Analyzing the Transcriptomes of Two Quorum-Sensing Controlled Transcription Factors, RcsA and LrhA, Important for Pantoea stewartii Virulence

The Gram-negative proteobacterium Pantoea stewartii subsp. stewartii causes wilt disease in corn plants. Wilting is primarily due to bacterial exopolysaccharide (EPS) production that blocks water transport in the xylem during the late stages of infection. EsaR, the master quorum-sensing (QS) regulator in P. stewartii, modulates EPS levels. At low cell densities EsaR represses or activates expression of a number of genes in the absence of its acyl homoserine lactone (AHL) ligand. At high cell densities, binding of AHL inactivates EsaR leading to derepression or deactivation of its direct targets. Two of these direct targets are the key transcription regulators RcsA and LrhA, which in turn control EPS production and surface motility/adhesion, respectively. In this study, RNA-Seq was used to further examine the physiological impact of deleting the genes encoding these two second-tier regulators. Quantitative reverse transcription PCR (qRT-PCR) was used to validate the regulation observed in the RNA-Seq data. A GFP transcriptional fusion reporter confirmed the existence of a regulatory feedback loop in the system between LrhA and RcsA. Plant virulence assays carried out with rcsA and lrhA deletion and complementation strains demonstrated that both transcription factors play roles during establishment of wilt disease in corn. These efforts further define the hierarchy of the QS-regulated network controlling plant virulence in P. stewartii.


Introduction
Pantoea stewartii subsp. stewartii (P. stewartii) is a Gram-negative rod-shaped, gamma-proteobacterium that belongs to the Enterobacteriaceae family containing plant-associate enterics (eg. Erwinia, Dickeya, and Pectobacterium spp.) and human enteric pathogens associated with plants (eg. Escherichia and Salmonella spp.). P. stewartii is the causative agent of Stewart's wilt in maize [1]. It is transmitted passively from the gut of the corn flea beetle, Chaetocnema pulicaria, to the plant leaves via the feeding process of the insect vector [2]. The symptoms of the disease include water-soaked lesions in the leaf when the bacterium is in the apoplast during the early stages of with 1.5% agar and P. stewartii strains were grown in either LB or Rich Minimal (RM) medium (1X M9 salts, 2% casamino acids, 1 mM MgCl 2 , and 0.4% glucose). The growth medium was supplemented with nalidixic acid (Nal, 30 μg/ml), ampicillin (Ap, 100 μg/ml), kanamycin (Kn, 50 μg/ml), chloramphenicol (Cm, 30 μg/ml), or streptomycin (Str, 100 μg/ml) as required (see Table 1). P. stewartii strains were grown at 30°C, while E. coli strains were maintained at 37°C.

Construction of markerless deletion mutant strains
Chromosomal deletions of lrhA and rcsA were constructed based on the Gateway system (Life Technologies, Grand Island, NY) and suicide vectors. Two 1kb fragments from upstream and downstream of the desired deletion region were first separately amplified using the primers designated in S1 Table. After that, a two-step PCR reaction was performed in which the upstream and downstream segments for the specific gene were joined together and primers 1kbUPF-attB1 and 1kbDNR-attB2 (S1 Table), specific to each construct, were added to facilitate the Gateway BP reaction (Life Technologies). The final PCR products were cloned into the pGEM-T vector (Promega, Madison, WI) for sequencing before being transferred to the Gateway plasmids using BP Clonase II enzyme mix and then to the suicide vector pAUC40 [19] using LR Clonase II enzyme mix (Life Technologies). The resulting plasmids were transformed into competent E. coli DH5α λ pir cells [25]. A tri-partite conjugation was used to transfer the suicide vector constructs into P. stewartii DC283 [26] using E. coli strain CC118λpir [27]

Strains Genotype and notes a References
Pantoea stewartii strains DC283 Wild-type strain; Nal r [26] ΔlrhA Unmarked deletion of lrhA coding sequence; Nal r This study ΔlrhA/lrhA + DC283 ΔlrhA with chromosomal complementation of lrhA and its promoter downstream of glmS; Nal r Cm r This study

ΔrcsA
Unmarked deletion of rcsA coding sequence; Nal r This study ΔrcsA/rcsA + DC283 ΔrcsA with chromosomal complementation of rcsA and its promoter downstream of glmS; Nal r Cm r This study

Construction of chromosomal complementation strains
Complementation strains were constructed by generating a chromosomal insertion of the promoter and coding regions of the target gene into the neutral region downstream of glmS on the P. stewartii chromosome using the pUC18R6K-mini-Tn7-cat vector system developed by Choi et al [30]. Specifically, primer sets with either EcoRI and XhoI or SacI and SpeI sites (S1 Table) were used to amplify the target regions and clone them into pGEM-T (Promega) for sequencing confirmation. DNA fragments of interest were moved into the pUC18R6K-mini-Tn7-cat vector using double digestion with EcoRI and XhoI or SacI and SpeI (New England BioLabs (NEB), Ipswich, MA), followed by ligation with T4 DNA ligase (NEB) and transformation into DH5α λpir or S17-1 λpir [31]. These transformants served as the donor in the conjugation process with the appropriate deletion strain of P. stewartii as the recipient. Colony PCR reactions using one primer in the inserted gene and two primers flanking the glmS region [32] were conducted to screen for the presence of the chromosomal insertion. DNA sequencing of appropriate PCR products was used to validate the integrity of the complementation strains.

Phenotypic capsule production assay
Wild-type, ΔrcsA and ΔrcsA/rcsA + strains were grown in LB supplemented with the appropriate antibiotics overnight at 30°C with shaking. The overnight cultures were used to inoculate fresh LB to an OD 600 of 0.05 and grown at 30°C to OD 600 of 0.2. The strains were then cross streaked on agar plates containing 0.1% casamino acids, 1% peptone, 1% glucose (CPG) and 1.5% agar [12]. Capsule production was assessed qualitatively after 48 hours of incubation at 30°C by visually comparing the surface appearance of the streaks.

Phenotypic surface motility assay
Swarming motility for the wild-type, ΔlrhA, and ΔlrhA/lrhA + strains was investigated under strict conditions to ensure a reproducible phenotype. Overnight cultures were diluted to an OD 600 of 0.05 in LB broth and grown to an OD 600 of 0.5. Five μl of cell culture at OD 600 of 0.5 were spotted directly onto the agar surface of LB 0.4% agar quadrant plates supplemented with 0.4% glucose [21], which were poured on the same day of the experiment. Plates were put at room temperature for 30 min to 1 hour before incubating them lid-up in a closed box with a flat bottom inside the 30°C incubator. Pictures of the plates were taken after 48 hours of incubation.

Transcriptome analysis methods
The RNA-Seq method for analyzing the transcriptome of the wild-type P. stewartii DC283 strain and the two strains each carrying a deletion of one of the two genes, lrhA or rcsA, has been previously published [15]. Briefly, RNA was extracted from duplicate samples of each strain separately grown in RM medium using a Qiagen (Valencia, CA) miRNeasy RNA

RNA-Seq Data analysis and qRT-PCR validation
Data in the form of fastq files was received and aligned to the P. stewartii DC283 version 8 draft genome from NCBI (NZ_AHIE00000000.1) using the Geneious 7.0 software with the default "low sensitivity" settings, which also counted the numbers of reads aligning to each protein coding gene (excluding the highly repeated genes annotated as "transposase" or "IS66 ORF2 family protein"). Microsoft Excel was then used to compute an expression level for each gene by simply normalizing the read counts for each gene to the total number of mapped reads per million mapped reads (RPM) for each sample. Then the ratio of the RPM expression values (averaged over the two biological replicates) was used to compare the wild-type and deletion strain gene expression levels. (The raw read counts and RPM normalized expression levels for each of the four samples are available in the NCBI GEO database: Accession # GSE69064.) This analysis was used to select genes for qRT-PCR confirmation and validation of the RNA--Seq data. The validation genes were required to have (1) a greater than a four-fold change in average expression (RPM) between the two strains (wild-type and deletion), (2) greater than 100 reads mapped to the reference coding sequence in at least one of the samples and (3) reproducible levels of expression (< two fold change) in duplicate trials. The Bioconductor R software package "DESeq" [33] was also used to analyze the raw read counts using a more sophisticated gene expression normalization and error model to calculate multiple testing adjusted pvalues to estimate the statistical significance of detected gene expression changes. The fold changes (DESeq foldchange) determined by this second method were very similar to our Microsoft Excel analysis for the genes with 4-fold or greater change and the adjusted pvalues (DESeq padj) for those genes selected for qRT-PCR validation were all less than 0.012 (Tables 2 and 3). The qRT-PCR method used for RNA-Seq validation has been previously described in Ramachandran et al [15]. Briefly, each strain was grown in the same manner as for RNA-Seq. RNA was extracted using a miRNeasy RNA extraction kit (Qiagen) and converted to cDNA using the ABI High Capacity cDNA Reverse Transcription kit (Thermo Fisher Scientific, Waltham, MA). Primers (S2 Table) designed using Primer Express software (ABI) were optimized and used to amplify~100 bp regions of each gene of interest to determine the abundance of each transcript. The Pfaffl method was used to compare the wild type versus mutant abundance of a transcript to determine the fold regulation [34].

GFP fusion construction and testing
A transcriptional fusion between the rcsA promoter and the gene for green fluorescent protein (GFP) was created using traditional molecular techniques. The rcsA promoter is located within the region -600 bp upstream of the annotated translation initiation codon [35]. The restriction sites EcoRI and KpnI were added to the 5' and 3' ends of the promoter sequence, respectively, through the PCR primers (S1 Table). The PCR-amplified promoter fragment was ligated into pGEM-T (Promega) and sequenced. After restriction digestion of the pGEM-T construct and pPROBE'-GFP-[tagless] vector [36], a ligation produced the final pPROBE'-GFP-[tagless] vector containing the rcsA promoter. E. coli DH5α was transformed with this plasmid construct which was then moved into wild-type P. stewartii DC283 via conjugation using a triparental mating with the pEVS104 helper plasmid. The conjugation plates were scanned using a Typhoon Trio Scanner (GE Healthcare, Pittsburgh, PA) set to use the blue laser to screen for GFP production. Subsequently, the ΔlrhA and ΔlrhA/lrhA + strains were conjugated to receive the same P rcsA pPROBE'-GFP-[tagless] vector creating the desired reporter strains.
The transconjugates were grown in RM medium supplemented with Nal and Kn, overnight to an OD 600 < 0.5 and then diluted in fresh RM to an OD 600 of 0.025. The cultures were allowed to grow at 30°C with shaking at 250 RPM to an OD 600 0.5. The GFP production was monitored in 96-well plates using a Tecan Infinite 200 (Durham, NC) set with 485 excitation and 535 emission filters. The three strains were each analyzed in triplicate for one experiment, and the three wells were averaged together to establish the mean fluorescence. The average fluorescence reading from the blank was subtracted from the average fluorescence of each strain tested to remove background signal due to the medium. The normalized fluorescence was then divided by the OD 600 for the sample to yield relative fluorescence readings/ OD 600 . The three average relative fluorescence readings/ OD 600 from three different experiments were averaged together and the overall standard error and two-tailed homoscedastic Student's t-test values calculated.

Plant virulence assay
The procedure for conducting the virulence assays with P. stewartii strains in Zea mays seedlings was adapted from von Bodman et al. [12] with some modifications. Sweet corn seedlings (Zea mays cv. Jubilee, HPS Seed, Randolph, WI) were grown in Sunshine mix #1 soil in an growth chamber (Percival Scientific, Inc., Boone, IA) at 28°C, 80% relative humidity, 16

Accession numbers
The read data for the pairs of duplicate samples for the P. stewartii wild-type, ΔrcsA, and ΔlrhA strains, have been deposited in the NCBI Sequence Read Archive (SRA) with accession numbers, GSM1691841, GSM1691842, GSM1691843, GSM1691844, GSM1691845 and GSM1691846, respectively. An Excel file summarizing the differential gene expression in total counts and normalized reads per million (RPM), using the P. stewartii DC283 version 8 NCBI gene annotations, has been deposited in the NCBI Gene Expression Omnibus (GEO) database (GEO Accession GSE69064).

Deletion of rcsA or lrhA impacts phenotypic outputs
Cross streaks of the P. stewartii wild-type, ΔrcsA and ΔrcsA/rcsA + strains on CPG agar demonstrated that deletion of rcsA yielded an easily visible decrease in the level of capsule production ( Fig 1A). Chromosomal complementation of rcsA restored capsule synthesis to levels similar to the wild type ( Fig 1A). Separately, the P. stewartii wild-type, ΔlrhA and ΔlrhA/lrhA + strains were analyzed for motility. Under the test conditions employed, the wild-type strain exhibited unidirectional expansion the majority of the time as described by Herrera et al [21] (Fig 1B), but sometimes a more symmetrical expansion was observed. The ΔlrhA strain exhibited a different phenotype with a noticeably smaller occupied surface area (Fig 1B). Two chromosomally complemented strains of the ΔlrhA strain were constructed because a smaller promoter region (~600 bp upstream of lrhA coding region) strain could not complement the deletion (data not shown) whereas the strain with almost the entire intergenic region upstream of lrhA gene (~921 bp upstream) could ( Fig 1B). This suggests that there are critical regulatory elements more than 600 bp upstream in the lrhA promoter region.

RNA-Seq analysis reveals the RcsA and LrhA regulons
RNA-Seq data was acquired in duplicate for the P. stewartii wild-type, ΔrcsA, and ΔlrhA strains to analyze the global impact of the two regulators. Each strain yielded 17 to 19 million reads that were mapped to the protein coding genes on the P. stewartii DC283 genome (NZ_A-HIE00000000.1). The normalized gene expression data (RPM) from the two trials was averaged and then the deletion strains were compared to the wild-type strain to determine the changes in gene expression when each transcription regulator was absent (Fig 2). Genes with decreased expression in the deletion strains were considered to be positively regulated, either directly or indirectly, by RcsA or LrhA, respectively, in the wild-type strain. Conversely, genes with enhanced expression in the deletion strains were considered to be negatively regulated by the presence of LrhA or RcsA in the wild-type strain. A conservative four-fold change in RPM gene expression level was used as a selection criterion for genes to be further analyzed. An additional DESeq analysis supported the initial RPM data analysis. RcsA activates 11 genes and represses seven genes four-fold or greater (Table 2) and LrhA activates three genes and represses 23 genes four-fold or greater ( Table 3). The most highly regulated RcsA-regulated genes are primarily related to capsule production, an activity previously shown to be under RcsA control [4]. The most highly regulated genes in the LrhA regulon are mostly hypothetical proteins or proteins with just putative gene function. Interestingly, a 3.04-fold repression of rcsA by LrhA was detected during the RNA-Seq analysis hinting at a possible coordination of regulation between RcsA and LrhA.

qRT-PCR validates the RNA-Seq data
Following the initial analysis of the RNA-Seq results, five target genes were selected from each of the putative RcsA and LrhA regulons to validate the RNA-Seq data via qRT-PCR. Some of the most highly regulated genes (Tables 2 and 3), which do not code for hypothetical proteins, were chosen for validation of the RNA-Seq data. The rcsA gene was also included in the analysis as its three-fold regulation by LrhA represented a potential feedback loop in the downstream QS system. The five genes tested for the RcsA regulon were: wceG2, wza, argC, CKS_3504, and CKS_2806. The five genes tested for the LrhA regulon were: CKS_3793, CKS_0458, CKS_5208, CKS_5211, and rcsA. The RNA-Seq data trends for the ΔrcsA and ΔlrhA strains were successfully validated by all five genes tested via qRT-PCR (Figs 3 and 4). Although there was variability in the absolute values of the fold changes between the RNA-Seq and qRT-PCR results, the regulation for a given gene shows similar trends (activation or repression) using both approaches. The qRT-PCR data also confirmed the initial observation from the RNA-Seq data that LrhA is repressing rcsA expression about three-fold. The interplay between these specific components of the P. stewartii QS system was thus further examined to build a more robust model of interactions.

RcsA is controlled by LrhA
A rcsA promoter-GFP reporter transcription fusion was used to analyze expression of rcsA in P. stewartii DC283 wild-type, ΔlrhA and ΔlrhA/lrhA + strains (Fig 5). These experiments were performed at an OD 600 of 0.5, the same growth conditions used for the RNA-Seq and qRT-PCR analysis. The relative fluorescence units (RFU) produced by the ΔlrhA strain were significantly more (p< 0.05) than either the wild-type strain or the ΔlrhA/lrhA + strain, suggesting LrhA represses rcsA either directly or indirectly. The results of the GFP assays further verified that LrhA negatively regulates rcsA, confirming the existence of a previously unrecognized feedback loop in the regulatory circuitry downstream of EsaR in the QS network; it is now recognized that both EsaR and LrhA negatively regulate rcsA.

RcsA and LrhA are involved in plant virulence
The virulence of the ΔrcsA and ΔlrhA strains was compared to the wild type and respective complemented strains in a xylem-infection system. At day twelve post-infection, plants infected with the wild-type strain exhibit typical symptoms of the wilting stage and had an average disease severity score of~3.5 while negative control plants infected with PBS were healthy with a score of~0. The ΔrcsA strain also did not cause any severe disease symptoms with a score of~0, while the ΔlrhA strain expressed an intermediate level of virulence with a score of~1.75 (Fig 6). The ΔrcsA/rcsaA + strain partially restored the virulence of P. stewartii, whereas the ΔlrhA/lrhA + strain fully complemented virulence. Both RcsA and LrhA clearly play roles in the pathogenicity of P. stewartii.

Discussion
Previous studies demonstrated that the quorum-sensing master regulator EsaR directly represses rcsA and directly activates lrhA transcription [15,35]. As anticipated, phenotypic studies examining the impact of deletion of the genes encoding RcsA or LrhA resulted in noticeable effects on the production of capsule and the motility of P. stewartii cells, respectively. These important second-tier transcription factors downstream in the QS regulon were further examined for their role in global gene regulation via RNA-Seq analysis of gene expression in each of the two deletion strains compared with the wild-type strain. A qRT-PCR analysis confirmed that five genes for each deletion were regulated in the same manner shown by the RNA-Seq data. This allowed for increased confidence in the RNA-Seq data and the ability to draw more solid conclusions about the transcriptomes and the downstream network of gene regulation controlled by RcsA and LrhA.
Many of the most highly activated genes found in the RcsA regulon were related to capsule production: wceG1, wceG2, wza, wzb, wzc, wceO, and wceF (Table 2). Previous work has demonstrated that RcsA directly activates the promoters of the cps gene cluster in both E. coli and P. stewartii [4,16,19]. In E. coli, RcsA is one of two colonic acid capsular polysaccharide transcriptional activators, it is also self-activating, degraded by Lon-proteases, and requires RcsB to activate cps genes [16,37]. Similarly, in P. stewartii RcsA has been shown to activate genes required for EPS production as well as to self-activate its own gene [16,35]. The RNA--Seq data also revealed additional genes in the P. stewartii RcsA regulon that are directly or indirectly suppressed by RcsA. Specifically, RcsA repressed several genes in the cys and arg operons more than four-fold ( Table 2). The cys genes are involved in sulfate activation, which leads to cysteine biogenesis in E. coli [38]. The genes cysD and cysN are in an operon and each encodes a subunit of the sulfate adenyltransferase complex [39,40]. The arg genes are involved in arginine biosynthesis in E. coli [41,42]. The three arg genes most regulated by RcsA are argB encoding acetylglutamate kinase, argC encoding N-acetyl-gamma-glutamylphosphate reductase, and argI encoding ornithine carbamoyltransferase. It is unclear why RcsA would repress the arginine and cysteine biosynthetic pathways and how this might relate to capsule production in P. stewartii. P. stewartii LrhA is 77% identical at the amino acid level to its E. coli counterpart. In E. coli, LrhA is a LysR-type transcriptional factor negatively controlling motility, chemotaxis, flagellar biosynstheis [20] and type 1 fimbrial expression [43]. However, unlike E. coli, where LrhA highly represses some genes related to flagellar function and chemotaxis by approximately 3-80 fold [20], the P. stewartii RNA-Seq data only indicated a~two-to three-fold level of repression. On the other hand, LrhA does repress expression of two putative fimbrial subunits (encoded by CKS_0458 and CKS_0459) more than four-fold in P. stewartii (Table 3 and Fig 4), while the E. coli K12 strain MG1655 genome (CP009685.1) does not contain homologues of these genes. Thus there are clear differences between the role of LrhA in E. coli and P. stewartii. Interestingly, the P. stewartii fimbrial subunit genes CKS_0458 and CKS_0459 were also previously found to be directly controlled by the master QS regulator EsaR [15]. This is another example of coordinated control of gene expression at more than one level in the P. stewartii QS regulatory network.
Other P. stewartii annotated genes regulated by LrhA include CKS_5208, encoding rhamnosyltransferase I subunit B (RhlB), which is repressed. This enzyme may be involved in surfactant production necessary to facilitate swarming motility. In addition, the gene most highly repressed by LrhA, CKS_5211, is located adjacent to CKS_5208 in the genome. Although originally annotated as a putative alpha/beta superfamily hydrolase/acyltransferase, CKS_5211 has very high homology with rhamnosyltransferase I subunit A (RhlA) of other sequenced strains of Pantoea. This further suggests an important role for LrhA in controlling surfactant production in P. stewartii. One annotated activated gene, CKS_3793, codes for cytochrome d ubiquinol oxidase subunit I, which presumably plays a role during aerobic respiration. The only other annotated gene found to be four-fold or more activated by LrhA was wceO, which is related to capsule synthesis and is also activated by RcsA, suggesting that its regulation is highly complex.
The RNA-Seq, qRT-PCR and transcriptional fusion experiments all confirmed the existence of a previously unknown feedback loop in the quorum-sensing network downstream of EsaR, with LrhA repressing RcsA. A developing model of this network suggests that there is tight coordinate control of key virulence factors in P. stewartii (Fig 7) [14,15,35] with the EsaR, RcsA, and LrhA proteins creating a coherent type three feed forward loop [44]. The repressive role exerted by EsaR on rcsA transcription appears to be reinforced by the negative regulation of rcsA exerted by the EsaR-activated gene product LrhA. Negative control of rcsA expression by LrhA likely helps repress capsule production in the bacterium until the correct temporal point in disease progression. Positive autoregulation of RcsA in P. stewartii [35] will result in higher levels of the protein that escape proteolysis by Lon [16], thereby reinforcing the signal to increase capsule production at high cell density [35]. Normally, QS tightly controls expression of rcsA so that the bacterium only produces high levels of capsule after migration to the xylem has occurred. If capsule is expressed too early, it can actually hinder the ability of P. stewartii to cause disease [4]. Alternatively, the ΔrcsA strain showed no sign of infection in the plant virulence assay. Thus if the bacterium is incapable of producing capsule it also cannot form a biofilm in the xylem. The complement of rcsA however only partially restored the virulence of the deletion strain, perhaps indicating how important precise fine-tuning of rcsA expression is to disease outcome. The ΔlrhA strain significantly decreased the virulence of P. stewartii, but not to the degree seen with the ΔrcsA strain. This may in part be due to the removal of one level of regulation of rcsA, but is also probably due to impacts on the adhesion and motility of the bacterium, an area of on-going research. Since RcsA controls capsule production in P. stewartii and capsule is a main virulence factor for the bacterium, it would make sense for it to be regulated at multiple points in the regulatory network to ensure that the bacteria successfully migrate to the xylem before capsule and biofilm production begins in the corn Expression from the rcsA promoter. A GFP reporter was used to measure levels of transcription from the rcsA promoter. Strains were grown to an OD 600 of 0.5 and the average fluorescence/OD 600 was measured. As indicated by an asterisk, expression from the ΔlrhA strain is significantly higher (p< 0.05) than either the wild-type strain or the ΔlrhA/lrhA + strain using a two-tailed homoscedastic Student's t-test. This indicates that LrhA normally represses expression of rcsA in the wild-type strain. Data represents three experimental samples analyzed in triplicate. Error bars denote standard error.   plant. The tight control of RcsA by QS and the downstream network of transcriptional regulators, including RcsA itself [35] and LrhA, help ensure the precise timing of disease progression.
Supporting Information S1