Genome-wide transcriptional analysis of the Arabidopsis thaliana interaction with the plant pathogen Pseudomonas syringae pv. tomato DC3000 and the human pathogen Escherichia coli O157:H7

Authors

  • Roger Thilmony,

    1. Department of Energy-Plant Research Laboratory,
    Search for more papers by this author
    • These authors contributed equally to this work.

    • Present address: USDA-ARS Western Regional Research Center, Albany, CA 94710, USA.

  • William Underwood,

    1. Department of Energy-Plant Research Laboratory,
    2. Genetics Graduate Program and
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Sheng Yang He

    1. Department of Energy-Plant Research Laboratory,
    2. Department of Plant Biology, Michigan State University, East Lansing, MI 48824, USA
    Search for more papers by this author

*(fax +1 517 353 9168; e-mail hes@msu.edu).

Summary

Pseudomonas syringae pv. tomato DC3000 (Pst) is a virulent pathogen that causes disease on tomato and Arabidopsis. The type III secretion system (TTSS) plays a key role in pathogenesis by translocating virulence effectors from the bacteria into the plant host cell, while the phytotoxin coronatine (COR) contributes to virulence and disease symptom development. Recent studies suggest that both the TTSS and COR are involved in the suppression of host basal defenses. However, little is known about the interplay between the host gene expression changes associated with basal defenses and the virulence activities of the TTSS and COR during infection. In this study, we used the Affymetrix full genome chip to determine the Arabidopsis transcriptome associated with basal defense to Pst DC3000 hrp mutants and the human pathogenic bacterium Escherichia coli O157:H7. We then used Pst DC3000 virulence mutants to characterize Arabidopsis transcriptional responses to the action of hrp-regulated virulence factors (e.g. TTSS and COR) during bacterial infection. Additionally, we used bacterial fliC mutants to assess the role of the pathogen-associated molecular pattern flagellin in induction of basal defense-associated transcriptional responses. In total, our global gene expression analysis identified 2800 Arabidopsis genes that are reproducibly regulated in response to bacterial pathogen inoculation. Regulation of these genes provides a molecular signature for Arabidopsis basal defense to plant and human pathogenic bacteria, and illustrates both common and distinct global virulence effects of the TTSS, COR, and possibly other hrp-regulated virulence factors during Pst DC3000 infection.

Introduction

Pseudomonas syringae strains collectively infect hundreds of taxonomically diverse plant species and cause disease symptoms ranging from leaf spots to stem cankers (Hirano and Upper, 2000). The P. syringae pv. tomato (Pst) strain DC3000 used in this study causes necrotic lesions that are often surrounded by chlorotic halos in susceptible tomato and Arabidopsis plants (Katagiri et al., 2002; Ma et al., 1991; Whalen et al., 1991). To successfully colonize plants, P. syringae strains have evolved a variety of virulence factors to subvert host defenses or to obtain nutrients. One common virulence mechanism is the hrp-gene-encoded type III protein secretion system (TTSS; He et al., 2004; Jin et al., 2003). The TTSS is used by P. syringae to inject >40 virulence effector proteins into the host cell (Chang et al., 2005; Collmer et al., 2002; Greenberg and Vinatzer, 2003; Nomura and He, 2005). Different P. syringae strains also produce a variety of phytotoxins (Bender et al., 1999). Although phytotoxins are generally not required for bacterial pathogenicity, they do enhance pathogen virulence in host plants (Bender et al., 1999). Pst DC3000, for example, produces a polyketide toxin, coronatine (COR), that is required for full virulence in Arabidopsis and tomato plants (Brooks et al., 2004; Ma et al., 1991; Mittal and Davis, 1995; Zhao et al., 2003). Emerging evidence suggests that production of the TTSS and COR is coordinately regulated in Pst DC3000 (Fouts et al., 2002; Penaloza-Vazquez et al., 2000), suggesting that mutations in the regulatory hrp genes (e.g. hrpS, hrpL, and hrpA) could affect the expression of both the TTSS and COR.

How TTSS effectors promote bacterial pathogenesis is poorly understood and remains an outstanding question in biology. In a previous study, we conducted microarray analysis examining the effects of wild-type Pst DC3000 and its hrpS mutant (defective in hrp regulon expression producing a non-functional TTSS), the Pst DC3118 COR mutant (defective in COR toxin production) and the CORhrpS double mutant (defective in the TTSS, COR toxin and potentially other hrp regulon-controlled virulence factors) on the expression of about 7200 Arabidopsis genes in the compatible Arabidopsis—Pst DC3000 interaction (Hauck et al., 2003). That study revealed that the TTSS is involved in biased suppression of Arabidopsis genes that encode putatively secreted proteins (Hauck et al., 2003). The TTSS-mediated suppression of these Arabidopsis genes is correlated with the ability of the TTSS and its effector AvrPto to suppress callose deposition (Hauck et al., 2003) and presumably other host responses that are associated with plant basal defense to hrp mutant bacteria (Bestwick et al., 1995) or to ‘pathogen-associated molecular patterns’ (PAMPs) such as bacterial flagellin or elongation factor Tu (EF-Tu) (Gómez-Gómez et al., 1999;Kunze et al., 2004). However, it remains to be determined what role TTSS effectors play in the transcriptional regulation of other host physiological processes important for pathogenesis.

The molecular mechanism by which COR facilitates Pst DC3000 virulence is also not well understood. COR shows a remarkable structural similarity to the plant hormone methyl jasmonic acid (MeJA), which is involved in defense and wound response signaling (Bender et al., 1999; Benedetti et al., 1995; Feys et al., 1994; Uppalapati et al., 2005). Identification of the Arabidopsis JA/COR perception mutant coi1, which is insensitive to both MeJA and COR, further suggests a similar mode of action of COR and MeJA (Kloek et al., 2001; Xie et al., 1998). The COI1 gene encodes a subunit of an SCFCOI1 complex (E3 type ubiquitin ligase) presumably involved in proteasome-mediated protein degradation (Devoto et al., 2002; Turner et al., 2002; Xu et al., 2002). Kloek et al. (2001) and Zhao et al. (2003) observed hyper-expression of a defense gene, PR-1, in P. syringae-inoculated coi1 Arabidopsis and tomato plants, respectively. Resistance in coi1 Arabidopsis plants is apparently caused by elevated SA-mediated defense, because a coi1 Arabidopsis mutant expressing the nahG gene, which encodes an SA-degrading enzyme, fails to restrict bacterial growth (Kloek et al., 2001). These observations suggest that COR is involved in the suppression of an SA-dependent host defense mechanism. However, the precise nature of the COR-targeted host defense and the contribution of COR to global host gene expression during bacterial infection have not yet been determined.

Recently, gene expression studies were performed using Arabidopsis suspension-cultured cells or seedlings treated with purified flagellin peptide (flg22), providing the first glimpse into the global gene expression changes during basal defense (Navarro et al., 2004; Zipfel et al., 2004). Whether the same set of Arabidopsis genes are regulated by flagellin and other pathogen-associated molecular patterns (PAMPs) during bacterial infection of intact plant leaves remains to be determined. Recent studies suggest that the TTSS and COR have overlapping functions in the suppression of some PR genes and induction of some wounding response genes (He et al., 2004; Zhao et al., 2003). The extent of such overlap, however, has not yet been determined at the whole-genome scale. Nor is it known how many of the PAMP-induced genes are subjected to transcriptional suppression by the TTSS and/or COR, respectively.

In this study, we conducted genome-wide gene expression analysis of Arabidopsis plants treated with defined Pst DC3000 and E. coli O157:H7 mutants to gain molecular insights into (i) the transcriptional changes associated with basal defense to live bacteria, which carry multiple PAMPs, (ii) the contribution of flagellin to the regulation of the basal defense transcriptome during infection, and (iii) the global effects of hrp-regulated virulence factors, primarily the TTSS and COR, on the basal defense transcriptome and other host physiological processes. Previous global gene expression analyses have been performed in mainly resistant Arabidopsis–Pst interactions (Tao et al., 2003) and other defense-related treatments (reviewed in Wan et al., 2002). We have conducted genome-wide analyses aimed at understanding the transcriptional responses of Arabidopsis plants to PAMPs, TTSS effectors and COR from the same bacterium during infection. The results should complement other studies and provide a useful resource for future study of Pst DC3000 pathogenesis in plants.

Results and discussion

Rationale and experimental design

All the experiments described were conducted using the Affymetrix (Santa Clara, CA, USA) ATH1 GeneChip and 4-week-old Arabidopsis leaves. For identification of PAMP-, TTSS- and COR-regulated genes, we used Pst DC3000, the Pst DC3118 COR mutant, and the Pst DC3118 CORhrpS double mutant bacteria, as described previously (Hauck et al., 2003). Because some of the strains have different growth potentials in plants, we used three different levels of bacterial inocula for gene expression profiling: low (1 × 106 bacteria/ml), moderately high (5 × 107 bacteria/ml) and high (1 × 108 bacteria/ml) (Table 1). At 1 × 106 bacteria/ml, Pst DC3000 reliably causes disease within 3–4 days, and is used in this and other laboratories as a standard disease assay (Katagiri et al., 2002). This inoculum level has also previously been used to identify hundreds of Pst-regulated genes (Hauck et al., 2003). Gene expression changes at this inoculum level were examined after a 24 h infection period (Figure S1). During this time period, Pst DC3000 and the Pst COR mutant multiply similarly to levels 40–50-fold higher than that of the PsthrpA, hrpS and CORhrpS mutants (data not shown). Therefore, to minimize the effects the different bacterial population numbers have on host gene expression, and to examine samples with early, synchronous responses, we also used inoculum concentrations of 1 × 108 and 5 × 107 bacteria/ml. In these cases, the tissues could be harvested earlier (at 7 and 10 h after inoculation, respectively) when development of basal defense-associated papillae is occurring (Hauck et al., 2003). At the inocula and time points chosen for various strains, high and comparable expression levels of a known PAMP-induced gene, Flagellin-Induced Receptor Kinase 1 (FRK1), were observed (Figure S2).

Table 1.  Arabidopsis genes reproducibly regulated by bacterial inoculation
Comparison TreatmentaRepressed genesbInduced genesbTotalExpression pattern identified
  1. aSummary data from plants 7 h after inoculation with 1 × 108 bacteria/ml (rows 1–8), 10 h after inoculation with 5 × 107 bacteria/ml (rows 9–11), and 24 h after inoculation with 1 × 106 bacteria/ml Pst bacteria (rows 12–17).

  2. bCriteria for gene selection are described in the text.

 1. Pst hrpA versus mock108, 7 h224253477PAMP regulation
 2. Pst hrpA fliC versus mock108, 7 h595100PAMP regulation
 3. Pst hrpA versus Pst hrpA fliC108, 7 h11920Flagellin-specific regulation
 4. E. coli O157:H7 versus mock108, 7 h96186282PAMP regulation
 5. E. coli TUV-86-2 fliC versus mock108, 7 h240218458PAMP regulation
 6. E. coli O157:H7 versus E. coli TUV-86-2 fliC108, 7 h000Flagellin-specific regulation
 7. Pst DC3000 versus mock108, 7 h9014951396Pseudomonas regulation
 8. Pst DC3000 versus Pst hrpA108, 7 h9044151319Pst virulence factor regulation
1270 7971901Non-redundant totals
 9. Pst CORhrpS versus mock5 × 107, 10 h197154351PAMP regulation
10. Pst COR versus mock5 × 107, 10 h8293991228Pseudomonas regulation
11. Pst COR versus Pst CORhrpS5 × 107, 10 h516275791Type III effector regulation
1018 5491537Non-redundant totals
12. Pst CORhrpS versus mock106, 24 h000PAMP regulation
13. Pst COR versus mock106, 24 h212297509Pseudomonas regulation
14. Pst COR versus Pst CORhrpS106, 24 h175253428Type III effector regulation
15. Pst DC3000 versus Pst COR106, 24 h347597944COR toxin regulation
16. Pst DC3000 versus mock106, 24 h6856801365Pseudomonas regulation
17. Pst DC3000 versus Pst CORhrpS106, 24 h7777001477Pst virulence factor regulation
 95510181929Non-redundant totals
172414302800Overall non-redundant totals

In total, we conducted eight pairwise comparisons for the 1 × 108 data set, three for the 5 × 107 data set, and six for the 1 × 106 data set. A summary of the number of genes that passed our selection criteria (see Experimental procedures) and were reproducibly differentially regulated in all three biological replicates for each comparison is shown in Table 1. These comparisons identified genes that are responsive to bacterial PAMPs of Psthrp mutants and E. coli O157:H7 as well as those that are regulated by Pst DC3000 TTSS effectors and/or COR toxin (Table 1, Table S1). Our global transcriptional profiling identified 2800 genes that were reproducibly differentially regulated. The expression data for these genes across all of our comparisons is presented in Table S1.

PAMP-induced transcriptional changes

In order to identify the transcriptional changes resulting from PAMP perception, we inoculated Arabidopsis with three different bacteria: Pst DC3000 hrpA mutant, Pst DC3000 CORhrpS mutants, and human pathogenic E. coli O157:H7. The Pst hrpA and CORhrpS mutants lack the ability to secrete TTSS effectors into the host and are unable to multiply significantly in plant host leaves or produce disease symptoms. Furthermore, the CORhrpS mutant is unable to produce COR toxin and similarly, the hrpA mutant is also probably defective in COR toxin production via a feedback mechanism. The hrpA mutant strain has repressed expression of the hrpS and hrpL genes, which globally control the expression of genes involved in the production of the TTSS, COR toxin, and possibly other virulence factors (Fouts et al., 2002; Wei et al., 2000). Therefore, in the hrpA and CORhrpS mutants, many, if not all, plant-specific virulence factors are not produced. E. coli O157:H7, a human pathogen, is also unable to multiply or cause any symptoms on Arabidopsis leaf tissue (data not shown) and is not expected to carry plant-specific virulence factors. We hypothesized that inclusion of these three bacterial strains in the analysis should give us a global view of common and distinct plant transcriptional responses through the perception of PAMPs displayed by human and plant pathogenic bacteria. To assess the contribution of flagellin perception in host gene regulation during bacterial infection, we examined Arabidopsis gene expression regulated by the hrpA fliC double mutant of Pst DC3000 (Hu et al., 2001) and a fliC mutant derivative of E. coli TUV86-2 (Gunzer et al., 1998).

We identified a set of 736 PAMP-responsive genes that were differentially regulated at least twofold in three biological replicates for at least one of the following three treatments with non-pathogenic bacteria: Pst hrpA versus mock 108 bacteria/ml, E. coli O157:H7 versus mock 108 bacteria/ml, and Pst CORhrpS versus mock 5 × 107 bacteria/ml. A tally of the number of Arabidopsis genes regulated in response to each of the three bacterial treatments is presented in Table 1. Of the 736 reproducibly regulated genes, 347 were induced and 389 were repressed. A list of the 736 PAMP-regulated genes with annotation and fold changes for each treatment is presented in Table S2. Because we are particularly interested in genes whose products may be involved in the induction or elaboration of basal defenses, we chose to focus our analysis primarily on the set of 347 PAMP-induced genes.

Hierarchical cluster analysis of PAMP-induced genes

To provide an overview of the regulation of the 347 PAMP-induced genes across our set of treatments and biological replicates and to visualize the regulation of these genes in response to inoculation with virulent Pst DC3000, we performed hierarchical clustering. The resulting cluster is displayed in Figure 1. The induction of Arabidopsis genes after inoculation with E. coli was strikingly similar to the induction elicited by the Pst hrp mutants. The correlation coefficients based on the PAMP-induced genes for each of the 1 × 108 and 5 × 107 comparisons are shown in Table S3. Although Pst hrpA inoculation resulted in the reproducible regulation of 195 more genes (that matched or exceeded our selection criteria) compared with inoculation with E. coli O157:H7 (Table 1), the expression profiles were remarkably similar for the two treatments, as shown in Figure 1, and their correlation coefficient was 0.79 (Table S3). We are particularly interested in identifying genes whose products may play a role in the plant basal defense response. Virulent Pst DC3000 is able to overcome or evade this host basal defense response and may do so by repressing or blocking the induction of host genes whose products are involved in the elaboration of these defenses. Therefore, we expected that some PAMP-induced genes would be either repressed or simply not reproducibly induced upon inoculation with Pst DC3000. The blue bar in Figure 1 illustrates a cluster of genes that exhibit an expression pattern consistent with a role in basal defense. A total of 201 of the 347 PAMP-induced genes were either repressed or not significantly regulated (average fold change < 1.5) in response to Pst DC3000 inoculation (Table S2). We propose that these genes may have a role in the Arabidopsis basal defense response to bacteria.

Figure 1.

Expression profile of 347 PAMP-induced Arabidopsis genes.
Each column represents a single biological replicate, each row represents a differentially expressed probe set. The expression of each gene is colored green for repression and red for induction as illustrated in the fold change color bar. Each set of three independent biological replicates comprising the nine comparisons displayed is labeled. Genes highlighted by the blue bar on the right are of particular interest as these genes are induced by PAMP perception, but are not reproducibly induced or are repressed by Pst DC3000 and/or Pst COR.

Transcription factors and signaling components are induced in response to PAMPs

Of the 347 PAMP-induced Arabidopsis genes, 86 encode known or predicted transcription factors and kinases. Thirty transcription factors are reproducibly induced in response to PAMP perception, 17 of which are repressed or not significantly regulated in response to DC3000 (Table 2a). Of these 30 PAMP-induced transcription factors, 12 belong to the WRKY family and eight of these are induced by PAMPs but not by Pst DC3000. WRKY22 (At4g01250) was previously shown to participate in the FLS2-mediated signaling cascade triggered by perception of the synthetic flagellin peptide flg22 (Asai et al., 2002). Additionally, six of these 12 WRKY transcription factor genes (shaded gray in Table 2a) were shown to be induced in Arabidopsis seedlings in response to treatment with flg22 (Zipfel et al., 2004). These data suggest that WRKY family transcription factors play an important role in regulating the Arabidopsis basal response to bacteria. Determining the target genes regulated by these PAMP-induced WRKY transcription factors should further our understanding of the processes involved in the plant basal defense response.

Table 2.  Selected PAMP-induced Arabidopsis genes
Array elementAGIDescriptionAverage fold change
hrpACORhrpSE. coliDC3000
  1. The average fold changes of each treatment compared to the mock control are shown. The hrpA, E. coli and DC3000 treatments were 7 h post-inoculation with 1 × 108 bacteria/ml, while the CORhrpS treatment was at 10 h post-inoculation with 5 × 107 bacteria/ml. Lines within the table denote the cut-off for PAMP-induced genes that are repressed or not significantly induced (less than 1.5-fold) by Pst DC3000. Genes below the line are also induced by DC3000. Gray shading indicates genes previously identified as induced by the bacterial PAMP flagellin.

(a) Transcription factors induced by PAMP perception
254231_atAt4g23810AtWRKY532.092.642.00−10.56
267385_atAt2g44380CHP-rich zinc finger protein, putative4.392.704.00 −6.06
246993_atAt5g67450AZF1 Cys2/His2-type zinc finger protein 12.964.492.58 −4.70
254159_atAt4g24240AtWRKY72.702.582.46 −3.25
255596_atAt4g01720AtWRKY472.303.032.09 −2.58
255547_atAt4g01920DC1 domain-containing protein, similar to CHP-rich proteins1.552.411.38 −2.46
260037_atAt1g68840AtRAV2 AP2 domain-containing protein1.951.182.14 −2.09
266010_atAt2g37430Putative C2H2-type zinc finger protein3.733.173.65 −2.00
251745_atAt3g55980Zinc finger (CCCH-type) family protein2.002.242.30 −1.59
267384_atAt2g44370CHP-rich zinc finger protein, putative21.617.6411.58 −1.52
253535_atAt4g31550AtWRKY112.461.742.64 −1.38
264616_atAt2g17740CHP-rich zinc finger protein, putative9.404.815.79 −1.38
255568_atAt4g01250AtWRKY225.922.525.40 −1.07
263797_atAt2g24570AtWRKY172.763.033.251.00
248306_atAt5g52830AtWRKY274.492.464.591.10
260783_atAt1g06160ERF subfamily ERF/AP2 transcription factor family3.911.321.381.41
249890_atAt5g22570AtWRKY381.414.191.821.48
255753_atAt1g18570AtMYB51 myb family transcription factor2.462.142.091.55
257919_atAt3g23250AtMYB15 myb family transcription factor5.662.642.961.74
253485_atAt4g31800AtWRKY181.783.171.622.00
250153_atAt5g15130AtWRKY724.294.925.792.52
267588_atAt2g42060CHP-rich zinc finger protein, putative6.817.464.812.58
260432_atAt1g68150AtWRKY97.133.654.393.25
261648_atAt1g27730AtZAT10 salt-tolerance zinc finger protein2.002.302.463.82
246214_atAt4g36990AtHSF4 putative heat shock transcription factor3.173.172.706.81
245252_atAt4g17500AtERF-1 ethylene response factor 12.581.292.0010.08
245976_atAt5g13080AtWRKY757.293.173.7311.58
252214_atAt3g50260DREB subfamily A-5 of ERF/AP2 transcription factor family1.872.581.7813.30
267140_atAt2g38250Putative GT-1-like transcription factor3.172.052.5213.93
247264_atAt5g64530No apical meristem (NAM) family protein2.706.203.5630.55
(b) Genes encoding proteins known or predicted to be involved in cell wall modification or secretion
255524_atAt4g02330Hypothetical protein, similar to pectinesterase3.253.732.00 −5.40
245052_atAt2g26440Putative pectinesterase2.142.961.74 −2.41
251895_atAt3g54420AtEP3 class IV chitinase2.704.923.03 −2.24
247954_atAt5g56870Beta-galactosidase, putative/lactase, putative2.241.232.70 −2.19
255595_atAt4g01700Chitinase, putative, similar to peanut type II chitinase3.036.062.64 −1.59
264669_atAt1g09630AtRABA1a Rab small GTPase1.702.241.52 −1.52
259173_atAt3g03640AtGLUC beta glucosidase1.953.911.74 −1.48
259443_atAt1g02360Chitinase, putative3.254.493.17 −1.45
253099_atAt4g37530Peroxidase, putative1.912.192.14 −1.05
251832_atAt3g55150Exocyst subunit EXO70 family protein14.255.539.19 −1.02
254543_atAt4g19810Glycosyl hydrolase family 18; similar to chitinase/lysozyme3.483.732.581.07
260556_atAt2g43620Glycosyl hydrolase family 19 (chitinase)1.296.351.551.07
246532_atAt5g15870Glycosyl hydrolase family 81; similar to beta-glucan-elicitor receptor2.352.462.301.10
265499_atAt2g15480Putative glucosyltransferase2.241.911.621.15
254468_atAt4g20460UDP-glucose 4-epimerase-like protein3.732.303.251.18
247327_atAt5g64120Peroxidase, putative3.033.403.561.87
254673_atAt4g18430AtRABA1e RAB GTPase4.814.492.761.70
254262_atAt4g23470Hydroxyproline-rich glycoprotein family protein1.322.521.352.14
264960_atAt1g76930AtEXT1/4 extensin hydroxyproline-rich glycoprotein1.592.761.382.14
255111_atAt4g08780Peroxidase, putative3.323.821.912.30
254314_atAt4g22470Putative hydroxyproline-rich glycoprotein4.813.034.002.89
259553_atAt1g21310Proline-rich extensin-like family protein, contains extensin-like region3.653.733.733.25
261474_atAt1g14540Anionic peroxidase similar to lignin-forming peroxidase6.355.666.35 3.32
249459_atAt5g39580ATP24a peroxidase11.859.198.77 4.39
245148_atAt2g45220Pectinesterase family protein6.203.404.70 4.92
247487_atAt5g62150Peptidoglycan-binding LysM domain-containing protein3.252.763.03 8.19
246228_atAt4g36430Peroxidase, putative11.063.4011.5 8.98
255110_atAt4g08770Peroxidase, putative54.449.4014.524.82

In addition to WRKY transcription factors, nine known or putative zinc finger transcription factors and four ERF/AP2 family transcription factors were induced by PAMPs. Two of the four ERF/AP2 family transcription factors and six of the nine zinc finger proteins were previously shown to be induced by flg22 (Zipfel et al., 2004). Interestingly, the ethylene-responsive AtERF1 transcription factor was found to be induced by PAMPs. The plant hormone ethylene has been shown to play a role in plant defense against pathogens, and AtERF1 was also found to be induced in response to flg22 treatment. However, our study shows that AtERF1, while reproducibly induced approximately twofold by PAMP perception, is even more strongly induced (approximately 10-fold) by the virulent pathogen Pst DC3000. These data suggest that ethylene signaling may play a role in the induction of basal defenses, but that the basal defense responses triggered by ethylene signaling are either insufficient to prevent Pst DC3000 pathogenesis or that Pst DC3000 blocks ethylene signaling or responses downstream of AtERF1 induction. This example illustrates the value of determining the expression profile of these genes in response to Pst DC3000 in addition to PAMPs rather than simply identifying genes that are induced by PAMPs.

In addition to transcription factors, a number of other signaling components were induced by PAMP perception. Particularly abundant among PAMP-induced genes were genes known or predicted to encode kinases. Fifty-six known or putative kinases were induced by PAMP perception (Table S4). Of these 56 PAMP-induced kinases, 31 were found to be induced by flg22 treatment (shaded gray in Table S4). Receptor-like kinases, including lectin family kinases, and wall-associated kinase-like proteins are the most highly represented classes of kinases induced by PAMPs. Wall-associated kinases (WAKs) and WAK-like kinases (WAKLs) have recently been implicated in the Arabidopsis response to pathogens. AtWak1 was shown to be induced by P. syringae and SA (He et al., 1999), and WAK1 was recently shown to interact with the glycine-rich protein AtGRP3 and to phosphorylate oxygen-evolving enhancer protein 2 (OEE2; Yang et al., 2003). Treatment of Arabidopsis protoplasts with purified AtGRP3 induced PR-1 gene expression, and inoculation of plants with avirulent Pst DC3000 enhanced WAK1-dependent phosphorylation of OEE2, suggesting that WAK1 and GRP3 may play a role in the response to pathogens. Additionally, the WAK-like kinase WAKL22 was recently found to be identical to the resistance gene Rfo1 that confers resistance to a number of isolates of the fungal pathogen Fusarium oxysporum (Diener and Ausubel, 2005). Although neither WAK1 nor WAKL22 were specifically induced by PAMPs in our study, the prevalence of WAK-like kinases among PAMP-induced genes, taken together with the recent results highlighted above, suggests that this class of proteins may play an important role in the plant response to pathogens.

Also of particular interest are a number of leucine-rich repeat (LRR)-containing receptor-like kinases. These receptor-like kinases may be candidates for pattern recognition receptors (PRRs) in Arabidopsis. Although Arabidopsis is known to perceive a number of bacterial PAMPs including flagellin, lipopolysaccharides and EF-Tu, only a single PRR for bacterial PAMPs has been identified (Felix et al., 1999; Kunze et al., 2004; Zeidler et al., 2004). FLS2, the PRR that mediates flagellin perception in Arabidopsis, is an LRR receptor kinase (Gómez-Gómez and Boller, 2000). Identifying additional PRRs in Arabidopsis is a high priority and should lead to new insights into the signaling pathways initiated by perception of bacterial PAMPs and the downstream basal defense responses that protect plants against a wide array of bacteria. These PAMP-induced LRR receptor-like kinases provide a starting point in the search for other Arabidopsis PRRs.

PAMPs induce genes that may be involved in secretion and cell wall modification

One of the most well characterized aspects of plant basal defense against pathogens is modification of the cell wall. Plants challenged with bacteria or fungi typically develop cell wall appositions referred to as papillae at the point of interaction with the invading pathogen. Formation of papillae is characterized by the deposition of electron-dense materials such as phenolic compounds, callose, and cell wall proteins such as hydroxyproline-rich glycoproteins. Callose is synthesized at the plasma membrane, but the rest of the components are synthesized within the cell and must be delivered to the proper location at the wall for deposition into papillae. Because of this, components of the plant secretory system are likely to play an essential role in basal defense against pathogens. Consistent with the predicted importance of cell wall modification and polarized secretion in basal defense, we found that a number of PAMP-induced genes are potentially involved in these processes. Table 2(b) lists 28 PAMP-induced genes that may have a role in cell wall modification or polarized secretion. Among these 28 genes are a large number of genes encoding putative hydrolase enzymes that may potentially play a role in reorganizing the structure of the cell wall during papilla formation. Additionally, four genes encoding proline-rich and hydroxyproline-rich glycoproteins are induced by PAMPs. Seven putative peroxidases are induced that may be involved in the oxidative cross-linking of proteins and polymers into the papilla structure. The peroxidase ATP24a (At5g39580) was previously shown to be up-regulated during the RPP4- and RPP8-mediated resistance responses to the oomycete pathogen Peronospora parasitica (Eulgem et al., 2004). Consistent with a role in oxidative cross-linking in the cell wall, all of these peroxidases are predicted to enter the secretory pathway based on TargetP analysis (http://www.cbs.dtu.dk/services/TargetP).

Strikingly, three genes encoding potential components of the secretory pathway are up-regulated by PAMP perception. Two members of the RabA family of small GTPases, RABA1a and RABA1e, were induced by PAMPs. Rab proteins are a class of small GTPases that also function in vesicle trafficking and regulate vesicle docking. Based on work in mammalian and yeast systems, this class of Rabs is predicted to participate in post-Golgi vesicle trafficking (Vernoud et al., 2003). Additionally, two RabA isoforms from pea were suggested to have a role in delivery of cell wall components to the plasma membrane (Nagano et al., 1995). Also, a member of the EXO70 family of proteins (At3g55150) was up-regulated in response to PAMPs. EXO70 family proteins are subunits of the exocyst protein complex that is required for exocytosis in yeast and is also suggested to play a role in docking of exocytotic vesicles (Cole et al., 2005). The EXO70 family member was previously found to be up-regulated in response to flg22 treatment (Zipfel et al., 2004); however, RABA1a and RABA1e represent novel PAMP-induced genes with functions related to secretion. Taken together, the up-regulation of these genes supports a role for targeted vesicle trafficking in the Arabidopsis basal defense response to bacteria, and identifies candidate secretory pathway components that may play a role in secretion of papilla components or other defense-related compounds.

Flagellin does not contribute uniquely to PAMP-induced transcriptional changes

Given the substantial overlap of the PAMP-induced genes identified by our microarray analysis with genes previously identified as induced by the synthetic flagellin peptide flg22 (131 of 347 PAMP-induced genes), we sought to determine the specific contribution of flagellin to the induction of these genes. To determine whether perception of flagellin makes a specific contribution to the transcriptional changes observed in Arabidopsis after bacterial inoculation, we compared the expression profiles of Arabidopsis plants inoculated with fliC mutant derivatives of Pst hrpA and E. coli O157:H7 to those of plants inoculated with the parent strains. fliC mutants are impaired in the production of the flagellin protein. Therefore, this comparison allows us to determine the specific contribution of flagellin perception to PAMP-induced transcriptional changes.

A fliC mutant of E. coli O157:H7 was not available; therefore, a fliC mutant of the O157:H7 derivative strain TUV86-2 was used. To validate the use of this strain, we performed microarray analysis to compare the expression profiles of plants inoculated with E. coli O157:H7 or E. coli TUV86-2 and found that the expression profiles of these plants were highly similar (data not shown). The expression profiles of the 347 PAMP-induced genes from plants inoculated with fliC mutant bacteria were similar to those inoculated with the parental, flagellin-producing strains (Figure 1). The correlation value for the Pst hrpA versus mock and Pst hrpA fliC versus mock comparisons is 0.93, and for the E. coli versus mock and E. colifliC versus mock comparisons it is 0.97 (Table S3). Consistent with the similarity apparent in the clustering analysis and the calculated correlation, we found that only 20 genes were reproducibly regulated more strongly by the flagellin-producing strains than by the fliC mutant derivatives (Table 1). No genes were regulated more strongly by E. coli O157:H7 than by the fliC mutant derivative. However, when we compared the expression profile of plants inoculated with Pst hrpA fliC to mock inoculation, we found that, using our selection criteria, 377 fewer genes were reproducibly regulated by hrpA fliC than by hrpA (Table 1). This is a result of our stringent selection criteria as genes were regulated similarly, but not as robustly, by Pst hrpA fliC. This suggests that, at least in the case of hrpA, flagellin perception may make a quantitative contribution to the PAMP-induced regulation of transcription. However, this quantitative contribution is relatively minor as only nine genes were reproducibly induced at least twofold more strongly by the flagellin-producing hrpA strain than by hrpA fliC (Table 1). These results suggest that, although perception of flagellin induces transcriptional changes in Arabidopsis, flagellin does not make a substantial unique contribution to the PAMP-induced transcriptional changes observed after bacterial inoculation. Arabidopsis can perceive other PAMPs such as LPS and EF-Tu, and our results suggest that perception of flagellin induces transcriptional changes that overlap almost completely with those induced by other PAMPs. Consistent with these results, multiplication of fliC mutants of Pst DC3000 and hrpA in Arabidopsis after vacuum infiltration is similar to that of the flagellin-producing parent strains (Figure S3). These results suggest that, although flagellin may make a minor quantitative contribution to PAMP-induced transcriptional changes, it is only one of multiple PAMPs recognized by plants and that different PAMPs induce overlapping plant responses to bacteria within leaf tissue. The recent finding by Zipfel et al. (2004) that multiplication of Pst DC3000 is enhanced after inoculation by dipping, but not after inoculation by vacuum-infiltration, suggests that perception of bacterial flagellin may make a unique contribution to defenses effective on the surface of the leaf, but does not make a unique contribution to elicitation of defenses when bacteria are infiltrated directly into the leaf intercellular spaces. Our data support this notion.

Analysis of the TTSS- and COR-regulated genes

To put the results of the PAMP gene regulation in perspective relative to a virulent pathogen capable of causing disease, and to identify host genes regulated by Psthrp-dependent TTSS effectors and COR toxin, we examined host gene expression using wild-type Pst DC3000 and the virulence-compromised mutants Pst COR and Pst CORhrpS at both low (1 × 106 bacteria/ml) and moderately high (5 × 107 bacteria/ml) inoculum levels (Table 1). Overall, we observed that relatively equal numbers of repressed and induced genes were identified in each comparison examined, with the exception of the 5 × 107 bacteria/ml Pst COR versus mock and Pst COR versus Pst CORhrpS comparisons which identified approximately twice as many repressed genes as induced genes. Together the 5 × 107 and 1 × 106 data sets identified 2439 host genes with Pst-responsive gene expression. Hierarchical cluster analysis of these 2439 reproducibly differentially regulated genes is shown in Figure 2. Of these 2439 genes, 1027 (42%) were reproducibly differentially expressed in both data sets (Table 1, Figure 2). Inspection of Figure 2 illustrates that even though 1412 of the genes were not contained in the lists of differentially regulated genes for both data sets (because they failed to surpass the selection criteria in both cases), many were regulated in a similar way. The overall reproducibility between the independent biological replicates is evident as indicated by the uniformity of the gene expression changes within the three biological replicates of each comparison (Figure 2). COR toxin was a potent regulator, inducing or repressing the expression of 944 genes 24 h post-inoculation (Table 1, row 4). The COR-induced genes comprise approximately 60% (597/1018) of the reproducibly induced genes (Figure 2, clusters A1 and A2, Table 1), while only 347 genes are COR toxin-repressed (Table 1). By comparing the samples inoculated with 5 × 107Pst COR and CORhrpS bacteria/ml (to minimize bacterial population differences), a total of 791 genes exhibiting hrp/TTSS effector-dependent regulation were identified (Table 1, row 11). These two strains lack COR and allow the identification of hrp/TTSS effector-dependent regulated genes in the absence of the effect of COR toxin. Two hundred and seventy five of these genes exhibited hrp/TTSS effector-dependent induction (Table 1, Figure 2, cluster B, Table S5). Upon visual inspection of Figure 2, it is clear that many of the differentially expressed genes are coordinately regulated in response to Pst DC3000 and the Pst COR mutant (repressed by both strains or induced by both strains compared to the mock control), while other sets of genes are regulated in opposing directions. This result illustrates the importance of using both hrp and COR mutant bacterial strains in discerning the specific effects of TTSS effectors and COR on host gene expression.

Figure 2.

The expression profile of the 2439 Arabidopsis genes reproducibly regulated by 1 × 106 and/or 5 × 107Pst bacteria/ml.
See Figure 1 legend for a description of the display. Groups of coordinately regulated genes are highlighted on the right and discussed in the text.

Independent validation of the microarray expression data was pursued using RNA blot analysis. A total of 12 repressed and 13 induced genes were selected, and cDNA probes for each gene were used for RNA blot hybridization. The results are shown in Figure S1. The RNA samples used for blot analysis were from independent biological samples (different from those used for Affymetrix chip hybridization) and included a time-course following inoculation with 1 × 106 bacteria/ml. The results confirmed the expression patterns observed with the microarray, and illustrate well that genes differentially expressed 24 hours post-inoculation in our microarray experiments may exhibit changes as early as 6 h and/or continue to be differentially expressed until at least 36 h after treatment.

Impact of Pst pathogenesis on host metabolism

An advantage of global gene expression profiling is the comprehensive view it can provide concerning the transcriptional regulation of genes involved in all aspects of plant physiology. In order to take advantage of this large data set, we utilized MapMan to visualize how Pst pathogenesis impacted the expression of genes involved in various host metabolic pathways and processes (Thimm et al., 2004; Usadel et al., 2005). Some of the most dramatic global differences are illustrated in the MapMan metabolism overview map displaying the 944 COR toxin-regulated genes and 791 hrp-regulated genes (see Tables S6 and S7 for these gene lists). The most obvious bias within the COR-regulated genes is the induction of many genes involved in secondary metabolism (Figure S3a). Genes that encode enzymes in the phenylpropanoid, terpanoid, anthocyanin and glucosinolate pathways are all COR-induced (Figure 3a, Figure S4a). This result is consistent with recent data demonstrating that JA and other jasmonates induce the expression of numerous genes that encode enzymes involved in secondary metabolism (Sasaki-Sekimoto et al., 2005; Taki et al., 2005). In contrast, hrp-dependent TTSS effectors are responsible for the repression of cell wall genes, particularly pectin esterases, as well as fatty acid biosynthesis (in the chloroplast) and a dramatic overall repression of photosynthesis (Figure 3b). Many genes encoding chloroplast-localized proteins involved in photosynthesis and the Calvin cycle, as well as genes involved in photorespiration and tetrapyrrole synthesis, are repressed by TTSS effectors (Figure 3b, Figure S4b).

Figure 3.

COR toxin and type III effectors target different aspects of plant metabolism.
(a) The MapMan ‘Metabolism Overview’ display created using the 944 COR toxin-regulated genes identified from the Pst DC3000 versus Pst COR comparison (24 h post-inoculation, 1 × 106 bacteria/ml).
(b) The MapMan ‘Metabolism Overview’display created using the 791 TTSS-regulated genes identified from the Pst COR versus Pst CORhrpS comparison (10 h post-inoculation, 5 × 107 bacteria/ml).
The average fold change of the three biological replicates is displayed as illustrated in the fold change color bar in the lower right of each panel (red is repressed, blue is induced).

Coronatine activates JA responses and secondary metabolism

Analysis of the data has also indicated that Pst challenge significantly targets several plant hormone signaling and stress-response pathways. The expression patterns observed for some genes involved in plant hormone signaling and stress responses are displayed in Figure 4.

Figure 4.

Pst coordinately regulates hormone and stress response genes.
The expression profile for 57 genes following Pst inoculation is shown. See Figure 1 legend for a description of the gene expression display. Gene annotation is listed on the right with background shading delineating a relationship to known pathways or responses. Orange, JA- or COR toxin-inducible; yellow, JA biosynthesis; red, secondary metabolism; green, ABA and abiotic stress response; violet, SA and defense responses; blue, basal defense.

Consistent with the results from an analysis of the effects of purified COR in tomato (Uppalapati et al., 2005), known JA/COR toxin-responsive genes including chlorophyllase (CLH1/CORI1 At1g19670; Benedetti et al., 1998; Tsuchiya et al., 1999) and CORI3 (At4g23600; Lopukhina et al., 2001) were induced 24 h post-inoculation in a COR-dependent manner. Interestingly, four genes encoding enzymes involved in jasmonate biosynthesis (LOX2, At3g45140; AOS, At5g42650; AOC1, At3g25760; OPR3, At2g06050; Bell et al., 1995; von Malek et al., 2002; Schaller et al., 2000) are also COR toxin-induced, suggesting activation of not only jasmonate responses but the host biosynthesis of jasmonates as well.

As suggested by the analysis using MapMan, host secondary metabolism is activated by COR toxin, including induction of the PAP1 transcription factor (At1g56650), a regulator of secondary metabolism including the production of anthocyanin pigments (Borevitz et al., 2000). As expected, CHS (At5g13930) and a putative anthocyanidin synthase (At2g38240) and UDP- anthocyanidin transferase (At4g27570) are strongly induced (Figure 4). The expression of PAL1 (At2g37040) involved in phenylpropanoid biosynthesis and GSTF11 (At3g03190) also displayed COR-dependent induction, consistent with their observed over-expression in PAP1 activation-tagged plants (Borevitz et al., 2000). Several other genes involved in phenylpropanoid metabolism also exhibited reproducible induction (F5H2, At5g04330; 4CL9, At1g20510; Eli3-2/CAD-B2, At4g37990), but, somewhat surprisingly, numerous others were repressed (PAL3, At5g04230; F5H1, At4g36220; 4CL1, At1g51680; COMT1, At5g54160; CAD1, At4g39330; Gachon et al., 2005). The physiological impact of this conflicting differential regulation on lignin biosynthesis is unclear.

Two genes encoding enzymes putatively involved in terpenoid biosynthesis (S-linalool synthase, At1g61120 and TPS03, At4g16740) and DHS1, the first enzyme of the shikimate pathway (At4g39980; Keith et al., 1991) are induced by COR toxin (Figure 4). The shikimate pathway produces precursors for the synthesis of numerous secondary metabolites as well as chorismate for the production of aromatic amino acids. One metabolic pathway coordinately regulated is that of tryptophan (Trp) biosynthesis. Pst coordinately induces three genes that encode enzymes that convert chorismate to Trp (Figure S5). Phosphoribosyl anthranilate isomerase (PAI1/PAI2, At1g07780/At5g05590, both represented by probe set 259770_s_at; He and Li, 2001) is induced in a COR toxin-dependent way, while ASA1 (At5g05730; Niyogi and Fink, 1992) and ASB1/ASB2 (At1g25220/At5g57890; Niyogi et al., 1993) are induced by both COR toxin and TTSS effectors (Figure S5). The coordinated induction of these genes is consistent with the induced expression of ATR1 (AtMyb34 At5g60890) and ATR2 (At5g46760), which are known to activate Trp and glucosinolate production in Arabidopsis (Celenza et al., 2005; Smolen et al., 2002). As the tryptophan pathway activators ATR1 and ATR2 are COR toxin-induced, we examined whether other genes that encode enzymes in this pathway were differentially regulated, but did not meet our stringent selection criteria. Indeed, five genes, PAT (At5g17990), TSA1 (At3g54640), TSB1/TSB2 (At5g54810/At4g27070; Radwanski et al., 1995), IGPS (At2g04400; Li et al., 1995) and a tryptophan synthase beta-like gene (At5g38530) were also induced following bacterial inoculation (Figure S5). Tryptophan is probably being used as a substrate for the production of glucosinolates and indole-3-acetic acid (IAA), as CYP79B2 (At4g39950; Zhao et al., 2002), CYP79B3 (At2g22330; Zhao et al., 2002) and NIT3 (At3g44320; Vorwerk et al., 2001) are reproducibly induced following inoculation with 1 × 106 bacteria/ml Pst DC3000.

The tryptophan pathway may also potentially contribute to the accumulation of the phytoalexin camalexin as PAD3 is a known Pst-induced gene (At3g26830; Zhou et al., 1999; Figure S5).

The synthesis of methionine-derived glucosinolates is also implied, as CYP79F1/CYP79F2 (At1g16400/At1g16410; Chen et al., 2003; Reintanz et al., 2001; Tantikanjana et al., 2004) and AOP2 (At4g03060; Kliebenstein et al., 2001) are induced by COR toxin. The expression pattern displayed in Figure S5 also illustrates how COR toxin and TTSS effectors can both synergistically induce the expression of some genes while antagonistically regulating the expression of other genes.

Pst induces ABA/abiotic stress response genes

COR toxin and, to some degree, TTSS effectors induce numerous ABA-responsive genes, including a significant number of genes implicated in ABA/abiotic stress-responsive signaling (Figure 4; Bray, 2002; Finkelstein and Rock, 2002). ABI1 (At4g26080; Leung et al., 1994) and ABI2 (At5g57050; Leung et al., 1997) are both induced in the absence of COR toxin at the 5 × 107 bacteria/ml inoculum level, suggesting that their expression is at least partly regulated by TTSS effectors (Figure 4). NCED3 (At3g14440; Iuchi et al., 2001), a 9-cis-epoxycarotenoid dioxygenase crucial for ABA synthesis, is also induced following Pst COR inoculation (Figure 4). NCED4 (At4g19170), a related family member, is repressed by virulent Pst DC3000 and Pst COR bacteria at the higher inoculum levels. A total of six transcription factors associated with ABA and abiotic stress responses (HB-12, At3g61890; HB-7, At2g46680; NAC3/ANAC055, At3g15500; RD26 ANAC072, At4g27410; ANAC019, At1g52890; MYC2/RD22BP1, At1g32640) are also induced by Pst infection (Figure 4) (Abe et al., 2003; Fujita et al., 2004; Greve et al., 2003; Olsson et al., 2004; Soderman et al., 1996; Tran et al., 2004). These results imply that ABA responses are being activated by Pst and that both TTSS effectors and COR toxin are involved in this activation.

Defense response-related gene expression

Of particular interest is the role COR toxin and TTSS effectors have in SA-mediated defenses as those responses are necessary for resisting pathogen attack. Several SA-responsive genes involved in pathogen defense (EDS5, At4g39030; ICS1, At1g74710; PAD4, At3g33960; EDS1, At3g48090; NIMIN-1, At1g02450) were induced in a TTSS-dependent manner (Figure 4) (Falk et al., 1999; Jirage et al., 1999; Nawrath et al., 2002; Weigel et al., 2005; Wildermuth et al., 2001). The Pst-induced expression of these genes is consistent with previous published results, but it was somewhat surprising that their induction is enhanced between 1.5- and 20-fold by TTSS effectors. The expression of these genes is either not reliably induced or induced 2–5-fold by Pst CORhrpS bacteria relative to the mock inoculated control in the 5 × 107 data set. This suggests that the host plant is recognizing PAMPs present on the Pst CORhrpS mutant and potentially also recognizing specific TTSS effectors and activating defense gene expression (Figure 4). Interestingly, COR toxin actually dampens by approximately fourfold the induction of at least one SA-related defense gene (NIMIN-1). In the samples examined, the defense response genes AIG1 (At1g33960; Reuber and Ausubel, 1996) and TONB (At2g32190; de Torres et al., 2003) are induced 6–20-fold.

In contrast to the induction observed for the previously discussed SA/defense-related genes, virulent Pst bacteria repress three genes associated with basal defense responses. FlagellinSensitive 2 (FLS2 At5g46330), which is required for the perception of the PAMP flagellin, is repressed by both COR toxin and TTSS effectors at high inoculum concentrations (Figure 4) (Gómez-Gómez et al., 1999; Zipfel et al., 2004). The Flagellin induced Receptor Kinase 1 (FRK1, At1g19190; Asai et al., 2002; de Torres et al., 2003) and WRKY22 (At4g01250; Asai et al., 2002) are induced by E. coli, Pst hrpA and Pst CORhrpS mutant bacteria at high inoculum levels, and that induction is blocked by TTSS effectors in the Pst COR mutant (Figure 4). COR toxin produced by Pst DC3000 also contributes to the repression of FRK1, WRKY22 and FLS2 expression at the 1 × 106 inoculum level. These results are consistent with the idea that TTSS effectors and COR toxin contribute to pathogenesis by suppressing basal defense responses.

Expression of auxin- and cytokinin-related genes is altered during Pst infection

It is clear that Pst virulence systems are modulating stress hormone response pathways (JA, SA and ABA), but plant host physiology also appears to be experiencing specific shifts in auxin and cytokinin responses as well. Thirty-one auxin-related genes are repressed while 13 other genes are induced following Pst inoculation (Figure S6). In most cases, the gene regulation appears to be partially dependent on both COR toxin and TTSS effectors. A few genes are also reproducibly differentially regulated in response to bacterial PAMPs present on E. coli and non-pathogenic strains of Pst as well. The 31 repressed genes include four AUX/IAA and 18 SAUR (small auxin up-regulated) genes, six auxin transporters (PIN3, At1g70940; PIN4, At2g01420; PIN7, At1g23080; AUX1, At2g38120; LAX1, At5g01240; LAX3, At1g77690) and ARF18 (At3g61830) (Hagen and Guilfoyle, 2002; Okushima et al., 2005; Swarup et al., 2004). AUX/IAA genes are rapidly induced by auxin and are hypothesized to act in a feedback loop repressing the auxin response signaling pathway (Tiwari et al., 2004). Five SAUR genes, three IAA-amino acid hydrolases (ILR1, At3g02875; ILL5, At1g51780; ILL6, At1g44350), three IAA-amido synthases (DFL2, At4g03400; GH3.3, At2g23170; GH3.12, At5g13320), IAA18 (At3g16500) and NIT3 are induced. NIT3 will synthesize IAA (Figure S5b), while IAA-amino acid hydrolases and amido synthases convert IAA from an amino acid conjugate to the free form (Staswick et al., 2005). Taken together, these results suggest that Pst is impacting host auxin signaling, potentially de-repressing the pathway, altering auxin movement and activating biosynthesis of the hormone. These results are consistent with the observation that Pseudomonas syringae strains express biosynthetic enzymes that can produce IAA from tryptophan (Buell et al., 2003; Glickmann et al., 1998). Pst DC3000 has been shown to elicit the accumulation of IAA in Arabidopsis leaves during infection, and activation of this pathway contributes to disease susceptibility (O'Donnell et al., 2003).

Auxin and cytokinin are involved in many important aspects of plant growth and development. Pst pathogenesis appears to not only target auxin signaling but also genes involved in cytokinin accumulation and signaling. Seven type A response regulators (ARR4, 5, 6, 7, 9, 15 and 16; D'Agostino et al., 2000; To et al., 2004) are repressed by COR toxin following inoculation with 1 × 106Pst DC3000 bacteria/ml (Figure S7). The other type A family members had low expression in the samples tested, and in most cases were scored as absent (data not shown). One type B response regulator (ARR14, At2g01760) was repressed by Pst, but in this case repression appears to be dependent on TTSS effectors and not COR toxin. None of the other type B family members were reproducibly differentially regulated. Cytokinin synthesis is mediated by the isopentenyl transferase gene family, and one member of that family, IPT3 (At3g63110), is COR toxin-repressed (Kakimoto, 2003a; Miyawaki et al., 2004). Two cytokinin oxidase genes (CKX4, At4g29740; CKX5, At1g75450; Werner et al., 2003) are induced by TTSS effectors and COR toxin respectively. The repression of IPT3, together with CKX4 and CKX5 induction, suggests a reduction in the levels of cytokinin within the leaves following inoculation. The type A response regulators are known to be rapidly induced by cytokinin, and their repression would be consistent with a fall in the cytokinin levels. Members of the type A response regulator family appear to be negative regulators of cytokinin signaling and may be part of feedback loop repressing induced responses (Kakimoto, 2003b; To et al., 2004). The coordinated repression of this gene family suggests that cytokinin signaling may be de-repressed by Pst during infection.

Conclusion

Although the Arabidopsis–Pst DC3000 interaction was initially developed as a model for studying disease resistance mechanisms (Whalen et al., 1991), this pathosystem is now also used widely for studying susceptible plant–pathogen interactions (Nomura et al., 2005). A comprehensive transcriptional analysis of this interaction using defined Pst DC3000 virulence mutants would, we thought, provide the community with a valuable resource for future understanding of this interaction at whole-plant, tissue, cellular and molecular levels. Indeed, this analysis revealed several interesting results. First, it appears that Arabidopsis plants respond similarly to PAMPs presented on live human and plant pathogenic bacteria, and that flagellin perception is not necessary for the regulation of PAMP-induced genes or bacterial multiplication in the context of Pst DC3000 infection of Arabidopsis under our experimental conditions in which bacteria are infiltrated directly into the leaf intercellular spaces. We found that an impressive number of PAMP-induced genes belong to transcription factors, signaling components, and cell wall- and/or secretion-associated proteins, suggesting coordinate regulation of genes involved in signaling and secretion with those that encode secreted products during basal defense. Second, use of Pst DC3000 mutant strains deficient in COR toxin (Pst DC3118) or COR and the TTSS (the hrpS COR mutant) enabled us to obtain genome-wide knowledge of both distinct and overlapping effects of these two important virulence factors in modulating host gene expression during actual bacterial infection. There is clear evidence for TTSS effector-mediated suppression of basal defense-associated genes. In contrast, many of the SA-regulated defense genes, including some of those associated with basal defense, are induced by TTSS effectors. The most dramatic effect of COR is the induction of JA-regulated genes, as expected, but COR also appears to have a major impact on the expression of genes involved in secondary metabolism. Finally, both TTSS effectors and COR contribute to the dramatic regulation of genes that are responsive to plant hormones, such as auxin, ABA and cytokinin.

We hope that the results described in this paper and the online supplementary materials will facilitate future study of Pst DC3000 pathogenesis in Arabidopsis using a variety of genomic, cell biology and biochemical approaches.

Experimental procedures

Plant growth, bacterial strains and bacteria enumeration

Arabidopsis thaliana Col-0 gl1 plants were grown in soil in growth chambers with a day/night cycle of 12 h/12 h, a light intensity of 100 μE, and a constant temperature of 20°C. Four-week-old plants were used for experiments. Each biological replicate contains plants that were grown, inoculated and harvested side-by-side in the growth chamber entirely independent of the other replicates (typically separated in time by several weeks or months). Bacteria for inoculation were grown in low-salt Luria–Bertani broth (Katagiri et al., 2002) to the mid-logarithmic phase at 30°C. Bacterial cultures were centrifuged at 2500 g for 15 min to recover bacteria, which were resuspended in sterile water to a final OD600 of 0.002 (equivalent to 1 × 106 bacteria/ml), 0.1 (equivalent to 5 × 107 bacteria/ml) or 0.2 (equivalent to 1 × 108 bacteria/ml). Dilution plating was used to confirm the number of bacteria present in the inoculum. The mock inoculum was water containing 0.004% Silwet L-77 surfactant (Osi Specialties, Friendship, WV, USA). Fully expanded leaves were vacuum-infiltrated with bacterial suspensions, and in planta bacteria were enumerated as described by Katagiri et al. (2002). The mean values of the bacterial populations are plotted, with the standard deviation displayed as error. The leaves of 10–15 plants grown in several different pots were harvested for each RNA sample used for microarray and RNA blot analysis. The regulation/secretion-defective PsthrpS mutant, the Pst DC3118 CORhrpS double mutant, the secretion-defective Pst hrpA mutant, E. coli O157:H7 and the E. coli TUV86-2 fliC mutant have been previously described (Berin et al., 2002; Hauck et al., 2003; Hayashi et al., 2001; Yuan and He, 1996). Inoculation with E. coli strains was staggered with the 108 data set such that replicate number 1 of inoculations with E. coli O157:H7 and TUV86-2 fliC was performed alongside replicate number 2 of the other 108 inoculations. To complete three replicates for each E. coli strain, we performed a final set of inoculations consisting of E. coli O157:H7, E. coli TUV86-2 fliC and mock inoculum. Therefore, in the publicly available data set in the NASCArrays database (http://affymetrix.arabidopsis.info/narrays/experimentbrowse.pl), the E. coli inoculations are annotated as replicates 2, 3 and 4 to correspond with the other 108 inoculation data to which they are directly compared.

RNA isolation and labeling

Total RNA was isolated from Arabidopsis leaves using the RNAgents total RNA isolation system (Promega, Madison, WI, USA). The RNA concentration was determined by absorbance at 260 nm, and its quality was evaluated by separation on 2% formaldehyde denaturing agarose gels. The total RNA was further purified using RNeasy minicolumns (Qiagen, Valencia, CA, USA). Biotinylated target complementary RNA (cRNA) was prepared from 16 μg of total RNA using the procedure described by the manufacturer of the GeneChip array (Affymetrix, Santa Clara, CA, USA). Briefly, a primer encoding a T7 RNA polymerase promoter sequence fused to (dT)24 (Genset Oligos, La Jolla, CA, USA) was used for double-stranded cDNA synthesis using SuperScript II reverse transcriptase (Invitrogen, Carlsbad, CA, USA). The resulting cDNA was transcribed in vitro using the BioArray High-Yield RNA Transcript Labeling Kit (Enzo Biochem, New York, USA) in the presence of biotinylated UTP and CTP to produce biotinylated target cRNA.

Affymetrix GeneChip hybridization and data collection

The labeled target cRNA was purified, fragmented, and hybridized to Arabidopsis ATH1 GeneChip arrays according to protocols provided by the manufacturer (Affymetrix) in a Hybridization Oven model 640 (Affymetrix). The GeneChips were washed and stained with streptavidin–phycoerythrin using a GeneChip Fluidics Station model 400, and then scanned with a Gene Array Scanner (Hewlett-Packard, Palo Alto, CA, USA).

Affymetrix GeneChip data analysis

The gene expression data were normalized and analyzed using the Affymetrix Microarray Program Suite (MAS 5.0 statistical algorithms and the data mining tool, version 2.0). Output from all GeneChip hybridizations was scaled globally such that its average intensity (TGT value) was equal to an arbitrary target intensity of 500, allowing comparisons between GeneChips. Data was then compared between sample chips from the same biological replicate producing a signal log2 ratio (SLR) calculated from the GeneChip fluorescence signal intensity data. The software was used to determine whether there was a genuine change in mRNA accumulation (change call, D for decrease, I for increase) and change call P-value. SLRs, change calls and P-values were determined for each bacterially inoculated sample compared with its corresponding mock control or bacterial mutant-treated sample.

Six pairwise comparisons were made within the 1 × 106 bacteria/ml data set, three pairwise comparisons for the 5 × 107 bacteria/ml data set and eight pairwise comparisons for the 1 × 108 bacteria/ml data set (see Table 1). We used significance analysis of microarrays (SAM; Tusher et al., 2001) to identify significant genes based on their differential expression between our set of samples. The SAM analysis feature contained in the TIGR Microarray Experiment Viewer version 3.1 (Saeed et al., 2003; http://www.tm4.org/mev.html) was used to conduct the analysis. The normalized signal log2 ratios for each replicate of each comparison were used for the analysis. Each inoculation level data set (7 h post-inoculation, 1 × 108; 10 h post-inoculation, 5 × 107; 24 h post-inoculation, 1 × 106) was independently analyzed using the multi-class design, with the default settings. The delta value was set at a level such that the median false discovery rate was less than 5%. This analysis identified a non-redundant list of 3864 probesets. Microsoft Access database management software (Microsoft, Redmond, WA, USA) was used to further filter and query the data. Reproducibly differentially expressed probe sets were selected from this list of 3864, based on a signal log2 ratio of at least −1.0, a gene expression change call of D (decrease) and P-value > 0.995 or signal log2 ratio of at least 1.0, change call of I (increase) and P-value of < 0.005 for all three biological replicates for at least one type of comparison. Probe sets that meet these rigorous selection criteria were further analyzed in detail. To our surprise, not a single probeset was reproducibly differentially regulated in the 1 × 106 bacteria/ml Pst CORhrpS versus mock comparison (Table 1). This result illustrates the stringency of our selection criteria, because one would expect that a few probesets might be reproducibly differentially expressed just by chance. Our failure to identify Pst CORhrpS (PAMP) differentially regulated genes following inoculation with 1 × 106 bacteria suggests that the level of bacteria in this treatment is insufficient to elicit robust, detectable changes in host gene expression 24 h post-inoculation, and supports our rationale of using use higher doses of bacteria to identify PAMP-regulated gene expression in these comparisons.

Hierarchical clustering and viewing was performed using either the Cluster version 3.0 and TreeView programs (Eisen et al., 1998; http://rana.lbl.gov/EisenSoftware.htm) or the TIGR Microarray Experiment Viewer version 3.1. The hierarchical clusters displayed in Figures 1 and 2 were analyzed using the Microarray Experiment Viewer with the default similarity metric settings (Euclidean distance) with the complete linkage clustering method selected. MapMan version 1.6.0 (Thimm et al., 2004; Usadel et al., 2005) was used for analysis of the functional classes and metabolic pathways affected following bacterial inoculation. The 944 COR toxin-regulated genes were identified from the 1 × 106 bacteria/ml Pst DC3000 versus Pst COR comparison, while the 791 TTSS-regulated genes were identified from the 5 × 107 bacteria/ml Pst COR versus Pst CORhrpS comparison using the criteria listed above. The SLRs for the three biological replicates in the selected comparison were averaged and displayed using the MapMan software onto the Metabolism Overview, Secondary Metabolism and Photosynthesis displays. The Nottingham Arabidopsis Stock Centre (NASC) web site (http://nasc.nott.ac.uk/) contains the raw microarray data in a MIAME-compliant format for all of the experiments described (reference NASCArrays-340).

RNA isolation and blot analysis

Total RNA was isolated from Arabidopsis leaves using the RNAgents total RNA isolation system (Promega). The RNA concentration was determined by absorbance at 260 nm, and RNA was separated on 2% formaldehyde denaturing agarose gels. The RNA was blotted onto Hybond N+ nylon membranes (Amersham, Piscataway, NJ, USA) using 10X SSC, and UV-crosslinked using a Stratalinker (Stratagene, La Jolla, CA, USA) with the auto-crosslink setting. The cDNA sequences of 15 differentially expressed genes identified by microarray analysis were used for RNA blot hybridization. Fourteen of the sequences were derived from ESTs acquired from the Arabidopsis Biological Resource Center stock center (Columbus, OH, USA). Upon receipt of the ESTs, they were sequenced to confirm their identity. The cDNA probe sequences were liberated from the cloning vector using restriction digestion or PCR amplification with vector primers. The following cDNA sequences were used for labeling and RNA blot hybridization: At3g16370 from EST 175O18T7, At1g12090 from EST 135D13T7, At2g38540 from EST 135H16T7, At1g29670 from EST 240F18T7, At2g19860 from EST E1C7T7, At1g03870 from EST 172M22T7, At3g16240 from EST 206H6T7, At1g33240 from EST 88L16XP, At2g41940 from EST 163F20T7, At5g61590 from EST 215C17T7, At1g72610 from EST 85C7T7, At2g10940 from EST 40D7T7, At5g24780 from EST 114D3XP, At1g52890 from EST 165P18T7, At1g43160 from EST 132C14T7, At5g26340 from EST 181G24T7, At2g24850 from EST 245M18T7, At3g45140 from EST 304E6T7, At2g17500 from EST 178E2T7, At1g62660 from EST 145A20T7, At4g27410 from EST 99D19T7, At4g02380 from EST E5E10T7, At4g02940 from EST 251H23T7, and At3g16390 from EST 215J18T7. A 1.15 kb cDNA sequence for CLH1/CORI1 (At1g19670) was amplified with RT-PCR using the following primers: 5′-CAGAATTCAACACAACTCTTTAATTATC-3′ and 5′-ATCTCGAGTAACAAATGTTTTGATCGAG-3′. The amplified product was cloned into the EcoRI and XhoI sites of pBluescript SK+ and sequenced. cDNA sequences from the above listed clones were labeled with 32P-dCTP using the Stratagene Prime-It II random primer labeling kit. RNA blot hybridizations were performed using PerfectHyb Plus hybridization buffer (Sigma, St Louis, MO, USA) following the manufacturer's protocol, and then washed with 0.5 x SSC and exposed to film for 6h to 4 days, depending on signal produced for each transcript.

Acknowledgements

We gratefully acknowledge Annette Thelen and the Genomic Technology Support Facility at Michigan State University for valuable assistance with the Affymetrix microarray analysis, Dr Thomas Whittam (Department of Microbiology, Michigan State University) and Dr Arlette Darfeuille-Michaud (Pathogenie Bacterienne Intestinale, Clermont-Ferrand, France) for providing E. coli strains, and Dr Thomas Whittam for providing lab space for work with human pathogenic bacteria. Arabidopsis EST clones used for RNA blot hybridization were obtained from the Arabidopsis Biological Resource Center. This work was supported by research grants from the US Department of Energy (DE-FG02-91ER20021) and National Institutes of Health (1R21AI060761-01) to S.Y.H., a postdoctoral fellowship to R.T. from the US Department of Agriculture, a US Department of Education GAANN Graduate Fellowship and a Michigan State University Plant Science Graduate Fellowship to W.U., and funds from the Center of Microbial Pathogenesis at Michigan State University to S. Y. H. and Thomas Whittam.

Ancillary