Genome-Wide Identification of HrpL-Regulated Genes in the Necrotrophic Phytopathogen Dickeya dadantii 3937

Background Dickeya dadantii is a necrotrophic pathogen causing disease in many plants. Previous studies have demonstrated that the type III secretion system (T3SS) of D. dadantii is required for full virulence. HrpL is an alternative sigma factor that binds to the hrp box promoter sequence of T3SS genes to up-regulate their expression. Methodology/Principal Findings To explore the inventory of HrpL-regulated genes of D. dadantii 3937 (3937), transcriptome profiles of wild-type 3937 and a hrpL mutant grown in a T3SS-inducing medium were examined. Using a cut-off value of 1.5, significant differential expression was observed in sixty-three genes, which are involved in various cellular functions such as type III secretion, chemotaxis, metabolism, regulation, and stress response. A hidden Markov model (HMM) was used to predict candidate hrp box binding sites in the intergenic regions of 3937, including the promoter regions of HrpL-regulated genes identified in the microarray assay. In contrast to biotrophic phytopathgens such as Pseudomonas syringae, among the HrpL up-regulated genes in 3937 only those within the T3SS were found to contain a hrp box sequence. Moreover, direct binding of purified HrpL protein to the hrp box was demonstrated for hrp box-containing DNA fragments of hrpA and hrpN using the electrophoretic mobility shift assay (EMSA). In this study, a putative T3SS effector DspA/E was also identified as a HrpL-upregulated gene, and shown to be translocated into plant cells in a T3SS-dependent manner. Conclusion/Significances We provide the genome-wide study of HrpL-regulated genes in a necrotrophic phytopathogen (D. dadantii 3937) through a combination of transcriptomics and bioinformatics, which led to identification of several effectors. Our study indicates the extent of differences for T3SS effector protein inventory requirements between necrotrophic and biotrophic pathogens, and may allow the development of different strategies for disease control for these different groups of pathogens.


Introduction
The type III secretion system (T3SS) is an essential virulence determinant of many bacterial pathogens that delivers effector proteins into host tissues. In plant pathogens such as Pseudomonas spp., Dickeya spp., Erwinia spp., Pantoea spp., Ralstonia spp., and Xanthomonas spp., the T3SS is required for bacterial pathogenicity on host plants and for elicitation of the hypersensitive response (HR), a localized programmed cell death, in non-host plants [1][2][3][4]. The T3SS genes of phytobacteria are induced in the plant apoplast, when in close contact with host cells or in nutritionally poor synthetic T3SS-inducing minimal medium, which is thought to mimic conditions within the plant apoplast [1,2,4].
Expression of the T3SS in bacteria is coordinately regulated by networks of transcription factors in response to environmental stimuli. Based on differences in regulation, T3SSs can be divided into two groups. The Group I system is found in P. syringae, Dickeya, Erwinia, and Pantoea spp., where T3SS genes are activated by HrpL, a member of the extracytoplasmic factor (ECF) family alternative sigma factor [5,6]. The Group II system is found in X. campestris and R. solanacearum, where transcription of T3SSassociated genes is regulated by members of the AraC family of transcriptional proteins [4,[6][7][8].
The T3SS regulatory pathway of D. dadantii 3937 (3937) has recently been elucidated [9][10][11][12][13]. Similar to E. amylovora and P. syringae, the T3SS of 3937 is regulated by a two-component signal system (TCS) HrpX/HrpY, which activates the gene encoding HrpS, a s 54 -enhancer binding protein. The HrpS protein is required for expression of the alternative sigma factor, hrpL. Presumably, HrpL of 3937 up-regulates several T3SS genes downstream in the type III secretion regulon. In 3937, the T3SS genes are also regulated by an alternative pathway in which the RsmA protein down-regulates T3SS genes by affecting the mRNA stability and translation of hrpL. Another TCS GacS/GacA activates an untranslated regulatory RNA rsmB, which sequesters RsmA and neutralizes its effect on hrpL [10,11,14].
Different modes of infection are used by phytopathogens to invade plant hosts. Biotrophic phytopathogens, such as P. syringae, keep the host plant cells alive and rely on living cells for growth and tissue colonization. Necrotrophic pathogens such as D. dadantii, on the other hand, kill plant cells upon tissue colonization by producing toxic proteins or plant cell wall degrading enzymes (PCWDEs), including pectinases [15][16][17][18]. Although it is becoming clear that phytopathogens may engage in either or both of these processes at different times of the infection process, for convenience the above, rather strict, definitions will be used throughout the manuscript [19].
To further expand our knowledge in this area, microarray and bioinformatics analyses were used to investigate HrpL-controlled genes within 3937 on a genome-wide scale, and the interaction of the HrpL protein with promoter regions of T3SS genes was further studied. In addition, the ability of the T3SS to translocate the 3937 effector protein DspA/E was demonstrated using a calmodulin-dependent adenylate cyclase (Cya) as a reporter.

Use of transcriptomics to identify HrpL-regulated genes
Although HrpL is considered to be the major sigma factor controlling many T3SS genes in several biotrophic phytopathogens, limited information is available describing the mechanism by which HrpL, the alternative sigma factor of a necrotrophic pathogen such as D. dadantii regulates T3SS gene expression. To explore the inventory of HrpL-regulated genes in 3937, the transcriptome profile differences between wild-type 3937 and a hrpL mutant WPP96 (Supplementary Table S1) were compared by microarray analysis. The microarray results are based on the geomean of ten normalized expression values derived from three biological replicates and two dye-swap experiments with a technical replicate for each array.
Based on our microarray assays, we identified 63 genes that were differentially regulated between 3937 and WPP96. Approximately 40% of these genes were up-regulated by HrpL with the arbitrary cut-off value of 1.5-fold difference (Table 1 and Table 2; Supplementary Fig. S1, S2). The HrpL-up-regulated T3SS genes are classified into two major categories; 1) T3SS substrate genes dspA/E, hrpA, hrpK, hrpN, and hrpW; and 2) T3SS apparatus genes hrpP, hrcQ, hrpG, and hrpF. In addition, orfC (a potential hrpW chaperone), and two proteins of unknown function located in the hrp cluster were also up-regulated by HrpL (Table 1). Other HrpL up-regulated genes identified in the microarray assay included those with functions related to regulation and chemotaxis, as well as others with unknown function (Table 1; Supplementary Fig.  S1). In our microarray assay, two thirds of the HrpL-downregulated genes have unknown functions, and the remainder are possibly involved in regulation, transportation, metabolism, or stress response (Table 2; Supplementary Fig. S1).
Our microarray data correlated well with the qRT-PCR results. Out of the 16 HrpL-regulated genes examined by qRT-PCR, only gene ASAP15704 (Bcr/CflA subfamily drug resistance transporter) exhibited a microarray expression pattern that did not correspond to the qRT-PCR value. ASAP15704 appeared to be up-regulated in the microarray assay. However, the qRT-PCR result showed down-regulation without significant difference (Table 1). Due to extremely low expression of the dspA/E gene in the hrpL mutant, Ct value in the hrpL mutant background was high (greater than 35) compared to Ct value of 27 in wild-type 3937 background, which causes the large qPCR value variation although the qRT-PCR also indicated that dspA/E is upregulated by HrpL (Table 1, 2). However, even dspA/E was not included in the qRT-PCR and microarray correlation calculation, there is still reasonably high correlation (R 2 = 0.63) between the microarray and qPCR results for the remaining 15 genes (Table 1, 2).

Bioinformatics-based determination of HrpL-regulated genes
To further characterize the set of D. dadantii HrpL-regulated genes, a hidden Markov model (HMM) was used to analyze the hrp box consensus sequence of 3937. Using 69 hrp boxes from four species as training sets (Supplementary Table S2), we identified 73 genes with a potential hrp box, although only 11 genes with HMMER bit scores .8.5 were considered as reliable (Supplementary Table S3).
Among the HrpL up-regulated genes identified in the microarray assay, only hrpA, hrpF, hrpJ, hrpK, hrpN, and hrpW were found to contain a hrp box sequence in their promoter regions (bit score . 8.5). Surprisingly, dspA/E was not included in this list of predicted hrp boxes, which may be due to the presence of a mispredicted CDS in the source genome that would obscure identification of a hrp box (Supplementary Table S3).
HrpL specifically binds to promoter regions of T3SS genes together with RNAP HrpL is suggested to form a complex with core RNA polymerase (RNAP) and bind to the hrp box sequence in the promoter regions of T3SS genes of phytopathogens. However, few reports have demonstrated a direct interaction between HrpL and RNAP at specific regulatory regions for T3SS genes. In our microarray assay hrpA and hrpN were found to be up-regulated by HrpL (Table 1). In addition, the HMM analysis (Supplementary  Table S3) identified putative hrp box sequences in the promoter regions of hrpA and hrpN (putative hrp box is underlined and listed in Supplementary Table S1).
To study the interaction of the HrpL protein with the promoter regions of hrpA and hrpN in 3937, an electrophoretic mobility shift assay (EMSA) was employed. RNAP (Epicentre Technologies, Madison, WI) and the His 6 -tagged HrpL (His 6 -HrpL) were incubated together with digoxigenin (DIG) labeled DNA fragments of hrpA or hrpN promoter regions containing hrp box sequences. Neither the His 6 -HrpL nor RNAP alone showed detectable binding to the hrp box DNAs of hrpA and hrpN (Fig. 1). However, mobility of the DNA fragments was retarded when hrp box DNA of hrpA or hrpN was incubated with RNAP and His 6 -HrpL ( Fig. 1), indicative of the requirement of RNAP in the binding of HrpL to these DNA fragments.
In addition, the specificity of RNAP-His 6 -HrpL binding to the promoter regions of hrpA and hrpN was demonstrated by a competition assay. Unlabeled DNA fragments of hrpA or hrpN promoter regions were able to compete with DIG-labeled hrpA or hrpN promoter regions to form RNAP-His 6 -HrpL -hrpA or RNAP-His 6 -HrpL -hrpN complexes (Fig. 1). Finally, when an internal DNA fragment of the hrpA or hrpN gene was used, it was unable to compete with the labeled hrpA promoter region (Fig. 1A) or labeled hrpN promoter region (Fig. 1B). These results are different from the competition effect observed with unlabeled promoter regions of T3SS genes (Fig. 1), and suggest specific binding between the RNAP-His 6 -HrpL complex and the promoter regions of hrpA and hrpN (Fig. 1).

The T3SS effector DspA/E is translocated into the plant host through a functional T3SS
Among all of the HrpL up-regulated genes identified in the microarray study (Table 1) and HMM bioinformatics prediction (Supplementary Table S3), dspA/E is the only gene confirmed to encode a potential T3SS effector in 3937 [24]. In several phytopathogens, T3SS effectors function inside the host cytosol and interfere with host defenses by targeting the host cytoskeleton and defense system [25]. While several studies have demonstrated that the T3SS in D. dadantii plays a role in pathogenicity during the initial stages of infection [20,21,23,26,27], there has been no direct evidence for the translocation of effectors by this T3SS to the plant cytosol.
The Cya protein is a calmodulin-dependent adenylate cyclase that has been widely used as a reporter to study the translocation of proteins by bacterial pathogens [28][29][30][31]. Cya is active only upon delivery into the host cell cytosol, where it elevates cyclic AMP levels [32]. To investigate the translocation of DspA/E from 3937 into plant host cells, the Cya protein from Bordetella pertussis was used as a reporter [32]. Cya was fused to the C-terminal end of full-length DspA/E, or to the first 315 amino acids of the protein, resulting in DspE-Cya and DspE 315 -Cya fusion proteins, respectively (Supplementary Table 1). The corresponding plasmids were   transformed into a 3937 outC mutant strain A1919, and an outC/ hrcV double mutant strain Ech141. outC encodes an inner membrane protein of the type II secretion system (T2SS), which is responsible for the secretion of PCWDEs, including pectinases. Mutations in outC therefore prevent secretion of these enzymes and, in turn, reduce or eliminate plant cell maceration [33]. hrcV encodes a structural component of the T3SS, and a mutation in this gene is expected to prevent translocation of T3SS-dependent proteins [34]. An outC mutant strain A1919 and an outC/hrcV double mutant strain Ech141 carrying constructs that encode either full-length DspA/E (pDspE-Cya) or the first 315 amino acids of DspA/E (pDspE 315 -Cya) fused to Cya were infiltrated into Nicotiana benthamiana, and the concentration of cAMP (picomoles of cAMP per microgram of protein, 6 standard errors) was determined 7 h post infiltration. Compared with the cAMP concentration (0.0360.01) for N. benthamiana leaves infiltrated with the T2SS/ T3SS double mutant strain Ech141 harboring pDspE-Cya, a significantly higher concentration of cAMP (0.9760.04; p = 0.018, t-test) was observed in N. benthamiana leaves infiltrated with the T2SS mutant A1919 carrying pDspE-Cya (Table 3). Similarly, the concentration of cAMP (0.8060.02) from N. benthamiana leaves infiltrated with A1919 (pDspE 315 -Cya) was significantly higher than the concentration (0.0460.04; p = 0.003, t-test) from Ech141 (pDspE 315 -Cya). These results demonstrate that DspA/E of 3937 is translocated into the cytosol of N. benthamiana in a T3SSdependent manner, and that the 315-amino-acid N-terminal fragment contains the translocation signal.
As a negative control, the N. benthamiana leaves infiltrated with buffer only had a cAMP level of 0.0260.00. In addition, Cya fusion proteins (p20514-Cya and p20422-Cya) were constructed with genes unrelated to T3SS (Supplementary Table S1). p20514-Cya and p20422-Cya contain full length genes that encode a putative cytoplasmic protein and the DNA-binding transcriptional regulatory protein FabR, respectively. No significant amount of cAMP was detected in Nicotiana benthamiana leaves infiltrated with the wild-type strain carrying p20514-Cya or p20422-Cya (0.0160.00 and 0.0160.01, respectively).

Discussion
Many hypersensitive response and pathogenicity (Hrp) outer proteins (Hops) and effectors have been discovered in biotrophic phytopathogens, including P. syringae [5,25,[35][36][37][38]. In this report, microarray analysis with the necrotrophic pathogen 3937 and its hrpL mutant demonstrated that the only known effector, dspA/E, is up-regulated by HrpL. Similarly, only a limited number of T3SSrelated genes were predicted to contain a hrp box based on an HMM search using promoter regions of genes identified in the microarray as being up-regulated by HrpL (Supplementary Table  S3). These genes include hrpK, hrpN, hrpF in the hrpF operon, hrpJ in the hrpJ operon, hrpW in the dspA/E operon, and hrpA in the hrpA operon (Supplementary Table S3; Supplementary Fig. S2). When applied to all 3937 intergenic regions, the HMM predicts only five additional hrp boxes with confidence (Supplementary  Table S3; Supplementary Fig. S2). Although the microarray assay results demonstrated more HrpL-downregulated genes than those upregulated by HrpL (Table 1, 2; Supplementary Fig. S2), none of the HrpL-downregulated genes had a hrp box HMM prediction with a high confidence level ( Table 2; Supplementary Table S4; Supplementary Fig. S2H-I). Nearly half of the HrpL-upregulated genes were annotated as T3SS associated genes in the 3937 genome and about half of the highly confident HMM predicted genes were also identified by microarray (Table 1; Supplementary  Table S4; Supplementary Fig. S2E-F). The correlation among T3SS genes identified by genome annotation, microarray, and HMM prediction indicates that HrpL majorly upregulates T3SS genes, and the genes down-regulated by HrpL may be due to indirect regulation (Table 1, 2; Supplementary Table S4; Supplementary Fig. S2).
In addition, ten T3SS associated genes in the 3937 genome were determined to be up-regulated by microarray with an arbitrary cut-off value of 1.5-fold change. The number of T3SS genes identified by microarray drops significantly from ten to six HrpL-upregulated genes if the array cut-off value is set at a 2-fold difference ( Table 1, Supplementary Table S4; Supplementary Fig.  S2). Although there may be some false-positive HrpL-regulated genes identified using a 1.5-fold cut-off value, the false-negative rate is likely to increase if we apply a 2-fold change cut-off value in this study. In contrast, the number of genes in common among the set of T3SS associated genes in the 3937 genome and the HMM predicted genes did not change when the confidence level of the HMM prediction increased (Table 1, 2; Supplementary Table S4; Supplementary Fig. S2). Therefore, we used a 1.5-fold cut-off value for microarray analysis and a confidence level (HMM score) of 8.5.
If it is to be assumed that all T3SS effectors are preceded by a HrpL binding site, then this would imply that 3937 contains very few potential effectors. An alternative explanation is that some lower confidence hrp box predictions are in fact valid HrpL binding sites. Although the reason for 3937 and related necrorophic pathogens to have so few predicted effectors compared to biotrophic pathogens is unknown, and with the proviso that more may yet be discovered, it is possible that this low number is associated with the different modes of infection employed by necrotrophic and biotrophic plant pathogens. In biotrophic phytopathogens, the T3SS appears to be the major virulence determinant and is required for the bacterium to disarm the host defense responses during bacterial invasion.
In D. dadantii, however, the pectinases secreted from the T2SS play a more important role in pathogenesis [17,39]. Rather than eliminating pathogenicity, most T3SS mutants of D. dadantii show a delayed maceration of plant tissues [21,23,27]. This suggests that the T3SS of D. dadantii functions to promote the initial establishment of the bacterium in host plants during the early stages of infection. This is supported by evidence that the T3SS genes of D. dadantii are expressed during the early stages of infection [9], together with a delay in the expression of pectate lyases [9,40]. However, this hypothesis remains to be tested for D. dadantii as recent transcriptome data from the related pathogen, Pectobacterium atrosepticum, suggest that the T3SS may be under quorum sensing control and, therefore, coordinately expressed later in infection together with the PCWDEs [41]. The finding that the hrp box sequence was mainly found in the T3SS genes identified in our microarray assay, suggests that many of the other genes identified in this array may be indirectly regulated by HrpL. Finally, although some T3SS genes e.g. dspF (ASAP19013) identified in the 3937 genome sequence (https://asap.ahabs. wisc.edu/asap/ASAP1.htm) were differentially expressed, they were below the cut-off value of 1.5-fold change and thus not included in Table 1.
HrpL is an RpoN-dependent ECF family sigma factor that regulates the expression of many T3SS genes, including those encoding effectors and assembly of the T3SS apparatus. It is suggested that HrpL associates with the core RNA polymerase and binds to the hrp box consensus sequence in the promoter regions of T3SS genes to initiate transcription [25]. In this study, a direct interaction between the HrpL protein and the promoter regions of hrpA (Fig. 1A) and hrpN (Fig. 1B) in 3937 was demonstrated by EMSA. The competition assays indicated that the binding between HrpL and the promoter regions of hrpA and hrpN was specific, suggesting that the HrpL protein directly regulates the transcriptional expression of these genes. In addition, the binding affinity of HrpL for the hrp box of different T3SS genes is different. For example, the binding of HrpL to the hrp box of hrpA (Fig. 1A) seems to be stronger than to that of hrpN (Fig. 1B). When added to a mixture of core RNAP and His 6 -HrpL sigma factor, the band intensity (Fig. 1A) was stronger using a hrpA labeled DNA probe than when the same amount (200 fmol) of hrpN labeled DNA probe was used (Fig. 1B). Interestingly, we observed a correlation between the binding affinity of HrpL and the expression level of HrpL-induced genes: the relatively higher affinity of HrpL for the hrpA promoter (Fig. 1A) renders hrpA expression over four-fold higher than that of hrpN based on the microarray data ( Table 1).
The calmodulin-dependent Cya protein is not secreted or translocated by the T3SS, [32], and is inactive inside the bacterial cytosol due to the absence of calmodulin. These properties make Cya a good reporter for T3SS translocation research in bacterial pathogens [28][29][30][31]. To avoid false positive results during our analysis of T3SS translocation as a consequence of plant tissue maceration by pectinases secreted from 3937, an outC mutant was used. This mutant is unable to secrete pectinases through the T2SS [33]. hrcV encodes an inner membrane structural component of the T3SS apparatus and is essential for secretion of effectors [34]. We used single mutants of either outC, or hrcV and a double mutant outC/hrcV to demonstrate that DspA/E from 3937 is delivered into the plant cytosol in a T3SS-dependent manner. In addition, a plasmid containing only the first 315 amino acids of DspA/E was used to show that the T3SS secretion signal is located in the first one-fifth of the N-terminus of the DspA/E protein. This latter result is consistent with other reports, where T3SS secretion signals have been found within the first 20-150 amino acids of T3SS substrates [42][43][44][45].
In summary, using microarrays, hrp binding site prediction, EMSA, and Cya translocation assays, we provide the genomewide study of HrpL-regulated genes in a necrotrophic phytopathogen. Specific binding between the alternative sigma factor HrpL and promoter regions of T3SS genes hrpA and hrpN was observed. Finally, in sharp contrast to the biotrophic pathogen P. syringae pv. tomato, only a single effector (DpsA/E) has so far been identified in 3937, which we have shown to be translocated by the 3937 T3SS into the plant cytosol, where it functions in virulence. The T3SS in this necrotrophic pathogen may thus play a minor role during infection compared to biotrophic pathogens, instead relying more on its ability to physically attack the plant through the production of PCWDEs.

Bacterial strains and plasmids
Bacterial strains, plasmids, and primers used in this study are listed in Supplementary Table S1. Wild-type 3937 and its mutant strains were stored at -80uC in 15% glycerol and grown on Luria-Bertani (LB) agar or T3SS-inducing minimal medium (MM) [9]. Plasmids were isolated from Escherichia coli DH5a using QIAprep Spin Miniprep Kit (Qiagen, Valencia, CA). Plasmids were introduced into bacterial strains by electroporation using a Gene Pulser Electroporation system (Bio-Rad, Hercules, CA). Antibiotics were used at the following concentrations: ampicillin, 100 mg/ mL; chloramphenicol, 50 mg/mL; kanamycin, 50 mg/mL; spectinomycin, 50 mg/mL. The hrpL mutant WPP96 was used in this microarray assay, and this mutant was complemented by plasmid phrpL in a previous study [12].

Construction of mutants
To construct the outC/hrcV double mutant, a 2119-bp BamHI-XhoI fragment containing the full hrcV ORF was amplified using primers hrcV-F and hrcV-R. This fragment was digested by PstI and re-ligated following removal of a 417-bp internal region of hrcV. The recombinant fragment was inserted into pWM91 at the BamHI and XhoI sites, resulting in the plasmid pWM91DhrcV (Supplementary Table S1). The plasmid was introduced into the outC mutant strain A1919, and the outC/hrcV double mutant was generated by homologous recombination and verified by DNA sequencing analysis.

RNA isolation and preparation of labeled cDNA
Wild-type D. dadantii 3937 and WPP96 were grown in T3SSinducing minimal medium for 6 h, and total RNA was isolated using a Bio-Rad Aurum Total RNA Mini Kit as described by the manufacturer (Bio-Rad, Hercules, CA). RNA was quantified using a ND-100 spectrophotometer (NanoDrop Technologies, Wilmington, DE), and the quality was assessed on an agarose gel.
RNA was labeled as previously described [46] with minor modifications. Briefly, 12 mg of total RNA was labeled in a 45-mL labeling reaction with 10 mL of a combination of 1 mL of spike RNA mix (GE Healthcare, Piscataway, NJ), 1.8 mL of 256 amino allyl deoxynucleoside triphosphate (dNTP) mixture [5 mL each of dATP, dCTP, dGTP (100 mM), 2 mL dTTP (100 mM), 6 mL aminoallyl dUTP, and 17 mL SDW], 4.5 mL of dithiothreitol (0.1 M), 2 mL of 1 mg/mL Oligo dT18 primers (Ambion, Austin, TX), and 2 mL of reverse transcriptase in reaction buffer (Ambion, Austin, TX). Primers and RNA were heated to 70uC for 10 min and then cooled on ice following incubation for 2 h at 42uC. The remaining RNA was denatured with 15 mL of 1 M NaOH and 15 mL of 0.5 M EDTA (pH 8.0) at 65uC for 10 min following neutralization with 15 mL of 1 M HCl.
Purification of cDNA was performed as described by the supplier (Qiagen, Valencia, CA). One microliter of the appropriate Cy dye ester suspended in dimethyl sulfoxide (GE Healthcare, Piscataway, NJ) and 2.0 mL of 1 M sodium carbonate were added to the purified cDNA in amber tubes and incubated for 1 h at room temperature in the dark. To the labeled cDNA, 3.0 mL of 4.0 M hydroxylamine hydrochloride was added and incubation was extended for an additional 30 min in the dark. The labeled targets were combined and mixed with distilled water and further purified with a QIAGEN MiniElute column (Qiagen, Valencia, CA). The labeling efficiency was estimated with 1 mL of the cDNA using a NanoDrop ND-100 spectrophotometer.
Hybridization to DNA microarrays and image analysis D. dadantii microarrays (AMADID No. 012716) were developed at the Scottish Crop Research Institute (SCRI), Dundee, UK, through Agilent Technologies, Inc. (Santa Clara, CA) based on the genome sequence of strain 3937. The microarray experiment was carried out as described [46] with minor modifications. Briefly, labeled cDNA was made up to a volume of 80 mL with distilled water, denatured at 98uC for 3 min, and then mixed with 25 mL of control target and 105 mL of 26 hybridization buffer (Agilent, Santa Clara, CA). The hybridization, wash, and dry were carried out in a Lucidea TM Slidepro hybridization chamber (GE Healthcare, Piscataway, NJ). After hybridization for 17 h at 60uC, the arrays were washed for 1 min each with wash solution I (66 SSPE [16 SSPE is 0.18 M NaCl, 10 mM NaH 2 PO 4 , and 1 mM EDTA, pH 7.7], 0.005% N-lauroylsarcosine), followed by wash solution II (0.066 SSPE, 0.005% N-lauroylsarcosine).
Arrays were scanned using a GenePix 4200A scanner (Molecular Devices, Sunnyvale, CA) with appropriate exposure settings for Cy3 (595 nm) and Cy5 (685 nm) at a 10-mm resolution, generating separate TIFF images. Signal intensity and ratios were generated using GenePix Pro 6 software (Molecular Devices, Sunnyvale, CA). Images were imported and aligned with clone position information (Agilent GAL file) using automated and manual grid alignment features. Median spot and individual median background (annulus setting) intensity values were extracted for each wavelength and imported into analysis software.

Microarray normalization and analysis
Microarray data with intensities reproducibly higher than background from 10 array results of three biological replicate experiments were selected for analysis. Following local background subtraction, the signal for each spot was normalized based on the geomean value of the median intensity of all the spots for each array as described [47]. Each ratio (hrpL mutant/3937 wild type) was converted to its log2 value for Significance Analysis of Microarrays (SAM) software analysis [48]. Genes showing differential expression in the microarray assays were determined by setting the number of falsely called genes to less than one for the significant gene list output. Significant genes compared to the geometric mean of the ratios from ten arrays were considered as HrpL-up-regulated if their level of transcription in the hrpL mutant WPP96 was no more than two-thirds of that observed in the corresponding 3937 wild type. Genes were considered HrpLdown-regulated if their level of transcription was at least 1.5-fold higher in the hrpL mutant WPP96 than that of wild-type 3937. The microarray data have been deposited into the GEO database with an accession number of GSE17185.

Real-time qRT-PCR
For independent confirmation of gene expression results obtained with oligonucleotide microarrays, 3937 and the hrpL mutant WPP96 were grown in T3SS-inducing minimal medium. Total RNA from the bacteria was isolated using the TRI reagent method (Sigma, St. Louis, MO) and treated with the Turbo DNAfree DNase kit (Ambion, Austin, TX). These RNA preparations were independent from those used for microarray hybridizations. An iScript cDNA Synthesis Kit (Bio-Rad, Hercules, CA) was used to synthesize cDNA from 0.5 mg of treated total RNA. The Real Master Mix (Eppendorf, Westbury, NY) was used for a real time PCR reaction to quantify the cDNA level of target genes in different samples. The rplU gene was used as the endogenous control for data analysis [10]. The primer pairs used in this study are listed in Supplementary Table S1. Reactions were run and data collected using the Opticon TM 2 system (Bio-Rad, Hercules, CA). All PCR experiments were performed in triplicate, and standard deviations were calculated.

HrpL binding site prediction by HMM
A training set of 69 HrpL binding sites extending from the 236 to 27 positions was compiled from literature sources (supplementary Table S2). These sequences were aligned using Jalview [49] and the resulting alignment used as input to construct hidden Markov models (HMMs) with the hmmbuild and hmmcalibrate packages of HMMer (http://hmmer.janelia.org/). The HMMer package hmmsearch was used to query test data sets and the target genome. Parameters for model construction (background and null models, HMMer version) were optimized by ten-fold crossvalidation on the 69 sequence dataset. An optimal model was indicated, with a bit-score classifier threshold of 8.5. This model and threshold were applied to intergenic regions derived from the 3937 genome sequence obtained from ASAP (http://asap.ahabs. wisc.edu/software/asap/) to predict candidate HrpL-binding sites.
An overnight LB culture of E. coli BL21 (DE3) containing His 6 -HrpL encoding plasmids phrpL-28a was subcultured into 100 mL of LB broth containing kanamycin. Bacterial cells were grown at 37uC with shaking until an OD 600nm of 0.5 was reached. Following addition of IPTG (isopropyl-beta-thio galactopyranoside) to a final concentration of 1 mM, the culture was incubated for an additional 4 h to induce the expression of the his 6 -hrpL gene. The bacterial cells were harvested by centrifugation at 22,0006g, resuspended in 20 mL of water, and lysed by freezing and thawing. The lysate was sonicated on ice and centrifuged. The insoluble pellet was washed twice with a washing buffer (25% [wt/vol] sucrose, 5 mM EDTA, 1% [vol/vol] Triton X-100 in phosphatebuffered saline) and suspended in 5 mL of denaturing solution (6 M guanidinium hydrochloride, 0.1 M Tris-HCl buffer [pH 8.0], 1 mM EDTA [pH 8.0], 0.03 M b-mercaptoethanol). Protein renaturation was initiated by 20-fold dilution in 0.2 M ammonium acetate (pH 8.5). After 48 h of incubation at room temperature, the insoluble protein was removed by filtration through a 1-mm Whatman filter paper and the soluble protein was precipitated with 55% saturated ammonium sulfate overnight. The precipitate was collected by filtration through a 0.45 mm Fisherbrand Nylon filter and solubilized in 2 mL of 10 mM Tris-HCl (pH 8.0). The concentration of purified HrpL protein was measured by the BCA protein assay kit (Bio-Rad, Hercules, CA).
The purified His 6 -tagged HrpL protein was confirmed by Western Blot using mouse anti-His antibody (GE Healthcare, Piscataway, NJ) as a primary antibody and goat anti-mouse Ig(H+L)-HRP (Southern Biotech, Birmingham, AL) as a secondary antibody followed by detection with an Immun-Star TM HRP Substrate Kit (Bio-Rad, Hercules, CA).

Electrophoretic mobility shift assay
Primers were designed to amplify fragments containing the hrp box in promoter regions of hrpA and hrpN. PCR products were labeled with DIG using the Biotin 39 end DNA labeling Kit (Pierce, Rockford, IL). In EMSA, a core RNAP (50 ng) (Epicentre Technologies, Madison, WI) was incubated on ice for 20 min with 400 ng of purified His 6 -HrpL protein. The core RNAP and His 6 -HrpL sigma factor mixture was added into eppendorf tubes containing the DNA probe (200 fmol), 1 mg Poly (dI?dC), DNA binding buffer, 50% Glyerol, 100 mM MgCl 2 , and 1% NP-40 (Pierce, Rockford, IL) to give a final volume of 20 mL. Following incubation at 28uC for 20 min, the samples were loaded onto a 6% native polyacrylamide gel and eletrophoresed in 0.56 TBE. The DNA probes were transferred to an N-positive nylon membrane and detected according to the manufacturer's protocol of chemiluminescent nucleic acid detection module (Pierce, Rockford, IL)

Construction of cya fusion plasmid
The cya fusions were constructed using the Gateway cloning technology (Invitrogen, Carlsbad, CA) as described [30]. In brief, the appropriate PCR products were cloned into the pENTR/SD/ D-TOPO vector (Invitrogen, Carlsbad, CA) to create the entry clone. The PCR products were then transferred by recombination into the destination vector, pCPP3234, which contains the cya gene without the start codon, to create an in-frame fusion of the target gene dspA/E and cya generating the constructs pDspE-Cya and pDspE 315 -Cya.

Quantification of cAMP in planta
Mutants derived from 3937 were infiltrated into leaves of N. benthamiana at a cell density of 0.3 (OD 600nm ). Leaf disks were harvested from the infiltrated areas 7 h post infiltration with a 1cm-diameter cork borer and frozen in liquid nitrogen. Frozen disks were prepared for cAMP quantification as described [30]. The cAMP content in samples was determined with the Correlate-EIA Direct Cyclic AMP Enzyme Immunoassay Kit (Assay Designs, Inc., Ann Arbor, MI), following the manufacturer's instructions. Protein concentration in the samples was determined with Bio-Rad protein assay kit (Bio-Rad, Hercules, CA). cAMP was quantified in triplicate for each sample, and reported in picomoles of cAMP per micrograms of protein with standard errors. This experiment was performed twice, and similar results were observed. Values are a representative of these two experiments.    Table 4). Array_All: all the HrpL-regulated genes from microarray experiment with a cut-off value of 1.5-fold (Table 1, 2). Array_UP: all the HrpL-upregulated genes ( Table 1). Array_DOWN: all the HrpL-downregulated genes ( Table 2). Array_2Fold: all the HrpLregulated genes with cut-off value of 2-fold changes (Bold ones in Table 1, 2). Array_UP_2: HrpL-upregulated genes with at least 2fold changes (Bold ones in Table 1). Array_DOWN_2: HrpLdownregulated genes with at least 2-fold changes (Bold ones in Table 2). HMM_All: all the genes predicted by HMM (Supplemental Table 3). HMM_SIG: HMM predicted genes with HMM score greater than 8.5 (Bold ones in Supplemental Table 3).

Acknowledgments
This work is dedicated to Professor Noel T. Keen. We thank Guy Condemine (Unité Microbiologie Adaptation et Pathogénie,UMR 5240 CNRS-UCB-INSA-Bayer CropScience) and Amy Charkowski (University of Wisconsin) for kindly providing us with the mutants A1919 and WPP96 respectively; Lisa M Schechter and Alan Collmer for providing plasmid pCPP3234 and advice on Cya assay; Eulandria Biddle for critical discussions and reading of the manuscript; and Nicole Perna (University of Wisconsin) for providing access to the annotated D. dadantii genome sequences (https://asap.ahabs.wisc.edu/asap/ASAP1.htm).