The response of the poplar transcriptome to wounding and subsequent infection by a viral pathogen

Authors

  • Caroline M. Smith,

    1. Department of Plant Sciences, University of Oxford, South Parks Road, Oxford OX1 3RB, UK;
    Search for more papers by this author
  • Marisa Rodriguez-Buey,

    1. Umeå Plant Science Centre, Department of Plant Physiology, Umeå University, SE-901 87 Umeå, Sweden
    Search for more papers by this author
  • Jan Karlsson,

    1. Umeå Plant Science Centre, Department of Plant Physiology, Umeå University, SE-901 87 Umeå, Sweden
    Search for more papers by this author
  • Malcolm M. Campbell

    Corresponding author
    1. Department of Plant Sciences, University of Oxford, South Parks Road, Oxford OX1 3RB, UK;
      Author for correspondence:Malcolm M. Campbell Tel: +44 1865 275135 Fax: +44 1865 275074 Email: malcolm.campbell@plants.ox.ac.uk
    Search for more papers by this author

Author for correspondence:Malcolm M. Campbell Tel: +44 1865 275135 Fax: +44 1865 275074 Email: malcolm.campbell@plants.ox.ac.uk

Summary

  • • The Populus–Poplar Mosaic Virus (PopMV) pathosystem is the best characterized of all forest tree–virus interactions. The details of the host response to this virus are completely unknown.
  • • The transcript abundance for approximately 10 000 Populus genes was simultaneously interrogated using spotted cDNA microarrays. Relative transcript abundance was compared for RNA extracted from Populus leaves that were untreated, mock-inoculated leaves that were wounded by leaf abrasion and inoculated leaves that were abraded and then infected by virus.
  • • Statistical analysis of the microarray data identified suites of genes that exhibited increased or decreased transcript abundance in response to wounding, systemic PopMV infection or both together. Genes implicated in programmed cell death, and cell wall reinforcement were a major feature of the wound response, whereas genes encoding metallothionein-like proteins, and proteins implicated in cell wall remodelling were a major feature of the PopMV response.
  • • The identification of wound- and PopMV-regulated genes opens the door for future studies aimed at testing specific hypotheses related to the mechanisms used by forest trees to contend with stress.

Introduction

Owing to their long lifespan and large size, individual forest trees present many more opportunities for infection by pathogens during their lifetime than do herbaceous annual plants (Tainter & Baker, 1996). The fact that forest trees are very apparent to pathogen attack, and that the woody perennial habit remains an evolutionarily successful life-history strategy, suggests that trees must possess robust and durable defence mechanisms. Investigations into the mechanisms that trees employ to cope with pathogen challenge has revealed some of the chemical defences that are used. However, in general, there is a paucity of knowledge pertaining to the breadth of molecular genetic programmes that underpin these and other defences.

The genus Populus provides excellent models to investigate the molecular genetic mechanisms that underpin pathogen defence responses in forest trees. In addition to containing some of the most economically important tree species in the Northern hemisphere, a member of the genus, Populus trichocarpa, has been adopted as the model tree species for forest biologists, and will be the first tree to have its genome sequenced (Brunner et al., 2004). The Populus genome sequence will enhance the substantial molecular toolkit that already exists for Populus species, including a comprehensive expressed sequence tag (EST) collection (Sterky et al., 1998; Bhalerao et al., 2003), and microarrays for transcriptome analysis (Andersson et al., 2004; Hertzberg et al., 2001). Furthermore, these tools can all be used in conjunction with useful pedigrees and detailed genetic maps that were developed based on decades of Populus breeding (Bradshaw et al., 1994; Bradshaw & Stettler, 1995; Villar et al., 1996; Frewen et al., 2000; Cervera et al., 2001). Such pedigrees have proven to be particularly useful in uncovering Populus loci that confer resistance to fungal pathogens (Villar et al., 1996). Most investigations into the defence mechanisms used by Populus have focused on fungal pathogens (Royle & Ostry, 1995; Newcombe, 1996; Villar et al., 1996; Legionnet et al., 1999; Ostry & Ward, 2003).

Other pathogens also impinge on the health of Populus, including viruses (Cooper, 1993; Nienhaus & Castello, 1989). The best characterized viral pathogen of Populus species is Poplar mosaic virus (PopMV) (Biddle & Tinsley, 1971a, 1971b, 1971c; Atkinson & Cooper, 1976; Brunt et al., 1976; van der Meer et al., 1980; Cooper & Edwards, 1981; Henderson et al., 1992; von Kontzog & Ebrahim-Nesbat, 1992; Cooper, 1993). It is described as a specialist virus, with a single-stranded, sense-strand RNA genome, whose natural host range is thought to be limited to Populus species (Cooper, 1993; Henderson et al., 1992). Electron microscopy analysis of PopMV particles suggests that it is a member of the genus Carlavirus (Brunt et al., 1976). This hypothesis is supported by the finding that the complete PopMV genome sequence shares extensive similarity with other carlaviruses (Smith & Campbell, 2004a).

The details of PopMV pathogenesis are not well characterized. The natural vector for transmission of PopMV is not known, although it is clear that the virus can be readily transmitted after wounding, allowing the virus to spread if plants are propagated by cuttings (Biddle & Tinsley, 1971c). PopMV was reported to induce chlorosis of the leaves of infected Populus (Biddle & Tinsley, 1971a; Atkinson & Cooper, 1976) but recent analyses suggest that this symptom is not necessarily diagnostic for the presence of the virus (Smith & Campbell, 2004b). Little is known about the long-term effects of PopMV on Populus growth and development, but a reduction in the growth of several nursery cultivars has been documented (Biddle & Tinsley, 1971c). There is growing evidence to suggest that spread of PopMV within infected Populus hosts is contingent on the genotype of the host (Cooper & Edwards, 1981; von Kontzog & Ebrahim-Nesbat, 1992) (Smith & Campbell, 2004b), but the basis for this is not yet known. To date, the molecular responses that are invoked in Populus following PopMV infection remain to be documented.

As a first step in characterizing the molecular responses of Populus to PopMV, we examined alterations in the activity of the Populus transcriptome following PopMV infection. Making use of a compatible Populus–PopMV pathosystem that was recently shown to permit systemic spread of the virus by 14 d post inoculation (Smith & Campbell, 2004b), we characterized the changes in the host transcript accumulation using spotted cDNA microarrays. Here we show that relatively modest, but highly reproducible changes in transcriptome activity are invoked by wounding and PopMV infection.

Materials and Methods

Experimental organisms

The virus isolate used in this work (PV-0341) was obtained from the Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ) virus collection (DSMZ, Braunsweig, Germany). The virus was propagated and maintained on Nicotiana megalosiphon. The Populus clone used in these studies was clone 52-226, derived from a P. trichocarpa ×Populus deltoides cross. The clone was generously provided by Dr David Ellis of CellFor Inc. (Vancouver, BC, Canada).

Plant growth conditions

In vitro-propagated 52-226 clones were grown on sterile, solid Murashige–Skoog medium (Murashige & Skoog, 1962) containing 1.5% w : v sucrose, 0.3% w : v Agar M, and 0.22% w : v Phytagel (Sigma, Dorset, UK). Plants were subcultured to fresh media every 6 wk. Plant material was grown at 21°C and light intensity at 500 lx in a controlled growth chamber (Vindon Scientific Ltd, Oldham, UK) under long-day conditions.

Poplar clones with well-formed root systems were transferred into pots containing a moist 3 : 1 mix of Levington Universal compost and Vermiperl vermiculite (Silverperl Ltd, Lincoln, UK). The plants were grown in a temperature controlled Sanyo Gallenkamp growth room (Sanyo Gallenkamp, Loughborough, UK) under long-day conditions (8 h dark/16 h light, 40 µmol m−2 s−1) at 22°C.

Propagation of the PopMV on N. megalosiphon

The virus was propagated on plants of N. megalosiphon. Plants of N. megalosiphon were grown from seeds in a Sanyo Gallenkamp growth room under long-day conditions as described earlier. The N. megalosiphon plants were infected as described for Populus. Infected leaf material was collected 2 wk post inoculation, flash frozen in liquid nitrogen and stored at −80°C.

Inoculation of Populus clones with PopMV

Before virus inoculation, plants were both shaded for 12 h and well watered, as these conditions have been used in virology experiments to enhance virus susceptibility and symptom development (Kado & Agrawal, 1972). The first fully expanded leaves were inoculated. The leaf to be infected was marked and dusted with Celite 545 (Sigma, Dorset, UK). To prepare the viral inoculum, 0.5 g of infected N. megalosiphon leaf material was ground in phosphate buffer and 10 l of solution was applied to each leaf for rub-inoculation. Mock inoculated (wounded) plants were inoculated with phosphate buffer only. Control plants that had not been abraded or treated with buffer were also maintained. After 1 min the Celite/virus inoculum was rinsed off with water to prevent excessive evaporation. Plants were then shaded overnight and leaves monitored for signs of infection. The infected leaf was collected 14 d post inoculation (dpi), flash frozen in liquid nitrogen, and stored at −80°C. The infected leaves were collected at the same time of day as they had been inoculated, exactly 14 dpi.

RNA preparation and quality control

RNA samples were prepared using 200 mg of leaf tissue from poplar clone 52-226 using the a previously described method (Haruta et al., 2001). Three RNA samples were pooled to create three samples each consisting of three leaves for each condition. For each leaf tissue sample an aliquot was kept and used in reverse transcriptase polymerase chain reaction (RT-PCR) analysis to check for viral infection. cDNA was synthesized from total plant RNA in a reverse transcription reaction using the Superscript· II System (Invitrogen, Paisley, UK) by the method detailed in the manufacturer's instructions. A 1-l aliquot of the cDNA was used as a template in a multiplex PCR reaction. One set of primers were specific for the PopMV virus and the other set amplified part of an endogenous gene, phytoene desaturase.

Microarray analysis

The cDNA synthesis, probe preparation, hybridization conditions, data collection and data analysis was done according to previously published methods (Andersson et al., 2004), with some modifications. The cDNA synthesis made use of 50 g of total RNA as template. The labelled cDNA was mixed with hybridization buffer to final concentrations of 5× standard saline citrate (SSC), 25% formamide, 0.2% sodium dodecyl sulphate (SDS) and 25 g of tRNA (Life Technologies, Paisley, UK) and 30 g of Oligo-dA (80-mer; CyberGene AB, Huddinge, Sweden). The samples were heated to 95°C for 3 min, chilled on ice for 30 s and applied to the ASP (automated slide processor, Amersham Lucidea SlidePro; Amersham Biosciences, Little Chalfont, UK) chambers containing the prehybridized slides. Prehybridization was done at 42°C for 30 min in the presence of prehybridization buffer (5× SSC, 0.1% bovine serum albumin (BSA) and 1× Denhardt's solution). The hybridization was carried out at 42°C for 15 h. Scanning of the slides was performed using the Packard Bioscience ScanArray 4000 (Packard BioScience, Billerica, MA, USA) software and hardware. Focusing was performed using a ‘line-scan’ across at least six good spots (blue or green intensity) with as high a magnification as possible. Each fluor was selected in turn and scanned to register the signals (Cy3 excited at 532 nm and Cy5 excited at 635 nm). The focus was adjusted to obtain the largest peaks possible without saturation. The laser power and Photomultiplier Tube (PMT) gain were also adjusted at this stage to obtain the cleanest and brightest image possible. Once the settings were adjusted, a high-resolution scan was obtained (5 µm).

Images were analysed using GenePix Pro 4.1 (Axon Instruments, Foster City, CA, USA) software. Individual spots were located, the Cy3 and Cy5 fluorescence intensity at each spot was measured, and the background signal intensities determined. Adaptive circles were employed to allow the spots to differ between 80% and 125% of the expected diameter of 100 µm. Only composite spots with a pixel intensity of more than 500 were used. Weak spots were flagged as ‘not found’ and artefacts such as dust and salts were flagged as ‘bad’. The data was extracted and stored as GPR (GenePix Results) files showing raw data and derived ratio measurements. GenePix Array List (GAL) files corresponding to the slide batch numbers were used to assign names and identifiers to each feature on the microarray. Raw data from the image analysis software was exported as GPR files and imported into the software R (Ihaka & Gentleman, 1996). R allowed the correction of slides for intensity and spatial dye biases using local regression. Pairs of intensity log ratios for each spot were generated. Lowess (locally weighted linear regression) normalization was used to normalize for dye bias and for background subtraction (Yang et al., 2002).

For multiple comparisons among slides, a false-discovery-rate multiple testing correction (Benjamini & Hochberg, 1995) was applied to generate adjusted P-values in GeneSpring software version 6.0 (Silicon Genetics, San Carlos, CA, USA). The false discovery rate that was chosen was less stringent than the family-wise error rate, but provided a good balance between the discovery of statistically significant changes in transcript abundance and protection against false positives (Benjamini & Hochberg, 1995). Only genes with reproducible (≥ 10 observations) expression differences with adjusted P-values < 0.05 were considered in our analysis. Hierarchical clustering was carried out using TIGR Multiexperiment Viewer (http://www.tigr.org/software/tm4/).

Experimental design

Clones of the P. trichocarpa × P. deltoides hybrid, 52-226, were propagated under sterile conditions and maintained in a growth cabinet until they were deployed in the infection experiment. Deployment involved transfer of rooting stock to soil, and gradual hardening of the plants in a climate-controlled growth room. The PopMV isolate (PV 0341) was propagated for the experiments using the propagation host N. megalosiphon. Leaf extracts from virus-infected N. megalosiphon were pooled, and used as the inoculum to infect hardened Populus clones.

Populus clones were inoculated at a site approximately 2 cm from the tip of the leaf, at a location equidistant between the leaf margins, on one of the two fully expanded leaves that were formed when the plantlets were first transferred from sterile growth medium to soil. Inoculation with the virus involved preparing the inoculated leaf by abrading it with Celite, and then applying the viral inoculum. One-third of the plants were abraded in this fashion and then inoculated with PopMV. These were referred to as the inoculated (I) plants. Data collected from I material provided an indicator of the composite response to both leaf abrasion and PopMV infection. In order to account for any response that was caused by the preinoculation abrasion, another one-third of the plants were abraded but not inoculated with PopMV, providing a mock-inoculated (MI) control. Any results obtained with the abraded MI tissue provided an indicator of the wound response. The remaining one-third of the plants had no treatment, providing a base-line control (C) for all of the experiments. Nine clonal replicates of Populus hybrid 52-226 were used for each experimental condition. The inoculated, mock-inoculated, and untreated control leaves were harvested 14 dpi, frozen in liquid nitrogen, ground to a powder and used for RNA extractions.

The RNA was extracted from individual leaves and quality verified electrophoretically. It was then analysed by RT-PCR to detect the presence of PopMV. Only the PopMV-inoculated samples contained PopMV RNA. The RNA samples from three leaves were pooled to generate one replicate. In this way, there were three replicates (each containing three plants) for each condition. These RNA samples were then used to generate the probe for the microarray analyses via indirect amino-allyl labelling of cDNA. These probes were then used in competitive hybridization experiments on the Populus spotted 13K cDNA microarray (Andersson et al., 2004). All conditions were competitively hybridized, and all were replicated in reciprocally labelled competitive hybridization experiments (‘dye-swaps’). The analysis involved 18 microarrays in total.

As its name implies, the Populus spotted 13K cDNA microarray had approximately 13 000 Populus cDNAs (13 490) spotted on a derivatized glass slide. The microarray is thought to represent between 9000 and 10 000 unigenes assembled from EST collections representing a number of different Populus tissues (Andersson et al., 2004). Each cDNA is spotted in duplicate on each slide to provide a control for hybridization. In the experiments described herein, the combination of microarrays and duplicate spotting provided 36 data points documenting the relative transcript abundance for each coding sequence. These data were then subjected to rigorous statistical analyses to determine trends in transcriptome activity in response to wounding and PopMV infection.

Results and Discussion

Highly reproducible changes in Populus transcriptome activity were observed in response to wounding or PopMV infection

Following statistical analysis of the data, two general trends were observed. First, the analyses revealed that there were highly reproducible changes in the pattern of relative transcript abundance. The relative expression levels of many genes had such a highly consistent pattern across all replicates that they were supported with a P-value that was less than 0.05; consequently, these were used in all subsequent analyses. The second major trend was that the relative expression levels that were observed between experimental conditions were moderate at best. For example, the maximum fold change increase in relative transcript abundance across all comparisons was 4.80 and the maximum fold change decrease was 0.15 (6.67-fold decrease). Fortunately, the robust nature of the data allowed comparisons to be made with genes that showed fold changes that were as low as 1.20, as these relative expression levels were still supported with a P-value that was less than 0.05.

It is not entirely surprising that the relative fold changes in transcript abundance were moderate to low. The experimental material was collected at single time-point that ensured that the inoculated leaves were completely PopMV-infected, but at a time that was likely beyond that which would document only a wound response. Previous studies had determined that inoculated clones of Populus hybrid 52-226 were definitely PopMV infected by 14 dpi (Smith & Campbell, 2004b), so this was chosen as the time-point for these experiments. It is possible that this time-point was beyond that which documents the maximum changes in transcript abundance in response to PopMV; however, it may allow a better discrimination of PopMV-mediated effects from those induced by wounding. While some of genes involved in the signal transduction pathways that are invoked by early infection, including those that respond to the oxidative burst associated with the early stages of pathogen attack (Wojtaszek, 1997), might have been missed, the time point should also bypass the strong early signalling events related only to wounding (Creelman et al., 1992; Titarenko et al., 1997; Schenk et al., 2000; Sugimoto et al., 2000, 2003; Li et al., 2001). In fact, as will be discussed later, there is evidence to suggest that ‘early’ pathogen responses are still being invoked, and that the repercussions of wounding are still clearly in evidence, even at 14 dpi.

Transcript abundance both increased and decreased in response to wounding and PopMV infection

On the basis of a P-value cut-off of 0.05, two groups of genes could readily be identified: those that exhibited an increase in transcript abundance in either of the treatments (MI or I) relative to C, and those that exhibited a decrease in transcript abundance between either MI or I and C. Overall, 2065 genes exhibited an increase in transcript abundance in the treatments relative to the control, whereas 741 genes showed a decrease in relative transcript abundance between either MI or I and C. Clearly, a greater proportion of Populus genes exhibited an increase in transcript abundance in response to mock inoculation and/or to PopMV infection compared with those for which transcript abundance decreased in response to these stimuli. Increased gene expression undoubtedly generates the suite of mechanisms that the plant uses to contend with the stimulus. In addition, the apparent simultaneous repression of a subset of the genes suggests that resource reallocation was taking place, remodelling growth and development in response to the stimuli.

From the two groups of genes described above, six subsets of genes were assembled. The first three subsets of genes included only those genes that exhibited an increase in relative transcript abundance between pairs of experimental conditions. For example, the first set of genes contained all of those that exhibited an increase in relative transcript abundance in MI conditions relative to the untreated C condition. This set of 947 genes represented all those whose transcript abundance increased due to wounding alone. The second set of genes comprised all of those that exhibited an increase in relative transcript abundance in the I leaves relative to the MI leaves. This set of 599 genes represented all those whose transcript abundance was increased by the presence of the virus, over and above any affect attributable to wounding. The third set of genes comprised all of those that exhibited an increase in relative transcript abundance in the I leaves relative to the C leaves. That is, the members of this set of 1301 genes showed an increase in transcript abundance over the control after the combined effects of wounding and PopMV infection.

The other three subsets of genes were similar to the first three except that they contained only those genes that exhibited a decrease in relative transcript abundance between pairs of experimental conditions. That is, the fourth set (310 genes) exhibited a decrease in transcript abundance from MI to C leaves (wounding effect), the fifth set (308 genes) a decrease from I to MI leaves (viral effect) and the sixth set (351 genes) a decrease from I to C leaves (viral and wounding effect).

In these experiments, the response of the host transcriptome to PopMV inoculation was a composite reaction to both wounding and PopMV infection. Therefore, it was important to delineate between those genes whose transcript abundance was altered primarily by wounding from those which were modulated by PopMV infection. As a first step in discriminating between these groups of genes, the data were analysed using Venn diagrams (Figs 1 and 2).

Figure 1.

Classification of genes with increased transcript abundance identified by microarray analyses. The lists of significantly upregulated cDNAs for each of the three experimental conditions – MI/C (wounding), I/MI (virus), I/C (wounding + virus) – were used in Venn diagram analysis. Overlapping intersections between conditions allowed the identification of sets of genes whose transcript abundance were increased as a result of wounding alone (a), virus alone (b) or increased with both wounding and virus (c). The transcripts considered in this analysis were those showing reproducible (= 10 observations) expression differences with adjusted P-values < 0.05.

Figure 2.

Classification of genes with decreased transcript abundance identified by microarray analyses. The lists of significantly downregulated cDNAs for each of the three experimental conditions – MI/C (wounding), I/MI (virus), I/C (wounding + virus) – were used in Venn diagram analysis. Overlapping intersections between conditions allowed the identification of sets of genes which were decreased as a result of wounding alone (a), virus alone (b) or decreased with both wounding and virus (c). The transcripts considered in this analysis were those showing reproducible (= 10 observations) expression differences with adjusted P-values < 0.05.

Groups of genes that were identified by Venn analysis were categorized using the Umeå Plant Sciences Centre/Munich Information Centre for Protein Sequences (UPSC-MIPS) classification scheme (Bhalerao et al., 2003). Each gene was assigned to a UPSC-MIPS category and the distribution of genes among the functional categories was examined (Figs 3 and 4). The entire cDNA set from the array was also classified to allow a comparison between the transcript distributions. The transcript profiles for all of the genes that met the filtering criteria were also subjected to hierarchical cluster analysis. Cluster analysis of the 2806 genes that met these criteria resolved 15 clusters with correlation coefficients equal to or greater than 0.80 (Fig. 5). The 15 clusters were assigned to seven groups (described later) and recorded as gene list tables with associated relative expression levels (Tables S1–S7, see Supplementary Material section for details).

Figure 3.

The transcripts which exhibited an increase in abundance were grouped into 20 functional classes according to Umeå Plant Sciences Centre/Munich Information Centre for Protein Sequences (UPSC-MIPS) classification scheme. The 13527 cDNAs from the Populus cDNA microarray were also grouped into functional classes (a) and show the genes in each functional group as a percentage relative to the total number of cDNAs on the slide. (b) The distribution of genes among functional classes whose transcript abundance is increased as a result of wounding alone. (c) The distribution of transcripts whose abundance increased as a result of Poplar mosaic virus (PopMV) infection only. (d) The distribution of transcripts whose abundance is increased by both wounding and PopMV. Differentially expressed transcripts were considered as those showing reproducible (10 observations) expression differences with adjusted P-values < 0.05.

Figure 4.

The differentially expressed repressed transcripts were grouped into 20 functional classes according to Umeå Plant Sciences Centre/Munich Information Centre for Protein Sequences (UPSC-MIPS). The 13527 cDNAs from the Populus cDNA microarray were also grouped (a) and illustrate the genes in each functional group as the percentage of their numbers relative to the total number. (b) The distribution of genes among functional classes whose transcript abundance is decreased as a result of wounding alone. (c) Transcripts whose abundance decreased as a result of virus only. (d) Transcripts whose abundance is decreased by wounding and virus. Differentially expressed transcripts were considered as those showing reproducible (= 10 hybridizations) expression differences with adjusted P-values < 0.05.

Figure 5.

Expression graphs of clusters obtained by hierarchical cluster analysis. All differentially expressed transcripts were log transformed and subjected to hierarchical clustering using the Euclidean distances metric and a correlation coefficient of 0.8. Fifteen clusters were identified. The three data points on the graphs correspond to the three experimental conditions MI/C (wounding), I/MI (virus), I/C (wounding + virus). Hierarchical cluster analysis was visualized out using TIGR Multiexperiment Viewer (http://www.tigr.org/software/tm4/).

A large subset of Populus genes exhibited an increase in transcript abundance in response to wounding

Overlapping intersection sets from Venn diagrams identified those genes where wounding alone was responsible for the overall increase in transcript abundance in inoculated leaves relative to the controls (Fig. 1a, Table S1). That is, the subsequent infection by PopMV did not have a statistically significant effect on the transcript abundance of these genes at the time that the leaf tissue was sampled. Hierarchical clustering on the basis of transcript profiles also resolved clusters containing a large number of putatively wound-induced genes (Fig. 5, Clusters 1, 6, 8 and 12). The genes within these clusters contained all those identified by the intersection set in the Venn diagram (Fig. 1a). The transcript profiles for these genes clearly show that change in transcript abundance is a result of wounding alone, as no positive change in transcript abundance could be attributed to PopMV infection (I/MI) (Fig. 5).

The genes for which transcript abundance increased in response to wounding was the largest of all the intersection sets identified. This large group of genes was dominated by those proposed to encode proteins designated as being related to the UPSC-MIPS ‘Metabolism’ category (Fig. 3). While 17% of the genes on the entire poplar microarray were assigned to the ‘Metabolism’ category, 23% of the genes that exhibited an increase in transcript abundance in response to wounding were assigned to this category (Fig. 3).

Among the most prominent genes for which transcript abundance increased in response to wounding were those that have been postulated to play a role in the modification of plant cell walls. For example, 25 genes encoding enzymes implicated in lignin biosynthesis were found within this group. This included genes encoding enzymes of general phenylpropanoid metabolism, such as phenylalanine ammonia-lyase (EC.4.3.1.5), cinnamate 4-hydroxylase (EC.1.14.13.11) and hydroxycinnamate:CoA ligase (EC.6.2.1.12), as well as genes encoding enzymes that are directly involved in the biosynthesis of the monomeric precursors of lignins, the monolignols, including caffeoyl-CoA O-methyltransferase (EC.2.1.1.104), hydroxycinnamoyl-CoA reductase (EC.1.2.1.44), 5-hydroxyconiferaldehyde O-methyltransferase (EC.2.1.1.68) and hydroxycinnamyl alcohol dehydrogenase (EC.1.1.194). In addition, genes encoding enzymes implicated in the polymerization of lignins, the laccases (EC.1.10.3.2) were also found within this group. Strikingly, four different genes encoding hydroxycinnamoyl-CoA reductase enzymes were observed in the subset that showed the greatest increases in transcript abundance in response to wounding alone. The prominent role played by hydroxycinnamoyl-CoA reductase in channelling carbon skeletons into lignin biosynthesis (Jones et al., 2001; Goujon et al., 2003) suggests that lignification may be a major response at the wound site, even at 14 d following wounding.

It is tempting to speculate that the transcript abundance of genes encoding lignin biosynthetic enzymes increased in response to wounding in order for the plant to regenerate xylem vessels that were damaged at the wound site. Xylem cells are frequently regenerated at wound sites to re-establish vascular continuity, and it may be that the increased transcript abundance of lignin biosynthetic genes contributes to the final stages of xylem regeneration in the wounded Populus leaves. Alternatively, the increased transcript abundance for genes encoding lignin biosynthetic enzymes may be related to the creation of a lignified barrier at the wound site that functions to prevent infection by opportunistic microbes, or to limit water loss. Consistent with the hypothesis that the wounded tissues are creating a barrier, another gene postulated to encode a protein that plays a role in cell wall reinforcement, hydroxyproline-rich glycoprotein (HRGP) (Cosgrove, 1997), also exhibited increased transcript abundance in response to wounding alone. Previously, HRGP gene expression has been shown to be induced in response to wounding (Ahn et al., 1996). This may be an ancient wound response, as it is also observed in alga (Ender et al., 1999).

Aside from genes encoding enzymes involved in the modification of the plant cell wall, genes encoding proteins assigned to the UPSC-MIPS ‘Cell Rescue’ category, including many involved in programmed cell death (PCD), exhibited increased transcript abundance in response to wounding alone (Fig. 3). This included 10 genes that encode proteolytic enzymes, such as a homologue of a senescence-associated cysteine protease, and a papain-like cysteine protease. Cysteine proteases have been hypothesized to be involved in PCD in several plant species, where they are thought to degrade specific proteins and thereby regulate the process of PCD (Kuriyama & Fukuda, 2002). In fact, protein turnover appeared to be a substantial feature of the wound response, as genes assigned to the UPSC-MIPS ‘Protein Synthesis’ category were proportionally under-represented relative to their abundance on the entire microarray (χ2 analysis, P < 0.001), and those assigned to the ‘Protein Fate and Folding’ category were over-represented (χ2 analysis, P < 0.001).

The wound-induced increase in relative transcript abundance for PCD-related genes is not entirely surprising. Following wounding, plants need to limit subsequent secondary effects that can occur at the wound, which include water loss and infection by opportunistic pathogens. As indicated above, this may involve the biosynthesis of a resilient, lignified cell wall. This process is frequently accompanied by localized PCD, which creates a zone of dead cells that also function as an effective barrier to water loss and pathogen colonization.

It was somewhat surprising that the wound response was still clearly identifiable on the basis of increased transcript abundance, even at 14 d after wounding. However, the types of genes that exhibited increases in transcript abundance, such as those related to lignification and PCD suggest that the wound response is in its final stages. This hypothesis is supported by the fact that, in this study, genes that previously have been implicated in early wound responses, such as transcription factors and jasmonic-acid induced genes (Creelman et al., 1992; Titarenko et al., 1997; Schenk et al., 2000; Sugimoto et al., 2000, 2003; Li et al., 2001), did not exhibit an increase in transcript abundance in response to wounding alone at the time point sampled.

Populus genes that exhibited a decrease in transcript abundance in response to wounding tended to be related to chloroplast function

Overlapping intersection sets from Venn diagrams also identified those genes where wounding alone was responsible for an overall decrease in transcript abundance in inoculated leaves relative to the controls (Fig. 2a, Table S2). As was the case for the first group of genes described above, PopMV infection after wounding did not have a statistically significant effect on the transcript abundance of these genes at the time that the leaf tissue was sampled. Hierarchical cluster analysis also resolved groups of genes whose transcript abundance pattern was consistent with being wound-repressed (Fig. 5, Clusters 2 and 3). There was complete correspondence between the genes found in the Venn intersection set and those identified on the basis of hierarchical clustering.

The group of genes that displayed a decrease in transcript abundance in response to wounding was small relative to the group where transcript abundance increased. Within this small group there was a striking degree of commonality in the function of the genes. Many of the genes in this group are predicted to encode proteins that are related to chloroplast function, including protochlorophyllide reductase, thioredoxin, thylakoid lumenal protein, ribulose bisphosphate carboxylase/oxygenase and a number of chloroplast ribosomal proteins. The decrease in transcript abundance of genes encoding chloroplast components would likely result in a concomitant decrease in chloroplast biogenesis and activity. Decreases in chloroplast function in response to wounding have been proposed to be a component of remobilization of resources away from the wounded leaf (Criqui et al., 1992; Reinbothe et al., 1993; Zhou & Thornburg, 1999). This remobilization presumably decreases the dependency of this leaf for photosynthesis, and results in a channelling of resources to other facets of the wound response, such as barrier biosynthesis, or to new growth.

Within the group of genes that were predicted to be wound repressed, the proportion of genes classified in the UPSC-MIPS ‘Protein Synthesis’ category were over-represented, including those encoding chloroplast precursors and ribosomal proteins. That is, wounding is predicted to repress many genes categorized as being involved in protein synthesis. This finding is entirely consistent with the observation that genes in this category were under-represented among those that were predicted to be wound induced (Fig. 3).

Taken together the data suggest that the ‘late’ wound response in Populus (i.e. 14 d after wounding) involved changes in cell wall biogenesis, cell death and protein turnover, with a concomitant decrease in the biogenesis of new proteins and photosynthetic functions, relative to untreated leaves. It is important to note that this prediction is based entirely on changes in transcript abundance, which may not necessarily reflect protein abundance in these functional categories. Nevertheless, the transcript abundance data raise several new and important testable hypotheses pertaining to how Populus responds to wounding.

Populus genes that exhibited an increase in transcript abundance in response to PopMV alone were stereotypical defence-response genes

Venn diagram analysis identified a group of genes that exhibited an increase in transcript abundance that was independent of wounding (Fig. 1b, Table S3). These genes represent those for which PopMV alone induced an increase in transcript abundance. These genes comprise two groups identified by hierarchical cluster analysis (Fig. 5): one where PopMV induced small increases in transcript abundance (Cluster 1) and one group where PopMV induced greater increases in transcript abundance (Cluster 11). This group of 137 genes included many that have been implicated in the pathogen defence response in other plant species. In keeping with this, when these genes were subjected to UPSC-MIPS classification (Fig. 3), the category that was the most over-represented relative to the proportion found on the whole array was the category defined as ‘Cell Rescue’ (from 17% on whole array to 25% in response to PopMV; statistically different by χ2 analysis, P < 0.001).

The most prominent subset of genes that exhibited an increase in transcript abundance in response to PopMV was a group predicted to encode metallothionein-like proteins. Of the 25 genes that exhibited the greatest increase in transcript abundance in response to PopMV, seven were predicted to encode metallothionein-like proteins. Similarly, a gene encoding a metallothionein-like protein is strongly induced in tobacco in response to tobacco mosaic virus (TMV) (Choi et al., 1996). While the exact function of the metallothionein-like protein encoded by the TMV-induced gene is not known, the protein functions like other metallothioneins to sequester metals (Suh et al., 1998). It has been hypothesized that the sequestration of metals by these proteins plays a role in modulating redox reactions in the plant, with knock-on effects on the generation of active oxygen species in the infected tissues (Choi et al., 1996). Metallothioneins have been shown to protect cells against the damaging effects of reactive oxygen species (You et al., 2002).

The prominence of genes that encode metallothionein-like proteins, among those for which transcript abundance increased in response to PopMV, suggests a central role for these proteins in the response to viral infection at 14 dpi. Given that the Populus clone used in these experiments appeared completely susceptible to PopMV, and that the virus had spread systemically throughout the plant by 14 dpi, it may be that the metallothionein-like proteins are functioning to contain an oxidative burst through the plant tissues as a protective mechanism. This hypothesis assumes that reactive oxygen species have been generated some time in advance of the increase in metallothionein transcripts, perhaps as part of an oxidative burst that can be induced by pathogen-derived signals (Wojtaszek, 1997). It is possible that the restriction of the oxidative burst also explains the ability of PopMV to systemically infect this particular Populus clone. That is, it may be that PopMV invokes the production of metallothioneins in order to suppress the oxidative burst that frequently comprises the first line of plant defence to pathogens (Lamb & Dixon, 1997). By keeping the host defence mechanisms in check, PopMV may evade host defences and spread systemically.

It is tempting to speculate that the role of metallothionein-like proteins may be related to the fact that viral infection can be inhibited by low concentrations of heavy metals. Recently, nontoxic concentrations of cadmium have been shown to inhibit virus movement in plants (Citovsky et al., 1998; Ueki & Citovsky, 2002). Given that metallothionein-like proteins can sequester heavy metals such as cadmium (Suh et al., 1998), it is possible that an increase in the abundance of metallothionein-like proteins would alter the local concentrations of heavy metals, either to inhibit the progress of the virus or, alternatively, perhaps to enhance the infection. The production of metallothionein-like proteins may be invoked by PopMV to protect it from the inhibitory effects of heavy metals. It will be interesting to determine whether the metallothionein-like protein response actually functions in a defence capacity, or whether it has been ‘captured’ by PopMV to enable it to infect its host in an unimpeded fashion.

Several genes predicted to encode heat-shock proteins (HSPs) also exhibited increased transcript abundance in response to PopMV. The expression of HSPs was induced in other plant–virus interactions (Aranda et al., 1996, 1999; Escaler et al., 2000; Whitham et al., 2003) and this induction may facilitate viral infection (Glotzer et al., 2000). While it remains to be determined what role, if any, is played by HSPs in the Populus–PopMV interaction, it is possible that they also facilitate viral infection in this pathosystem.

In addition to metallothioneins and HSPs, seven genes that are predicted to encode proteins with roles in signalling and activation of defence mechanisms against pathogens also exhibited increased transcript abundance in response to PopMV infection. One of these genes shares significant sequence similarly with the Ve resistance locus. The Ve locus encodes a cell-surface, leucine-zipper glycoprotein, which functions as a component of the signal transduction cascade involved in resistance to the fungal wilt pathogen Verticillium dahliae (Kawchuk et al. 2001). It is possible that the related Populus gene encodes a protein that functions similarly in host–pathogen signalling, and facilitates the manifestation of a defence response. In keeping with this hypothesis, 11 genes encoding several pathogenesis-related (PR) proteins also exhibited increased transcript abundance in response to PopMV infection. It is rather striking that the transcript abundance for these genes is still increased so late after infection, given that the virus has clearly spread throughout the entire leaf by 7 dpi (Smith & Campbell, 2004b) and that PR protein gene expression is generally induced within days of infection. It may be that the later accumulation of PR protein transcripts in this particular host–pathogen interaction results from the fact that it is a compatible interaction. A delayed defence response is a hallmark of a compatible reaction (Tao et al., 2003).

Interestingly, one gene that exhibited an increase in transcript abundance in response to the virus encodes a β-1,3-glucanase, an enzyme that is predicted to hydrolyse callose. The deposition of callose, and its degradation by β-1,3-glucanase, has been implicated in the restriction of virus movement. For example, antisense inhibition of a tobacco β-1,3-glucanase reduced the movement of TMV, whereas overexpression of the β-1,3-glucanase allowed greater virus spread (Bucher et al., 2001). The deposition of callose is thought to constrict the neck region of plasmodesmata, thereby preventing viral movement; therefore, the hydrolytic activity of the β-1,3-glucanase reduces this constriction and permits virus spread. Increased β-1,3-glucanase transcript abundance was also observed in Arabidopsis in response to a number of RNA viruses (Whitham et al., 2003). This gene may be targeted for induction by RNA viruses, such as PopMV, in order to permit virus movement within the host. The mechanism by which viruses mediate this increase in transcript abundance is unknown.

Populus genes involved in cell wall integrity exhibited a decrease in transcript abundance in response to virus infection

A group of genes that exhibited a decrease in transcript abundance independent of wounding was identified (Fig. 2b, Table S4). This set of 99 genes represents those for which PopMV alone induced a decrease in transcript abundance. Analysis of transcript abundance profiles of these 99 genes revealed two clusters whose profiles were consistent with being repressed by virus infection alone (Fig. 5). Within these two clusters were a number of transcripts known to be involved in cell wall modification.

A number of genes encoding extensins exhibited a statistically significant decrease in transcript abundance as a result of virus infection. Extensins are a family of cell wall-localized hydroxyproline-rich glycoproteins that have been proposed to play a role in development, wound healing and plant defence (Showalter, 1993). Biochemical evidence supports the existence of an extensin–pectin cross-linked network forming a highly impenetrable barrier (Qi et al., 1995). Decreased extensin deposition, with an ensuing decrease in cell wall cross-linking, might prevent the formation of this barrier and thereby facilitate pathogen spread.

Functional classification of the 99 genes whose transcript abundance was shown to decrease as a result of virus infection revealed that one of the classes that was over-represented was ‘Control of Cellular Organization’ (Fig. 4). In accordance with the UPSC-MIPS classification, a number of genes whose transcript abundance was decreased were predicted to encode enzymes involved in the biosynthesis of cell wall components such as pectin, including several genes encoding pectin methylesterase (PME). Control of cellular organization is important for virus infection as virus movement from cell-to-cell occurs via the plasmodesmata. Maintenance of cell structure is therefore imperative for the virus to be able to infect systemically. The repression of transcripts encoding genes involved in cell restructuring may be the result of the virus modifying the host's transcriptome in a manner favourable for systemic spread. Alternatively, the host may be attempting to limit the virus infection by decreasing the synthesis of PME, which has been shown to be important in the systemic spread of TMV (Chen et al., 2000; Chen & Citovsky, 2003). Viruses encode movement proteins (MPs) that are proposed to interact with the plasmodesmata to allow the virus to move from cell to cell (Rhee et al., 2000). Recent studies have shown that the MPs can interact with PME, and in mutants with reduced expression of PME virus spread was retarded (Dorokhov et al., 1999; Chen et al., 2000; Chen & Citovsky, 2003). Thus, the reduction in transcript abundance for PME might reduce the synthesis of this enzyme in an attempt by Populus to reduce the possibility for interaction between PME and the PopMV MPs, and thereby restrict virus movement.

In addition to transcripts related to cell wall restructuring, there was also a decrease in transcript abundance of those thought to play a role in the manifestation of defence responses. These included genes encoding an elicitor-inducible cytochrome p450 and PR proteins, and a number of genes that have been shown to be stress-responsive in other systems. The decrease in transcript abundance of genes that are thought to restrict virus infection indicates that the presence of the virus altered the host's response in a way that is likely to favour virus spread. It remains to be determined how this occurs, but it raises the intriguing possibility that PopMV reorients host transcriptome activity in a manner that is conducive to systemic spread, as has been observed for other plant and animal viruses (Aranda & Maule, 1998).

Hierarchical cluster analysis revealed genes where wounding and PopMV had an additive effect on host transcript accumulation

The overlapping section at the centre of each Venn diagram represents a set of genes that were either induced by the combination of wounding and virus (Fig. 1c, Table S5) or repressed by both wounding and virus (Fig. 2c, Table S6). These groups of 77 and 17 genes, respectively, represent an overlap of the wound and virus response. Two UPSC-MIPS classes were over-represented in the group of genes that exhibited an additive increase in transcript abundance due to both wounding and virus infection: ‘Cell Rescue’ and ‘Metabolism’ (Fig. 3) (χ2 analysis, P < 0.001). The genes in this group were found in two hierarchical clusters (Fig. 5). Cluster 9 contained 11 of the genes that were induced by wounding and further induced by the virus, and the remaining 66 genes were found in cluster 1. Thirty of these genes are predicted to be involved in a generalized stress response that may be triggered by abiotic and biotic stresses (Cheong et al., 2002; Kreps et al., 2002). These genes include those predicted to encode metallothionein-like proteins, aluminium-induced proteins, ubiquitin, and the histone protein, H1C. Moreover, five genes encoding proteins implicated in defence response-related calcium signalling were found within this group of 77 genes, including calcineurin B and calmodulin-binding protein (Takezawa, 2000). In addition to genes involved in generalized stress responses, six genes predicted to be pathogen-responsive were included within this group. This included a gene related to Sn-1 from bell pepper, which encodes a protein that has been shown to be wound inducible and is thought to participate in the early disease resistance response (Pozueta-Romero et al., 1995; Osmark et al., 1998).

The transcripts whose abundance was increased on both wounding and virus infection represent a convergence of different signalling pathways activated in response to different stresses. The overlap of genes involved in the response to both wound and pathogen attack has been reported (Cheong et al., 2002; Durrant et al., 2000; Reymond et al., 2000). A generalized stress response that allows the plant to respond to a number of stresses would be advantageous in protecting against opportunistic pathogen infection following wounding, and water loss at the wound site.

The number of transcripts exhibiting an additive decrease in abundance in response to the combined effect of wounding and virus infection was low. The UPSC-MIPS classification of these 17 genes showed that the ‘Metabolism’ category was over-represented by 13% as a proportion of the total 13K array (Fig. 3d) (χ2 analysis, P < 0.001). Hierarchical cluster analysis showed that all these genes were found within one cluster (Fig. 5, Cluster 7). In accordance with the UPSC-MIPS classification, a large proportion of the genes in this cluster are implicated in cell wall restructuring. This suggests that wounding and virus infection may result in a shift in resource allocation that reduces the requirement for cell wall restructuring enzymes. This may be related to the PCD programme that is invoked by wounding.

Hierarchical cluster analysis revealed genes where wounding and PopMV had antagonistic effects on host transcript accumulation

Hierarchical cluster analysis highlighted transcripts where wounding and PopMV infection had opposite effects on the abundance of the transcript. This antagonistic effect was masked by Venn diagram analysis but was clearly illustrated following cluster analysis (Fig. 5, Table S7). Transcripts whose abundance was increased by wounding and decreased by virus were contained within clusters 8 and 14. The genes found within this group are predicted to be involved in pathogen defence. These included chitinases, WRKY transcription factor and binding proteins, and a CCCH-type zinc-finger protein. This small group of transcripts included those that were increased as a result of wounding, but were decreased by the virus. While suppression of within-plant wound signalling by virus infection has been documented (Preston et al., 1999), the mechanisms by which viruses achieve this have not been characterized. Detailed analysis of the transcripts identified within this group may provide some elucidation of this mechanism.

Conclusion

To date, virtually nothing is known about how forest trees respond to viral pathogens. This has certainly been true for the Populus–PopMV pathosystem. Given the range of responses that could be invoked in Populus in response to PopMV, it will be important to have some specific, well-formulated hypotheses to direct future research. The transcriptome analysis conducted here provides insights into specific changes that may be taking place in Populus metabolism, growth and development in response to PopMV infection. Specific groups of genes have been identified that exhibited statistically significant differential transcript abundance in response to wounding and systemic viral infection. The nature of the proteins encoded by these genes should guide future hypotheses aimed at dissecting the mechanisms used by long-lived woody perennial plants to contend with the diversity of stresses they encounter throughout their lifetimes.

An important caveat in transcriptome analyses is that transcript abundance is suggestive of, but does not necessarily directly reflect, what is taking place at the level of the proteome, the cell, or the whole plant. Nevertheless, future work can focus on the specific groups of genes described herein to determine their precise role in the host response. For example, the transcriptome analysis suggests that it might be fruitful to investigate the role of PCD and lignification in the Populus wound response, and the role of metallothionein-like proteins and PME in the response to PopMV. Future work could focus on the dynamics of the expression of these genes by examining transcript abundance at time-points both before and after systemic infection. The robust, statistically supported data obtained in the highly replicated microarray experiment described herein provide an impetus to undertake focused analysis of a smaller number of genes.

Given that a substantial forward and reverse genetics toolbox is coming on line for Populus (Brunner et al., 2004), it will soon be possible to make use of the data described herein to test the roles of specific candidate genes in mounting stress responses in a long-lived woody perennial plant. In addition, future work can test hypotheses that aim to determine the extent to which changes transcript abundance invoked or repressed by PopMV are actually advantageous for the virus, and elucidate how PopMV may guide such changes. Furthermore, the suites of genes for which transcript abundance increased or decreased in response to wounding or PopMV infection provide useful starting points to identify the upstream signalling components involved in the perception and transduction of wound and virus signals in woody plants. Moreover, the availability of the Populus genome sequence should allow the identification of regulatory DNA motifs that may play key roles in the modulation of the specific groups of genes, or regulons, described herein. Such analyses will provide useful insights into factors that impinge on the health of forest tree species.

Acknowledgements

The authors are grateful to Dr David Ellis (CellFor Inc., Vancouver, BC, Canada) for generously providing the Populus clones used in this study. Gratitude is also extended to Andreas Sjödin for assistance with the statistical treatment of the data. We are also most grateful for the thorough and useful constructive criticism provided by two anonymous reviewers. This work was supported by funding from the Knut and Alice Wallenberg Foundation (The Swedish Centre for Tree Functional Genomics) and the Kempe Foundation (M. R-B.). Funding was also provided by the Biotechnology and Biological Sciences Research Council, UK, with supplementary funding from CellFor Inc., Canada. The experiments were carried out under a Department of Environment, Fisheries and Rural Affairs, UK, plant health licence (PHL 225/3895).

Supplementary Material

The following material is available as Supplementary material at http://www.blackwellpublishing.com/products/journals/suppmat/NPH/NPH1151/NPH1151sm.htm

Table S1 Predicted identities of genes that exhibited an increase in relative transcript abundance in response to wounding alone (MI/C, I/C)

Table S2 Predicted identities of genes that exhibited a decrease in relative transcript abundance in response to wounding only (MI/C, I/C)

Table S3 Predicted identities of genes that exhibited an increase in relative transcript abundance in response to virus (I/MI, I/C)

Table S4 Predicted identities of genes that exhibited a decrease in relative transcript abundance in response to virus (I/MI, I/C)

Table S5 Predicted identities of genes that exhibited an increase in relative transcript abundance in all three competitive hybridisation conditions (MI/C, I/MI, I/C)

Table S6 Predicted identities of genes that exhibited a decrease in relative transcript abundance in all three competitive hybridisation conditions (MI/C, I/MI, I/C)

Table S7 Predicted identities of genes that exhibited increased relative transcript abundance in response to wounding and a decrease by virus

Ancillary