Genome-wide expression profiling Arabidopsis at the stage of Golovinomyces cichoracearum haustorium formation.

Compatibility between plants and obligate biotrophic fungi requires fungal mechanisms for efficiently obtaining nutrients and counteracting plant defenses under conditions that are expected to induce changes in the host transcriptome. A key step in the proliferation of biotrophic fungi is haustorium differentiation. Here we analyzed global gene expression patterns in Arabidopsis thaliana leaves during the formation of haustoria by Golovinomyces cichoracearum. At this time, the endogenous levels of salicylic acid (SA) and jasmonic acid (JA) were found to be enhanced. The responses of wild-type, npr1-1, and jar1-1 plants were used to categorize the sensitivity of gene expression changes to NPR1 and JAR1, which are components of the SA and JA signaling pathways, respectively. We found that the infection process was the major source of variation, with 70 genes identified as having similarly altered expression patterns regardless of plant genotype. In addition, principal component analysis (PCA) identified genes responding both to infection and to lack of functional JAR1 (17 genes) or NPR1 (18 genes), indicating that the JA and SA signaling pathways function as secondary sources of variation. Participation of these genes in the SA or JA pathways had not been described previously. We found that some of these genes may be sensitive to the balance between the SA and JA pathways, representing novel markers for the elucidation of cross-talk points between these signaling cascades. Conserved putative regulatory motifs were found in the promoter regions of each subset of genes. Collectively, our results indicate that gene expression changes in response to infection by obligate biotrophic fungi may support fungal nutrition by promoting alterations in host metabolism. In addition, these studies provide novel markers for the characterization of defense pathways and susceptibility features under this infection condition.

Obligate fungal biotrophs must establish compatible interactions with their hosts to survive. These organisms have evolved unique strategies to extract nutrients from infected living cells, including the assembly of specialized structures located in intimate contact with host cells. However, it is still unclear how these parasites are able to avoid plant defense activation (Panstruga, 2003). Past research efforts have typically focused on the study of mechanisms underlying fungal nutrition (Voegele, 2006), providing relatively less information regarding host conditions leading to the establishment of compatibility (for review, see Huckelhö ven, 2005;O'Connell and Panstruga, 2006). Recently, high-throughput transcriptome analyses have helped define the responses triggered in plant tissues upon recognition of these fungal pathogens. These studies were performed under conditions promoting activation of defenses involving either host resistance (Caldo et al., 2004;Eckey et al., 2004;Eulgem et al., 2004;Zierold et al., 2005;Michel et al., 2006) or nonhost resistance (Zimmerli et al., 2004;Bruggmann et al., 2005). Under conditions leading to compatibility, basic gene expression changes have been examined in Vicia faba (Wirsel et al., 2001), barley (Hordeum vulgare; Gjetting et al., 2004), and Brassica oleracea (Casimiro et al., 2006). Except for these reports, however, few studies have examined plant defense programs and global host metabolic features leading to the establishment of compatibility.
The interaction between Arabidopsis (Arabidopsis thaliana) and Golovinomyces cichoracearum (formerly Erysiphe cichoracearum) is an ideal system for exploring the compatibility between plants and obligate biotrophs. G. cichoracearum UCSC1 causes powdery mildew disease on some Brassicaceae and Cucurbitaceae species (Adam et al., 1999;Braun et al., 2002), and can establish either incompatible or compatible interactions with different Arabidopsis accessions (Adam and Somerville, 1996). Resistance against different Golovinomyces species is mediated by monogenic traits (Xiao et al., 1997(Xiao et al., , 2001Adam et al., 1999) such as that conferred by the RPW8 locus, which encodes a small novel protein conferring resistance against a wide range of powdery mildews (Xiao et al., 2001(Xiao et al., , 2005. In contrast, the conditions promoting compatibility have been less well characterized. The progressive differentiation of fungal infection structures is believed to require diverse host target molecules that are collectively called susceptibility factors (Heath, 2000;Parniske, 2000;Schulze-Lefert and Panstruga, 2003). Researchers have hypothesized that the absence of one or more of these molecules would lead to a durable and recessively inherited resistance against all pathogen species. Forward genetic studies on the Arabidopsis-G. cichoracearum interaction led to the isolation of six recessive loci necessary for successful fungal proliferation (pmr1-pmr6), four of which have been cloned to date. PMR2 is an ortholog of the MLO gene from barley, which encodes a modulator of plant defenses and cell death (Bü schges et al., 1997;Panstruga and Schulze-Lefert, 2003;Consonni et al., 2006). PMR4 encodes a callose synthase and lack of this gene product mediates resistance through deregulation of the salicylic acid (SA) pathway (Nishimura et al., 2003). PMR5 encodes a protein of unknown function whereas PMR6 encodes a pectate lyase-like protein and plants lacking PMR5 or PMR6 display alterations in cell wall composition (Vogel et al., 2002(Vogel et al., , 2004. Most of these PMR genes are constitutively expressed and are expected to constitute susceptibility components. The only host gene identified to date that may encode a susceptibility factor and becomes induced by fungal interaction is EDR1, which encodes a MAPKKK acting as a negative regulator of the SA and ethylene (ET) pathways (Frye and Innes, 1998;Frye et al., 2001;Tang et al., 2005a). Two other host products affecting resistance to G. cichoracearum are those encoded by EDR2 and EDR3. EDR2 is a novel protein expressed in all tissues and organs that may regulate defenses through lipid signaling (Tang et al., 2005b;Vorwerk et al., 2007), whereas EDR3 belongs to the dynamin-related protein family and is at least partially located in mitochondria (Tang et al., 2006). Further studies will be required to identify other nonconstitutive plant genes involved in the establishment of compatibility with G. cichoracearum.
During interactions with virulent pathogens, plants accumulate defense gene transcripts in a transient (Tao et al., 2003) or delayed manner (Frye and Innes, 1998). The major pathogen-inducible defense pathways are regulated by SA, jasmonic acid (JA) and ET, or complex networks interconnecting these defense pathways (Kunkel and Brooks, 2002;Glazebrook et al., 2003). Previous studies have partially analyzed the participation of the SA and JA/ET pathways in compatible interactions between Arabidopsis and G. cichoracearum. Exogenous JA or the SA analog BTH were found to protect susceptible plants against fungal infection (Maleck et al., 2002;Zimmerli et al., 2004;Glazebrook, 2005). Plants with mutations in the SA or JA signaling pathways (e.g. pad4, eds5, eds14, eds15, sid2, npr1-3, NahG, and coi-1) are hypersusceptible to G. cichoracearum and its closest relative Golovinomyces orontii (formerly Erysiphe orontii; Reuber et al., 1998;Dewdney et al., 2000;Ellis et al., 2002aEllis et al., , 2002b. In addition, plant mutants with constitutive or inducible activation of the SA-(pmr4, edr1) or JA/ET-(wrky70, cev-1) dependent defenses are resistant to G. cichoracearum (Frye and Innes, 1998;Ellis and Turner, 2001;Nishimura et al., 2003;Li et al., 2006). However, it is still unknown whether under compatible infection conditions these defense pathways are either not activated or inactivated by pathogen effectors. The use of defense pathway markers such as NPR1 and JAR1 would help to elucidate this point. NPR1 is an ankyrin-repeat protein that stimulates plant defenses through the SA-dependent pathway by interaction with transcription factors. This interaction is impaired in npr1-1 (nonexpressor of pathogenesis related genes) mutant plants (Cao et al., 1997;Fan and Dong, 2002;Rairdan and Delaney, 2002). JAR1 is involved in the JA adenylation that is apparently required for JA function and the jar1-1 (JA resistant 1) mutation reduces the impact of the jasmonatedependent defense pathway (Staswick et al., 1992(Staswick et al., , 1998(Staswick et al., , 2002, which becomes activated upon interaction with G. cichoracearum (Zimmerli et al., 2004), making jar1-1 plants useful to reveal responses associated with jasmonate-dependent pathway in this susceptible state.
The haustorium formation constitutes a relevant step for analysis of Arabidopsis gene expression changes during compatible interactions with G. cichoracearum. The haustorium is a structure that develops in the extracellular space of host epidermal cells, and expands to increase the surface contact with the invaded cell following invagination of a specialized hypha. Haustorial structures are only built by obligate biotrophs and their formation is considered a selective advantage for survival because they function as feeding organs for transporting sugar, amino acids, and water to the fungal tissues (Adam et al., 1999;Szabo and Bushnell, 2001), and they are likely to potentiate fungal virulence by helping release effectors to the host cell (Szabo and Bushnell, 2001;Panstruga, 2003;Ellis et al., 2006). In this sense, cells that already support the establishment of haustoria due to prior infection by compatible isolates of Erysiphales, become more susceptible to subsequent infections by nonhost fungi. Over the past decade, numerous studies have reported on the structural and physiological features of haustorial development, mostly regarding differentiation and nutrient transport (O'Connell and Panstruga, 2006). However, the molecular characteristics of host cell responses at this particular infection step are not well understood.
Here, we sought to describe qualitative and quantitative changes in the global gene expression profiles of susceptible Arabidopsis plants supporting the development of G. cichoracearum haustoria. We analyzed the features of compatibility at this infection stage, and further evaluated the contribution of the SA-and JA/ET-dependent defense signaling pathways in the pathogen-induced responses by comparing responses in infected wild-type, npr1-1, and jar1-1 plants. Our findings collectively contribute to knowledge regarding early host cell alterations generated in response to attack by this virulent obligate biotrophic fungus.

Experimental System
We used microarray analysis to characterize global transcriptional changes occurring in Arabidopsis leaf tissues infected with G. cichoracearum at early stages of haustorium formation. This step involves the generation of the fungal feeding organ and is critical for pathogen proliferation. To test the impact of NPR1and JAR1-defense signaling pathways at this infection step, we examined gene expression changes in wildtype, npr1-1, and jar1-1 plants treated with the virulent fungal isolate UCSC1 (Adam et al., 1999). We collected samples from infected tissues 18 h postinfection (hpi) because our observation of fungus-treated tissues indicated that the number of haustoria detected at 18 hpi was approximately 90% of that found at 24 hpi ( Fig.  1A), when the G. cichoracearum haustorium is fully developed in Arabidopsis (Adam and Somerville, 1996;Zimmerli et al., 2004). At 18 and 24 hpi, the abundance of primary hyphae, secondary hyphae, and haustoria did not significantly differ among the three tested genotypes (Fig. 1, A and B). At 96 hpi, conidiation was slightly lower in npr1-1 plants than in jar1-1 plants (Fig. 1C). However, from 6 to 8 d postinfection (dpi) onward, npr1-1 and jar1-1 plants displayed the same or even higher levels of susceptibility than wildtype plants (data not shown).
We performed four independent replicates for each plant genotype and isolated samples from uninfected (T0) and infected (T18) leaf tissues to obtain a set of 24 samples (four replicates of three genotypes, each under two infection states) for cDNA microarray analysis.
cDNA Microarray Analysis mRNA was isolated from the 24 samples and used to synthesize labeled cDNAs, which were hybridized into 24 microarrays as indicated in Supplemental Table  S1 and ''Materials and Methods''. Initially, we analyzed the technical and biological variability of the observed gene expression changes by determining the Pearson's correlation coefficient (r) for data obtained at T18 and T0, in plants from each genotype across the four replicates. Low correlation was found among overall responses (r 5 0.12-0.19; average 0.16) suggesting that the large majority of genes were insensitive to infection. Differentially expressed genes (discussed in detail in the next section) were defined as those showing an absolute standardized difference in expression between T18 and T0 $2.5 SD from zero in at least one replicate. When these differentially expressed genes were assessed for correlation, we observed a high level of reproducibility of responses among the four independent replicates (r 5 0.49-0.60; average 0.54). Based on these results, we used the whole dataset for the following studies.
Although our cytological observations did not suggest that either of the mutants showed enhanced susceptibility at 18 hpi ( Fig. 1), we tested for genotypespecific differences at the molecular level by calculating the r-values obtained for all combinations of paired samples. For each gene, the mean values from four replicates of normalized data were considered and Figure 1. G. cichoracearum development on susceptible wild-type (wt), npr1-1, and jar1-1 mutant Arabidopsis plants. A, Abundance of haustoria on the plant epidermal cells at 18 and 24 hpi. Percent values are relative to germinated conidia and represent the average 6 SD of 15 infected leaves per plant. Similar results were obtained in three independent experiments. B, Bright field images illustrating fungal development at 18 hpi on representative plants of each genotype. Arrows indicate secondary germ tubes, and arrowheads indicate haustoria formed from primary germ tubes. Scale bars, 10 mm. C, Abundance of mature conidiophores per field at 96 hpi assessed by microscopic observation (2003). Values represent the average 6 SD of 10 infected leaves per plant. Similar results were obtained in three independent experiments. The asterisk (*) indicates statistically significant differences at P , 0.001. either all data contained in the slides (10,268 ESTs; Table I), or only pathogen-regulated genes (936 ESTs; Table II) were included in the test. Significant correlations (P , 0.0001) were observed for all comparisons between T0 samples of different genotypes (Tables I  and II) indicating that the npr1-1 and jar1-1 mutations did not introduce major qualitative differences in host basal gene expression. In contrast, fungal infection triggered major changes in the gene expression profiles of plants of all three genotypes. The expression levels of most genes were not altered in a coordinated way, as indicated by low or null correlations for the T0 to T18 comparisons (Table I). However, the T0 and T18 samples were fairly strongly negatively correlated when the analysis was performed only with genes displaying significant up-or down-regulation (first three columns/bottom three rows, Table II). Interestingly, significant correlations were found for all pairs of T18 samples (last three columns/bottom three rows, Tables I and II) indicating that lack of NPR1 or JAR1 did not significantly affect the large majority of the gene expression changes observed at 18 hpi.
Arabidopsis Genes Responding to G. cichoracearum Independently of the NPR1-or JAR1-Mediated Signaling Pathways Plant genes showing infection-mediated changes independent of the NPR1-and JAR1-transduction pathways could conceivably modulate basal defenses or features of host compatibility (e.g. cell death inhibition, defense inhibition, fungal nutrition). In an effort to identify such genes, we sought genes showing expression changes in the same direction (induction or repression) in infected plants of all three genotypes. We applied a multivariate ordination approach (PCA), to a matrix of standardized mean differences between T0 and T18 for each plant type. The results from this analysis indicated that the first principal component (PC1), which expressed the effect of infection, was the dominant source of variability (76%). We then selected the 1% of genes showing the highest differences between T18 and T0 samples, yielding 144 genes (67 induced and 77 repressed; Supplemental Table S3). To increase the confidence in detection of gene expression changes, we applied two other statistical analyses, significant analysis of microarrays (SAM) and ANOVA, and selected genes that consistently showed expression changes in all three analyses. SAM identified 217 differentially expressed genes (116 induced and 101 repressed; Supplemental Table S4) whereas ANOVA identified 223 such genes (117 induced and 106 repressed; Supplemental Table S5).
The results obtained from the three statistical methods are shown in Figure 2. Similar fractions of differentially expressed genes were selected by each method, 281 genes were selected by at least one method (145 induced, 136 repressed), and 70 genes were selected by all three methods (36 induced, 34 repressed). Northern-blot assays were performed to Table I. Pearson correlation coefficients calculated over all the intensity data from each slide The Pearson correlation coefficients were calculated after obtaining the fold-change average ratios for T0/CR (time 0 hpi versus common reference) and T18/CR (time 18 hpi versus common reference) for each of the indicated dataset pairs. The numbers in the body of the table correspond to the r-values. The first three columns/bottom three rows indicate genes showing lack of correlation for expression levels at T0 versus T18 among the three plant genotypes. The first three columns/top three rows, and last three columns/bottom three rows indicate significant positive correlations of gene expression levels among the three genotypes in uninfected (T0, top left corner) or infected (T18, bottom right corner) samples (P , 0.0001).  Table II. Pearson correlation coefficients among the infection-responsive genes The uncentered Pearson correlation was used to calculate correlation coefficients among the 936 genes showing differential expression between 0 hpi (T0) and 18 hpi (T18) in all three plant genotypes. The first three columns/bottom three rows indicate genes having a significant negative correlation between gene expression levels at T0 versus T18 for the three genotypes, showing consistent changes on gene expression (either induction or repression). The first three columns/top three rows and last three columns/bottom three rows indicate high significant positive correlations of gene expression levels among the three genotypes in uninfected (T0, top left corner) or infected (T18, bottom right corner; P , 0.0001).  Table III. The largest group of genes included those affecting basal metabolism (12 induced, 12 repressed), with other notable groups including genes affecting transcription and signal transduction (seven induced, 11 repressed), energy generation (five induced genes), cellular transport (three induced, one repressed), and defense (two induced, three repressed).
Host Genes Sensitive to NPR1-and JAR1-Mediated Signaling We looked for genes sensitive to infection through the NPR1-or JAR1-dependent pathways, by seeking those genes showing differential responses in wildtype plants versus npr1-1 or jar1-1 mutants. We applied PCA-based analysis to discriminate and further subtract the influence of PC1 (infection). We found that the ordination space created by the PC2 and PC3 together represented 24% of the total variability (see ''Materials and Methods''). The genes ordered in this space are presented in Figure 3A, which shows the approximately 1% of ESTs furthest from the origin. Among them, PC2 distinguishes jar1-1 with a positive coefficient from wild-type and npr1-1 with negative contributions. Genes induced during infection in wildtype and npr1-1 plants but not in jar1-1 plants, appear on the negative side of the PC2 axis. PC3 distinguishes npr1-1 from wild type and jar1-1. Genes induced in wild-type and jar1-1 plants but not in npr1-1 plants are located on the negative side of the axis. This analysis allowed the selection of 74 genes. To determine whether these genes shared common expression patterns, we performed cluster analysis. Four gene clusters were obtained based on their common expression profiles (Groups 1-4, Fig. 3B). The identity and functional classification of genes from these four groups are shown in Table IV. Group 1 includes 22 genes showing infection-induced activation in wild-type plants, slight repression in jar1-1 plants, and no change in npr1-1 plants (Fig. 3B). Group 2 includes 17 genes repressed by infection in wild-type plants and slightly induced in infected jar1-1 and npr1-1 plants. Group 3 contains 17 genes displaying differences in infected jar1-1 versus wild-type or npr1-1 plants and Group 4 contains 18 genes showing differences in infected npr1-1 versus wild-type or jar1-1 plants. Thus, Groups 3 and 4 seemed to include genes responding to JAR1-and NPR1-dependent signaling at the time of haustorium formation, as supported by the presence of PDF1.1, which is known to be sensitive to JA (Glazebrook et al., 2003 and references therein) in Group 3, and that of ICS1 (isochorismate synthase), which is sensitive to SA (Wildermuth et al., 2001) in Group 4.
Northern-blot analysis was then used to confirm the behavior of 10 genes from Groups 3 and 4 (five from each group) in naïve and fungus-infected tissues from wild-type and mutant plants. We chose genes having identical basal expression levels in naïve plants of all three genotypes and differences in the transcript abundance in one or the other infected mutants ( Fig. 4B; Supplemental Table S2), because they may respond to signals generated upon infection. Consistent with our microarray data, these genes failed to show infectionmediated induction or repression in one of the mutants. The responses of four from these 10 genes are shown in Figure 4B.
These studies allowed us to identify genes apparently sensitive to the JA or SA pathways that require JAR or NPR1 for infection-mediated activation such as At4g17090 (glycosyl hydrolase like; GH14) and At3g16640 (tumor-related gene; TURP), respectively (Fig. 4B).

Genes Modulated by Balances between the JA and SA Pathways
Next, we analyzed possible causes for the differential expression of genes in Groups 3 and 4 in a given mutant. There seemed to be two possibilities. First, virulent G. cichoracearum could stimulate gene expression changes through the NPR1-and JAR1-dependent pathways. In this case, the absence of gene activation in one mutant could be due to lack of activation functions sensitive to this pathway. Alternatively, because both pathways display mutual antagonistic interactions (Schenk et al., 2000;Kloek et al., 2001;Glazebrook et al., 2003;Spoel et al., 2003), the absence of gene activation could be due to active signaling by the opposite pathway. In this case, repression or lack of induction of genes from Group 3 in infected jar1-1 plants could be due to exacerbation of SA-sensitive responses, whereas down-regulation of genes from  Table III. Arabidopsis genes differentially regulated during the early interaction with G. cichoracearum LS, Light-sensitive genes according to Smith et al. (2004). Bold type indicates genes that are induced or repressed by the large majority of other biotic stresses (e.g. bacterial pathogens, nonhost bacteria, and bacterial and fungal elicitors including LPS and flg22, fungal necrotrophs and biotrophs, according to data obtained from www.bbc.botany.u-toronto.ca). Underlined type indicates genes that appear to be specifically induced or repressed by G. cichoracearum or G. orontii. Group 4 in infected npr1-1 plants could be caused by enhanced flux through the JA pathway.

At2g24360
Ser/Thr/Tyr kinase, putative Signal transduction At2g34420 Chlorophyll a/b-binding protein/LHCII type I (LHB1B2) Energy At4g35100 Plasma membrane intrinsic protein (aquaporin) Cellular transport At5g02500 Heat shock cognate 70-kD protein 1 (HSC70-1; HSP70-1) Cellular rescue, defense, and virulence At2g40940 Ethylene receptor, subfamily 1. ERS1 Cellular rescue, defense, and virulence Ribulose bisphosphate carboxylase small chain 1A Energy At1g21550 Calcium-binding protein, putative Signal transduction At5g11030 Aberrant lateral root formation 4, ALF4 Morphogenesis At3g01540 DEAD box RNA helicase (DRH1) Transcription At5g01520 Zinc finger (C3HC4-type RING finger) family protein, similar to MTD2 Unknown At3g02420 Expressed protein Unknown At3g19390 Cys proteinase, putative Protein degradation At5g44120 12S seed storage protein (CRA1) Metabolism At5g63840 a-Glucosidase, putative Metabolism At1g30510 Ferredoxin-NADP(1) reductase, putative Cellular transport At4g21720 Expressed protein Unknown At1g07350 Transformer Ser/Arg-rich ribonucleoprotein, putative Transcription At4g39540 Shikimate kinase family protein Metabolism At1g35320 Expressed protein Unknown At3g03380 DegP protease, putative Protein degradation At4g21860 Met sulfoxide reductase domain-containing protein Unknown [CBS-domain-containing protein]) was insensitive to SA (Fig. 5A) To further characterize the expression patterns of the subset of genes from Groups 3 and 4 that are sensitive to both SA and JA, we analyzed the behavior of TURP (Group 3) and GH14 (Group 4) in response to other treatments. Because infection-mediated induction of TURP requires functional JAR1 to be induced by fungal treatment (Fig. 4B), we tested whether this gene could be induced by treatment of wild-type tissues with exogenous JA. As shown in Figure 5B, JA had no effect on the basal expression levels of TURP. We thus hypothesized a possible mechanism for TURP regulation (Fig. 5A) wherein TURP induction in fungus-infected wild-type tissues is signaled through JAR1 via release of SA-mediated gene repression. To evaluate this possibility, we monitored the expression level of TURP in wild-type plants exposed to an avirulent race of Pseudomonas syringae pv tomato, which strongly stimulates the SA pathway. As expected, TURP was repressed in this condition (Fig.  5C). In addition, TURP was also repressed by bacterial treatment of jar1-1 plants but not npr1-1 plants. These results may indicate that either SA-mediated TURP repression requires functional NPR1, or that the enhanced JA levels found in npr1-1 plants (Spoel et al., 2003) released the proposed SA-mediated repression.
The Group 4 gene GH14 displayed behavior equivalent to that described for TURP, but involving SA and JA in opposite roles. Our results revealed that GH14 was not induced by exogenous SA (Fig. 5B) and electronic-northern (e-northern) data reported by others (Botany Department, University of Toronto; http://bbc.botany.utoronto.ca/) indicated that this gene was repressed under conditions that strongly stimulate the JA pathway (infection by Botrytis cinerea and Phythophthora infestans). Interestingly, we found that treatment with avirulent bacteria did not modify GH14 expression in wild-type plants, whereas this treatment induced the gene in jar1-1 plants and slightly repressed its expression in npr1-1 plants. These results are consistent with a regulatory mechanism involving negative control of GH14 by SA, which may in turn release JA-mediated gene repression (Fig. 5A).

Conserved cis-Elements among Genes from the Same Category
Gene expression profile studies allow genes to be clustered based on similar expression patterns in response to the same treatment. Because it is likely that some genes within a cluster will contain analogous regulatory information, we looked for common cis- Tonoplast intrinsic protein, g-TIP, or MIP water channel Cellular transport At3g06050 Mitochondrial matrix localized peroxiredoxin. Cellular rescue, defense At4g15080 Zinc finger (DHHC type) family Unknown At3g43670 Copper amine oxidase, putative Metabolism At2g22660 Similar to Gly-rich protein Unknown At3g27770 Expressed protein Unknown At1g21550 Calcium-binding protein, putative Signal transduction At1g06720 Expressed protein Transcription At1g51830 Leu-rich repeat protein kinase Signal transduction At1g49750 Leu-rich repeat family protein Signal transduction At5g14920 Similar to GA-regulated protein 1 precursor GASA3 Metabolism, defense At1g10660 Expressed protein Unknown At1g65980 Peroxiredoxin type 2, putative Defense At3g09770 Zinc finger (C3HC4-type RING finger) family protein Metabolism, defense At5g47560 Tonoplast malate/fumarate transporter Cellular rescue and transport regulatory elements common to the identified gene clusters. We were unable to find sequences recognized by transcription factors associated with previously reported responses to pathogens, such as binding motifs for ERF/AP2, R2R3MYB, TGAbZIP, Whirly, WRKY, ORCA, EIN3/EIL, MYB1, and ABI3 (Plant-CARE database). However, we did identify a few ciselements previously described as being involved in diverse defense responses (Table V). Among them, the GAAGAAGAA motif was present in the promoters of all the genes showing infection-mediated expression, regardless of the plant genotype. This motif is recognized by the TL1 element, which provides NPR1dependent regulation for genes encoding proteins involved in the secretory pathway during defense responses (Wang et al., 2005). Another motif, the GTCCA sequence, was abundant in the promoters of genes showing infection-mediated repression, regardless of plant genotype. This motif was previously described as being overrepresented in the promoters of genes repressed by chitin oligomers . In the promoters of genes from Groups 3 and 4, the [A/T]GTGACG motif was highly represented. This motif was previously identified in JA-responsive gene promoters from barley (http://intra.psb.ugent. be:8080/PlantCARE). Interestingly, this sequence forms the core of the as-1 element found in the promoter of genes responding in a SA-dependent, NPR1-independent fashion (Uquillas et al., 2004). We additionally found several apparently novel motifs that have not yet been described as targets of known transcription factors (according to searches in the PlantCARE, TRANSFAC, and PLACE databases). Among them, the TATGTG and the TTTTT-CTTCTTC motifs were conserved in the promoters of genes that were induced or repressed by fungal infection, respectively, regardless of genotype. The promoters of members of Group 3 contained the CCCACC motif that is identical to the consensus sequence for the animal transcription factor KLF6 (Kruppel-like factor 6), which is a tumor suppressor gene. Another putative regulatory sequence, GAAGTGATAG, was also abundant in the promoters of Group 3 genes, whereas the majority of Group 4 genes contained the motifs, ACAAGAAAA-AAA and GTTATA, in their promoter sequences.

DISCUSSION
The Arabidopsis Transcriptome at the Time of

G. cichoracearum Haustorium Formation
We herein explored the molecular features of plant susceptibility to infections caused by obligate biotrophic fungi by examining changes in the Arabidopsis transcriptome during a compatible interaction with G. cichoracearum. Our studies were performed at the stage of haustorium formation because this is a critical stage of reproduction in these parasites (Mendgen and Hahn, 2002;Schulze-Lefert and Panstruga, 2003). To further characterize the pathogen-induced gene expression changes, we categorized them with respect to their sensitivity to the SA-and JA-dependent pathways.
In our studies, RNA was extracted from whole infected leaf tissues. Considering that only epidermal cells come into contact with the fungus, the responses of infected cells may be diluted by those of uninfected cells, leading to modest changes in overall gene expression levels. To overcome this limitation and increase the confidence for detection of true expression changes, we evaluated the fold-change data by combining supervised (SAM and ANOVA) and unsupervised (PCA) statistical analyses and selected only genes showing changes in all three methods. Thus, we expected to rescue a subset of genes having the most robust expression changes following infection. Total RNA (10 mg/line) isolated from healthy or infected tissues from wild-type (wt), npr1-1 (n), or jar1-1 (j) plants, sampled at the indicated hpi, was analyzed using cDNA probes for the indicated genes. A, Induction (top part) or repression (bottom part) of genes corresponding to the Venn diagram shown in Figure 2 corroborated the results obtained from our microarray assays. B, Expression pattern analysis of genes from Groups 3 and 4, those that could not be induced by infection in jar1-1 or npr1-1 plants (top) and those that were repressed under the same conditions (bottom).
To confirm our findings, we randomly chose 10 genes from this subset and subjected them to northern-blot analysis, corroborating our microarray results (Fig. 4).

Changes Independent from NPR1 and JAR1
Seventy NPR1-and JAR1-insensitive infectionresponsive genes were identified based on their identical behaviors in infected wild-type, jar1-1, and npr1-1 plants. Using a free e-northern facility (Botany Department, University of Toronto; http://bbc.botany.utoronto. ca/), we evaluated how these genes responded to other biotic stresses including virulent and avirulent bacteria, necrotrophic and biotrophic fungi, nonhost pathogens, and bacterial and fungal elicitors. We found that nearly half of the genes altered during the interaction with G. cichoracearum were also affected in a similar direction and to a similar extent by at least one other pathogen or treatment. A subset of eight induced and nine repressed genes responded similarly to the majority of the stresses. Among them, the induced genes included those encoding UDP-glycosyl transferase (At2g30140); microtubuleassociated protein (At4g29950), cytochrome P450 (At2g34500), Glc-6-P dehydrogenase (At5g40760), syntaxin SYP122 (At3g52400), and three unknown proteins (At4g21850, At4g01880, and At1g11200). The subset of repressed genes included those encoding starch branching enzyme (At2g36390), germin-like protein AtGER3 (At5g20630), thiamine biosynthesis protein (At2g29630), pyruvate dikinase (At5g26570), peptidylprolyl isomerase (At1g18170), ribosomal protein L13 (At1g78630), membrane channel related protein (At2g28900), low- Figure 5. Northern-blot experiments showing the sensitivity of genes from Groups 3 and 4 to exogenous SA or JA. A and B, RNA was extracted from wild-type plants treated with two different concentrations of SA or JA at 24 h posttreatment (hpt). C, RNA was isolated from wild-type, npr1-1, or jar1-1 plants infiltrated with Pseudomonas syringe pv tomato DC3000 harboring the avrRpm1 gene (5 3 10 6 cfu/mL; avr) or 10 mM MgCl 2 (mock). The following genes were analyzed by northern blotting: CBS, CBS-domain-containing protein (At5g10860); TURP, tumorrelated protein (At3g16640); PGM, phosphoglycerate-mutase-like protein (At3g50520); DI, disulfide isomerase-like protein (At5g18120); ICS1, isochorismate synthase 1 (At1g74710); GH14, glycosyl hydrolase family 14 (At4g17090); F-Box, FBL6/EBF1 F-box-domain-containing protein (At2g25490); AHR, alkyl hydroperoxide-reductase-like protein (At3g06050); MIP, major intrinsic protein aquaporin (At2g36830). Hybridization with rRNA probes was used as a loading control. Probes for the genes PR1 and PDF1.2 were used as controls for the effects of SA and JA. temperature and salt-responsive protein LTI6A (At4g30660), and an unknown protein (At5g05740). Interestingly, only seven of the examined genes were specifically regulated by the biotrophic fungi G. orontii (http://bbc.botany.utoronto.ca) and G. cichoracearum (this work). Among them, five were induced (chlorophyll a oxygenase [At1g44446], putative peroxisomal transport protein [At1g04530], aquaporin PIP2A [At3g53420], glutathione peroxidase 1 [At2g25080], and a signal transducer of phototropic response [At2g30520]), whereas two were repressed (nodulin from MtN3 family [At5g23660] and ABI3 interacting protein 1 [At5g63780]). Additional studies will be required to assess if these genes, which appear to be specifically modified by G. orontii and G. cichoracearum, are involved in the regulation of plant defenses and/or susceptibility conditions. Plant susceptibility factors are thought to be involved in controlling the early stages of interactions with obligate parasites (Schulze-Lefert and Vogel, 2000;Mendgen and Hahn, 2002;Panstruga, 2003). In Arabidopsis, a few genes capable of enhancing G. cichoracearum pathogenesis have been described. Among them, some of the PMR genes (PMR2, PMR5, and PMR6) function independently of the SA-and JA-dependent defense pathways. As expected, these PMR genes were not detected in our experiments because they have constitutive expression. Similarly, the EDR1 gene encoding a MAPKKK was not selected by our assays because it becomes induced late in the infection process (Frye et al., 2001). We believe that some of the genes identified in this work are likely to encode susceptibility factors and that some of them could be sensitive to NPR1 and/or JAR1. In other pathosystems, pathogenmediated regulation of genes encoding susceptibility factors such as NHO1 (Kang et al., 2003) and Bax-inhibitor 1 (Sanchez et al., 2000) have been shown to depend on the JA-signaling cascade.
To examine overall alterations in the host that were specifically stimulated by G. cichoracearum, we disregarded the 17 genes affected by the majority of other biotic stresses. Among the 53 remaining genes, the highest fold-change levels involved genes regulating host energy generation. The changes in these genes suggest a net accumulation of chlorophyll binding proteins (At2g05070, At3g08940, and At2g34430) and chlorophyll a oxygenase (At1g44446) in infected tissues. Because epidermal cells colonized by the fungus lack chloroplasts, these genes must be induced in nearby uninfected cells, probably neighboring mesophyllic tissues. Previous studies have shown that expression of photosynthetic genes is stimulated by G. orontii infection (http://bbc.botany.utoronto.ca/) but is severely repressed by nonhost mildews at 18 hpi (Zimmerli et al., 2004). A local increase of photosynthetic functions is in agreement with the phenomenon of ''green islands'' produced in compatible interactions between other Erysiphales and their hosts. In this case, the infected tissues increase their photosynthetic rate subsequently altering the source-sink balance (Truernit et al., 1996).
A large set of genes (23) modulating host metabolism showed important alterations upon infection and some of them are compatible with conditions favoring fungal nutrition. Concerning sugar metabolism, we found strong activation of genes encoding for trehalose-6-P synthase (At1g70290 and At1g23880), which importantly affects carbon assimilation in plants (Eastmond and Graham, 2003). Other general pathogen-induced transcriptional changes, such as up-regulation of the gene encoding Glc-6-P dehydrogenase (At5g40760) and repression of the gene encoding the starch branching enzyme (At2g36390), may help increase the availability of simple sugars at infection sites. Interestingly, at 18 hpi we did not detect changes associated with carbon assimilation, such as up-regulation of sugar transporters found to occur in this interaction several days after infection (Fotopoulos et al., 2003). Conversely, by SAM and ANOVA tests we did detect at this time of infection modest induction of At3g11900, encoding an amino acid transporter (data not shown). These results are in agreement with the notion that transportation of nutrients from plant cells to fungal cells through the haustorium is active at 18 hpi.
The naïve and infected samples analyzed in this study were excised at two different day times, corresponding to 3 (T18) and 9 (T0) hours after light onset (12 and 6 PM, respectively). We evaluated if the 70 genes selected as sensitive to G. cichoracearum displayed diurnal changes according to data reported by Smith et al. (2004). Genes At4g33120, At2g05070, At2g34430, At5g19190, and At1g12760 could not be analyzed in this way because they were absent in these studies. We observed that 25 of these genes were insensitive to light whereas the reminding 40 displayed expression differences at 2 to 4 and 8 h after light onset (samples equivalent to T18 and T0, Table V. Conservation of putative regulatory elements in the promoters of coregulated Arabidopsis genes altered during G. cichoracearum infection Results of promoter sequence analysis for genes that were induced or repressed in all three infected genotypes (36 and 34 genes, respectively), as well as genes belonging to Groups 3 (17 genes) and 4 (18 genes), showing differences in SA-or JA-dependent responses, respectively. respectively). The 40 light-sensitive genes were indicated in Table III. Most of these genes changed in the same trend (induction, repression) in the Smith et al. (2004) experiment and in ours. However, genes such as At5g45820, At4g26670, At3g07650, At2g40080, and At1g56300, displayed quantitative or kinetic differences between both experiments (not shown), suggesting that pathogens may alter the light-mediated regulation of these genes. It is important to note that because the Smith et al. (2004) study used plants grown under different light/dark cycles from our study, as well as samples isolated at slightly different daytimes, the exclusive effect of light on the selected genes, under the infection condition here evaluated, cannot be deduced from direct comparison of both sets of data. Additional experiments will be required to determine how light and pathogens independently affect these genes.
The high abundance of light-sensitive genes among those selected in Table III was not unexpected. Light influences host cellular processes, such as the photosynthetic activity, antioxidant defenses, carbohydrate catabolism among others, and it is likely that these processes may be affected by fungal infection. Genes sensitive to both, light and pathogens, have been characterized in several species including those encoding for ascorbate peroxidase from rice (Oryza sativa; Agrawal et al., 2003), lipoxygenase from maize (Zea mays; Nemchenko et al., 2006), and DEA1 from tomato (Solanum lycopersicum; Weyman et al., 2006).
Haustorial development constitutes a key stage for fungal propagation involving the intimate contact between the invader and the host. In the interaction here analyzed, about 90% of the infection attempts are successful. Thus, it is expected that gene expression changes occurring at this time reflect the establishment of susceptibility, as well as the activation of PAMPtriggered immunity (PTI), which proved to be insufficient to limit fungal propagation. Suppression of PTI by fungal effectors may also take place under this condition, as reported for compatible interactions with powdery mildews where PTI suppression may facilitate the formation of haustoria (Caldo et al., 2006).
As mentioned before, SAM, ANOVA, and PCA tests detected 70 genes responding to G. cichoracearum. As expected, this group included genes sensitive to several other biotic stresses (17 genes indicated in Table III) that may participate in PTI modulation. Activation of a gene encoding for phospholipid hydroperoxide glutathione peroxidase (At2g25080), which may display chloroplastic antioxidant activities, was also detected. In addition, all three tests pointed out a down-regulation of genes encoding germin-like protein AtGER3 (At5g20630), lowtemperature and salt-responsive protein LTI6A (At4g30660), and heat shock protein Hsp20a (At1g06460). This finding supports the notion that fungal obligate biotrophs may avoid host defenses (Mendgen and Hahn, 2002;O'Connell and Panstruga, 2006). It is important to note that selection of expression changes by combination of SAM, ANOVA, and PCA tests may be a reliable way of identifying robust alterations, although it may not be useful to detect mild expression changes occurring under this condition. The activation of PAL1 (At2g37040), CAD4 (At3g19450), and NHO1 (At1g80460), genes perceived by SAM test (Supplemental Table S4) but not by all three methods, may illustrate the latter kind of responses.
Collectively, robust expression changes of genes insensitive to NPR1 and JAR1 during the interaction with G. cichoracearum at 18 hpi suggested that compatibility conditions are already established for fungal propagation at this time. We found changes compatible with activation of defenses including PTI as well as suppression of host defenses. We also found changes that seemed to indicate that the pathogen plays an active role in forcing the host's metabolism to enhance photoassimilation and amino acid levels at the infection sites. Furthermore, our results suggest that photosynthesis may be increased in uninfected cells located near the penetration sites, thus providing substrates for fungal feeding through the haustoria.
Genes Sensitive to SA and JA Cytological observations indicated that under our experimental conditions, the initial steps of G. cichoracearum proliferation proceeded similarly in jar1-1, npr1-1, and wild-type plants. Similar abundance of haustoria and infection structures developed at 18 to 24 hpi in the mutant and wild-type plants. We did observe a slight reduction in the number of conidiophores in npr1-1 plants versus jar1-1 plants at 96 hpi, although these differences were not sustained by 6 to 8 dpi. This finding may suggest that the increased JA levels found in naïve npr1-1 plants (Spoel et al., 2003) may help delay the initial stages of fungal propagation.
SA, JA, and ET accumulate in response to different pathogens, leading to specific changes in gene expression. Under several infection conditions, the SA and JA/ET pathways display mutual repression mechanisms (Schenk et al., 2000;Kloek et al., 2001;Glazebrook et al., 2003;Spoel et al., 2003;Bostock, 2005). These pathways regulate host responses in a variety of ways depending on the timing and magnitude of JA/ET and SA accumulation, as well as cross-talk between these pathways and with other yet-unknown regulatory mechanisms (De Vos et al., 2005). For interactions between Arabidopsis and G. cichoracearum, SA-dependent responses are known to be induced at 4 dpi (Frye and Innes, 1998; for review, see Glazebrook, 2005), but these responses have not previously been analyzed during the early stages of infection. Other studies have shown that early activation of PDF1.1, PDF1.2, and PDF1.3 in this system takes place at 18 hpi and decreases by 24 hpi (Zimmerli et al., 2004).
We sought to characterize JAR1-and NPR1-dependent responses at the time of haustoria formation. Analysis of the bi-and tridimensional spaces generated by the PCs in a PCA identified four groups of genes with common behaviors, including a group of JAR1sensitive genes that were repressed or not induced in infected jar1-1 plants (Group 3) and a group of NPR1-sensitive genes that were repressed or not induced in infected npr1-1 plants (Group 4). The identification of these two groups indicated that there were some changes in NPR1 and JAR1 signaling in the host transcriptome at 18 hpi. Consistent with this observation, measurements of JA and SA levels at 18, 24, and 48 hpi showed that levels for both hormones were elevated in infected leaves. In this period, these hormones undergo up to 5-fold increase above their basal levels (Supplemental Fig. S1). We used e-northern analysis (http://bbc.botany.utoronto.ca/) to evaluate how genes from Groups 3 (17 genes) and 4 (18 genes) responded to other biotic stresses. Eight genes from Group 3 were repressed by avirulent bacteria suggesting that SA plays a role in their repression. In contrast, four other genes from this group displayed the opposite response (induction by avirulent bacteria). We further found that 10 genes from Group 4 were repressed by necrotrophic pathogens known to induce the JA pathway, whereas only one of them was induced by these pathogens. These results support the belief that Groups 3 and 4 include genes sensitive to SA and JA, respectively. Among the genes categorized in these two groups, only PDF1.1 (Group 3) and ICS1 (Group 4) are classical markers of the JA and SA pathways, respectively, whereas the remainder constitute additional targets of these signaling cascades. In the future, these genes may contribute to the characterization of cross-talk points between the JA and SA signaling cascades in other plant-pathogen interactions. In addition, genes from Group 3 may be used to evaluate the extent and timing of JA-pathway induction in this interaction, along with the basis for its eventual downregulation. These findings, along with a previous study showing that constitutive activation of the JA/ET pathway occurring in the cev1 mutant leads to enhanced resistance to G. cichoracearum (Ellis et al., 2002), seem to suggest that this pathway signals effective defenses against the fungi, but becomes repressed by fungal virulence products in infected wild-type plants.
The enhancement of SA and JA levels detected at 18 hpi in this compatible interaction (Supplemental Fig. S1) is consistent with the interplay of the SAand JA-dependent defenses at this stage of infection. Some of the genes from Groups 3 and 4 were found to be sensitive to both SA and JA, suggesting that their expression levels might be determined by the balance between these pathways. To examine this hypothesis, we characterized the transcriptional features of one gene from each group, TURP (Group 3) and GH14 (Group 4). Our results suggested that the expression of TURP may be controlled by two negative regulatory steps involving both SA and JA, as shown in Figure  5A. Interestingly, in wild-type tissues, TURP is induced upon fungal infection but repressed upon treatment with avirulent bacteria. The reason for this differential response is not yet known but it could be related to differences in the balance between the JA and SA pathways and the timing of accumulation of these components under each condition. In this sense, our findings suggest that the JA-stimulated responses in plants treated with avirulent bacteria may not be as robust as the SA-mediated responses. This is in agreement with a previous report that activation of VSP2 and PDF1.2 was not observed in these tissues, even though they contained increased JA levels (De Vos et al., 2005). Alternatively, other factors specifically controlling defenses against virulent fungi or avirulent bacteria could account for the observed differences. Similarly, the GH14 gene from Group 4 appears to be controlled by two negative regulatory steps involving JA and SA, as shown in Figure 5A. Treatment with avirulent bacteria did not alter the expression of GH14 in wild-type plants, whereas it induced the gene in jar1-1 plants and slightly repressed its expression in npr1-1 plants. Accumulation of JA is expected to occur beginning at 3 hpi in Pseudomonas-treated plants (De Vos et al., 2005). Under this condition GH14 may be down-regulated by JA because infected jar1-1 mutants show activation of this gene (Fig. 5C). The repression of GH14 found in infected npr1-1 plants could be caused by enhanced JA levels in the mutant (Spoel et al., 2003) or by the requirement of NPR1 for execution of the SA-mediated gene repression.
Another member of Group 4, the gene encoding ICS1, has been extensively characterized in the context of pathogenesis (Dewdney et al., 2000;Wildermuth et al., 2001). We here found that basal ICS1 expression is completely abolished upon infection. Because only epidermal cells are infected by the fungus, this result may suggest that ICS1 expression could be regulated at the systemic level. Our data showed that ICS1 is insensitive to exogenous JA and is down-regulated in infected npr1-1 plants. The latter observation is inconsistent with previous reports showing that ICS1 is negatively regulated by NPR1 in G. cichoracearuminfected plants at 7 dpi (Wildermuth et al., 2001). This disparity may be due to the difference in the analyzed time points or it may indicate that an NPR1-sensitive repressor of ICS1 functions at the late stages of infection.
We herein report the first use of genome-wide expression analysis to characterize Arabidopsis genes specifically regulated during G. cichoracearum haustorium formation. We found uncharacterized putative regulatory elements in the promoters of coregulated Arabidopsis genes altered during G. cichoracearum infection, functional characterization of which require further analysis. Our findings may contribute to the discovery of novel plant genes altered by the pathogen to establish successful infection, although future studies will be required to examine the participation of these gene products in pathogenic virulence strategies, including plant defense repression. Our results also suggest new cross-talk points between the SA-and JA/ ET-dependent networks in plant-pathogen interactions. Collectively, this work is an important first step toward understanding the molecular mechanisms involved in the establishment of compatible interactions with obligate biotrophs.

Plant Infections
Four-week-old wild-type, npr1-1, and jar1-1 plants were inoculated with a G. cichoracearum UCSC1 conidia, using a settling tower as previously described (Adam et al., 1999), with high-sporulating infected squash leaves being tapped over the surface of 4-week-old Arabidopsis plants. The inoculated plants were kept in the dark under 100% humidity for 1 h and then placed into growth chambers. Fungal development was monitored by trypan blue staining and optical microscopy, as previously described (Adam et al., 1999). Leaves of healthy and infected plants isolated at 18 hpi (when haustorium development was at 90% of maximum) were collected and shockfrozen in liquid nitrogen. Four independent replicates of each experiment were collected, each on a different date.

Treatments of Plants with SA and JA
Plants were treated with SA or JA according to protocols described by Schenk et al. (2000). Briefly, 5-week-old Col-0 plants were either spayed with 1 or 5 mM SA (Sigma Aldrich) solutions or treated with methyl jasmonate (MJ; 95%; Sigma Aldrich), as follows. Aliquots of 400 mL of 0.5% or 1.0% v/v MJ ethanolic solutions were applied on cotton balls and allowed to evaporate into a 20-L sealed container holding the plants, giving maximum concentrations of 0.42 or 0.84 mmol MJ/L, respectively.

SA and JA Determination
The endogenous SA and JA levels present in naive and infected Arabidopsis leaves were determined from near 500 mg of plant samples.
Extraction and quantification processes were performed using jasmonic acid and deuterated salicylic acid as internal standards and HPLC electrospray tandem mass spectrometry technology, according to standardized procedures (Ross et al., 2004).

RNA Isolation
For microarray hybridization, total RNA was isolated from frozen leaf tissues using the TRIzol Reagent (Invitrogen) according to the protocol recommended by the Arabidopsis Functional Genomics Consortium (http://www. arabidopsis.org/info/2010_projects/comp_proj/AFGC/RevisedAFGC/site2RnaL. htm#isolation). The RNA samples were treated with RQ1 DNAse (Promega) and poly-A RNA was isolated from 300 to 400 mg of total RNA using the Oligotex Mini Prep kit (QIAGEN). For northern-blot hybridization, total RNA was isolated from rosette leaves (Alvarez et al., 1998) and 10 mg of RNA was blotted and hybridized with the appropriate EST probes.

Microarray Hybridization
Labeled cDNAs were synthesized from each sample and hybridized onto 24 microarrays as described in Supplemental Table S1, where the access number of each slide is provided. Four biological replicates were performed and for each experiment two sets of cDNAs were hybridized onto 11.5 K arrays and another two onto 15 K arrays. For each of the four replicates, six slides (three genotypes under two conditions) were hybridized with a common reference consisting on pooled samples from wild-type, npr1-1, and jar1-1 cDNAs, as described in Supplemental Table S1. The common reference was labeled with Cy3, whereas Cy5 was used to label either the T0 or T18 samples on the same slides. The normalized values corresponding to Cy5/Cy3 spot intensity ratios from each slide are shown in Supplemental   Table S2. The raw intensity data and all array details can be accessed through http://genome.www5.stanford.edu. Microarray hybridization was performed using the CIW-MSU joint array protocol (http://www.arabidopsis.org/info/ 2010_projects/comp_proj/AFGC/RevisedAFGC/AFGC_Protocols_Dec_2001L. pdf) as described previously .

Acquisition, Transformation, and Processing of Microarray Data
The hybridization signal for each channel (Cy3 and Cy5) was read using a ScanArray Lite scanner (Packard Biochip Technologies) and spot signals were analyzed using the Genepix 3.0 software (Axon Instruments). Grids were predefined and manually adjusted to ensure optimal spot recognition, and spots with dust or locally high background were flagged as bad. ESTs spots that were not flagged and had a regression correlation greater than 0.6 were selected for further analysis (Gollub et al., 2003). The raw data were deposited into the Stanford University microarray database (http://genome-www5. stanford.edu/microarray/SMD; Sherlock et al., 2001) under the identifiers given in Supplemental Table S1.
The resulting values were consolidated in spreadsheet files (Excel; Microsoft), and imported into the GeneSpring 6.0 microarray analysis software (SiliconGenetics). Further normalizations of the values were performed per spot (data channel divided by control channel both previously corrected by background) and per chip. We applied within-sample normalization to correct for intensity dependencies using LOWESS adjustments of the relationship between expression level (M) and intensity (A). We also performed betweensample normalization by dividing the expression levels of each array by their 50th percentile). Log 2 transformations of the normalized fold-change ratios (Cy5 channel/Cy3 channel) were then analyzed as described below and in the Results section.
For SAM analysis, when a two-class unpaired SAM test with 1,000 permutations was applied to our data (11,250 spotted ESTs) for comparison of T0 to T18 samples, 486 ESTs were selected as significant (367 induced and 161 repressed) showing 1.7-fold differential expression, threshold of foldchange value is 0.71, false significant-median number is 2.11, and false discovery rate is 0.43%. To increase the stringency of selection, we selected only ESTs exhibiting a T0 to T18 fold-change ratio #0.5 or $2 on at least two of the three genotypes. This analysis yielded 217 ESTs, (116 induced and 101 repressed genes; Supplemental Table S4).
ANOVA analysis was used to evaluate whether the changes observed in average gene expression were due to the infection response, the plant genotype, or an interaction between these two factors. The original matrix of 11,250 EST spots was reduced to 10,286 when those ESTs having more than two missing data points in each treatment were discarded. We then estimated the mean expression value from the 2 to 4 data points available for each EST. Our ANOVA results corroborated our finding that the observed gene expression changes were primarily due to infection, with several genes yielding P-values ,0.0001. In contrast, the plant genotype and the interaction effects had smaller contributions to the observed changes (Supplemental Table S5). Selection based on P-values with a cutoff of 1/1,000 (one false positive gene per 1,000 analyzed) identified 225 differentially expressed genes (227 ESTs), 117 induced and 106 repressed by infection (Supplemental Table S5).
To generate an ordination space in which it was easer to visualize the relationship between genes and treatments and to select differentially expressed genes, PCA was applied to the matrix of standardized mean differences of the expression values at T0 and T18. The technical details of the implementation of PCA in the context of this application are presented in Supplemental Data Analysis S1. Then, PCA was used not to reduce the number of dimensions of the problem, but rather to look for new axes that might allow us to discriminate ESTs showing consistent and different responses to infection in wild-type, npr1-1, and jar1-1 plants. The eigenvalues and eigenvectors of the spectral decomposition of the matrix D, which were used to interpret the PCA analysis, are presented in Supplemental Table S6. Interpreting the meaning of principal components according to eigenvector coefficients (factor loading) and its relevance according to its eigenvalues revealed that the PC1 expressed the global effect of the infection, which is the dominant source of variability (76%). Furthermore, the extreme values of PC1 correspond to genes that were under-and overexpressed with respect to the uninfected condition in plants of all three genotypes. The genes repressed during infection were located on the positive side of the axes, whereas induced genes were located at the negative extremes. We selected the 1% of genes located furthest to the right (induced) and left (repressed) with respect to the center of the ordination space (indicating no change). We chose 1% to select a manageable number of genes and to reduce the possibility of false positives. This analysis yielded 144 ESTs: 67 induced and 77 repressed (Supplemental Table S3).
Cluster analysis was performed by applying the k-centroid algorithm on the matrix of standardized differences. The number of EST clusters was determined by plotting a measure of heterogeneity within clusters along with the number of clusters. We then used the slope change criteria on this graphical representation of the reduction of heterogeneity as a function of the number of clusters (data not shown) to select the appropriate number of groups of genes with different patterns of differences between treatments.

Conserved cis-Regulatory Elements in the Identified Gene Clusters
To identify conserved cis-elements, the sequences 1,000 bp upstream of the ATG of the promoter regions of the gene groups identified in this work were analyzed using the MEME (http://meme.sdsc.edu/meme/website/meme. html), Motif-finder (www.arabidopsis.org), CREDO (http://mips.gsf.de/ proj/regulomips/), and POBO (http://ekhidna.biocenter.helsinki.fi:9801/ pobo) programs. We looked for conserved motives not exclusively present in genes sensitive to light. We therefore analyzed all genes within a group to further select those motives only present in both, light-sensitive and lightinsensitive genes. The elements reported in Table V satisfied this criterion. Note that only 16 out of the 20 genes containing the ''GTCCAA'' motif are light sensitive, and that Groups 3 and 4 contain four and six light-sensitive genes, respectively (not shown).

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Accumulation of JA and SA in G. cichoracearuminfected tissues.
Supplemental Table S1. Experimental design of the microarray slide hybridization.
Supplemental Table S2. Normalized gene expression data for all 24 samples.
Supplemental Table S3. Genes responding to infection selected by PCA.
Supplemental Table S4. Genes responding to infection selected by SAM analysis.
Supplemental Table S5. Genes responding to infection selected by ANOVA analysis.
Supplemental Table S6. Eigenvalues of the spectral decomposition of the standardized differences (T0 versus T18) matrix between uninfected (0 hpi) and infected (18 hpi) samples on the expression of 10,286 genes.
Supplemental Data Analysis S1.