experiment, thus maintaining a constant level of infestation. Aphids were confined within a 2 cm diameter clip cage on each leaf. Leaves on control plants were brushed on the upper surface with a fine paintbrush to simulate manipulation and also enclosed within a clip cage. Control and experimental plants were placed in a randomised block design and maintained at 22°C (light intensity ∼100 μmole m−2 s−1, relative humidity ∼63%) for 2 or 36 h. Experiments were always started at 08:00 h, at the start of the light period, to minimise any effects of diurnal variation. Clip cages and aphids were removed and the 20 treated or control leaves were collected and immediately frozen. This procedure was repeated 12 times, six times for the 2 h and 6 times for the 36 h time points, producing 24 batches of leaf samples for subsequent RNA extraction.
RNA isolation
Total RNA was extracted from leaf samples using Qiagen® RNeasy Plant Mini Kits according to the manufacturer's recommended protocol. Each replicate sample consisted of 20 leaves. However, due to technical limitations, these were extracted as leaf pairs and the RNA samples pooled for subsequent expression profiling.
Expression profiling
cDNA, and subsequently cRNA, was synthesised from total RNA (5 μg) for each sample according to the protocol previously described by Zhu et al. (Reference Zhu, Budworth, Han, Brown, Chang, Zou and Wang2001). Each labelled cRNA sample was hybridised to an individual Arabidopsis whole genome ATH1 GeneChip (Affymetrix) (Zhu et al., Reference Zhu, Budworth, Han, Brown, Chang, Zou and Wang2001). Design features of this custom chip are reviewed by Zhu (Reference Zhu2003).
Gene expression analysis
The expression data were subjected to per chip normalisation (to the 50th percentile) and per gene normalisation (to the median) using GeneSpring 6.2 (Silicon Genetics: www.silicongenetics.com). The data were then filtered on flags (present only in at least six out of the 12 samples). Secondarily, a two-fold change in expression filter was applied. Analysis of variance on log transformed expression values was carried out separately for treated versus control for each time point with GeneSpring, applying the Benjamini & Hochberg (Reference Benjamini and Hochberg1995) method for controlling the false discovery rate.
Results and discussion
Following 2 or 36 h of exposure to aphids, only 27 genes from the 24 K in the whole Arabidopsis genome showed significantly altered expression in the leaf on which the aphid was feeding. After 2 h, one gene was up-regulated and one down-regulated, while after 36 h, 23 genes were up-regulated and two were down-regulated (fig 1). There were no genes in common between 2 and 36 h. The number of genes affected by aphid feeding was markedly lower than that reported in a study of M. persicae infesting Arabidopsis for 48 and 72 h, where 832 genes were reported to be up-regulated, and 1349 down-regulated (De Vos et al., Reference De Vos, van Oosten, van Poecke, van Pelt, Pozo, Mueller, Buchala, Metraux, van Loon, Dicke and Pieterse2005). In our study, nine of the genes that were up-regulated after 36 h were also reported to be up-regulated in the De Vos study. These included the lectin-like protein gene At3g16530 and two of the three glutathione transferases (At1g02930 and At2g02930) (table 1). Interestingly, one gene, At5g55730, coding for a fasciclin-like protein, was down-regulated in both studies. The Arabidopsis gene chip has recently been used to analyse the effect of the phloem-feeding whitefly, Bemisia tabaci (Gennad) (Kempema et al., Reference Kempema, Cui, Holzer and Walling2007). This study seemingly identified 700 genes up-regulated and 556 down-regulated. Only 17% of these were common to the De Vos study and one gene (PAD4) was common with the present study. Currently, there are few published studies using full genome arrays to investigate plant-pest interactions, the current status of which has been recently reviewed (Thompson & Goggin, Reference Thompson and Goggin2006).
Fold change in Arabidopsis genes showing significant changes in expression following 2 h (open bars) or 36 hours (closed bars) infestation of rosette leaves by 4th instar nymphs of Myzus persicae. Each leaf was infested with ten aphids; two leaves per plant were infested and mRNA was extracted from infested leaves only. Significant changes in expression were determined by Genespring analysis of the six independent replicates. Bars represent average±SD of fold change in expression of the six, paired, independent biological replicates. The horizontal line represents no change in expression.
Putative annotation of Arabidopsis genes showing significant changes in expression following 2 or 36 h infestation of mature rosette leaves by 4th instar nymphs of Myzus persicae. Each leaf was infested with ten aphids; two leaves per plant were infested and mRNA was extracted from infested leaves only. Average spot fluorescence (arbitrary units) of the six independent replicates is shown, plus the significance of difference in fluorescence of control and infested leaves determined by paired Students t-test. Note that these 27 genes were identified as having significantly altered expression by genome wide analysis using Genespring (see text for further details).
While common features between these data sets may be viewed as encouraging, the lack of adequate biological replication (one and two, respectively), and resultant higher numbers of genes identified in both studies, make it difficult to determine the significance of the overlap. It is possible that the longer exposure times and heavier infestation (72 h and 40 reproducing aphids in the De Vos et al. (Reference De Vos, van Oosten, van Poecke, van Pelt, Pozo, Mueller, Buchala, Metraux, van Loon, Dicke and Pieterse2005) study and 100 whitefly for 21 days in the Kempema et al. (Reference Kempema, Cui, Holzer and Walling2007) study) may account for the much higher numbers of genes showing altered expression (2181 in the De Vos et al. (Reference De Vos, van Oosten, van Poecke, van Pelt, Pozo, Mueller, Buchala, Metraux, van Loon, Dicke and Pieterse2005) study and 1256 in the Kempema et al. (Reference Kempema, Cui, Holzer and Walling2007) study, compared to 27 in the present investigation). It must also be likely that our statistical stringency and a bias towards higher transcript levels may have led to type II errors. However, it is more likely that the inadequate biological replication and inherent lack of statistical power is the major contributing factor and that at least three (Lee et al., Reference Lee, Kuo, Whitmore and Sklar2000) and probably more than five (Pavlidis et al., Reference Pavlidis, Qinghong and Stafford Nobel2003) biological replicates are required to achieve stable results. Analysing one chip per treatment or not accounting for the false discovery rate using our data yielded a list of genes showing different expression levels similar in size to those of De Vos et al. (Reference De Vos, van Oosten, van Poecke, van Pelt, Pozo, Mueller, Buchala, Metraux, van Loon, Dicke and Pieterse2005) but with no statistically significant differences. Such comparison highlights the importance of an appropriate level of replication and statistical analysis in transcriptomics experiments. The need for higher numbers of replicates is also demonstrated by the high number of genes showing significantly altered expression following nematode infestation when using only two biological replicates (Jammes et al., Reference Jammes, Lecompte, de Almida-Engler, Bitton, Martin-Magniette, Renou, Abad and Favery2005). It is, therefore, vital that the emerging MIAME protocols take account of replication to ensure comparability between different transcriptomics experiments and also to encourage standardisation of statistical analysis. Without such rigour, comparison between different studies remains speculative at best.
Since aphids are specialist herbivores targeting the phloem sieve element, it is reasonable to expect that changes in gene expression may be phloem-specific. Currently, interest is focusing on the role of forisomes in such phloem-specific defence in legumes (Fabaceae) (Will & van Bel, Reference Will and van Bel2006). Alternatively (or additionally), alterations in the nutritional quality of the sieve element sap may have a role in plant-aphid interactions (Sandström et al., Reference Sandström, Telang and Moran2000; Hunt et al., Reference Hunt, Pritchard, Bennett, Zhu, Barrett, Allen, Bale and Newbury2006). In this study, we extracted mRNA from whole leaves, where changes in gene expression will be biased toward higher transcript numbers. By comparison, phloem constitutes a small fraction (<0.5% of total leaf volume) of total plant tissue (Sjölund, Reference Sjölund1997). In the present study, where changes in gene expression may be restricted to a single cell type, such low transcript numbers may well be undetected. In this regard, none of the genes identified in this work were homologous with those previously reported in either phloem-enriched protein (Barnes et al., Reference Barnes, Bale, Constantinidou, Ashton, Jones and Pritchard2004) or cDNA (Doering-Saad et al., Reference Doering-Saad, Newbury, Bale and Pritchard2002, Reference Doering-Saad, Newbury, Couldridge, Bale and Pritchard2006) libraries.
Despite the cell-specific site of aphid feeding, EPG experiments reveal other interactions as the aphid locates the sieve element. Thus, changes in gene expression induced by aphid infestation may be responses of the plant during the pathway phase of the interaction. Alternatively, gene expression changes may be induced by the insect to manipulate the plant so as to produce a more favourable food source (Prado & Tjallingii, Reference Prado and Tjallingii1997) or to overcome plant defence (Urbanska et al., Reference Urbanska, Tjallingii, Dixon and Leszczynski1998).
The annotation of the Arabidopsis genome provides an objective way to assign function to the genes that have been identified. However, since the annotation is not complete, it is still useful to analyse altered expression profiles subjectively, combining the putative function (if known) with current knowledge of the physiology and biochemistry of aphid feeding. Taking this approach, the gene transcripts identified in the present study can be placed into a number of functional classes.
Oxidative stress
A common feature of plants under stress is the induction of genes associated with oxidative stress. Aphids are known to induce oxidative stress (Ni et al., Reference Ni, Quisenberry, Heng-Moss, Markwell, Sarath, Klucas and Baxendale2001; Moran et al., Reference Moran, Cheng, Cassell and Thompson2002); for example, reactive oxygen species were increased in tomato (Solanum lycopersicum L.) during infestation by the potato aphid, Macrosiphum euphorbiae (Thomas) (de Ilarduya et al., Reference De Ilarduya, Xie and Kaloshian2003). Four genes with significant up-regulation at 36 h fit into this category. The first, a FAD-Binding domain-containing protein similar to a reticuline oxidase precursor, is a documented oxidative stress-responsive gene with a role in oxidoreductase activity and possible antimicrobial activity (Custers et al., Reference Custers, Harrison, Sela-Buurlage, Van Deventer, Lageweg, Howe, Van Der Meijs, Ponstein, Simons, Melchers and Stuiver2004; Kim et al., Reference Kim, Kim, Park and Lim2005). A blue-copper binding protein, also up-regulated after 36 h, is known to be locally induced in response to wounding as a likely mechanism of increased resistance to oxidative stress (Yang et al., Reference Yang, Im, Chung and Cho2002). The third gene was an ankyrin repeat family protein. Ankyrin proteins have a documented role in plant disease resistance and regulation of antioxidation metabolism (Yan et al., Reference Yan, Wang and Zhang2002; Dong, Reference Dong2004). Antioxidant enzymes are likely to be important in sequestering free radicals produced by physical damage (Dillwith et al., Reference Dillwith, Berberet, Bergman, Neese, Edwards and McNew1991). Spontaneous DNA mutations may occur as a potential knock-on effect of free radical production and, under such conditions, the up-regulation of DNA repair enzymes can be observed (Filkowski et al., Reference Filkowski, Kovalchuk and Kovalchuk2004). This may explain the observed increase in expression of a nucleoside diphosphate-linked moiety transcript annotated with a function involved in DNA repair.
Likely elicitors of oxidative stress produced by aphids include cell wall fragments and oligosaccharides that are the products of physical damage caused during stylet penetration activity (Moran et al., Reference Moran, Cheng, Cassell and Thompson2002). Aphid saliva may also be an elicitor involved in the alteration of plant oxidative conditions (Dillwith et al., Reference Dillwith, Berberet, Bergman, Neese, Edwards and McNew1991; Miles & Oertli, Reference Miles and Oertli1993; Miles, Reference Miles1999; Walling, Reference Walling2000). Therefore, induction of some of these genes suggests that the aphid may be altering the oxidative status of the plant.
Defence during pathway phase
Glutathione transferases (GSTs) are a class of well-described enzymes involved in xenobiotic detoxification, mediating specific conjugation and peroxidase reactions with glutathione (Wagner et al., Reference Wagner, Edwards, Dixon and Mauch2002; Dixon et al., Reference Dixon, Skipsey, Grundy and Edwards2005; Edwards et al., Reference Edwards, Brazier-Hicks, Dixon and Cummins2005). Aphid feeding after 36 h induced significant up-regulation of three glutathione transferases, GST1, GST2 and GST16. GST1 was the most strongly induced, with 4.6-fold increased mean expression in the infested plants compared with the control. Strong induction in response to wounding, low temperature, high salt and aluminium treatments has previously been demonstrated (Yang et al., Reference Yang, Kim, Kim, Guh, Kim and Cho1998; Ezaki et al., Reference Ezaki, Suzuki, Motoda, Kawamura, Nakashima and Matsumoto2004). GST2 induction in response to a range of biotic and abiotic stresses is also well documented, including following exposure to copper and hydrogen peroxide (Lieberherr et al., Reference Lieberherr, Wagner, Dubuis, Metraux and Mauch2003; Smith et al., Reference Smith, Deridder, Guo, Seeley, Regnier and Goldsbrough2004). Regulation is known to occur via ethylene and salicylic acid signalling pathways (Zhou & Goldsbrough, Reference Zhou and Goldsbrough1993; Lieberherr et al., Reference Lieberherr, Wagner, Dubuis, Metraux and Mauch2003; Smith et al., Reference Smith, Nurizadeh, Peer, Xu, Bandyopadhyay, Murphy and Goldsbrough2003; Mang et al., Reference Mang, Kang, Shim, Kim, Park, Kim, Bahk and Kim2004). Eliciting factors from the aphid that may induce such genes are likely to include a range of glycoproteins, oligosaccharides and enzymes secreted as constituents of the gelling saliva (Miles, Reference Miles1999; Cherqui & Tjallingii, Reference Cherqui and Tjallingii2000). These genes could also be included in the oxidative stress category, as reduced glutathione has redox buffering power, which may confer cellular protection from this type of stress (Ogawa, Reference Ogawa2005).
Induction of a cytochrome P450 gene is also consistent with plant defence in response to aphid stylet penetration activity. While there are two broad classes of these enzymes, both have a defensive role, being involved in either detoxification of xenobiotics or biosynthesis of a range of molecules associated with defence and signalling (Li et al., Reference Li, Schuler and Berenbaum2002). A gene coding for an esterase family protein was up-regulated 2 h post-infestation. Classically, esterases play a role in detoxification; significant increases in esterase activity have previously been demonstrated in barley (Hordeum vulgare L.) in response to Diuraphis noxia (Kurdj.) feeding and hypothesised to be related to toxic and oxidative stress imposed by the aphid (Ni & Quisenberry, Reference Ni and Quisenberry2003).
Cell wall modification
Two genes with annotations associated with cell wall modification showed significant differential expression at 2 and 36 h. These genes are likely to be associated with the physical damage to cell walls caused by the intercellular passage of the stylets and the profuse number of cell punctures during the pathway phase. The first of these genes codes for an extracellular dermal glycoprotein, known to be an inhibitor of endo-glucanases (York et al., Reference York, Qin and Rose2004). This function may confer a plant defence response by blocking cell wall degrading enzymes, such as pectinase, present in the watery saliva (Cherqui & Tjallingii, Reference Cherqui and Tjallingii2000). Early up-regulation at 2 h is consistent with EPG data, showing that most pathway (and, thus, physical damage) occurs during this time. A homologue in carrot has been shown to play an important role in plant defence and signal transduction during wounding (Satoh et al., Reference Satoh, Sturm, Fujii and Chrispeels1992; Shang et al., Reference Seki, Narusaka, Ishida, Nanjo, Fujita, Oono, Kamiya, Nakajima, Enju, Sakurai, Satou, Akiyama, Taji, Yamaguchi-Shinozaki, Carninci, Kawai, Hayashizaki and Shinozaki2004).
A gene for a fasciclin-like arabinogalactan-protein was down-regulated at 36 h. Arabinogalactan-proteins are a family of hyperglycosylated, hydroxyproline-rich proteins found at the surface of plant cells (Borner et al., Reference Borner, Lilley, Stevens and Dupree2003) Up-regulation in response to salt stress has previously been documented (Lamport & Kieliszewski, Reference Lamport and Kieliszewski2005). The reversed expression observed in this study may be consistent with the type of stress imposed. Arabinogalactan-proteins may have a role in reducing the number of pectin cross-linkages in the cell wall with the result of increasing porosity (Lamport & Kieliszewski, Reference Lamport and Kieliszewski2005). This would facilitate cell wall penetration by the aphid, and it can be speculated that down-regulation of this gene represents active facilitation of feeding by aphids.
Potential phloem defence
Plant defence responses may also be specifically targeted at mechanical damage in the sieve element. A gene for a UDP-glucosyl transferase family protein involved in the glycosylation of aglycones (xenobiotics, hormones and defence-related metabolites) (Li et al., Reference Li, Schuler and Berenbaum2002) was up-regulated at 36 h. Some UDP-glucosyl transferases have been shown to be involved in callose synthesis (Hong et al., Reference Hong, Zhang, Olson and Verma2001). Callose deposition at the sieve plate is a common defence response to sieve element disruption and may be linked with sieve element damage caused by the aphid stylets (Will & van Bel, Reference Will and van Bel2006).
A legume lectin family protein gene was up-regulated 3.9-fold after 36 h. While lectins (carbohydrate-binding proteins), are known to play a role in defence responses against a range of bacteria, fungi, viruses and other plant antagonists, there is evidence that some lectins are phloem-specific (Vijayan & Chandra, Reference Vijayan and Chandra1999; Dinant et al., Reference Dinant, Lark, Zhu, Vilaine, Palauqui, Kusiak and Thompson2003). Aphid growth and fecundity is known to be affected by jackbean (Canavalia gladiata Jacq) and snowdrop (Galanthus nivalis L.) lectins (Stoger et al., Reference Stoger, Williams, Christou, Down and Gatehouse1999; Sauvion et al., Reference Sauvion, Charles, Febvay and Rahbe2004; Vasconcelos & Oliveira, Reference Vasconcelos and Oliveira2004).
Also up-regulated at 36 h was a gene for a syntaxin protein (AtSyp122) involved in vesicle trafficking of the defence-related molecules, SNAREs (soluble N-ethylmaleimide-sensitive factor adaptor protein receptor) (Sanderfoot et al., Reference Sanderfoot, Assaad and Raikhel2000). The modified plasmodesmata, which connect sieve elements to companion cells, are thought to be one of the major pathways by which macromolecular signals are transported in the plant (Oparka, Reference Oparka2004). It is possible that syntaxin may be involved in this type of trafficking. However, syntaxin receptors may be localised to various membranes throughout the plant to induce a defence response (Bassham et al., Reference Bassham, Gal, Conceicao and Raikhel1995). Indeed, a calcium-dependent reaction of the AtSyp122 gene, elicited by a bacterial flagellin peptide, leads to exocytosis of hypothesised defence-related proteins in cultured Arabidopsis cells (Nühse et al., Reference Nühse, Boller and Peck2003).
Calcium-dependent signalling
Calcium is an important message in the transduction of developmental and environmental signals, including biotic and abiotic stresses (Reddy & Reddy, Reference Reddy and Reddy2004). At 36 h, two genes with functions associated with calcium homeostasis showed significant up-regulation, including a gene for calcium-binding EF hand family protein and a calmodulin-binding family protein (At4g33050). Calmodulin is a calcium-binding protein that can bind to, and regulate, an assembly of different proteins. Calmodulin-related signalling genes in the TCH family were up-regulated by aphid infestation after 72 h (Moran et al., Reference Moran, Cheng, Cassell and Thompson2002). A defence role has been implicated in downy mildew infestation (León et al., Reference León, Rojo, Titarenko and Sanchez-Serrano1998; Bergey & Ryan, Reference Bergey and Ryan1999; Eulgem et al., Reference Eulgem, Weigman, Chang, McDowell, Holub, Glazebrook, Zhu and Dangl2004). Indeed, At4g33050 may be a target of whirly transcription factors that regulate defence gene expression (Desveaux et al., Reference Desveaux, Marechal and Brisson2005).
Signalling
A transcript for the PAD4 gene was up-regulated at 36 h; PAD4 may operate to integrate signals from the salicylate and jasmonate defence pathways (Zhou et al., Reference Zhou, Tootle, Tsui, Klessig and Glazebrook1998; Pegadaraju et al., Reference Pegadaraju, Knepper, Reese and Shah2005). PAD4 is implicated in plant defence since a loss-of-function mutant in Arabidopsis for the PAD4 gene supported a higher aphid population (Pegadaraju et al., Reference Pegadaraju, Knepper, Reese and Shah2005). It is also part of a suite of genes involved in plant disease resistance against Pseudomonas syringae (pv) and Peronospora parasitica (Pers) (Glazebrook et al., Reference Glazebrook, Zook, Mert, Kagan, Rogers, Crute, Holub, Hammerschmidt and Ausubel1997; Feys et al., Reference Feys, Moisan, Newman and Parker2001). The salicylic acid (SA) signalling pathway may mediate defence induction in A. thaliana under feeding pressure by M. persicae (Moran et al., Reference Moran, Cheng, Cassell and Thompson2002). This is consistent with observations of M. persicae and M. euphorbiae on tomato, which induced strong up-regulation of the salicylic pathway PR and P4 genes (Fidantsef et al., Reference Fidantsef, Stout, Thaler, Duffey and Bostock1999). However, the role of the SA is not yet resolved, and aphid reproduction was unaffected on mutant Arabidopsis plants in which SA signalling was compromised (Pegadaraju et al., Reference Pegadaraju, Knepper, Reese and Shah2005). In Arabidopsis leaves, aphid feeding did not induce any measurable changes in SA levels (De Vos et al., Reference De Vos, van Oosten, van Poecke, van Pelt, Pozo, Mueller, Buchala, Metraux, van Loon, Dicke and Pieterse2005).
A short-chain dehydrogenase/reductase family protein was significantly up-regulated at 36 h post-infestation. This protein has a function in cellular differentiation and signalling (Kallberg et al., Reference Kallberg, Oppermann, Jornvall and Persson2002). Its signalling role is likely to be important in the transduction of stress elicitors.
Carbohydrate metabolism
A transcript annotated ‘accumulation of photosystem one (APO1)’ was up-regulated at 36 h. This gene has a role in chloroplast cluster complex accumulation (Amann et al., Reference Amann, Lezhneva, Wanner, Herrmann and Meurer2004). When the aphid is feeding, it effectively becomes a new solute sink within the plant (Douglas, Reference Douglas2003) so that its up-regulation may be associated with alterations in carbon partitioning, either through aphid-induction to manipulate the plant to produce a more favourable diet, or as a passive response to the aphid becoming more of a sink on the plant.
Transcription
Five percent of the Arabidopsis genome is known to be involved in transcriptional regulation coding for more than 1500 transcription factors (Riechmann et al., Reference Riechmann, Heard, Martin, Reuber, Jiang, Keddie, Adam, Pineda, Ratcliffe, Samaha, Creelman, Pilgrim, Broun, Zhang, Ghandehari, Sherman and Yu2000). Several transcripts differentially expressed at 36 h were annotated with functions involved in regulation of gene expression. The first of these genes, coding for a no apical meristem family protein, has similarities with NAC-domain proteins that are known to have transcription factor activity (Ooka et al., Reference Ooka, Satoh, Doi, Nagata, Otomo, Murakami, Matsubara, Osato, Kawai, Carninci, Hayashizaki, Suzuki, Kojima, Takahara, Yamamoto and Kikuchi2003). A number are reported to be involved in responses to a range of biotic and abiotic stresses, including Sclerotinia sclerotiorum (Lib.), flea beetle feeding, wounding, drought and low temperature (Seki et al., Reference Shang, Shibahara, Hanada, Iwafune and Hirano2002; Hegedus et al., Reference Hegedus, Yu, Baldwin, Gruber, Sharpe, Parkin, Whitwill and Lydiate2003; Delessert et al., Reference Delessert, Kazan, Wilson, Van Der Straeten, Manners, Dennis and Dolferus2005; Mahalingam et al., Reference Mahalingam, Shah, Scrymgeour and Fedoroff2005). Also up-regulated was a gene for a zinc-binding family protein, part of the PLATZ1 class of zinc-dependent DNA-binding proteins which repress transcription of A/T-rich sequence motifs (Nagano et al., Reference Nagano, Furuhashi, Inaba and Sasaki2001). Finally, a cytidine deaminase, down-regulated at 36 h, may have a role in RNA turnover, catalysing the hydrolysis of cytidine into uridine and ammonia (Faivre-Nitschke et al., Reference Faivre-Nitschke, Grienenberger and Gualberto1999).
Unknown
Five transcripts (At1g27020, At2g18680, At2g18690, At3g12740 and At5g52750) were significantly up-regulated at 36 h, but literature and BLAST searches revealed unknown functions. These transcripts could perhaps be of most interest, as they could provide information about previously uncharacterised defence-related processes.
Conclusions
This study has identified candidate plant genes that are strongly implicated in the aphid-plant interaction and provides sound target genes for future projects to study their function in more detail. These targets will inform the development of novel defence strategies (Moran et al., Reference Moran, Cheng, Cassell and Thompson2002). Future work will aim to localise cellular expression of each of these genes and study detailed changes in expression using RT-PCR. Examination of aphid performance on knockout mutants will provide further information about the function of these individual genes. Finally, the small number of genes that we have confidently identified will facilitate focused systems biology studies examining the co-regulation of these genes within gene networks.
Acknowledgements
The authors thank the Biotechnology and Biological Sciences Research Council for funding the studentship for Clare Couldridge and Syngenta for both funding and technical support with the microarrays. We acknowledge the helpful comments of two anonymous referees.
