SCMV experiment: within-time-point analysis

The majority of significantly differentially expressed sequences was identified for time point T2 (0.69% of the 4578 observations of all differential genes for a given time point), followed by T4 (0.24%), T3 (0.19%), T9 (0.13%), T5 (0.10%), and T1 (0.04%) (Figure 1). The number of expressed sequences was significantly different between T2 and all other time points at a significance level of 1%, and between T1 and T4 at a significance level of 5% (McNemar test).

In the whole set of 27468 observations a similar distribution of up-regulated genes was found for all four near-isogenic genotypes. When considering only significantly differentially expressed genes, the majority of genes were up-regulated in genotype F7 ^SS/RR, carrying the Scmv1 resistance locus (Figure 2). Except for 16.9% of genes, the folds of change of significantly differentially expressed genes were below 2-fold (Figure 3).

Six pair-wise genotype-contrasts were considered. Most significantly differentially expressed genes summarised over all time points were found between F7 ^RR/RR - F7 ^SS/RR (18 genes), followed by F7 ^RR/RR - F7 ^RR/SS (11 genes), and F7 ^SS/SS - F7 ^RR/RR (9 genes) (Table 1). Out of the 28 different genes showing differential expression, 13 were in common among two pair-wise contrasts, 7 among three and 2 among four pair-wise contrasts. None of the genes showed common differential expression among five or all six pair-wise comparisons.

Two genes (605018B04.x1 and 605018B03.x1) were most commonly significantly differentially expressed within time points and for different genotype pairs, 12 and 7 times respectively (Table 1, see Additional file 1).

SCMV experiment: between-time-point analysis

14070 out of 45780 observations showed significant differential expression between time points at the level of p ≤ 0.05. Most significantly differentially expressed genes were found in comparisons of time point T1 with all other time points (Figure 4).

The majority of genes were up-regulated in T1, with significantly differentially expressed genes surpassing all observations in this time point by more than double (Figure 5).

All 28 genes from the within-time-point analysis were also significantly differentially expressed in the between-time-point analysis (see Additional file 2).

Comparison to previous SCMV studies

In order to maximise the chance to identify putative candidate genes involved in resistance to SCMV and MDMV, pre-selected SCMV candidate genes were spotted on the array utilised for this study. Shi et al. [18, 19], in their studies on SCMV infected maize reported 302 and 497 differentially expressed genes when utilising macro- and microarray approaches, respectively. 451 of those genes were successfully amplified and included in our microarray experiments. The remaining about 40% of genes included on our custom microarray comprised resistance candidate genes and resistance gene analogues. For the 65 sequences, differentially expressed within-time points, 80% were derived from the pre-selected genes. When considering redundancy of detecting the same gene within time points repeatedly, 4.4% (21 genes) of the pre-selected genes, but only 1.6% resistance gene analogues (7 genes) showed differential expression.

MDMV experiment: within-time- point analysis

Only two genes were significantly differentially expressed in the MDMV experiment at a FDR level of p ≤ 0.05 within time points. The two genes were significantly differentially expressed at two different time points and all were up-regulated in F7 ^SS/SS. One of the two genes (605018B04.x1) was also significantly differentially expressed within time points in the SCMV experiment. The fold of change did not exceed 3-fold for all three significant gene × time point combinations (data not shown).

GO description for 605018B04.x1 was binding activity (metallothionein-like protein), whereas no GO assignment but homology to a solanesyl diphosphate synthase was found for the second gene (947026D04.x1) http://www.tigr.org/.

MDMV experiment: between-time-point analysis

Forty-six percent out of 7260 observations showed significant differential expression between time points at a FDR p level ≤ 0.05. The majority of up-regulated differentially and significantly differentially expressed genes were found for T1 (see Additional file 3). Distribution of genes regarding their fold changes is shown in Additional file 4.

The two genes (605018B04.x1, 947026D04.x1) identified in the within-time-point analysis as significantly differentially expressed were also significantly differentially expressed in the between-time-point analysis (see Additional file 5).

Comparison to previously published data

The purpose of this study was to identify and validate genes involved in resistance response to SCMV and MDMV. In previous SCMV experiments [18, 19], a set of candidate genes was identified to show significant differential expression between near-isogenic SCMV resistant and susceptible inbred lines. These genes, together with resistance gene analogs (RGAs) were spotted on our cDNA SCMV array. Twice as many genes based on earlier studies showed differential expression as compared to RGAs, indicating usefulness of pre-selection and reliability of microarray approach.

Comparison of SCMV and MDMV experiments

The MDMV experiment was set up to compare response of isogenic lines containing Scmv1 and/or Scmv2 regions from the resistant FAP1360A inbred line to related but different viruses. Comparative studies of related viruses displaying common symptoms in the same host offer an opportunity to link changes in global gene expression to specific symptoms and to identify common genes involved in resistance responses [22]. It was assumed in the experimental design that F7 ^RR/RR demonstrates full resistance to both SCMV and MDMV [20]. However, the finding of only very few significantly differentially expressed genes within time points in the MDMV compared to the SCMV experiment supports the reliability of our SCMV results, because the MDMV experiment compared only susceptible, while the SCMV experiment resistant and susceptible isogenic genotypes.

Comparison of array and qRT data

One of the most important issues after performing microarray experiments is validation of their results. Preferentially significantly differentially expressed genes within time points for comparisons of F7 ^RR/RR with F7 ^SS/RR or F7 ^RR/SS were considered. The 26S ribosomal RNA gene was chosen because of its putative map position on chromosome 6 and high fold change, whereas the gene putatively expressing a metallothionein-like protein was selected due to its high fold change and differential expression at all time points. The gene putatively expressing a 14-3-3-like protein GF14-6 and the S-adenosylmethionine synthetase 1 were selected due to their expression pattern, and the germin-like protein 4 due to its map location on chromosome 3.

The genes coding for metallothionein-like protein, S-adenosylmethionine synthetase 1, and germin-like protein 4 were confirmed by qRT experiments to be differentially expressed at significant levels. The germin-like protein showed very high PCR efficiencies, a likely result of template quantity, presence of inhibitor or high RNA purity (data not shown). Differential expression of the 26S ribosomal RNA gene and the second S-adenosylmethionine synthetase 1 gene was confirmed, but only in one of the four biological replications. Possible reasons of these findings are a) false positive results in microarray experiments, b) pooled performance (biological and technical replications) of array data as compared to analysis of pooled technical but separate biological replications for qRT, or c) low fold change values from microarray experiments (despite of significance), which might be difficult to reproduce by other methods if close to the significance threshold. Similar findings were reported by Czechowski et al. [23] and Dallas et al. [24], who indicated that only genes with higher expression levels from microarray experiments (> 1.5 folds of change) are likely to be validated by qRT-PCR.

Molecular mechanisms of plant response reaction to virus invasion

Candidate genes identified in this study were annotated according to Maize Gene Ontology Assignment to three major groups: genes encoding catalytic activity (oxidoreductase and hydrolase activities), genes with molecular function unknown, and genes encoding transporter activity (electron transporter activity). Catalysis provides chemical energy required for maintenance of living cells and is of particular importance for the plant while delimiting pathogen invasion. Catalytic activity of oxidoreductase (redox reactions) has been speculated to be crucial for the survival of host plants, since it is required for energy transduction, operation of many anabolic and catabolic pathways, nutrient assimilation, and for defence against disease organisms [25].

One of the functions of redox activity is the formation of hydrogen peroxide (H₂O₂), which generates a response to pathogen attack and enhances cells lignification and/or structural protein polymerization, thus producing a mechanical barrier for the invader that have not yet entered the symplasm. A rapid oxidative burst of H₂O₂ production has been previously reported as a result of pathogen invasion [26–29]. Moreover, Apostol et al. [30] speculated that H₂O₂ production may be part of a signal transduction mechanism for coordination of cellular defences. Furthermore, redox reactions depend on electron supply and active oxygen species. Therefore, electron transfer activity is thought to be unseparable from defence mechanisms [31].

Additional comparative gene annotation was based on results obtained by Whitham et al. [21] on Arabidopsis infected with five distinct viruses, including a mosaic potyvirus. Three homologous genes (pathogenesis-related protein, alcohol dehydrogenase and glutathione S-transferase) were identified in the classes cell rescue, defence, death, and ageing of Arabidopsis, thus indicating the reliability of microarray technology for detection of pathogenesis-related genes.

Association of map positions of differentially expressed candidate genes with Scmv1 and Scmv2

A virus resistance gene needs to be expressed before pathogen invasion, in order to enable a rapid response after infection. Its expression may increase further after virus attack. In previous QTL experiments [1], Scmv1 (as QTL) was detectable at each scoring time point after inoculation, whereas Scmv2 became first detectable and induced at later scoring stages. Assuming that both Scmv1 and Scmv2 are single genes, clustering of differentially expressed genes in the Scmv1 and Scmv2 genomes regions could either be due to linkage drag of genes located in the polymorphic regions in isogenic line contrasts without effect on SCMV resistance (caused by heterozygosity), or clustering of genes involved in SCMV resistance in the Scmv1 and/or Scmv2 regions. Assuming that each of the two donor segments is 40 cM long, both regions would represent 5% of the total genome (80 cM out of 1600 cM average maize genome size). The percentage of candidate genes (one gene each) falling into either the Scmv1 (6.00–6.01) or Scmv2 bins (3.04–3.05) is 8% (2 out of 24 putative map locations for the 14 mapped differentially expressed genes, Table 2), which is not significantly different from the 0-hypothesis tested by the X² test (no clustering of differentially expressed genes). Thus, no evidence of clustering of differentially expressed genes in the Scmv1 and Scmv2 regions was found, which also means, that differential gene expression of genes due to linkage drag was limited. Moreover, finding 2 out of 24 gene locations in agreement with the Scmv1 and Scmv2 genome locations is not significantly different from expectations based on probability theory. Thus, colocalization of differentially expressed genes is only a weak indicator for candidacy of being Scmv1 or Scmv2.

Time course data

In contrast to fungi and bacteria, viruses are directly transferred by specific vectors into host cells. The infection cycle includes virus disassembly, RNA translation and replication, new viral particle assembly, and movement. The time required for these processes may vary and be virus/host dependent.

Immediate response of plants against virus attack is obligatory for fast activation of defence mechanisms [32]. Significant differential expression of majority of genes in our study one hour post inoculation, dropping down to about one-third twelve hours post inoculation may be a result of such rapid responses of host plant to viral infection. The anticipation of detected genes in mechanical stress responses proved insignificant in applied mock control experiment.

To arque our statement we present the potyvirus study of Maule et al. [33], where induction of genes related to pathogenesis (putative protein targeting, virion assembly and trafficking) and to general stress responses was detected immediately after inoculation, similar to the study of Love et al. [34]. Furthermore, Marathe et al. [35], detected robust plant resistance responses at transcriptome level to potyvirus infection at early (at least three hours post inoculation) time points. Changes in gene expression due to responses initiated by specific interactions between virus and host proteins, for instance potyvirus coat protein VPg with plant eIF4 initiation factor have also been reported [36]. Moreover, rapid silencing and blockade of viral protein expression by modified RNAi was observed within one-two hours post inoculation [37]. In contrast, however Yang et al. [38] concluded that changes in gene expression depend on virus type and its accumulation (threshold of viral RNA and proteins) in infected tissues hence occur rather at later stages post inoculation. Similarly, Whitham et al. [39] stated that actual transcriptional changes depend on the progress of viral infection.

The conflicting observations of different research groups might be due to host-virus system specificity and need to be studied in more detail, before generalizations can be made.

Candidate genes and their involvement in signal transduction pathways

Metallothioneins are known to be involved in metal binding/metabolism and detoxification reactions in animals and yeasts [40, 41]. Slightly modified functions of metallothioneins have been reported in plants, where their increased expression was observed in senesced leaflet and abscission zones, under ethylene induction, or as a consequence of mechanical wounding when infecting tobacco with TMV virus was obsereved [42–44]. Moreover, a possible role of metallothioneins in controlling intracellular redox potential and activation of oxygen detoxification, a common strategy used by plants after pathogen invasion was suggested by Hamer [40]. Finding of metallothionein-like protein expressed in all time points (including mock control at significant fold of change) and for both viruses may suggest its differential expression as a cause of mechanical wounding. However, its participation in pathogen control cannot be ruled out.

S-adenosylmethionine synthetase is a key enzyme involved in generation of S-adenosylmethionine from methionine. S-adenosylmethionine is a major methyl donor in plants involved in polyamin and ethylene biosynthesis as well as in methylation reactions modifying lipids, proteins and nucleic acids [45–47]. Ethylene plays an important role in various plant disease resistance pathways. It has originally been considered a stress hormone due to its synthesis induced by stress signals, such as mechanical wounding, chemicals and metals, drought, extreme temperatures, and pathogen infection [46, 48]. Some pathogens can induce plant defence responses via activation of the ethylene signal transduction pathway, whereas plants deficient in ethylene signalling may show either increased susceptibility or increased resistance [49, 50]. Alternatively, methylation of the fully susceptible F7 ^SS/SS genotype might be reduced as revealed in our study by the upregulation of S-adenosylmethonine synthetase isoforms (946126A02.y1, 1091032B12.y1 a and b, 946063C12.y1) in all other genotypes. Resistance to SCMV and MDMV might depend on the methylation status of the plant, relating to post-transcriptional gene silencing mediated by HEN1 like methyltransferase [37].

Germins are water-soluble proteins expressed during seed germination in very young seedlings of wheat and barley. In mature leaves they are induced in response to pathogen attack [51]. In plants other than wheat and barley, sequences related to germins are termed "germin-like". Germins and germin-like proteins were isolated from hot pepper during resistance response to bacterial and viral infection [52, 53]. Pathogen response functions of germins were discovered with the identification of germin as an oxalate oxidase generating hydrogen peroxide. H₂O₂ is a catalyst of cell-wall reinforcement (oxidative cross-linking) and a basis for defence reactions in higher plants. The specific-pathogen-response OXO transcript was found in the wall of barley mesophyll cells six hours after inoculation with powdery mildew [54]. However, it is still unclear if germin-like proteins also have oxalate-oxidase activity and if their biological function is comparable to germins [55–59].

A common feature of expression patterns of genes in the resistant genotype F7 ^RR/RR is lack of signs of oxidative damage (downregulation of class III peroxidases and germins), whereas partially resistant and susceptible genotypes showed upregulation of genes controlling production of hydrogen peroxide. Furthermore, oxidative damage could affect chloroplasts, perturbing their proper function as shown for response to plum pox potyvirus [60]. The analysis of SCMV cylindrical inclusion (CI) virus protein (NP_734137), known to be involved in virus replication and cell-to-cell movement [61], with the ChloroP 1_1 CBS tool http://www.cbs.dtu.dk/services/ChloroP/ revealed a possible chloroplast transit peptide of 65 amino acids [62]. This could explain upregulation of the 14-3-3 proteins (Zm06_09h07_R) in genotypes other than the fully resistant genotype F7 ^RR/RR. 14-3-3 proteins are known to target the transit peptide to the chloroplast, where it will be cleaved upon entrance as shown for other potyvirus [63].

The putative antiviral function of 26S ribosomal RNA gene in ribosome depurination and blocking of translation of viral genetic materials was reported by Taylor et al. [64]. Other genes putatively related to pathogenesis, coding for example a calcium dependent protein kinase, cytochrome c, a zinc finger protein, a peroxidase precursor, an ubiquitin-conjugating enzyme, 40S ribosomal protein or eukaryotic translation initiation factor 4E, were identified in our microarray assay. The involvement of calcium dependent protein kinase (CDPK) in defence signalling has been investigated in transgenic tobacco cell cultures, where CDPK was found to be activated in Avr9/Cf-9 gene-for-gene-dependent signal transduction, as well as in tobacco leaves after Avr9/Cf-9 interaction and hypoosmotic stress [65, 66]. Its participation in signal transduction pathways in Arabidopsis infected with cucumber mosaic virus was demonstrated by Marathe et al. [35]. Cytochrome c has been previously reported to give an apoptotic-like response under Agrobacterium infection of maize suspension cells and was called a cell death inducer released from mitochondria during ROS-induced programmed cell death in plants [67–69]. Furthermore, zinc finger proteins were found to be induced by various types of stresses (for example ethylene treatment), under viral and fungal inoculation [38, 70]. An Arabidopsis zinc-finger protein encoded by the LSD1 gene acts as a negative regulator of hypersensitive response to restrict the spreading of cell death [71]. Class III peroxidases (plant-specific oxidoreductase) participate in lignification, suberization, wound healing and defence against pathogen infection. Increased mRNA levels of POX and its increased activity was previously reported in plants upon mechanical wounding in various plants [72, 73], [74–77]. Finally, altered expression of ubiquitin-conjugating enzyme, 40S ribosomal protein, and eukaryotic translation initiation factor 4E were identified under infection with various potyviruses [33, 38, 35, 78].

Microarrays

Raw intensity and background values generated by Array Vision, version 8.0 (Imaging Research Inc., St. Catharines, Canada) were utilized for data analysis. The main interest was to determine the expression patterns of pair-wise contrasts between genotypes at the same time point (within-time-point analysis), whereas contrasts of a genotype at two different time points were of secondary interest (between-time-point analysis). Locally weighted scatterplot smoothing (LOWESS) regression was performed to adjust for differences within an array.

The following linear mixed model was fitted:

where y _ijkl is the log2-signal intensity, g _i fixed effect for genotype, t _j fixed effect for the time point, a _k random effect for the array, d _l fixed effect for the dye, (g*t) _ij genotype and time point interaction and (g*d) _il genotype and dye interaction. The calculations were performed with the SAS System for Windows, Version 9.1.

Pair-wise contrasts between different genotype*time combinations in the SCMV experiment were estimated, considering only contrasts between genotypes within one time point and contrasts of one genotype at different time points. The corresponding FDR adjusted p-values and fold changes were determined. Least square means of genotype*time were calculated, i.e., the value of a certain genotype at a specific time point averaged over the other effects. The degrees of freedom for the tests were calculated according to the containment method. SAS (Institute Inc. (1999): SAS/STAT User's Guide, Version 8. Cary, NC). For MDMV data analysis the same linear model was fitted but separate variance terms for mock control and normal data were specified.

Blastn analysis in TIGR Unique Gene Indices http://www.tigr.org/plantgenomics/htdocs/blast_servers.html for maize was performed in order to reveal the putative function of unknown sequences from Arabidopsis thaliana, barley, maize, rice, rye and wheat, with a cut off e-value of 10 (Ros et al., 2004). Additional blastn analyses were performed in MIPS http://mips.gsf.de/ and IRGSP http://rgp.dna.affrc.go.jp/IRGSP/ databases for gaining maximum information about the genes of interest.

The calculations of significances for the number of genes between time points were calculated using the McNemar exact test (SAS System for Windows, Version 9.1).

QRT-PCR

where E _target is the PCR efficiency for the target gene and E _ref is the PCR efficiency for the endogenous reference. PCR efficiencies (E = 10 ^(-1/slope) -1), were derived from calibration data of serially diluted RNA: 100%, 50%, 10%, 5%, 1%, 0.5%, 0.1% and water. ΔCt_target and ΔCt_ref values were determined as described by Dilger et al. [80]. Baseline and threshold values were adjusted manually if necessary, as recommended by Applied Biosystems http://www.appliedbiosystems.com/support/tutorials/pdf/performing_rq_gene_exp_rtpcr.pdf.