During a compatible interaction, root-knot nematodes (Meloidogyne spp.) induce the redifferentiation of root cells into multinucleate nematode feeding cells (giant cells). Hyperplasia and hypertrophy of the surrounding cells leads to the formation of a root gall. We investigated the plant response to root-knot nematodes by carrying out a global analysis of gene expression during gall formation in Arabidopsis, using giant cell-enriched root tissues. Among 22 089 genes monitored with the complete Arabidopsis transcriptome microarray gene-specific tag, we identified 3373 genes that display significant differential expression between uninfected root tissues and galls at different developmental stages. Quantitative PCR analysis and the use of promoter GUS fusions confirmed the changes in mRNA levels observed in our microarray analysis. We showed that a comparable number of genes were found to be up- and downregulated, indicating that gene downregulation might be essential to allow proper gall formation. Moreover, many genes belonging to the same family are differently regulated in feeding cells. This genome-wide overview of gene expression during plant–nematode interaction provides new insights into nematode feeding-cell formation, and highlights that the suppression of plant defence is associated with nematode feeding-site development.
Sedentary endoparasitic nematodes are plant parasites that interact with their hosts in a remarkable manner. These obligate biotrophic pathogens establish and maintain permanent feeding cells, which are essential for their growth and reproduction. The molecular signals inducing the redifferentiation of plant cells into feeding cells remain unclear, but nematode secretions injected through the stylet are thought to play a central role in parasitism (Davis et al., 2004). Root-knot nematodes of the genus Meloidogyne have evolved strategies enabling them to infest thousands of plant species, probably by manipulating fundamental key elements of plant cell development. Second-stage juvenile (J2) nematodes penetrate just behind the root tip and migrate intercellularly towards the vascular cylinder, without causing damage to the root cells. Each J2 then selects five to seven parenchymatic cells, transforming into feeding cells that function as specialized sinks supplying the nematode with its nutrient requirement until reproduction. These feeding cells enlarge and become multinucleate through synchronous nuclear divisions in the absence of cytokinesis, and are called giant cells (Huang, 1985; Jones and Northcote, 1972). The fully differentiated giant cells are large and may contain more than 100 polyploid nuclei that have also undergone extensive endoreduplication (Wiggers et al., 1990). The hyperplasia and hypertrophy of the surrounding cells leads to formation of the typical root gall.
During the formation of these large feeding cells, root-knot nematodes induce cell-cycle activation (de Almeida Engler et al., 1999) and major essential rearrangements in the cytoskeleton architecture (de Almeida Engler et al., 2004). Giant cells expand by isotropic growth, and may reach a final size about 400 times that of root vascular cells. Mature giant cells act as transfer cells for the feeding nematode and are metabolically active, as shown by the breakdown of the large vacuole and their dense granular cytoplasm containing many organelles (Jones, 1981).
Microarray technology makes it possible to generate large-scale information about patterns of gene expression during plant–nematode interactions. The completion of the Arabidopsis genome sequence (Arabidopsis Genome Initiative, 2000) allows the development of a complete microarray for Arabidopsis thaliana with specific gene-sequence tags (GST) for all known or predicted genes found in the genome sequence, minimizing cross-hybridization between related genes (Crowe et al., 2003; Hilson et al., 2003). The specificity of the nucleic acids spotted is of particular importance for genes belonging to gene families, accounting for around 65% of all Arabidopsis genes. The complete Arabidopsis transcriptome microarray (CATMA) chips contained 24 576 GSTs corresponding to 22 089 genes, including 21 612 AGI (Arabidopsis Genome Initiative)-predicted genes and 477 Eugene-predicted genes (Thareau et al., 2003). CATMA chips were used recently to study the expression of 441 members of the Arabidopsis pentatricopeptide repeats (PPR) gene family (Lurin et al., 2004) and to determine the function of cullin 3A (Dieterle et al., 2005).
We report here a large-scale host gene-expression profiling study of a compatible interaction between a root-knot nematode and a plant carried out by a microarray strategy, using CATMA biochips. A time-course study using giant cell-enriched root tissues identified 3373 genes that displayed an altered expression pattern during feeding-cell formation. Quantitative PCR analysis and promoter GUS fusion examination confirmed the changes in mRNA levels observed in the microarray experiment. This global investigation highlighted the importance of key biological processes, major gene-expression patterns, and the unsuspected role of gene downregulation underlying the formation of giant cells.
To identify genes involved in the development of feeding cells induced by Meloidogyne incognita, we performed a genome-wide expression profiling of Arabidopsis in response to nematode infection. Thus galls from in vitro-infected Arabidopsis plants were dissected 7, 14 and 21 days post-inoculation (dpi). Root fragments without apical and lateral root meristems from uninfected plants were used as reference tissues, as nematodes infect the vascular tissue. In addition, previous studies have demonstrated that root meristems and giant cells display similarities in patterns of gene expression (Favery et al., 1998; Mazarei et al., 2003). Figure 1a shows the cytological differences between harvested uninfected and infected root tissues at the different time points used for differential analyses. Cytological observations show that at 7 dpi giant cells are induced and expanding, and are multinucleate with a dense cytoplasm. At 14 dpi the epidermal and cortical layers drop off, mitosis or DNA synthesis significantly decreases, but gall and giant cells continue to expand. At 21 dpi a mature gall contains fully expanded and differentiated giant cells. To prevent bias due to developmentally related gene expression, we compared 7-, 14- and 21-dpi galls with non-meristematic root fragments on plants at the same developmental stages (comparisons 1–3, respectively, Figure 1b). In order to follow the levels of expression on differentially expressed genes throughout gall development, changes in gene expression were also investigated by directly comparing 7-dpi with 14-dpi galls, and 14-dpi with 21-dpi galls (comparisons 4 and 5, respectively). A biological replicate and a reverse-labelling technical replicate were performed for each comparison. Genes with differential expression were selected by statistical analysis, using a cut-off P value of 0.05, as described in experimental procedures. All the data obtained from these hybridizations have been submitted to Array Express (accession number E-MEXP-233).
Time-course analysis indicates distinctive features during gall formation
Statistical analysis of comparisons 1–3 using a P value cut-off <0.05 after Bonferroni correction revealed 3373 genes displaying significant differential expression at some point during nematode infection. Thus 15 % of the 22 089 Arabidopsis genes represented on the CATMA chip displayed changes in mRNA levels in response to root-knot nematode infection. The strongest upregulation observed in galls was a 15-fold increase in expression for genes encoding a pectate lyase (At4g24780), a β-expansin EXPB3 (At4g28250) and a pyruvate decarboxylase-1 (At4g33070). The strongest downregulation (by a factor of nine) was observed for genes encoding an unknown expressed protein (At3g21520) and two defence-related genes, encoding patatin (At2g26560) and germin-like protein (At4g14630). In total, 139 genes displayed changes in expression levels by a factor greater than four, and 17 genes displayed changes in expression level by a factor greater than eight. Figure 2 illustrates the global distribution of the 1606 induced genes and the 1742 downregulated genes in galls at 7, 14 and 21 dpi, using uninfected root tissue as reference. In addition, 25 genes initially displayed induction followed by repression (or vice versa) during this time-course study (data not shown). Venn diagrams of experiments 1–3 showed that 10–23% of the induced or repressed genes were specifically regulated at only one time point during gall formation. The number of genes co-expressed at 14 and 21 dpi was greater than that co-expressed at 7 and 14 dpi, or 7 and 21 dpi. In addition, direct comparisons of galls at 7, 14 and 21 dpi (experiments 4 and 5) showed that most of the regulated genes (86%) are differentially expressed in galls only between 7 and 14 dpi (data not shown). Thus gene expression during early gall formation (7 dpi) differs from that at later stages (14 and 21 dpi), suggesting that specific biological processes may be required for gall initiation.
Temporal gene expression in galls and microarray data validation by qPCR
Hierarchical clustering on expression ratios obtained at the three time points after M. incognita infection was used to identify common expression kinetics among differentially expressed genes (Figure 3a). From the tree obtained, eight clusters of genes with co-ordinate expression were identified (Figure 3b). Clusters II and VIII include genes with constant responses, induced or repressed, during the time course, and identified groups of genes induced or repressed in a time-dependent manner in galls. Clusters III and V grouped transiently regulated genes. Clusters I and VI contained genes that responded early during gall formation, whereas clusters IV and VII contained genes displaying late responses to infection by root-knot nematodes.
We used quantitative RT–PCR (qPCR) to validate the microarray data, by looking at the expression profiles of 50 genes expressed in galls and uninfected root fragments over time. Using comparative Ct (Cycle threshold) method, the quantitative PCR study validated 84% of the DNA array hybridization results and confirmed the specific gene-expression profiles in galls (Table 1). Changes in gene expression were validated even when expression was increased or decreased by less than a factor of two in our microarray analysis. For example, we confirmed the constant downregulation in galls of the tonoplastic aquaporin TIP1;1 (Höfte et al., 1992); the lipoxygenase LOX1 (Melan et al., 1993); and the 7-dpi downregulation of the formate dehydrogenase gene FDH (Olson et al., 2000). We also validated specific expression patterns, such as the activation in 7-dpi galls followed by its repression at 21 dpi of a Heterodera schachtii-induced aminotransferase (At2g24850, Puthoff et al., 2003). Despite the small increase in mRNA levels of cyclin D3;2 and the cyclin-dependent kinase CKS2 observed on microarrays, we could demonstrate the early upregulation of these genes. Finally, we also showed the activation of genes in 7-, 14- and 21-dpi galls encoding expansin EXPA1 and EXPB1, formin AtFH10 and pyruvate decarboxylase PDC1. Considering the high validation rate of our microarray data, they appear to reflect the actual temporal gene-expression changes occurring in galls.
Table 1. Microarray data validation by quantitative RT--PCR. Function TIGR and functional category refer to automatic Arabidopsis annotations according to AGI number from the Institute for Genomic Research and the Munich Information Center for Protein Sequences, respectively. BCN, Arabidopsis genes induced in response to beet cyst nematode (Heterodera schachtii) (1Puthoff et al., 2003; 2Vaghchhipawala et al., 2001; 3Hermsmeier et al., 2000). The ΔΔCT_7, ΔΔCT_14, ΔΔCT_21 and CATMA_7, CATMA_14, CATMA_21 columns indicate quantitative RT–PCR and complete Arabidopsis transcriptome microarray (CATMA) results, respectively, in a log2 ratio obtained for comparisons between galls induced by Meloidogyne incognita and uninfected root fragments (references) at 7, 14 and 21 dpi. A statistical cut-off, P < 0.05 after Bonferroni correction (see colour code), was used to determine which genes were differentially expressed in galls. Positive ratio indicates that the gene is induced in galls (red boxes); negative ratio that the gene is repressed in galls (green boxes). VALID, number of qPCR validation for a gene for the three time points.
7 dpi galls/reference
14 dpi galls/reference
21 dpi galls/reference
Characterization of the biological processes involved in gall development
To explore the biological processes in which the differentially regulated genes are involved, we have classified genes according to functional categories of the Arabidopsis MIPS Functional Catalogue (Ruepp et al., 2004). The functional groups with the highest number of genes were those involved in metabolism in accordance with the giant cell function as nutrient sinks for the nematode (Figure 4; Table S1). However, similar numbers of genes involved in metabolism were induced and repressed. Genes involved in cell-wall metabolism appeared to be more induced than repressed, partially due to the upregulation of all genes identified as encoding pectate lyases and the two major classes (A and B, formerly α and β) of cell wall-loosening expansins (Table 1 and Table S2).
Many putative transcription factors were up- or downregulated during infection. The families with the highest number of members were MYB, Zinc finger, AP2/EREBP and WRKY. Other transcription factors, such as MADS-box, bZIP, bHLH, NAC, homeobox and Scarecrow, were also represented. Interestingly, 23 genes encoding transcription factors were shown to be specifically differentially expressed at 7 dpi (Table S1).
In contrast, several biological processes have been shown to display induction or repression during gall development. Most of the genes involved in cell-cycle regulation and DNA processing, energy, and protein synthesis have shown to be induced in galls (Figure 4; Table 1 and Table S1). Thus all 102 genes encoding 40S and 60S ribosomal proteins were upregulated in galls, and 71 of these genes showed specific upregulation in galls at 7 dpi. In contrast, a high proportion of the genes identified as being involved in cell rescue and defence, cellular communication and cellular transport were repressed, particularly at 14 and 21 dpi (Figure 4; Table 1 and Table S1). The defence-related group included genes that were previously shown to be induced during other plant–pathogen interactions, and encode, for instance, PR proteins (e.g. chitinase, PR-4); enzymes involved in the biosynthesis of phenylpropanoids (e.g. phenylalanine ammonia lyase); or a lipase-like protein that is important for salicylic acid signalling (PAD4). For cellular transport, analysis of the 25 members of the aquaporin gene family represented on CATMA chips showed that seven genes were downregulated, including those encoding tonoplast and plasma membrane intrinsic proteins TIP and PIP. Three aquaporin genes – two PIP and one NIP (Nod26-like intrinsic protein) – were upregulated (Table 1 and Table S2). These differences in transcriptional regulation within a gene family highlight the value of microarrays containing gene-specific sequences. In addition, the time-course analysis made it possible to distinguish biological processes involved during early and late plant responses to M. incognita infection.
Plant gene regulation in giant cells
We generated transgenic Arabidopsis plants containing constructs in which the GUS reporter gene was fused to the promoters of one downregulated gene, At1g73260, encoding a trypsin protease inhibitor (TPI); and one upregulated formin gene, AtFH10. Histochemical analysis of GUS expression after nematode infection and in uninfected roots showed that temporal and spatial GUS localization were consistent with the microarray results (Figure 5). In pTPI:GUS plants, GUS expression showed that the promoter was active throughout the root of uninfected Arabidopsis, whereas it was repressed in galls at 14 and 21 dpi. A weak GUS expression remained in epidermal cells of the gall. In contrast, strong GUS staining was observed in the galls of pAtFH10:GUS plants at 7, 14 and 21 dpi. No expression of pAtFH10 was observed in uninfected non-meristematic root fragments, except at the bases of some lateral roots. Sections of galls at 14 dpi revealed GUS expression in giant cells and in the surrounding cells (Figure 5b), and no GUS expression was found in the gall cortex. Our microarray time-course analysis allowed us to identify plant genes differentially expressed in giant cells.
The identification of genes regulated in nematode feeding sites represents a major challenge in understanding how nematodes alter root development to generate and maintain giant cells. In order to take steps towards identifying the key processes involved in giant cell induction, a genome-wide gene expression analysis was conducted. We used the CATMA to monitor changes in mRNA levels of approximately 22 089 genes. Dissected galls at different time points after inoculation were compared with uninfected root fragments. A set of 3373 genes has been identified that display significant changes in expression during gall development, from which actual temporal gene expressions have been validated by qPCR. This large proportion (15%) of genes displaying a differential expression pattern reflects the complexity of nematode feeding-site ontogenesis. So far, only a small number of expression-profiling studies have been carried out using compatible plant–pathogen interaction systems (Tao et al., 2003; van Wees et al., 2003; Whitham et al., 2003).
The proportion of genes upregulated in galls identified by microarray analysis is consistent with results obtained in a large promoter trap study of Arabidopsis lines (Favery et al., 1998, 2004). Although only 10 genes had previously been identified as being repressed in galls (Gheysen and Fenoll, 2002; Wang et al., 2003), our present study clearly illustrates that nearly similar numbers of genes were found to be up- and downregulated, indicating that gene downregulation may be essential to allow proper gall formation induced by M. incognita. Changes in host gene expression have also been documented during a cyst nematode infection. Microarray analysis of whole roots infected with cyst nematodes, using one-third of the predicted number of Arabidopsis genes, identified 128 genes displaying changes in expression during successful nematode parasitism (Puthoff et al., 2003; Table S3). As the genes identified are often expressed not only in feeding sites, but also in other parts of the root, particularly in root tips (Favery et al., 1998; Mazarei et al., 2003), the use of a dissected feeding site and meristemless, uninfected root fragments, rather than whole roots, may account for the discrepancy observed among the number of genes identified. Indeed, the use of dissected root fragments to enrich tissues of interest has been shown to be very useful in transcript profiling for early lateral root initiation (Himanen et al., 2004).
A hierarchical cluster analysis carried out for the 3373 differentially expressed genes clearly identified the temporal regulation of a fraction of these genes, reflecting the co-ordinated expression of genes involved in the various molecular processes that occur during gall development. Such transcriptional reprogramming must require a sophisticated regulatory system. Several types of transcription factor appear to orchestrate the extensive changes observed in gene expression. Some of the transcription factors identified here, such as MADS box, homeobox, NAC and Scarecrow proteins, have been associated mainly with the control of plant development (Bolle, 2004; Gong et al., 2004). Members of the MYB, AP2/EREBP, WRKY and bZIP transcription factor families have been shown to be implicated in controlling gene expression in various abiotic and biotic stress signal transduction pathways (Chen et al., 2002; Eulgem, 2005). In nematode-infected tissues, 17 of the 21 WRKY genes identified are downregulated, while accumulation of WRKY transcripts appears to be a general characteristic of plant defence in response to pathogens (Eulgem, 2005).
Interestingly, our global analysis indicates that successful establishment of the root-knot nematode is associated with a suppression of plant defence mechanisms. The importance of plant defence suppression during pathogenesis has been highlighted in many recent studies. Indeed, in planta development of obligate biotrophic and hemibiotrophic fungi is associated with a phase of active suppression of plant defence (Bouarab et al., 2002; Waspi et al., 2001). Recent data reveal that phytopathogenic bacteria use type III secreted effector proteins, toxins and other factors to interfere with host defences (Hauck et al., 2003; Abramovitch and Martin, 2004). Accordingly, successful pathogens seem to have evolved specialized strategies to suppress plant defence response and generate susceptibility in host plants. Defence suppression also appears to play an important role in symbiotic plant–microbe interactions. The NopL effector of Rhizobium sp. NGR234 suppresses PR gene expression when expressed in tobacco or Lotus japonicus (Bartsev et al., 2004).
A major reprogramming of plant metabolism also appears to occur throughout giant cell formation. We observed that a large number of metabolic and energy-related genes are not only upregulated, but also downregulated in galls. The specific upregulation in 7-dpi galls of 71 genes encoding 40S and 60S ribosomal proteins clearly suggests increased levels of protein synthesis during giant cell initiation.
Large amounts of water and solutes are transported from the xylem, through the cell-wall ingrowths of the giant cells, probably via water channels, facilitating passage across biological membranes (Gheysen and Fenoll, 2002). A previous study reported downregulation of the Arabidopsis γTIP1;1 gene, encoding a tonoplast protein, in galls (Goddijn et al., 1993), whereas later studies reported upregulation of the tobacco TIPTobRB7 (Opperman et al., 1994) and the M. truncatula NIP NOD26 gene (Favery et al., 2002) in giant cells and galls, respectively. Using this array containing gene-specific sequences, we showed that genes from the same family may be differently regulated, possibly justifying conflicting results. We have observed that three of the Arabidopsis aquaporin genes were upregulated (one NOD26-like and two plasma membrane PIPs) and seven (three PIPs and four TIPs, including TIP1;1) were repressed. The fine regulation of the transcriptional control of aquaporin genes in nematode feeding cells may account for the several functions proposed for these proteins in growth control, water transport and cell osmoregulation (Maurel and Chrispeels, 2001).
During giant cell initiation, acytokinetic nuclear divisions occur mainly during the first 7 days after root-knot nematode infestation in A. thaliana roots (de Almeida Engler et al., 1999). In our microarray analysis, cell-cycle genes, including a cyclin-dependent kinase (CKS-2) and a mitotic cyclin (cyclin D3;2), have been found to be upregulated in galls at 7 dpi. During giant cell development, DNA endoreduplication is observed within the nuclei. Genes regulating endoreduplication in Arabidopsis, such as CPR5 (Kirik et al., 2001), are shown here to be activated in galls at 14 and 21 dpi.
A striking feature during the formation of giant cells is their exceptional increase in cell size. Cytoskeleton reorganization in giant cells is certainly a key component during isotropic cell growth (de Almeida Engler et al., 2004), as indicated by the activation of a membrane-anchored formin (AtFH6) involved in actin nucleation (Favery et al., 2004). Microarray analysis, followed by promoter GUS fusion assays, showed the upregulation of another formin in giant cells, the AtFH10 gene. Expansion of developing giant cells may require specific cell-wall modifications such as cell-wall loosening. Even though cell wall-modifying enzymes are known to be secreted by nematodes during migration (Davis et al., 2004), plant genes encoding cell wall-modifying enzymes, such as endo-β-1,4-glucanases and pectin acetylesterase, have been shown to be differentially expressed during giant cell formation (Goellner et al., 2001; Vercauteren et al., 2002). In our microarray analysis, we identified members of several subclasses of cell wall-modifying enzymes that were induced or repressed during feeding site development. These included cell-wall disassembly enzymes such as β-xylosidases, β-1,4-endoglucanases, xyloglucan endotransglycosylases, polygalacturonases, and expansins. Expansins are cell wall-loosening proteins that induce stress relaxation and extension of the plant cell wall without hydrolytic breakdown of its major components (Lee et al., 2001). Among the 31 Arabidopsis genes encoding the two major classes of expansin (http://www.bio.psu.edu/expansins), we identified seven expansin A genes and two expansin B genes strongly upregulated in galls. The specific downregulation of two expansin-like genes at 21 dpi suggests a particular function for these proteins. We have also shown that all identified members of the pectate lyase gene family are activated in response to M. incognita infection. These enzymes, together with polygalacturonase and pectin methylesterase, are involved in pectin degradation. Reducing the pectin content of the cell wall may also increase nutrient availability to the nematode, as suggested for the plant gene PMR6 during the Arabidopsis–powdery mildew interaction (Vogel et al., 2002).
Using microarray analysis to characterize genes specifically regulated during gall development is an initial step towards a better understanding of the complex molecular mechanisms involved in giant cell formation. However, for full comprehension of the role in feeding cell development of the genes we identified, it is imperative to combine our data with those obtained from a gene-expression analysis at the cellular level, from knockout mutant characterization, and from biochemical data. Our genome-wide overview of gene expression during plant–nematode interactions provides novel insights into feeding cell formation. Comparison of our data with future findings from genome-wide expression profiling applied to other systems, such as feeding cells induced in roots by cyst nematodes, or susceptible responses to less-related pathogens, should allow better discrimination of specific and common features of the plant–root-knot nematode interaction.
Plant material and nematode infection
Arabidopsis thaliana plants (ecotype Wassilewskija) were grown on Gamborg B5 medium (Sigma-Aldrich, St Louis, MO, USA) containing 1% sucrose, 0.8% agar (plant cell culture tested, Sigma). Plates were inclined at an angle of 60° to allow the roots to grow along the surface. For nematode infection in vitro, 100 surface-sterilized, freshly hatched J2 of M. incognita were added on each 3-week-old seedling as described previously (Sijmons et al., 1991). The plates were kept at 20°C with a 16-h photoperiod. For the transcriptome study, galls were individually hand-dissected from 10 plates of 15 infected plants at stage R6 according to Boyes et al. (2001) at 7, 14 and 21 dpi. As reference samples, uninfected, non-meristematic root fragments, without apical and lateral root meristems, were dissected from seedlings grown under the same conditions. About 1500 galls or control uninfected root fragments were collected from each time point, frozen in liquid nitrogen immediately after excision, pooled per sample, and stored at −80°C until use. Each sample was replicated.
The microarray analysis was performed with the CATMA array containing 24 576 GSTs corresponding to 22 089 genes, including 21 612 AGI-predicted genes and 477 Eugene-predicted genes, from A. thaliana (Crowe et al., 2003; Hilson et al., 2003). The GST amplicons were purified on Multiscreen plates (Millipore, Bedford, MA, USA) and resuspended in TE-DMSO at 100 ng μl−1. The purified probes were transferred to 1536-well plates with a Genesis workstation (TECAN, Männedorf, Switzerland) and spotted on UltraGAPS slides (Corning, NY, USA) using a Microgrid II (Genomic Solutions, Huntingdon, UK). The current CATMA version printed at the URGV consists of three metablocks, each composed of 64 blocks of 144 spots. A block is a set of spots printed with the same print-tip. In these arrays, a print-tip is used three times to print a block in each metablock.
Five comparisons were analysed in this time-course analysis. In comparisons 1–3, each array was hybridized at the same time with cRNA from dissected galls, and the corresponding uninfected non-meristematic root fragments were labelled with Cy3 and Cy5 fluorescent dyes, respectively. Comparisons 4 and 5 correspond to hybridization with cRNA from dissected galls at 7 and 14 dpi, and cRNA from dissected galls at 14 and 21 dpi, respectively. For each comparison we performed a repeat using a second set of samples. In addition, to avoid dye bias and gene-specific dye bias, a dye-swap experiment was carried out: the cRNAs labelled with Cy3 in the first array was labelled with Cy5 in the second (Martin-Magniette et al., 2005). Therefore four arrays were used for each comparison assay.
RNA was extracted from these samples using TRIzol extraction (Invitrogen, Carlsbad, CA, USA) followed by two ethanol precipitations, then checked for RNA integrity with an Agilent bioanalyser (Waldbroon, Germany). cRNAs were produced from 2 μg total RNA from each pool with the Message Amp aRNA kit (Ambion, Austin, TX, USA). Then 5 μg cRNAs were reverse transcribed in the presence of 200 u SuperScript II (Invitrogen), cy3-dUTP and cy5-dUTP (NEN, Boston, MA, USA) for each slide, according to Puskas et al. (2002). Samples were combined, purified and concentrated with YM30 Microcon columns (Millipore). Slides were pre-hybridized for 1 h and hybridized overnight at 42°C in 25% formamide. Slides were washed in 2 × saline sodium citrate (SSC) + 0 1% sodium dodecyl sulphate (SDS) 4 min, 1 × SSC 4 min, 0.2 × SSC 4 min, 0.05 × SSC 1′ and dried by centrifugation. Twenty hybridizations (10 dye-swaps) were carried out. The arrays were scanned on a GenePix 4000A scanner (Axon Instruments, Foster City, CA, USA) and images were analysed using genepix pro 3.0 (Axon Instruments).
Statistical analysis of microarray data
The statistical analysis was based on two dye-swaps (four arrays each containing the 24 576 GSTs and 384 controls). For each array, the raw data comprised the logarithm of median feature pixel intensity at wavelengths 635 nm (red) and 532 nm (green). No background was subtracted. In the following description, log-ratio refers to the differential expression between the two tissues analysed: either log2(red/green) or log2(green/red), according to the experimental design. An array-by-array normalization was performed to remove systematic biases. First, features that were considered by the experimenter to be badly formed (e.g. because of dust) were excluded (flagged −100 in the genepix software). Then we performed a global intensity-dependent normalization using the Loess procedure (Yang et al., 2002) to correct the dye bias. Finally, on each block the log-ratio median was subtracted from each value of the log-ratio of the block to correct a print-tip effect.
To determine differentially expressed genes, we performed a paired t-test on the average of the log-ratio per dye-swap. That enabled us to take into account the technical and biological variability, and to give more importance to biological variability. Consequently the number of observations per spot varies between one and two, thus it is not adequate to calculate a gene-specific variance. For this reason we assumed that the variance of the log-ratios was the same for all genes, by calculating the average of the gene-specific variance. In order to assess this assumption, we excluded spots with a variance too small or too large (the number of excluded spots, approximately 80, varied for each differential analysis) before the average. Raw P-values were adjusted by the Bonferroni method, which controls the family-wise error rate (Ge et al., 2003).
Real-time quantitative RT–PCR
Real-time quantitative RT–PCR was carried out using the ABI PRISM 7900HT Sequence Detection System (Applied Biosystems). Specific primers for each gene selected were designed from the GST sequences (Crowe et al., 2003) using primer3 software (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi) (Table S4). For each gene amplified, a standard curve was generated from duplicate series of five template dilutions to test PCR efficiencies. For quantification, the template tested was cDNAs from 7, 14 and 21 dpi galls. The reference template was cDNAs from uninfected non-meristematic root fragments. PCR was conducted in duplicate in the presence of 1 ng cDNA, 1.2 μl of each primer 2.5 μm, 5 μl Sybr GREEN mastermix, and distilled water to a final volume of 10 μl. PCR conditions were as described above with 10 min at 95°C and 45 cycles at 95°C for 10 sec, 55°C for 10 sec and 72°C for 30 sec. The results were standardized by comparing the data with nine reference Arabidopsis genes that remain constant under different treatment conditions. The quantification of gene expression was performed using the comparative CT method.
Annotations were downloaded from The Arabidopsis Information Resource (TAIR) (http://www.arabidopsis.org). Functional categories of differentially expressed genes were obtained from the Munich Information Center for Protein Sequences (http://mips.gsf.de/proj/funcatDB). For analysis of gene-expression data across the different experiments, hierarchical clustering was performed with the genesis software (http://genome.tugraz.at) using default parameters. The data obtained in these microarray experiments have been submitted to Array Express (accession number E-MEXP-233; http://www.ebi.ac.uk/arrayexpress).
Promoter fusion, histochemical localization of GUS activity and microscopic analyses
The A. thaliana line pTPI was identified using a promoter trap screen of the INRA Versailles T-DNA lines collection (Bechtold et al., 1993) for genes regulated in feeding cells induced by M. incognita, as previously described (Favery et al., 1998). A fragment approximately 1.2 kb upstream of the start ATG of the AtFH10 (At3g07540) gene was obtained by PCR and fused to the GUS reporter gene in the pKGWFS7 vector (Karimi et al., 2002) using the Gateway technology, according to the manufacturer's instructions (Invitrogen). The identity and integrity of the promoter fragments were confirmed by sequencing. Arabidopsis plants were transformed by the floral dip method (Clough and Bent, 1998). Primary transformed plants were selected on kanamycin plates and transferred in soil for multiplication. Two-week-old promoter-GUS seedlings were inoculated as described above, and assayed histochemically for GUS activity (Favery et al., 1998).
Galls and root fragments were dissected, fixed in 1% glutaraldehyde and 4% formaldehyde in 50 mm sodium phosphate buffer pH 7.2, dehydrated, and embedded in Technovit 7100 (Heraeus Kulzer, Wehrheim, Germany) as described by the manufacturer. Sections were mounted in DPX (VWR International Ltd, Poole, UK) and observed with a Zeiss Axioplan 2 microscope (Zeiss, Jena, Germany).
We thank Liudmilla Chelysheva, Xavier Sarda (Biogemma) and Jean-Luc Evrard (CNRS) for helpful discussions, Gilbert Engler for critical reading of the manuscript, and Anne-Valérie Gendrel (URGV) for assistance with quantitative RT–PCR. We also thank INRA Versailles for providing the Arabidopsis T-DNA tagged line, and Mansour Karimi (Plant Systems Biology, VIB University of Gent, Belgium) for the pKGWFS7 vector. This work was supported by the French National Institute for Agronomic Research (INRA) and GENOPLANTE contracts AF2001032 and NO2001040. F.J. was supported by a fellowship from the Ministère de la Recherche et l'Enseignement Supérieur. This paper is dedicated to the memory of Marcel Favery.