Association of the transcriptional response of soybean plants with soybean mosaic virus systemic infection

Systemic infections of plant viruses result from the complex molecular interplay between the host plant and the invading virus (Maule et al., 2000; Golem & Culver, 2003; Whitham et al., 2003; Pompe-Novak et al., 2005). The magnitude of physiological and phenotypic changes in the host during viral infection suggests the involvement of a large number of host genes (Golem & Culver, 2003; Whitham et al., 2006). Thus, the intimate interaction between a plant virus and its host is complicated by the systemic nature of infection and global alterations in host gene expression (Maule et al., 2002; Whitham & Wang, 2004). Dissecting the plant gene-expression network that occurs in response to virus infection should assist in a better understanding of the infection process and ultimately in the control of plant viruses. To achieve this, microarray technology has been adopted to profile global gene expression of the host (mainly model plant species) infected with several positive-sense RNA viruses (Golem & Culver, 2003; Whitham et al., 2003; Ishihara et al., 2004; Marathe et al., 2004; Pompe-Novak et al., 2005; Senthil et al., 2005; Dardick, 2007; Shimizu et al., 2007; Yang et al., 2007).

Soybean mosaic virus (SMV) is a member of the genus Potyvirus in the family Potyviridae, which is the largest plant virus family. SMV is the most prevalent viral pathogen of soybean [Glycine max (L.) Merr.] in the world. Infection by SMV usually causes yield losses of between 35 and 50 % under natural field conditions and of up to 50–100 % in severe outbreaks (Arif & Hassan, 2002; Liao et al., 2002). SMV has a single-stranded, positive-sense [ss(+)] RNA genome approximately 9600 nt in length (Jayaram et al., 1992). Whilst the virus itself has been relatively well studied, the molecular mechanisms underlying soybean responses to SMV infection remain poorly understood.

In this work, we report analysis of gene expression in SMV-infected soybean plants during the course of infection. The most pronounced changes in gene expression occurred when viral RNA accumulation reached the highest level and the leaf underwent moderate symptom development. The SMV-induced and -repressed transcripts were classified based on their putative functions and clustered according to their expression patterns. A cross-sequence comparison of the transcriptional response to SMV and to other ss(+) RNA viruses identified transcripts with similar changes in gene expression. Cumulatively, the gene-expression analysis presented in this study reveals correlations between host gene-expression changes and disease-symptom development in soybean.

Plant materials, virus inoculation and RNA extraction.
Soybean [G. max (L.) Merr.] Williams 82 plants were maintained in a greenhouse at 80 % relative humidity with a day/night regime of 16 h light at 24 °C followed by 8 h dark at 16 °C. A Canadian SMV isolate (strain G2) was used as the inoculum. The unifoliate leaves of 10-day-old soybean plants were rub-inoculated with SMV and control plants were mock-inoculated. The first trifoliate leaf immediately above the SMV-inoculated unifoliate leaf was harvested at 7, 14 and 21 days post-inoculation (p.i.) for RNA extraction. Leaf-to-leaf and plant-to-plant variation was minimized by pooling the total RNA isolated from the first trifoliate leaf of at least 25 plants for each time point. Three independent biological replicates for each time point were performed. All of the mock- and SMV-inoculated soybean plants used in the microarray experiment were randomized and grown in different locations within a greenhouse under the same growth conditions. Total RNA isolation from soybean leaves was carried out by using TRIzol reagent (Invitrogen). The isolated total RNA was purified by using RNeasy Mini columns (Qiagen). SMV was detected by RT-PCR and quantified by Northern blot analysis.

Microarray probe labelling, hybridization, scanning and data analysis.
Soybean cDNA microarrays (18K A series) containing 18 613 soybean cDNAs of low redundancy were obtained from Dr Lila Vodkin (University of Illinois, Urbana–Champaign, IL, USA). The GEO accession number of these expressed sequence tags (ESTs) is GPL3015[NCBI GEO] . Two separate cDNA-labelling reactions (Cy3 and Cy5), one for the SMV-infected leaf and the other for the mock-inoculated leaf, were carried out. In total, 18 microarray slides were used for this study: six array hybridizations, including three reciprocal labelling experiments from three independent biological replications of either SMV-infected or mock-inoculated leaf, at each of the three time points. Dye swaps were performed to ensure that the results were not biased by dye effects. Total RNA (10 µg), isolated from each of the virus-infected and mock-inoculated leaf tissues at different time points, was labelled with either Cy3 or Cy5 fluorescent dye by using a CyScribe post-labelling kit (Amersham Biosciences). Synthesized probes were purified by CyScribe GFX (Amersham Biosciences) and hybridized to soybean chips following the manufacturer's protocol. Subsequent processing of the slides was essentially as described by Moy et al. (2004). The quantified data extracted from the 16-bit TIFF images from ArrayVision v. 6.0 (Imaging Research) were background-subtracted and then analysed with GeneSpring microarray-analysis software version 7.3 (Silicon Genetics), where LOWESS normalization was used to correct for any spatial or intensity-dependent biases within each array. Data analysis was essentially as described by Senthil et al. (2005). In brief, the mean normalized signal-intensity values for each transcript were calculated from six replicate hybridizations (three biological replicates and a dye-swap hybridization for each biological replicate) for each time point. log₂ ratios were then calculated as described in the GeneSpring microarray-analysis instruction manual, where the mean of normalized signal-intensity values from the virus-infected samples was divided by the mean of respective values from control samples (Cy5-infected/Cy3-control or Cy3-infected/Cy5-control in a dye-swap experiment). The fold changes of differentially regulated transcripts in virus-infected samples compared with the control samples were calculated based on these ratios. To select transcripts with significant changes during virus infection, P values (P≤0.05) derived from ANOVA were adjusted by using the multiple testing correction of Benjamini & Hochberg (1995) with a 5 % false-discovery rate (FDR), corresponding to a Q value of ≤0.05 (Storey & Tibshirani, 2003). This multiple testing correction procedure is used to correct for the occurrence of false positives and to maximize the likelihood of finding significant gene sets. In addition to these statistical criteria, we searched for transcripts that showed significant up- (≥2.0-fold) or down- (≤–2.0-fold) regulation at at least one time point. Expression profiles from each time point were clustered based on their similarity in expression pattern by using a hierarchical average linkage clustering algorithm and Pearson correlation distance metric, implemented in the GeneSpring v. 7.3 software. A gene tree heat map was also built based on this analysis.

Functional categorization of transcripts.
The 5' and 3' sequences of each differentially expressed soybean unique sequence of expressed sequence tag (uniEST) were queried against the non-redundant (nr) protein database by using the BLASTX algorithm (Altschul et al., 1997). In addition, we identified the protein in the Dana Farber Cancer Institute (DFCI) Soybean Gene Index database (), using the highest BLASTX score to identify the function of each nr uniEST. Significantly induced and repressed soybean transcripts were grouped into functional groups by performing searches using annotated soybean transcripts against the Arabidopsis MIPS (Munich Information Center for Protein Sequences) functional classification scheme to create a custom set of functional gene categories and subcategories. Transcripts with no matches in the database were labelled as unknown. Both χ² and Fisher's exact tests were carried out to confirm the significance on the representation of the numbers of genes in each functional category at each time point (P≤0.05). Statistical measurement to calculate the significant category was performed as described by Draghici et al. (2003).

Cross-sequence comparison of transcripts regulated differentially in soybean infected with SMV and in plants infected with other ss(+) RNA viruses.
The EST sequences of soybean transcripts regulated differentially in SMV-infected leaf tissues were cross-compared by BLAST-searching against the sequences of the transcripts that were significantly differentially expressed in response to infection by other ss(+) RNA viruses (Golem & Culver, 2003; Whitham et al., 2003; Ishihara et al., 2004; Marathe et al., 2004; Dardick, 2007; Yang et al., 2007). Sequence searches were performed by using the TBLASTX algorithm with default settings. The resulting BLAST output for each transcript was then parsed for high-scoring pair (HSP) and E value. Any hits with the existence of HSP≥100 and E≤10^–20 were indicative of significant similarity (Rubin et al., 2000).

Northern hybridization.
Three independent Northern blots were conducted to validate the microarray data. Total RNA (10 µg) was separated and transferred to a nylon membrane. RT-PCR products derived from primers (given in Supplementary Table S1, available in JGV Online) were used as a probe. The probes were labelled with [α-³²P]dCTP by using a Ready-To-Go DNA-labelling kit (Amersham Biosciences). A phosphorimager and QualityOne quantification v. 4.2 software (Bio-Rad) were used for visualization and quantification of radioactive signals.

Symptoms of SMV infection and viral RNA accumulation in soybean
The unifoliate leaf (the first true leaf) of 10-day-old soybean plants was inoculated mechanically with SMV. At 7 days p.i., the earliest symptoms appeared, analogous to mild mosaic in the first trifoliate leaf (Fig. 1a). Moderate mosaic and vein-clearing symptoms were evident in the trifoliate leaf at approximately 14 days p.i. Symptoms of the trifoliate leaf turned to severe mosaic and mild mottling at about 21 days p.i. (Fig. 1a). Accumulation of viral RNA was detected by Northern hybridizations in the first trifoliate leaf at 7 days p.i., reached its peak level at 14 days p.i. and then decreased by 1.7-fold (58 % of the maximum viral RNA level) at 21 days p.i. (Fig. 1b).

(56K): Fig. 1. Symptom development and SMV detection in systematically infected soybean trifoliate leaves at different times post-inoculation (days p.i.). (a) The left panel shows symptom development in the first trifoliate leaf of soybean plants in which the unifoliate leaf was inoculated mechanically with SMV and the right panel shows the corresponding leaf from mock-inoculated control plants. (b) Northern blot hybridizations showing viral RNA accumulation in the infected trifoliate leaf. The actin gene (GenBank accession no. AI494739) was used as an internal control to normalize the differences in total RNA samples loaded. The relative amount of SMV genomic RNA after normalization is given as a percentage.

Experimental design and data normalization for consistency of the microarray data
Reliability of the microarray data was assessed by normalizing the mean signal intensities of mock-inoculated samples against all replicates for each time point on a scatter plot (data not shown). A high correlation coefficient (0.90–0.93) among mock-inoculated samples for each time point was observed, indicating low technical and biological variability. In addition, variation differences between dye-swap experiments from mock-inoculated samples resulted in a correlation coefficient of 0.90–0.91 for each time point, indicating that dye effects were insignificant.

Identification of soybean transcripts regulated differentially over time
Gene-expression data were analysed by one-way ANOVA to identify differentially expressed transcripts. By using stringent criteria with a 5 % FDR that corresponds to a Q value of ≤0.05, we identified approximately 4.8 % nr transcripts (894 of the 18 613 soybean transcripts printed on the chip) that were significantly differentially expressed (induced or repressed) in response to SMV infection at three different time points. These significantly differentially regulated transcripts were filtered further by using a 2.0-fold increase or a –2.0-fold decrease in signal intensity. At 7 days p.i., 62 nr transcripts were induced and 46 were repressed by SMV; at 14 days p.i., 154 were induced and 90 repressed; at 21 days p.i., 100 were induced and 45 repressed. This final filtering resulted in 316 transcripts (representing 273 nr transcripts) that were induced by ≥2.0-fold, and 181 transcripts (representing 173 nr transcripts) that were repressed by ≤–2.0-fold, as illustrated by Venn diagrams (Fig. 2a, b). Comparison analysis between the 273 induced and the 173 repressed nr transcripts identified 168 transcripts that were significantly induced only at one time point but not significantly repressed at other time points, 68 transcripts significantly repressed only at one time point but not significantly induced at other time points and 105 transcripts significantly induced and repressed at different time points (Fig. 2c).

(15K): Fig. 2. Venn diagrams showing the differential distribution of induced (≥2.0-fold) and repressed (≤–2.0-fold) transcripts in SMV-infected leaf tissues at three different time points. A statistical cut-off with a 5 % FDR corresponding to a Q value of ≤0.05 was used to determine whether transcripts were differentially expressed. Number of transcripts shown in the non-overlapping sectors represents unique significant transcripts at each time point, whereas transcripts within the intersection represent the number of shared significant transcripts. (a) nr transcripts induced significantly by ≥2.0-fold at at least one time point (n=273); (b) nr transcripts repressed significantly by ≤–2.0-fold at one time point at least (n=173); (c) overlapping of 273 and 173 nr transcripts that were induced and repressed significantly by SMV infection, respectively.

Functional classification of soybean transcripts induced and repressed in SMV-infected leaf tissues
Soybean transcripts induced or repressed during SMV infection were classified according to the predicted functions by using BLASTX searches against the NCBI nr database. These transcripts were grouped further, essentially following the method of the Arabidopsis MIPS functional classification scheme. With the information gathered from the MIPS Arabidopsis database, we were able to assign putative functions to 63.2 % of the 18 613 transcripts on the slide, whilst the remaining 36.8 % (6854 transcripts) with no matches in the database were classified into the unknown category (Table 1). uniESTs were assigned with a putative function only if the best match had an E value of ≤10^–20. Based on the putative functions, the differentially expressed transcripts were classified into 11 functional categories (Fig. 3; see Supplementary Tables S2, S3 and S4, available in JGV Online). In four instances, subcategories were created to interpret the data better. χ² and Fisher's exact tests suggested that the majority of transcripts are associated with defence, chromatin regulation and cytoskeleton reorganization, protein synthesis and translation, metabolism (including cell wall-related and sulphur-assimilation transcripts) and development/storage proteins (including hormone-, chloroplast- and photosynthesis-related transcripts), comprising approximately 60 % of all differentially regulated transcripts changed significantly at a P value of ≤0.05 (Table 2; Fig. 3).

Table 1. Assignment of up- and downregulated transcripts that change in expression by ≥2- or ≤–2-fold in each functional class

(33K): Fig. 3. Functional distribution of soybean transcripts significantly induced and repressed in SMV-infected soybean leaf tissues at each time point. Bars represent the number of transcripts downregulated (≤–2.0-fold) or upregulated (≥2.0-fold) in each functional category at each time point. Text in bold indicates subcategories created under a major functional category.

Table 2. Statistical analysis to determine the enriched functional categories at each time point post-infection using χ2 and Fisher's exact tests

To visualize expression profiling of the transcripts at all three time points, hierarchical clustering was employed, using the 273 induced and 173 repressed transcripts (Fig. 4). A gene tree heat map was built by using the Pearson correlation distance metric with an average linkage clustering algorithm. Horizontal time points, representing differences in expression of transcripts, were plotted against the vertical grouping, corresponding to expression similarity. Six distinct groups of expression pattern were identified from the clustering analysis. A complete list of genes induced or repressed in different cluster groups, along with their expression levels and putative functions, is provided in Supplementary Table S2 (available in JGV Online).

(70K): Fig. 4. Hierarchical clustering of significantly differentially expressed soybean transcripts. Gene tree heat map consists of 273 up- and 173 downregulated transcripts. Expression values are colour-coded, with red indicating upregulation (>1.0) by SMV infection, green indicating downregulation (<1.0) and black indicating no change (1.0) in expression. The intensity of colour represents the degree of gene-expression level. A complete list of genes in each cluster is given in Supplementary Table S2 (available in JGV Online).

Clustering analysis also revealed that the greatest changes in gene expression took place in soybean leaves with moderate mosaic symptoms (14 days p.i.), whilst fewer differentially expressed transcripts were identified at the earlier or later stages when infected leaves displayed mild or severe mosaic symptom development. Thus, the magnitude of host gene-expression changes correlated with the amount of viral RNA detected in the leaves. These observations may reflect specific host responses to active virus replication and spread, apparently peaking at 14 days p.i. and declining thereafter. Another intriguing aspect is the large shift in host gene-expression patterns observed between 14 and 21 days p.i., coinciding with the onset of more severe disease symptoms in the infected leaves.

Confirmation of microarray data by Northern hybridization
As most gene changes occurred at 14 and 21 days p.i., nine transcripts either up- or downregulated significantly at 14 or 21 days p.i. were selected for Northern hybridizations to validate the microarray data. Overall expression patterns of all nine transcripts analysed by Northern blotting were consistent with those obtained by microarray hybridizations (Fig. 5; Supplementary Fig. S1, available in JGV Online). The relative expression ratios of the transcripts analysed by Northern hybridizations were lower than those from microarray hybridizations. This could be attributed to methodological variations, for example, normalization to actin expression in Northern blotting compared with a more global normalization method used in microarray hybridizations (Taniguchi et al., 2001; Yao et al., 2004).

(31K): Fig. 5. Confirmation of relative expression levels of the transcripts selected from the microarray analysis with Northern blot hybridizations. Nine transcripts were chosen to validate the microarray data. The GenBank accession number of each transcript is provided. The y-axis indicates the fold change of the transcripts from microarray (indicated by black bars) and Northern (indicated by grey bars) hybridizations. The x-axis represents time post-inoculation (days). Signals for Northern hybridizations are from three biological replicates carried out in this study.

Associations of differentially expressed transcripts in general metabolism with virus infection and symptom development
To gain insight into the spatial association of significantly differentially regulated transcripts with SMV infection and symptom development, we examined transcripts related to general metabolism. Lipoxygenase (LOX) came to our attention immediately, as seven of eight transcripts encoding LOX enzymes were repressed at 7 days p.i., with a mean of –1.7-fold hybridization intensity (see Supplementary Table S3, available in JGV Online). Expression levels of these LOX transcripts increased by a mean of 1.2-fold at 14 days p.i. and 2.6-fold at 21 days p.i. (Supplementary Table S3). LOX is a dioxygenase that catalyses the hydroperoxidation of polyunsaturated fatty acids. Induction of LOX activity has been consistently considered as a pathogen-induced defence response, such as in potato infected by a fungal pathogen (Kolomiets et al., 2000). In virus–plant interactions, Aranda et al. (1996) showed that the level of LOX1 transcript was depleted significantly in pea embryo cells undergoing pea seed-borne mosaic virus replication, and recovered at a shorter distance behind the infection front when virus replication declined. Based on the above study, LOX is probably associated with virus infection. However, past microarray studies on soybean have shown that LOX transcripts change dramatically under diverse treatments (Thibaud-Nissen et al., 2003; Moy et al., 2004). Therefore, it is also possible that the expression changes of this gene family may simply result from biotic stress.

Another interesting metabolic transcript repressed at 7 days p.i., but induced at 14 days p.i., is a transcript encoding callose synthase (GenBank accession no. AW278243). This enzyme is deposited as a layer between the plasma membrane and the cell wall (Jacobs et al., 2003; Nishimura et al., 2003; Ueki & Citovsky, 2005). Beffa et al. (1996) showed that the increased deposition of callose in and around tobacco mosaic virus (TMV)-induced lesions in β-1,3-glucanase-deficient tobacco plants resulted in decreased susceptibility to TMV. In the current study, repression of callose synthase at 7 days p.i. and upregulation of this gene at 14 days p.i. may be associated with a delayed resistance response in soybean (see Supplementary Table S3, available in JGV Online).

An intriguing subset of transcripts that were induced substantially at 14 days p.i. includes enzymes critical in the Krebs cycle, e.g. NADP-dependent malate dehydrogenase, cytosolic malate dehydrogenase and acetyl-CoA carboxylase; the oxidative pentose phosphate pathway, e.g. phosphoenolpyruvate/phosphate translocator; amino acid synthesis, such as amine oxidase, adenosine 5-phosphosulphate reductase, adenosylhomocysteinase and S-adenosylmethionine decarboxylase; sugar metabolism, e.g. sucrose synthase; and carbohydrate synthesis, e.g. β-amylase 1, mannosyloligosaccharide 1,2-α-mannosidase, GDP-mannose 3,5-epimerase and glutamate carboxypeptidase. Previous studies have suggested that the appearance of disease symptoms such as chlorosis on cucumber mosaic virus (CMV)-infected leaf tissues in cucumber is correlated with increased glycolysis, phosphoenolpyuruvate phosphate, respiration and starch accumulation, decreased photosynthesis and reduction in total protein synthesis (Técsi et al., 1996). The accumulation of sugar and starch in virus-infected plants is correlated with symptom development, such as altered coloration in leaf, leaf distortion, chlorosis, mosaic and stunted phenotype (von Schaewen et al., 1990; Herbers et al., 1997). Based on the above findings, it is possible that host gene-expression changes related to sugar, starch and amino acid metabolism constitute part of the host response to infection and that these changes may contribute to symptom development in soybean.

Repression of transcripts related to hormone, anti-oxidative metabolism, cell wall, sulphur assimilation, chloroplast and photosynthesis with SMV infection and symptom development
A wide range of symptoms in virus-infected plants can be linked to plant hormones, such as auxin or gibberellin (Padmanabhan et al., 2005, 2006; Culver & Padmanabhan, 2007). Interaction between the TMV replicase and auxin/indole acetic acid proteins has been shown to affect auxin-mediated pathways and contribute to disease symptoms in Arabidopsis (Padmanabhan et al., 2005, 2006). Transcripts encoding auxin-repressed protein were downregulated significantly in potato plants infected with potato virus Y at 7 days p.i. (Pompe-Novak et al., 2005). In rice plants infected with rice dwarf virus, the expression of a gibberellin-biosynthetic enzyme, ent-kaurene oxidase, was repressed, leading to the appearance of dwarf phenotype (Itoh et al., 2004; Zhu et al., 2005). In this study, five auxin-associated transcripts and one gibberellin-associated transcript were repressed in SMV-infected soybean leaf tissues at 14 or 21 days p.i. (see Supplementary Table S4, available in JGV Online). It is not clear whether these hormone-related gene changes also contribute to SMV symptom development in soybean.

Reactive or activated oxygen species have been suggested to be key mediators of local and systemic resistance responses in incompatible plant–pathogen reactions and to be involved in symptom development and pathogenesis in compatible plant–virus interactions (Sandermann, 2000; Hernández et al., 2004). In this study, the expression of two anti-oxidative metabolism-related ESTs encoding catalase (CAT; GenBank accession no. AW349008) and glutathione S-transferase (GST; GenBank no. AW472161) was suppressed significantly in the SMV-infected leaf at 14 days p.i. Another key detoxification gene, superoxide dismutase (SOD), was downregulated significantly at 7 days p.i. In contrast, two ESTs (GenBank accession nos AW471843 and AI938378) encoding peroxidase (PO) were induced significantly at 7 days p.i. (by about 1.7-fold) and 14 days p.i. (by >4-fold). Consistent with these transcriptional profiles, Zhuang et al. (1993) reported that SOD activity decreased with an increase in PO activity in the SMV-infected soybean leaf. A decrease of CAT and SOD activity and an increase of PO activity were also found in many other compatible host–virus interactions, such as in Phaseolus vulgaris infected with white clover mosaic virus (Clarke et al., 2002). Upregulation of PO transcripts and downregulation of CAT, SOD and GST transcripts may induce an oxidative stress in the early infection process. Such an anti-oxidative metabolism imbalance may be associated with the progression of SMV infection and symptom development, as suggested for the plum pox virus (PPV)–peach interaction (Hernández et al., 2004).

The other interesting changes observed in this study were the significant repression of cell wall-related genes. This gene group, containing several hundred different structural proteins and cell wall-related enzymes, is known to be a major determinant of cell morphogenesis in plants (Milioni et al., 2001; Huckelhoven, 2007). In SMV-infected leaves, 10 cell wall-related genes encoding proteins for reassembly (six genes), matrix polymers (two genes) and expansins (two genes) (see Supplementary Table S4, available in JGV Online) were downregulated significantly at 14 days p.i. Downregulation of cell wall-related transcripts during SMV infection may be associated with a reduction in cell wall cross-linking that contributes to disease-symptom development. This is consistent with the recent finding that cell wall-related transcripts were downregulated significantly in Arabidopsis infected with turnip mosaic virus (TuMV) (Yang et al., 2007) and in rice infected with rice dwarf virus (Shimizu et al., 2007). In these two studies, repression of cell wall-related transcripts was correlated with symptom development.

Chlorosis is usually associated with susceptible interactions in which virus replicates and moves throughout the plant. Chlorotic symptoms appear as yellowed areas in expanded leaves. In fully developed leaves, virus is largely confined to light-green areas, where it seems to interfere with chloroplast structure, function and/or development (Culver et al., 1991). In this study, 18 chloroplast and four photosynthesis transcripts belonging to the chloroplast and photosynthesis functional category were downregulated in the SMV-infected leaf at 14 days p.i. (see Supplementary Table S4, available in JGV Online). Recently, Yang et al. (2007) showed that a large fraction of the metabolic genes encoding chloroplast- and photosynthesis-related proteins were repressed significantly in Arabidopsis plants infected with TuMV. Downregulation of photosynthesis-related genes is also correlated with the development of infection symptoms, such as chlorosis, stunting or mosaic, in plants (Técsi et al., 1994, 1996; Maule et al., 2002; Pompe-Novak et al., 2005; Espinoza et al., 2007).

In this study, 10 transcripts involved in sulphur assimilation and utilization were upregulated significantly at 14 days p.i., followed by downregulation at 21 days p.i. These transcripts encode enzymes such as methionine synthase, adenosine 5-phosphosulphate reductase, adenosylhomocysteinase and adenosylmethionine decarboxylase, which play a major role in sulphur uptake. Recently, Yang et al. (2007) also reported that TuMV infection suppressed expression of genes involved in the sulphur-uptake pathway, affecting plant growth and development. Taken together, these results suggest that gene-expression changes in hormone, cell-wall biogenesis, chloroplast, photosynthesis and sulphur-assimilation pathways may contribute to symptom development in SMV-infected soybean plants.

Association of upregulated transcripts in protein synthesis and disease resistance with SMV infection
Translational regulation is a critical component of the cellular response to a variety of types of stress, such as viral infection, nutrient deprivation and heat shock. In the current study, 11 differentially regulated transcripts related to protein synthesis and translation were identified. These transcripts, including two translation-elongation factors and nine ribosomal transcripts, were affected slightly at 7 days p.i., but upregulated substantially at 14 days p.i. (see Supplementary Table S3, available in JGV Online). Similar induction of ribosomal genes was also observed in Nicotiana benthamiana infected with PPV (Dardick, 2007) and in Arabidopsis infected with TuMV (Yang et al., 2007). It is not known whether the increased expression of these ribosomal proteins is a simple stress response to compensate for the host cell that may lack sufficient translation components to maintain its viability, because many such components are hijacked by the virus for its genome translation and replication.

Transcripts encoding proteins related to defence and virulence were significantly over-represented at 14 and 21 days p.i., but not at 7 days p.i. (Table 2), suggesting a general delayed defence in the SMV-infected leaf. Of the 24 upregulated defence-related transcripts, a subset of 17 defence-related transcripts, such as pathogenesis-related (PR) protein 3 (chitinase), GST10, HSP, SOD and PO, were either downregulated or slightly affected at 7 days p.i., but substantially upregulated at 14 or 21 days p.i. (see Supplementary Table S3, available in JGV Online). These transcripts have been found to be vital in disease signalling, plant defence and stress responses (Cardinale et al., 2002; Marathe et al., 2004; Garcia-Brugger et al., 2006; Whitham et al., 2006). It is possible that, at late infection stages, the soybean plant responds to SMV infection by expressing defence-related genes, as in the case of other virus-infected plants (Whitham et al., 2003; Senthil et al., 2005; Shimizu et al., 2007). Collectively, these data suggest that there is a delayed host defence response; the immune reponse in soybean plants is not activated until the relatively late stage of infection.

Association of a common set of induced transcripts in general stress- and defence-related categories with infection by SMV and other ss(+) RNA viruses
To examine whether the diverse ss(+) RNA viruses have the ability to elicit common gene-expression changes, a cross-sequence comparison analysis was performed. The uniEST sequences of 894 significantly differentially expressed (either up- or downregulated, regardless of fold change) soybean transcripts identified in this study were BLAST-searched against all of the differentially regulated transcripts identified in potato in response to infection by three distinct fruit tree viruses, namely PPV, tomato ringspot virus (ToRSV) and Prunus necrotic ringspot virus (Dardick, 2007), and in Arabidopsis in response to infection by CMV (Ishihara et al., 2004; Marathe et al., 2004), TuMV (Yang et al., 2007), TMV (Golem & Culver, 2003) and five positive-sense plant viruses including TuMV, oilseed rape mosaic virus, turnip vein clearing virus, potato virus X and CMV (Whitham et al., 2003). As a result, 107 unique transcripts were identified (see Supplementary Table S5, available in JGV Online).

The 107 unique transcripts were classified into four major groups. Group A contains 43 soybean transcripts induced by SMV infection whose sequences match those of transcripts induced by infection with other ss(+) RNA viruses. Six of them encode ribosomal proteins (three 40S and three 60S ribosomal proteins). Group B includes 34 soybean transcripts induced during SMV infection, but these transcripts are repressed in plants infected by other ss(+) RNA viruses. For example, a transcript encoding plasma membrane ATPase (GenBank accession no. BE020672) was induced 1.3-fold by SMV infection at 21 days p.i., but was downregulated 1.5-fold by CMV infection at 6 h p.i. Three histone proteins related to chromatin regulation were induced approximately 1.2-fold by SMV infection at 7 days p.i. and were downregulated approximately 1.2-fold by ToRSV infection at 14 days p.i. Group C contains 19 SMV-induced soybean transcripts, each of which shares sequence similarity with more than one transcript induced or repressed in plants infected with other ss(+) RNA viruses. One example is a transcript induced by SMV infection at 14 days p.i. encoding heat-shock protein 70 kDa (Hsp70) that shares sequence similarity with four Arabidopsis transcripts (AT5g02500, AT3g09440, AT3g12580 and AT5g02490) induced by TuMV infection at 10 days p.i. (Yang et al., 2007). Group D consists of 11 SMV-induced soybean transcripts. In this group, more than one SMV-induced soybean transcript matches the sequence of a transcript induced or repressed by infection with other ss(+) RNA viruses. For example, the sequences of two photosystem I transcripts (GenBank accession nos AI461105 and AI495711) induced by SMV infection at 21 days p.i. match the sequence of an Arabidopsis transcript (At4g12800) induced by CMV infection at 12 h p.i. (Ishihara et al., 2004).

Among the 107 unique soybean transcripts, 62 (approx. 58 %), 14 (13 %), 12 (approx. 11 %) and 35 (33 %) transcripts share sequence similarity with transcripts regulated differentially by infection with other viruses identified by Ishihara et al. (2004), Marathe et al. (2004), Yang et al. (2007) and Dardick (2007), respectively (see Supplementary Table S5, available in JGV Online). As diverse plant RNA viruses may elicit a general stress and defence response to plant virus infection (Whitham et al., 2003), we directed our searches specifically to a common subset of such transcripts. Indeed, soybean transcripts involved in general stress and defence, such as LRR protein kinase (GenBank accession no. AW186515), pyruvate kinase (AW830175), protein phosphatase 2C (AW830157), protein kinase (AW831515), calmodulin (AI441176), peroxidase (AI496108), 2-Cys peroxiredoxin (AI443769), hypersensitive-induced response protein (AT3G01290) and universal stress protein (USP; AI735896), were identified from this analysis. It seems that, despite the differences in symptoms, hosts and viruses, there is a commonly shared stress and defence response in plants to the infections of ss(+) RNA viruses. Interestingly, all of these transcripts were upregulated by SMV infection at 14 or 21 days p.i., but not at 7 days p.i., further suggesting a delayed defence response in SMV-infected soybean plants.

In conclusion, high-throughput microarray analysis used in this study permitted us to draw some general associations between SMV infection and gene-expression changes in the SMV-infected leaf. Even though large sets of informative data were generated, our study is limited in certain ways. First, the total RNA samples derived from SMV-infected soybean leaves contained a mixture of uninfected, infected and post-infected cells, and therefore spatial and temporal information on the patterns of gene expression was not optimized. Second, certain zones in the SMV-infected leaf tissue might contain little or no viral RNA, commonly referred to as dark-green islands (Atkinson & Matthews, 1970; Moore & MacDiarmid, 2006). These zones may dilute or mask expression changes in the infected cells. Despite these limitations, this analysis suggests clearly that a number of genes are probably associated with SMV-compatible infection and symptom development, and activation of defence-like genes seems not to occur until virus accumulation reaches its highest level. Such a delayed defence response may be critical for SMV to establish its systemic infection. Further targeted functional studies of these differentially expressed transcripts, particularly those involved in defence, cell wall, sulphur assimilation, hormone, chloroplast and photosynthesis pathways, will provide new insights into SMV–soybean interactions.

The authors wish to thank Dr Vaino Poysa (AAFC) for soybean seeds, Dr Lila Vodkin (University of Illinois, Urbana–Champaign, IL, USA) for the microarray slides, Drs Mark Gijzen, Taiyun Wei, Dinah Qutob (AAFC) and Martina Stromvik (McGill University, Montreal, QC, Canada) for critical reading of the manuscript and Alex Molnar for photography. This work was supported in part by Ontario Soybean Growers, the AAFC Canadian Crop Genomics Initiative and the National Science and Engineering Council of Canada.

Footnotes

^,†, Alla G. Gagarinova^1,2 †Present address: Department of Molecular Genetics, University of Toronto, Toronto M5S 1A8, Canada.

The NCBI Gene Expression Omnibus (GEO) accession numbers for the microarray data from this work are GSE9824, GPL6258, GSM247941, GSM247942 and GSM247943.

A supplementary figure showing confirmation of microarray data by Northern hybridizations, as well as five supplementary tables and references cited therein, are available with the online version of this paper.