Early transcriptomic events in microdissected Arabidopsis nematode-induced giant cells

Authors

  • Marta Barcala,

    1. Facultad de Ciencias del Medio Ambiente, Universidad de Castilla-La Mancha, Avenida de Carlos III s/n, 45071 Toledo, Spain
    Search for more papers by this author
    • Present address: Centro de Biología Molecular ‘Severo Ochoa’-CSIC, Universidad Autonoma de Madrid, 28049 Madrid, Spain.

  • Alejandra García,

    1. Facultad de Ciencias del Medio Ambiente, Universidad de Castilla-La Mancha, Avenida de Carlos III s/n, 45071 Toledo, Spain
    Search for more papers by this author
  • Javier Cabrera,

    1. Facultad de Ciencias del Medio Ambiente, Universidad de Castilla-La Mancha, Avenida de Carlos III s/n, 45071 Toledo, Spain
    Search for more papers by this author
  • Stuart Casson,

    1. Integrative Cell Biology Laboratory, School of Biological and Biomedical Sciences, Durham University, Durham DH1 3LE, UK
    Search for more papers by this author
    • Present address: School of Biological Sciences, University of Bristol, Woodland Road, Bristol, BS8 1UG, UK.

  • Keith Lindsey,

    1. Integrative Cell Biology Laboratory, School of Biological and Biomedical Sciences, Durham University, Durham DH1 3LE, UK
    Search for more papers by this author
  • Bruno Favery,

    1. INRA, UMR, 1301, 400 route des Chappes, F-06903 Sophia Antipolis, France
    2. CNRS, UMR 6243, 400 route des Chappes, F-06903 Sophia Antipolis, France
    3. Université de Nice Sophia Antipolis, UMR 1301, 400 route des Chappes, F-06903 Sophia Antipolis, France
    Search for more papers by this author
  • Gloria García-Casado,

    1. Unidad de Genómica, Centro Nacional de Biotecnología-CSIC, Campus Universidad Autónoma de Madrid, 28049 Madrid, Spain
    Search for more papers by this author
  • Roberto Solano,

    1. Unidad de Genómica, Centro Nacional de Biotecnología-CSIC, Campus Universidad Autónoma de Madrid, 28049 Madrid, Spain
    Search for more papers by this author
  • Carmen Fenoll,

    1. Facultad de Ciencias del Medio Ambiente, Universidad de Castilla-La Mancha, Avenida de Carlos III s/n, 45071 Toledo, Spain
    Search for more papers by this author
  • Carolina Escobar

    Corresponding author
    1. Facultad de Ciencias del Medio Ambiente, Universidad de Castilla-La Mancha, Avenida de Carlos III s/n, 45071 Toledo, Spain
    Search for more papers by this author

  • The data obtained with the GCs RNA hybridizations have being submitted to Array Express (E-MEXP-2300).

For correspondence (fax +34 925268840; e-mail Carolina.escobar@uclm.es).

Summary

Root-knot nematodes differentiate highly specialized feeding cells in roots (giant cells, GCs), through poorly characterized mechanisms that include extensive transcriptional changes. While global transcriptome analyses have used galls, which are complex root structures that include GCs and surrounding tissues, no global gene expression changes specific to GCs have been described. We report on the differential transcriptome of GCs versus root vascular cells, induced in Arabidopsis by Meloidogyne javanica at a very early stage of their development, 3 days after infection (d.p.i.). Laser microdissection was used to capture GCs and root vascular cells for microarray analysis, which was validated through qPCR and by a promoter-GUS fusion study. Results show that by 3 d.p.i., GCs exhibit major gene repression. Although some genes showed similar regulation in both galls and GCs, the majority had different expression patterns, confirming the molecular distinctiveness of the GCs within the gall. Most of the differentially regulated genes in GCs have no previously assigned function. Comparisons with other transcriptome analyses revealed similarities between GCs and cell suspensions differentiating into xylem cells. This suggests a molecular link between GCs and developing vascular cells, which represent putative GC stem cells. Gene expression in GCs at 3 d.p.i. was also found to be similar to crown galls induced by Agrobacterium tumefaciens, a specialized root biotroph.

Introduction

Plant endoparasitic nematodes interact with their hosts in a subtle and specialized manner. Detailed information is still lacking on the molecular events crucial for their establishment and reproduction. Infection by both cyst and root-knot nematodes (RKN) entails the development of specialized feeding cells from the vascular cylinder (syncytia and giant cells, respectively). Giant cells are very large, with a dense cytoplasm indicative of an active metabolism, multiple polyploid nuclei, and a thickened, profusely invaginated cell wall. Root tissues surrounding the GCs lead to a typical swollen root structure, the gall (Gheysen and Fenoll, 2002).

It has been suggested that nematode secretions play an essential role during the genetic re-programming events occurring in the cells that will differentiate into GCs (Williamson and Hussey, 1996). Processes such as the cell cycle and cell wall growth, as well as hormone profiles, are altered in galls. Most of the available information on GC formation has been obtained through expression studies of promoter activation, and by using various differential expression techniques (Caillaud et al., 2008a). Microarray technologies have been applied mainly to whole galls containing GCs and the surrounding tissue (Bar-Or et al., 2005; Hammes et al., 2005; Jammes et al., 2005; Fuller et al., 2007). Wang et al. (2003) attempted to overcome this limitation through microaspiration of GC contents, but this technique is limited to large GCs at late infection stages, when most of the gene re-programming events leading to cell differentiation are likely to have already occurred. The first microarray analysis from isolated GCs by laser capture microdissection (LCM) has been recently performed in tomato (Portillo et al., 2009), but not in Arabidopsis, which contains smaller root cells. In contrast, microaspiration and LCM have been successfully applied to syncytia, the cyst-nematode feeding cells, in soyabean and Arabidopsis (Ramsay et al., 2006; Ithal et al., 2007b; Klink et al., 2007; Szakasits et al., 2009). Thus, to date, no data on microarray-based transcriptome analyses of Arabidopsis isolated GCs are available.

In the current study, we combined LCM with microarray hybridization to study specifically the GC transcriptome at very early infection stages (3 days after infection (d.p.i.)). Both techniques have been combined to obtain information on transcriptomic differences between GCs and uninfected root vascular cells in Arabidopsis. The results reveal the distinctiveness of the GC transcriptome in comparison with that of the hand-dissected galls, and identify similarities with their putative root vascular stem cells and with Agrobacterium tumefaciens-induced crown galls.

Results

Sample collection and LCM

To identify genes expressed in the initial stages of GC differentiation, infection progress and root growth were closely monitored to accurately select 3 d.p.i. galls and collect equivalent uninfected root segments without lateral roots and meristems (Figure 1a). Galls in the primary root were hand-dissected 72 h after T0 (Figure 1b,c). Equivalent galls and control root samples were collected either for the LCM of GCs or for the gall transcriptome analysis.

Figure 1.

 Plant material used for laser capture microdissection.
(a) Uninfected roots. Color labeling indicates root length at each 12-h monitoring point.
(b) Infection stage of A. thaliana with Meloidogyne javanica (black arrow) considered as T0 (nematode inside the root tip, but no root swelling detected).
(c) A representative 3 d.p.i. gall used for LCM. Arrows in (a) and (c) indicate segments collected for LCM.
(d, f) Cryosections of 3 d.p.i. galls and uninfected roots used for LCM. White arrows identify GCs.
(e, g) Sections after a laser beam shot. Microdissected areas are enclosed by the black circle. Scale bar in (a) represents 1 cm, 0.1 mm in (b) and (c), 50 μm in (d) and (e) and 20 μm in (f) and (g).

Three-d.p.i. GCs were clearly visible in cryosections (with highly dense cytoplasms, probably related to plasmolysis; Figure 1d). GCs and the vascular cylinder tissues, limited by the endodermis, from uninfected root segments, were isolated by LCM from gall and control root sections, respectively (Figure 1d–g); RNA was extracted for microarray hybridization and amplified as described in Experimental procedures.

Transcript profiling of very early developing Arabidopsis giant cells

Whole genome transcriptional profiles of 3 d.p.i. GCs and the vascular cylinder of uninfected primary root segments were obtained using ‘two-color’ 70-mer microarrays. Of the 29 000 Arabidopsis unigenes in the chip, 1161 genes (4%) were differentially expressed for a false discovery rate cut-off (q-value) <0.05 (see Experimental procedures) after correction for the multiple testing. The majority of the differentially expressed genes (DEG) were down-regulated (850 versus 310 up-regulated genes). Fold-change (Fc) values ranged from −24 to 20, corresponding to a peroxidase (PER20) and an expansin (EXPA6) transcript, respectively. A high number of genes showed strong (>4) induction or repression levels (Figure 2a). For simplicity, the terms induced/up-regulated, and repressed/down-regulated, would be used to mean transcript levels higher or lower than the control, respectively throughout the text.

Figure 2.

 Comparison of the DEG (q-value <0.05) in galls and GCs ranged by fold-change (Fc) values.
(a) DEG in GCs, yellow bars; common DEG in GCs and galls, blue bars. The number of ‘GCs distinctive genes’ increased considerably in the lower Fc ranges.
(b, c) Functional classification of ‘distinctive DEG in GCs and galls’ by MapMan, respectively. DEG in GCs and galls were compared relative to the control vascular cylinder cells of uninfected primary root segments and to uninfected primary root segments, respectively. Main BINs that exhibited a different behaviour in terms of expression profile compared with all the other remaining BINs after Wilcoxon rank sum test, considered with a P-value <0.05, are indicated with a star, subcategories are listed in Table S3.

To obtain an overall view of the processes that were altered during this early stage of GC development, classification of the DEG was performed using MapMan (Usadel et al., 2005). Most categories showed a higher number of down-regulated than up-regulated genes, indicating that the strong trend towards gene down-regulation was distributed among the different functional categories. Only the ‘DNA’ and ‘Cell’ categories included a high proportion of up-regulated genes (Figure 3a). Accordingly, they showed P-values <0.05, indicative of significant differential profiles in comparison to the rest of the categories (Table S3). Consistent with other studies of differential expression in galls (reviewed in Caillaud et al., 2008a), the up-regulated group included mostly genes encoding chromatin structure maintenance or remodeling proteins, such as histones H4 and H3, and cytoskeleton, microtubule associated proteins (Table 1b).

Figure 3.

 Functional classification of DEG in GCs (q-value <0.05) versus vascular cylinder cells of uninfected primary root segments by MapMan categories.
(a) The number of genes (Y-axis), differentially expressed in 3 d.p.i. GCs included in each category (BIN). ‘Not assigned’ category (402 entries) is not represented.
(b) Subcategories from ‘metabolism.’ Main BINs that exhibited a different behaviour in terms of expression profile compared with all the other remaining BINs after Wilcoxon rank sum test, considered with a P-value <0.05, are indicated with a star, subcategories are listed in Table S3.

Table 1.   DEG versus vascular cylinder cells of uninfected primary root segments included within MapMan categories
(a)TotalDownUp
Total DEG from GCs1161851310
‘Distinctive DEG from GCs’1041 (90%)797 (94%)244 (79%)
Total DEG from galls547193354
Common DEG from galls and GCs120 (10%)54 (6%)66 (21%)
(b) Functional categoryIDq- valueLog2 valueDescription
  1. Asterisk indicates ‘GCs distinctive genes’ defined as non-co-expressed with the equivalent 3 d.p.i. galls transcriptome. Log2 values are indicated. All genes were considered as DEG with a cut off q-value <0.05 (see Experimental procedures).

Cell wallAt1g027300.0422.4Cellulose synthase-like d5 (CSLD5)
At1g695300.0273.25Expansin (EXPA1)
At2g06850*0.0272.07Endoxyloglucan transferase (EXGT-A1)
At2g289500.0274.39Expansin (EXPA6)
At3g531900.0352.29Pectate lyase family protein
At4g302700.041.96Endoxyloglucan transferase
At4g39350*0.026−2.12Cellulose synthase 2 (CESA2)
At5g03760*0.042−2.4Glycosyl transferase family 2 protein
MetabolismAt1g03310*0.0262.8Starch debranching enzyme
At1g24807*0.027−1.68Anthranilate synthase beta subunit
At1g371300.0271.28Nitrate reductase 2 (NR2)
At1g65060*0.026−2.034-coumarate-coa ligase 3 (4CL3)
At1g74020*0.042−1.01Strictosidine synthase 2 (SS2)
At2g14750*0.026−2.29Adenosine phosphosulfate kinase 1 (AKN1)
At2g29130*0.0261.6Laccase 2 (LAC2)
At2g36390*0.0331.27Starch branching enzyme class
At2g37040*0.037−3.06Phenylalanine ammonia-lyase 1 (PAL1)
At3g19450*0.039−1.43Cinnamyl-alcohol dehydrogenase4 (CAD4)
At3g53260*0.035−2.33Phenylalanine ammonia-lyase 2 (PAL2)
At3g55120*0.026−3.16Chalcone-flavanone isomerase (CHI)
At4g26510*0.0470.72Uracil phosphoribosyltransferase
At4g27070*0.038−1.52Tryptophan synthase, beta subunit 2 (TSB2)
Hormone metabolismAt1g10470*0.026−1.5Response regulator 4 (ARR4)
At1g19050*0.027−1.57Response regulator 7 (ARR7)
At1g19220*0.041.62Auxin-responsive factor 11 (ARF11)
At1g19850*0.0264.11Auxin-responsive factor 5 (ARF5)
At2g17820*0.04−0.82Histidine kinase (AHK1)
At2g22670*0.0361.15Auxin-responsive protein 8 (IAA8)
At2g34650*0.043.04Protein kinase PINOID (PID)
At2g47430*0.027−1.81Cytokinin-responsive histidine kinase (CKI1)
At3g44870*0.0263.59S-adenosyl-l-methionine:carboxyl methyltransferase family protein
At3g48100*0.031−2.54Two-component responsive regulator 5 (ARR5)
At5g60450*0.039−1.38Auxin-responsive factor 4 (ARF4)
StressAt1g240200.0413.24Bet v I allergen family protein
At1g46264*0.041.12Heat-shock transcription factor (HSFB4)
At1g75030*0.032−1.17Thaumatin-like protein 3
At2g14610*0.026−4.07Pathogenesis-related protein 1 (PR1)
At3g04720*0.026−4.15Pathogenesis-related protein 4 (PR4)
At3g109850.027−1.82Wound-responsive protein 12 (WI12)
At3g46230*0.041.5917.4 kDa heat-shock protein class (HSP17.4)
At4g09940*0.03−1.43Avirulence-responsive family protein
At4g28240*0.027−1.19Wound-responsive protein-related
At5g12140*0.036−1.21Cystatin (CYS1)
At5g39730*0.034−1.81Avirulence induced gene 2 like protein (AIG2L)
At5g59720*0.0272.3618.1 kDa heat-shock protein class (HSP18.1)
RedoxAt1g11530*0.046−2.38Thioredoxin-related protein
At2g160600.0272.43Non-symbiotic hemoglobin 1 (GLB1)
At2g31570*0.027−2.08Glutathione peroxidase 2 (GPX2)
At2g43350*0.032−1.05Glutathione peroxidase 3 (GPX3)
At3g15360*0.033−1.01Thioredoxin M-type 4 (TRX-M4)
At5g60640*0.048−0.97Protein disulfide isomerase-like (PDIL1–4)
TranscriptionAt1g29280*0.041−1.47WRKY (WRKY65)
At3g11280*0.033−2.5Myb family transcription factor
At3g46130*0.033−1.15Myb family transcription factor (MYB48)
At3g56400*0.0472.77WRKY (WRKY70)
At4g174600.0294.3Homeobox-leucine zipper protein 1 (HAT1)
At5g15830*0.044−2.22Bzip transcription factor family protein
At5g22220*0.0351.11E2F transcription factor b (E2Fb)
At5g59780*0.042−2.03Myb family transcription factor (MYB59)
DNA/CellAt1g07820*0.0282.04Histone H4
At1g09200*0.0393.21Histone H3
At1g14690*0.0381.03Microtubule associated protein (MAP65–7)
At1g441100.0413Cyclin A1;1
At1g47210*0.0262.22Cyclin A3;2
At2g38720*0.0461.72Microtubule associated protein (MAP65–5)
At3g50070*0.0450.8Cyclin D3;3 (CYCD3;3)
At3g53750*0.0411.64Actin 3 (ACT3)
At3g54560*0.0380.85Histone H2A (HTA11)
At3g62030*0.0370.87Cyclophilin 4 (ROC4)
At4g020600.0261.56Prolifera protein (PRL)
At5g12910*0.0321.03Histone H3
At5g27670*0.0471.8Histone H2A (HTA7)
At5g33300*0.0361.36Chromosome-associated kinesin-related
At5g59690*0.0272.39Histone H4
At5g65460*0.0382.85Kinesin motor protein-related
TransportAt2g43330*0.0262.05Inositol transporter 1 (INT1)
At3g16240*0.0271.39Delta tonoplast intrinsic protein 1 (δ-TIP1)
At3g24300*0.04−2.42Ammonium transporter 1 (AMT1.3)
At3g26520*0.041−2.72Gamma tonoplast intrinsic protein (γ-TIP2)
At3g45680*0.026−2.15Proton-dependent oligopeptide transport
At4g35100*0.0271.94Plasma membrane intrinsic protein (PIP3A)
At5g43350*0.026−2.82Inorganic phosphate transporter (PHT1)
ProteinAt1g10940*0.0480.74Serine/threonine protein kinase (ASK1)
At1g75950*0.044−1.02E3 ubiquitin ligase (SKP1)
At2g20450*0.0430.8760S ribosomal protein L14 (RPL14A)
At3g43430*0.026−1.62Zinc finger (C3HC4-type RING finger) family
At3g61160*0.0271.77Shaggy-related protein kinase beta
At4g200700.03−2.4Peptidase M20/M25/M40 family protein
At4g27960*0.04−1.84Ubiquitin-conjugating enzyme E2 (UBC9)
At5g05110*0.037−1.56Cysteine protease inhibitor
At5g09770*0.0360.94Ribosomal protein L17 family protein
MiscellaneousAt2g35380*0.026−4.61Peroxidase 20 (PER20)
At3g223700.026−2.63Alternative oxidase 1A (AOX1A)
At4g13310*0.026−2.06Cytochrome P450 (CYP71A20)
At4g31500*0.026−2.15Cytochrome P450 (CYP83B1)
At4g399500.033−1.96Cytochrome P450 (CYP79B2)

The functional group with the highest number of DEG was ‘Metabolism’, with 20 up-regulated and 127 down-regulated genes (Figure 3a), mainly represented by ‘Secondary metabolism (SM)’ with a P-value <0.05 (Figures 3b and S5a, Tables 1 and S3). Other categories with a high number of DEG were ‘RNA’ and ‘Protein’, also with expression profiles significantly different to the rest of the categories (Figure 3a, Table S3). ‘Stress’ included a very high proportion of down-regulated genes associated with biotic and abiotic stress (Figure S1). However, heat-stress response genes, as those coding heat-shock proteins and a heat-shock transcription factor were up-regulated, which supports our previous results (Barcala et al., 2008; Escobar et al., 2003; Table 1). In ‘Hormone physiology’, DEG related to cytokinin and ethylene (ET) responses were mostly down-regulated. No DEG related to jasmonate (MJ) and only one related to salicylic acid (SA) were differentially expressed (Figure S1). A high proportion of the DEG were ‘Not assigned’, including annotated genes not yet correlated to any ontology (151) and genes of unknown function or coding hypothetical proteins (251).

Validation of microarray data

qPCR for several of the up- and down-regulated genes, using RNA from hand-dissected 3 d.p.i. galls versus uninfected control root segments (Table S1) was performed. Most of the genes with a high induction level in the microarray (e.g. EXPA6 and PR4) showed the same trend when analysed by qPCR (Table 2). However, relative transcript abundance was much lower when using RNA from hand-dissected galls than from LCM-dissected GCs in the microarray hybridization. Only one gene (HAT1) was not validated. Therefore, 10 out of 11 tested genes confirmed the microarray data. Further validation was obtained by analysis of a promoter trap T-DNA line corresponding to SRS5 (At1g75520; Favery et al., 2004), an up-regulated member of the SHORT INTERNODES protein family (Kuusk et al., 2006). In agreement with the microarray results, GUS activity was detected in 3 d.p.i. GCs and surrounding vascular tissue (Figure 4a,b), whereas uninfected plants showed GUS activity only in emerging lateral roots (Figure 4c). Therefore, the activation pattern of this promoter-trap line was fully consistent with the LCM-microarray data. In addition, two genes (PRL and CYP83B1; Table 2) were also validated by qPCR with RNA from the LCM GCs versus vascular cylinder cells from the same three independent replicates used for the microarray. PRL was confirmed to be upregulated and CYP83B1, a ‘GCs distinctive gene’, down-regulated, similarly to the results of the microarrays hybridization.

Table 2.   Comparison of qPCR and microarray expression data for selected genes
IDGene nameLog2
2−ΔΔCt GallMicroarray
  1. RNA from hand-dissected 3 d.p.i. galls versus uninfected primary root segments or RNA from LCM GCs versus control vascular cylinder cells of uninfected primary root segments was used for qPCR, as indicated.

At1g19850MP1.484.10
At1g37130NR21.491.28
At1g44110*CYCA1;10.923
At1g47210*CYCA3;20.692.22
At2g28950EXP63.284.39
At3g04720*PR4−2.33−4.15
At3g11280*MYB−0.46−2.5
At3g22370AOX1A−1.87−2.76
At3g26520*γ-TIP2−0.72−2.72
At3g48100*ARR5−1.01−2.54
  2−ΔΔCt GCMicroarray
At4g02060PRL1.11.55
At4g31500*CYP83B1−3.8−2.14
Figure 4.

 Validation of microarray data.
Promoter-trap line DZD14 analysed after Meloidogyne javanica infection for validation of the up-regulated gene At1g75520.
(a) Dark-field micrograph semithin-section (6.5 μm) of 3 d.p.i. gall after GUS staining. Purple precipitate indicates GUS expression in GCs (white arrow) and vascular surrounding tissue.
(b) Three-d.p.i. gall-containing root showing GUS blue signal in the gall.
(c) Uninfected roots showing GUS activity in lateral root primordia. Scale bars: 50 μm (a), 0.2 mm (b, c).

Comparative transcript profiles of GCs and galls

We next compared our GC transcriptome data with that of the hand-dissected galls. Galls transcripts were compared with the uninfected primary root segment samples. The samples were processed using the same methods used for LCM GCs, i.e. the same tissue preparation, equivalent isolation of RNA and the same amplification, hybridization steps and data processing.

Interestingly, only 120 genes out of 1161 DEG in GCs were shared with those of the gall transcriptome (Table 1a). A high number of down-regulated genes in GCs were not detected as DEG in galls (q-value <0.05), particularly the genes with Fc ranging –1 to –3 (Figure 2a). Thus, an apparent dilution effect of GC transcripts was observed when RNA from hand-dissected galls was used in the hybridization, especially for the repressed genes. The same effect was also seen for the genes with Fcs ranging from 1 to 3 (Figure 2a). The Fc values of most of the co-expressed genes were lower in galls than in LCM GCs, suggesting that the relative contribution of GC transcripts in galls was low. In addition, only 11 genes showed opposite regulation in galls as compared with that of the GCs (Table S2).

The non-co-expressed genes either from galls or GCs were named either ‘Gall distinctive’ or ‘GC distinctive’ respectively, and were further classified in functional categories by MapMan. Whereas in galls most of the distinctive genes were upregulated among all the categories except in ‘Photosynthesis’, the GC distinctive genes were predominantly down-regulated in most of the categories, except in ‘DNA’ and ‘Cell’ (Figure 2b,c). Interestingly, the categories of ‘SM’ and ‘Biotic stress’ included a high proportion of ‘Gall and GC distinctive genes’, especially in GCs, with them repressed in these cells, but up-regulated in galls (Figure S2a–d). However genes related to cell wall modification, such as expansins (EXPA6, EXPA1 and EXPA2), were mostly up-regulated co-expressed genes (Table S2). Of the cell cycle genes, only CYCLIN A1 was co-regulated, while CYCLIN D3;1 was gall distinctive and CYCLIN D3;3 was GC distinctive (Tables 1 and S2). Interestingly, only eight out of the more than 100 genes encoding transcription factors differentially expressed in GCs were co-regulated between GCs and galls. Hence, most of the DEG-coding transcription factors in the category of ‘RNA’, with a significantly different expression profile (P < 0.05) to the rest of the ‘GC distinctive categories’, were either ‘GC or gall distinctive’ (Figure 2, Table S3). In particular, the MYB and AP2/EREBP families were mostly up-regulated in galls and down-regulated in GCs (data not shown).

Meta-analysis of hormone responses and GCs

To explore transcriptional similarities between GCs and some physiological processes, comparisons were conducted with Arabidopsis transcriptome data downloaded from the GENEVESTIGATOR database (Zimmermann et al., 2004) (see Experimental procedures).

Few studies have explored the possible involvement of different hormones, including cytokinins, auxins and ethylene during gall and GC development (Glazer et al., 1986; Hutangura et al., 1999; de Meutter et al., 2003, 2005; Lohar et al., 2004). Our goal was to determine whether genes whose expression changed in response to hormones, also showed differential regulation in GCs. A meta-analysis was performed by clustering global gene expression data from 10 hormonal treatments (ABA, aminocyclopropane-1-carboxylic acid (ACC), brassinolide (BL), ET, gibberellic acid (GA3), indolacetic acid (IAA), MJ, SA and zeatin), including one treatment for vascular cell differentiation (brassinosteroids and boric acid, BL/H3BO3).

Hierarchical clustering (HCL) revealed that the global gene expression pattern most closely related to developing GCs corresponded to a BL/H3BO3 treatment that induces suspension cells to differentiate towards xylem elements (Kubo et al., 2005; Figure 5a). All other hormone treatments clustered into a single group clearly separated from the 3 d.p.i. GCs and the BL/H3BO3 experiments. Despite the similarity between the two experiments, non-co-expressed genes outnumbered co-expressed genes (Figure 6, Table S4). Gene classification in the MapMan categories (Figure S3a) revealed that the classes ‘Cell’, ‘Redox regulation’ and ‘Signaling’ contained the highest number of co-regulated genes. Among these were genes coding cytoskeletal proteins, cyclins and receptor kinases (Table S4). A major difference was found in genes involved in sugar modification and starch synthesis, such as sucrose synthases and starch debranching and branching enzymes, induced in GCs but repressed in BL/H3BO3 (Tables 1 and S4). Genes from the ‘SM’ category, coding enzymes of the phenylpropanoids, lignins, simple phenols and shikimate biosynthetic pathways involved in plant defences as well as in xylem differentiation, showed an overall down-regulation in GCs, but were mostly upregulated during tracheid development (Figure S3b, Table S4).

Figure 5.

 Similarities among 3 d.p.i. GCs and other Arabidopsis transcriptomes.
TMev HCL clustering of microarray data from (a) hormone-related experiments and (b) biotic interactions. Both analyses identify one transcriptome as the most closely related to 3 d.p.i. GCs (enclosed in a black box). The numbers within the tree correspond to the bootstrap values after 1000 iterations.

Figure 6.

 Comparison of the most closely related transcriptomes identified by clustering.
Venn diagrams showing numbers of common and non-common DEG among 3 d.p.i. GCs (Meloidogyne javanica), crown galls (Agrobacterium tumefaciens) and xylem-differentiating suspension cells (BL/H3BO3). The analysis identifies 280 putative regulated genes (50 up- and 230 down-regulated), distinctive to GCs, when compared with crown galls and in vitro differentiating vascular cells.

Transcriptome meta-analysis of biotic interactions

Also using meta-analysis, we compared global gene expression responses during different biotic interactions to those occurring in GCs. Arabidopsis transcriptome data from interactions with pathogenic bacteria (A. tumefaciens and Pseudomonas syringae), necrotrophic (Alternaria brassicicola, Phytophthora infestans and Botrytis cinerea) and biotrophic (Erysiphe cichoracearum and Erysiphe orontii) pathogenic fungi, a cyst nematode (Heterodera schachtii), an arbuscular mycorrhiza (Gigaspora rosea), a lepidopteran (Pieris rapae), a thrip (Franklinville occidentalis) and an aphid (Myzus persicae) were clustered by HCL analysis. The most similar profiles to the GCs were in crown-gall tissue formed by A. tumefaciens. The remainder, including H. schachtii, was grouped together in a cluster that was relatively dissimilar to GCs (Figure 5b).

Crown galls (Deeken et al., 2006) and 3 d.p.i. GCs showed a high number of co-expressed genes distributed in most subcategories, with similar transcriptional regulation-mediated changes in the cytoskeleton, cell cycle, chromatin remodeling and cell wall (Figure S4a, Table S5). The transcriptional patterns of transcription factors such as MYBs and WRKYs, and 8 of 21 genes in the category ‘Hormones’ either related to auxin signaling or auxin up-regulated genes were also common. In contrast, all genes in the cytokinin category were up-regulated by A. tumefaciens, but down-regulated in GCs. However, genes in ‘Metabolism’, ‘Protein’ and ‘Stress’, including nearly all of the ‘SM’ genes, were up-regulated by A. tumefaciens but repressed in GCs (Figures S4a,b and S5a,b, Table S5).

GC-distinctive genes?

Analysis of the transcriptomes of crown galls from A. tumefaciens, in vitro differentiating vascular cells (DVCs) and GCs shows that most of the genes that were common to both GCs and DVCs were down-regulated (Figure 6), which is in contrast to those shared by crown galls and DVCs, generally up-regulated. At the intersection of the three experiments ‘Not Assigned’, ‘Signaling’ (including mostly receptor kinases, mainly LRRs such as BAM1), ‘Protein degradation’ and ‘RNA regulation of transcription’ (including several transcription factors, such as LBD41, WRKY21 and IAA8) were highly represented. In contrast, the highest number of non-shared genes among the three experiments corresponded to down-regulated genes in GCs, distributed in ‘Not assigned–unknown’, ‘RNA’ (encompassing mainly transcription factors, especially from the AP2/EREBP and MYB families) and unclassified, as well as ‘Proteins’, ‘Stress’ and ‘Secondary metabolism’ (Tables S4 and S5).

Discussion

Down-regulation of gene expression occurs early during
GC differentiation

Knowledge of the regulation of gene expression in infected roots mainly corresponds to galls at the mid- to late-infection stages (Ramsay et al., 2004; Bar-Or et al., 2005; Hammes et al., 2005; Jammes et al., 2005). Galls are voluminous and complex multicellular structures containing only five to eight GCs. The contribution of GCs to the gall is relatively small, particularly in the early stages of development (Wang et al., 2003). Microcapillar-aspired and hand-dissected tomato GCs contain GC-enriched material (Wilson et al., 1994; Wang et al., 2003). In both cases, GCs were collected at very late developmental stages (25 d.p.i. and 1–2 months after infection, respectively) when key gene expression reprogramming events that lead to GC differentiation are likely to have already occurred. Very recently, LCM has been used to isolate GC RNA at early infection stages only in tomato (Fosu-Nyarko et al., 2009; Portillo et al., 2009). In contrast, data on cyst nematode feeding cells (syncytia)-specific transcriptional changes at different infection stages have been obtained by combining LCM and microarray analysis (Ithal et al., 2007b; Klink et al., 2007) and microaspiration in Arabidopsis (Szakasits et al., 2009). In this study, we provide novel insights into the Arabidopsis GC transcriptome at a very early developmental stage (3 d.p.i.) through direct comparison with the transcriptome of vascular cells from equivalent segments of uninfected roots.

Differential hybridization of Arabidopsis microarrays containing 29 000 single ORF sequences revealed 1160 DEGs (Table 1a), 850 down-regulated and 310 up-regulated. This figure contrasts with the differential transcriptome analysis performed in Arabidopsis galls at early- to mid-infection stages, although numbers of up- and down-regulated genes were similar (Jammes et al., 2005). GCs also differ from syncytia, which showed a higher number of up-regulated than down-regulated genes (Puthoff et al., 2003; Ithal et al., 2007a; Szakasits et al., 2009). On the other hand, microarray analysis from whole galls at late infection stages (21 d.p.i.) showed that two-thirds of the DEG were down-regulated (Fuller et al., 2007). The comparison between the transcriptome of the 3 d.p.i. GCs and the hand-dissected galls, indicates that most of the ‘GC distinctive genes’ (94%) were down-regulated, with representatives in most of the functional categories (Figure 2). Our data strongly suggest that down-regulation mechanisms might be essential in Arabidopsis, at least for the initial stages of GC differentiation, as previously suggested for galls (Goddijn et al., 1993; Jammes et al., 2005). Hence, the first evidence for a putative gene silencing mechanism through miRNAs during cyst–nematode infection has been recently provided (Hewezi et al., 2008). According to these data, the ‘GC distinctive genes’ from biotic stress and SM were mostly repressed in contrast to the ‘Gall distinctive’ distributed in the same categories (Figure S2). Microarray hybridization results were confirmed by qPCR and with a promoter trap line (Figure 4, Table 2).

Comparison of the transcriptomes of GCs and hand-dissected whole galls allowed the identification of more than 1000 ‘GC distinctive’ and 547 ‘Gall distinctive’ genes (Tables 1 and S2) at 3 d.p.i. Likewise, data for hand-dissected 7 d.p.i. Arabidopsis galls (Jammes et al., 2005) shared only 94 co-regulated genes out of the ca. 1000 that were DEGs in the GC transcriptome samples. Similarly, other genes up-regulated in galls, such as LEA14, ENOD40, EXPA5, ARR5 and RHA1/ATRAB5A (Gheysen and van der Eycken, 1996; Vercauteren et al., 1998; Favery et al., 2002; Lohar et al., 2004; Gal et al., 2006), were not detected in GCs by either in situ histochemical techniques or promoter–GUS fusions. These data confirm the molecular distinctiveness of the GC transcriptome within the gall, indicate the strong dilution effect of the GCs transcripts in the whole galls RNA and demonstrate the importance of obtaining the GC profile.

Plant defence and secondary metabolism shut-down in GCs

A distinctive feature of the GCs transcriptome was the general down-regulation of genes classified in the ‘SM’ and ‘Stress’ groups (Figures 3, S1, S2 and S5a). Most genes involved in the phenylpropanoid pathway, typically part of a defence mechanism against pathogen attack (Dixon et al., 2002), were repressed in 3 d.p.i. GCs, including PAL1, PAL2, 4CL3 and CAD (Table 1). The nematode-secreted chorismate mutase and calreticulin may be directly involved in plant defence suppression (Doyle and Lambert, 2003; Jaubert et al., 2005), supporting our results. Cyst nematode-infected roots show a general up-regulation of ‘SM’ and ‘Stress’ genes (Ithal et al., 2007a; Puthoff et al., 2007; P. Urwin, unpublished results), a pattern conserved at early- and mid-infection stages of soybean microdissected syncytia (Ithal et al., 2007b). Although defence gene expression was repressed in 5 d.p.i. Arabidopsis microaspirated syncytia (Szakasits et al., 2009), the ‘SM’ down-regulation might be a trait specific to GCs. Flavonoids accumulate in galls, but not in GCs, in parallel with the activation of three chalcone synthase promoters (Hutangura et al., 1999). PRP-coding genes are induced in hand-dissected tomato galls (Bar-Or et al., 2005; Table S2), but repressed in 3 d.p.i. GCs. Arabidopsis mutants deficient in flavonoid synthesis confirm a non-critical role of this pathway for RKN establishment, as reproduction was successful in these mutants (Wuyts et al., 2006).

Most WRKY, MYB, AP2/EREBP and bZIP transcription factor family members, induced by biotic or abiotic stresses including nematodes (Eulgem, 2005; Li et al., 2006; Eulgem and Somssich, 2007; Grunewald et al., 2008) were down-regulated in GCs, but not in galls (Figure 2, Table S2). Similarly, defence and PRP genes directly or indirectly regulated by SA and jasmonic acid (Lorenzo and Solano, 2005; van Loon et al., 2006) were not differentially expressed in 3 d.p.i. GCs (Figure S1). These data indicate that transcriptional plant defence responses were almost totally repressed in early stages of GC development. The coordinated down-regulation of specific gene sets involved in ‘SM’ seems to be a distinctive key feature of GCs. In this respect, host defence suppression during compatible and symbiotic interactions has also been reported, though not directly in the infected plant cells (Mithofer, 2002; Jammes et al., 2005; Nomura et al., 2005).

A complex modulation of plant-hormone pathways in early GCs

Two early auxin response genes coding IAA-amido synthases, which conjugate amino acids to auxins in vitro and participate in free-auxin balance in GCs (Glazer et al., 1986), two auxin responsive factor (ARF) genes and one auxin-responsive protein gene (IAA8) were induced in 3 d.p.i. GCs. Transcripts of two Aux/IAA factors, presumably repressors of auxin responsive transcription (Leyser, 2006; Quint and Gray, 2006), were down-regulated (Table 1, Figure S1). This might reflect the detection of an increase in auxin concentration in GCs, and indicates an important role for IAA during formation of RKN feeding cells, as suggested (Hutangura et al., 1999; Goverse et al., 2000b). RKN could also contribute to this local auxin increase by producing additional auxins (de Meutter et al., 2005). Our data are consistent with a transient and local increase in either auxin sensitivity or auxin accumulation in GCs as proposed (Karczmarek et al., 2004; Grunewald et al., 2009). Few genes related to ethylene signaling were activated in 3 d.p.i. GCs, and all genes related to ethylene biosynthesis were down-regulated, as were ethylene-responsive PRs (Table 1, Figure S1). Accordingly, ethylene was not detected in galls at 1–2 d.p.i. (Glazer et al., 1983). Although ethylene could be critical at the mid-stages of infection during gall enlargement and during cyst–nematode interaction (Glazer et al., 1983; Goverse et al., 2000b; Wubben et al., 2001), transcriptional patterns observed at 3 d.p.i. do not support a crucial role for ethylene-activated pathways during early GC differentiation. Additionally, genes encoding negative regulators of cytokinin signaling, such as TYPE-A RESPONSE REGULATORS (ARR4, ARR5, ARR7), and members of the two-component receptor system (CKI1 and AHK1) (Table 1, Figure S1; Hwang and Sakakibara, 2006) were all down-regulated in our study. This finding is in partial agreement to the ARR5 promoter activation in Lotus japonicum galls but not inside the GCs (Lohar et al., 2004). Further studies involving genes in hormone biosynthesis and action are needed to elucidate their putative role during GC differentiation and/or gall formation.

GC initiation involves re-entry into the cell cycle

Two of the earliest events in GC development are the increasing number of nuclei and activation of several genes related to cell cycle progression at feeding sites (Goverse et al., 2000a; Gheysen and Fenoll, 2002). Our transcriptome analysis in 3 d.p.i. GCs corresponds to a state in which cells have re-entered the cell cycle: genes encoding cyclins specifically expressed at early G1- and S-phases (CYCA1,1, CYCA3,2 and CYCD3,3) were induced (Table 1). As in 7 d.p.i. galls (Jammes et al., 2005), none of the mitotic cyclin genes (e.g. seen as CYCA2,1 and CYCB1,1 promoter:GUS fusions; de Almeida Engler et al., 1999; Niebel et al., 1996) were induced in 3 d.p.i. GCs. The cause of this discrepancy is currently unclear.

Our results indicate that 3 d.p.i. GCs were no longer arrested at G0 phase, as an E2Fb gene, which can promote both G1–S and G2–M transitions (Magyar et al., 2005) was induced (Table 1). E2F factors prepare cells for DNA replication by inducing the transcription of genes coding for components of the pre-replication complex, such as the replication licensing factors MCMs (minichromosome maintenance protein) (da Ines et al., 2007). PROLIFERA, a member of the MCM gene family, was also induced in 3 d.p.i. GCs and in syncytia (Huang et al., 2003) (Table 1), together with other S-phase specific genes, coding helicases and histones. As in galls (Jammes et al., 2005; Table S2), GCs accumulate transcripts from genes encoding histones H4 and H3.1, associated with proliferation in cell suspensions (Menges and Murray, 2002) or with replication (Okada et al., 2005), as well as transcripts from H2As such as HTA11 (Yi et al., 2006; Table 1).

Transcriptional changes in genes coding cytoskeletal proteins and cell wall modifying enzymes, including microtubule-associated kinases, syntaxins, kinesins, expansins, xyloglucan endotransglycosylases and pectate lyases, were mostly shared with galls (Table S2), except for the down-regulation of those involved in cell wall synthesis and lignification (Table 1). This agrees with the dramatic cellular changes observed during GC ontogeny (Huang, 1985; Van der Eycken et al., 1996; de Almeida Engler et al., 1999, 2004; Caillaud et al., 2008b). It may reflect events very early in GC development, where cytoskeletal changes are mostly associated with the cell cycle and vesicle trafficking, and cell wall modifications are associated with cell enlargement.

GC founder cells

The expression pattern of cell cycle-related genes in 3 d.p.i. GCs corresponds with the onset of a new differentiation program, requiring re-entry into the cell cycle. We found a high degree of similarity between the transcriptomes of 3 d.p.i. GCs and suspension cells differentiating into xylem parenchyma and vessel elements (Kubo et al., 2005) (Figures 5, 6 and S3, Table S4). This suggests that the ‘Root stem cells’ that start a new gene expression program could be vascular pro-elements. This possibility is supported by existing histological evidence that identified metaxylem, protoxylem or xylem parenchyma cells as the initial cells that develop into GCs (Christie, 1936; Dropkin and Nelson, 1960; Bird, 1961; Niebel et al., 1993; Williamson and Hussey, 1996; Bird and Koltai, 2000). Further evidence comes from studies expressing a nematode-secreted chorismate mutase in plants, suggesting that nematodes could inhibit the final differentiation of root vascular cells (Doyle and Lambert, 2003).

The similarities between both transcriptomes provide a molecular evidence that root cells differentiating to xylem elements could be the ‘Stem GCs’. In this respect, proliferating tracheids have recently been described around GCs (Hoth et al., 2008). The strong global gene repression observed in developing GCs, which includes a high number of non-common genes with the BL/H3BO3 experiment, might be part of a mechanism to interfere with a differentiation program already established in those founder cells. In particular, we found that genes related to ‘SM’ (Figures S3 and S5, Table S4) were mostly induced in tracheid differentiating cells, but repressed in GCs, while the opposite was true of genes associated with the regulation of sugar synthesis and starch-modifying enzymes (Tables 1 and S4). This situation could reflect a coordinated down-regulation of several pathways, to confer on GCs a unique molecular identity suited to their transformation into a metabolic sink, which is in contrast to tracheid differentiation. On the other hand, co-regulated genes that were found in both experiments to be involved in cytoskeleton changes, cell cycle or signaling, including kinesins, cyclin D3 (Table 1), or receptor kinases such as LRRs (At3g46330) (Figure S3, Table S4), may reflect a utilization, by early-forming GCs, of transcriptional networks already programmed in the initial xylem parenchyma or vessel cells.

Transcriptional similarities to other plant pathogen interactions

Among several biotic interactions, A. tumefaciens-induced crown galls showed the highest number of co-regulated genes that were similar to 3 d.p.i. GCs. The unexpected low similarity with cyst nematode-infected roots could be due, in part, to the lack of biological replicates or the different infection time (21 d.p.i.) for the cyst nematode analysis; also, syncytia were not isolated, while GCs were subjected to LCM. Interestingly, we see few similarities with the transcriptomes of soybean LCM 5 d.p.i. syncytia (Ithal et al., 2007b); and in Arabidopsis, of the 200 most up-regulated and repressed genes from microaspirated 5 d.p.i. syncytia, only 27 were similarly regulated with those of the Arabidopsis 3 d.p.i. GC transcriptome (Table S6). The most abundant, co-regulated genes among cyst nematodes, crown galls and GCs were those involved in cell wall modification and cytoskeleton changes (Figures 6 and S4, Tables S5 and S6), suggesting a common reprogramming of genes related to cell shape changes and enlargement. Co-ordinated regulation of genes in GCs, syncytia and crown galls may reflect shared characteristics. All three induce plant cells to function as nutritional sinks, providing water and nutrients to the pathogens. At the cellular level, they induce changes in cell cycle and cell morphology, most probably through alterations in the mediating mechanisms (Escobar and Dandekar, 2003; Ditt et al., 2006; Citovsky et al., 2007; Caillaud et al., 2008a). Future comparisons between syncytia and GC transcriptomes using the same microarray platform, at equivalent early infection stages, may provide relevant information about the common and dissimilar pathways used by both nematodes to differentiate these specialized cells.

A high number of genes related to chromatin remodeling were co-regulated between GCs and crown galls, probably related to the fact that S-phase entry is an essential feature for both tissues (de Almeida Engler et al., 1999; Citovsky et al., 2007; Table S5). However, the major differences were related to ‘SM’; that includes genes associated with cell wall thickening and biotic stress. In contrast to crown galls, a general down-regulation was observed in early developing GCs (Figure S5, Table S5). A. tumefaciens can also suppress plant defences at later stages of infection, but a general up-regulation of defence-related genes was observed at 24 h post inoculation (hpi) in tobacco (Veena et al., 2003) and at 48 hpi in Arabidopsis (Ditt et al., 2006).

Concluding remarks

The relevance of obtaining GC-specific transcript profiles, by direct isolation of the GC RNA through LCM, is shown in the transcriptional patterns of the many genes that were up-regulated in galls, but not in GCs. Our global transcriptional data and the meta-analysis comparing it to other transcriptomes suggest that down-regulation of particular stress-defence genes and several transcription factors, as well as many genes of still unknown function, might be key events leading to a unique differentiation programme unlike other biotrophic interactions. Similarities, such as the regulation of receptor kinases among crown galls, GCs and differentiating vascular cells, could reflect common transduction pathways. Furthermore, similarities encountered in the initial steps of vascular cell differentiation may explain the GCs ontogeny at a molecular level, while their differences may indicate important modifications leading to the distinctive GC identity. Future research on the ‘specific or distinctive GC-regulated genes’ may help clarify the role of key cellular regulators in GC formation and provide important clues about the mechanisms by which this unique differentiation process is induced by RKN. This information should provide new targets to engineer nematode resistance or tolerance in plants.

Experimental procedures

Plant growth and infection

An average of 10 Arabidopsis thaliana Col-0 seeds per plate were surface sterilized and sown in modified Gamborg B5 as described (Escobar et al., 2003). Plates were kept at 4°C for 2 days, and transferred to a growth chamber (Percival AR-66LX) at 25–26°C, 60% relative humidity and a long-day photoperiod. Four days later, half of the plates containing Arabidopsis seedlings were inoculated with 10–12 freshly hatched Meloidogyne javanica J2 per root tip. Plants were carefully examined every 12 h under a Leica Mz125 stereomicroscope to establish a penetration and infection timeline, resulting in a maximum error of 12 h when assessing gall age. Galls were collected 3 d.p.i. To reduce sample variation, more than 20 independent infection experiments were conducted and at least 10 plates were infected for each experiment.

Sample preparation for LCM

Fixation and embedding protocols for cryosectioning were based on Nakazono et al. (2003) and performed mainly as described in Portillo et al. (2009). Longer incubation steps (3 h) were made during cryoprotective treatment with 10 and 15% sucrose solutions, followed by an over-night infiltration in 34.3% (w/v) sucrose, 0.1% (w/v) safranin-O, 0.01 m in PBS buffer pH 7.4. Prior to embedding, samples were rinsed once in the final sucrose solution, without safranin, to eliminate excess dye. Cryosectioning was as described in Portillo et al. (2009).

LCM and RNA extraction

LCM and RNA extraction were performed as described in Portillo et al. (2009). Only clearly identified GCs were microdissected. Caps with microdissected GCs were obtained from galls of different plants and from different infection events. To account for biological variation, RNA from 11 different caps was extracted independently and pooled in three separate sets, each one consisting of total RNA from 205 GCs. Each pool, which was considered an independent biological replicate, was hybridized independently along with its corresponding uninfected control sample. For the control samples, vascular cylinder tissue of uninfected primary root sections was microdissected. RNA extracted from 12 caps was pooled from 305 microdissected control root tissue sets.

Microarray hybridization and statistical analysis

RNA amplification, labeling, slide hybridization, statistical analysis and normalization of data were basically performed as described in Adie et al. (2007). RNA from each biological replicate, either from the hand-dissected galls or the LCM GCs, was firstly converted to cDNA and divided into two aliquots: one for microarray analysis and the other for microarray data validation. For microarray analysis one aliquot was subjected to amplification (two rounds) with a MessageAmp aRNA amplification kit (Ambion, http://www.ambion.com/). Quality of aRNA transcripts was assessed with a 2100 Bioanalyzer (Agilent Technologies, http://www.home.agilent.com/).

Three biological replicates were independently hybridized using Superamine TeleChem slides containing more than 29 000 spots corresponding to the Arabidopsis thaliana synthetic 70-mer oligonucleotides set, version 3 from Qiagen-Operon obtained from Dr David Galbraith (University of Arizona). Background correction and normalization of expression data were performed using LIMMA (Smyth and Speed, 2003; Smyth, 2004). LIMMA is part of Bioconductor, an R language project (Ihaka and Gentleman, 1996). First, the dataset was filtered based on the spot quality. A strategy of adaptive background correction was used that avoids exaggerated variability of log ratios for low-intensity spots. For local background correction the ‘normexp’ method in LIMMA to adjust the local median background was used. The resulting log ratios were print-tip loess normalized for each array (Smyth and Speed, 2003). To have similar distribution across arrays and to achieve consistency among arrays, log-ratio values were scaled using as scale estimator for the median absolute value (Smyth and Speed, 2003). Linear model methods were used for determining differentially expressed genes. Each probe was tested for changes in expression over replicates by using an empirical Bayes moderated t-statistic (Smyth, 2004). To control the false discovery rate, P-values were corrected by using the method of Benjamini and Hochberg (1995). q-values indicate significance after correction for multiple testing controling the false discovery rate as indicated before. The expected false discovery rate was controlled to be <5%. Gene annotation source was Ath_AGI_TAIR7.m02.

Clustering and meta-analysis of data

Hierarchical clustering (HCL) of significantly up- and down-regulated genes was performed using the TIRG MultiExperiment Viewer programme (TMev; http://www.tm4.org/mev; Saeed et al., 2003, 2006). Pearson uncentered metric distance and a complete linkage were used with several available microarray experiments that were downloaded from the GENEVESTIGATOR website (http://www.genevestigator.ethz.ch/; Zimmermann et al., 2004). Genes that showed significant (q-value <0.05) differential expression in the GC array were selected. For those identifiers, the expression data (log2 values) of each experiment used for the comparisons were downloaded from the GENEVESTIGATOR database. Similarities or differences were established by comparing the log2 ratios among experiments with TMev. To ensure a normal distribution with a mean of 0 and a standard deviation equal to 1 for these log2 values, a z-score transformation was performed. For each gene, the mean and standard deviation of the log2 values from all the different analysed treatments were calculated and single log2 values (of each gene and treatment) were transformed by subtracting the mean and dividing by the standard deviation. A few loci of our DEG were not represented in the ATH1 22 k Affymetrix platform, and were excluded.

Microarray validation

A minimum of 100 hand-dissected galls per experiment, monitored as previously described, were collected and immediately frozen in liquid nitrogen. Total RNA from three independent experiments was extracted with Tri-reagent (AB Gene, http://www.abgene.com/), cleaned through columns (RNeasy Plant Mini Kit, Qiagen, http://www1.qiagen.com/) and subjected to in-column DNase treatment (RNase-free DNase set, Qiagen). Eluted RNA was quantified using a NanoDrop spectrophotometer. Typically, 50 ng from each sample (galls and uninfected root segments) were used for cDNA synthesis with the Archive cDNA kit (Applied Biosystems, http://www3.appliedbiosystems.com/). Primer Express software (Applied Biosystems) was used for primer design (Table S1). GADPH was used as an internal standard. qPCR was performed with SYBR-Green technology (Power SYBR-Green, Applied Biosystems) in an Applied Biosystems 7500 apparatus by using the relative quantification assay (2−ΔΔCt method). Arabidopsis transgenic plants from a promoter trap collection (http://www-ijpb.versailles.inra.fr/en/sgap/equipes/variabilite/crg), T-DNA line DZD14 (INRA-Versailles Genomic Resource Center, France), were infected and monitored as described above. Three-d.p.i. galls were subjected to GUS analysis, fixed and sectioned as described in Barcala et al. (2008).

Acknowledgements

We thank Manuel Pérez for his bioinformatic support. CE acknowledges grants from the Fundación Ramón Areces and the Ministerio de educación (MED; AGL-2004–08103-C02–02 and AGL2007–60273), CF from the Junta de Comunidades de Castilla-La Mancha (JCCM, GC-02–011, the Ministerio de Ciencia e Innovación (MICINN, CSD2007-16057)) and RS, from the MED (BIO2004–02502, BIO2007–66935, GEN2003–20218-C02–02, CSD2007–00057-B) and from the Comunidad de Madrid (GR/SAL/0674/2004). MB was a recipient of a FPU fellowship from the MED, JC of an FPI from the MED and AG from the JCCM. KL and SC acknowledge funding from the UK BBSRC. BF was supported by an innovative grant from the INRA division ‘Plant Health and Environment’.

Ancillary