Pseudomonas corrugata strain CFBP 5454 and the derivative mutants used in this study are listed in Table 1. They were routinely cultured at 28°C on either Nutrient Agar (NA, Oxoid, Milan, Italy) supplemented by 1% D-glucose (NDA), or Luria-Bertani (LB) agar (Oxoid, Milan, Italy) (Table 1). The pcoR-mutant strain, designated GL2 is a Tn5 mutant (pcoR76::Tn5) (Licciardello et al., 2007); the rfiA mutant (GLRFIA strain) was obtained by insertional mutagenesis using the conjugative suicide vector pKNOCK-Km (rfiA::pKnock) (Licciardello et al., 2009). The complemented mutant strains used in phenotypic tests are listed in Table 1.

Antibiotics were added as required in the following final concentrations: tetracycline, 40 μg ml^-1; gentamicin, 40 μg ml^-1, and kanamycin, 100 μg ml^-1.

For transcript profiling by RNA-seq, mid-logarithmic phase cells grown on nutrient broth (NB, Oxoid, Milan, Italy) were used to inoculate Improved Minimal Medium (IMM) (Surico et al., 1988) at an OD₆₀₀ = 0.05, and incubated in static conditions at 28°C (Scaloni et al., 2004; Licciardello et al., 2012; Strano et al., 2015). In each experiment three separate batch cultivations were performed for each bacterial strain. The Gram-positive bacterium Bacillus megaterium ITM100 and the yeast Rhodotorula pilimanae ATCC 26423 were used as bioindicators of CLP production according to Lavermicocca et al. (1997).

RNA from WT P. corrugata CFBP 5454 as well as GL2 (pcoR76::Tn5) and GLRFIA (rfiA::pKnock) mutants were extracted from cells grown at the early stationary phase (t = 40 h, OD₆₀₀ = 8.9) in IMM at 28°C. Samples from three replicates of each strain grown on separate days and different batches of medium were collected. The cultures were fixed using RNA^TM Protect Bacterial Reagent (Qiagen Inc.) in a ratio of 2 ml of reagent per 1 ml of bacterial culture. Centrifugation was used to pellet the cells (5000 rpm, 4°C, 20 min), and RNA was extracted in RNase-/DNase-free water using the RNeasy Mini Kit (Qiagen Inc.). Total RNA was quantified using micro-spectrophotometry (Nanodrop^TM 2000C, Thermo Scientific^TM, Waltham, MA, United States). The RNA quality was estimated using an Agilent 2100 Bioanalyzer and RNA samples with an RNA Integrity Number (RIN) above 8.0 were selected.

Libraries were prepared for sequencing according to the manufacturer’s instructions (Illumina). Single-end 51 nucleotide sequence reads were obtained using the Illumina HiSeq2000 system at Parco Tecnologico Padano (Lodi, Italy), processed with Casava version 1.8. Raw sequencing reads were quality controlled using FastQC v.0.10.1 and processed with Trimmomatic v.0.32 to remove sequencing adapters and low-quality bases. High-quality filtered reads were aligned against the P. corrugata genome (ATKI01000000). Bowtie v2.2.2 software was used to perform the alignments and generate the corresponding BAM files.

Aligned reads were processed using HTSeq v0.6.1 to extract read counts over the annotated genes for the genome provided. For all samples, the number of raw reads mapping to each gene was normalized based on the total number of input reads (non-rRNA and non-tRNA reads) for that sample. This normalization procedure enabled gene-expression patterns to be compared across strains, within and between experiments. Reads that partially overlapped a gene contributed to its total raw read value. Only genes that had an average of >10 reads in the three replicates for the WT in comparison with the mutants were considered for further analyses.

The read counts for each sample were imported into R and processed using the Bioconductor package EdgeR. Counts values were normalized using the Trimmed Mean of M-values (TMM) method and statistical comparisons of expression levels across different groups were performed using the EdgeR exact test method. For the further analyses, genes with a false-discovery rate of ≤0.05 were selected. We relied on the top 243 differentially expressed genes without any fold change cut-off.

The RNA-Seq data were submitted to the Sequence Read Archive (SRA) under accession number SRP128274.

RNA was extracted from tomato cv. Bacio plants previously inoculated with P. corrugata CFBP 5454. Tomato plants were grown in nursery flats. After germination and during the trials, plants were maintained in a growth chamber with a 16 h/8 h photoperiod and a temperature of 26°C. Tomato plants were pin-pricked on the stem at the axil of the first true leaf with bacterial cells from 48-h culture on NDA (Licciardello et al., 2007). Four days after inoculation, 5 cm of stem portions including the inoculation site was sampled and stored at -80°C. Pools of four stems for each bacteria-inoculated plant were ground in liquid nitrogen and 100 mg of powder processed for total RNA extraction with the RNeasy Plant minikit (Qiagen Inc.), according to the manufacturer’s instructions.

Quantitative Real-Time PCR (qPCR) was performed on 13 genes of the CLP cluster (crpC, grsb_1, grsB_2, dhbF_3, dhbF_4, syrD2, bepE_1, mefA, arpC, crpD, pcoA, pcoB, oprM_3). Three genes belonging to the biosynthetic cluster of alginate (algD, algG, algI) were selected for validation too. Nucleotide FASTA sequences were retrieved from the P. corrugata CFBP 5454 genome and used to design the primer sets useful for qPCR. Primers were designed with Beacon software (Premier Biosoft International Ltd., Palo Alto, CA, United States) and validated by BLAST (Altschul et al., 1990) in order to minimize the mispriming sites in other genomic loci (Supplementary File 1).

After treatment of the RNA samples with DNAse I (Invitrogen, Life Technologies, Italy), 1 μg of total RNA (from three different independent extractions) was used for cDNA synthesis with Superscript III (Invitrogen, Life Technologies, Italy) according to the manufacturer’s protocol. Samples in which reverse transcriptase had not been added were used as negative controls.

Reactions were conducted with the BioRad iQ Cycler and the SYBR^® Select Master Mix for CFX (Applied Biosystem, Life Technologies, Italy) according to the manufacturer’s protocols. To correct small differences in template concentration, the 16S rRNA gene was used for normalization (Conte et al., 2006). Analysis of the dissociation curve ensured that a single product was amplified. cDNA synthesis reaction was performed at 95°C for 15 s, and at 58–64°C for 1 min (for annealing temperatures see Supplementary File 1).

Data were analyzed using the comparative Ct method, wherein the Ct values of the samples of interest were compared to the Ct values of a control. All the Ct values were normalized versus the 16S rRNA gene. The relative expression (RE) values were calculated by the formula RE = 2-[ΔCt(Wt) - ΔCt(mutant)] (Livak and Schmittgen, 2001).

Genome comparative analysis and gene cluster visualization were performed using the Integrated Microbial Genomes & Microbiomes (IMG/M) system¹. The antiSMASH software pipeline (Blin et al., 2013) was used for the automated identification of secondary metabolite biosynthesis clusters. The number of genes differentially regulated was shown in a Venn diagram. Graphical representation of the relationship between intensity (LogCPM) and difference (Log2FC) of transcripts between P. corrugata CFBP 5454 (Wt) versus GL2 and GLRFIA derivatives mutants was done graphically represented using the DEPICTViz, Differential Expression, and Protein InteraCTions Visualization tool (Lima et al., 2016). Functional annotation, which categorizes genes into functional classes, was performed by Gene Ontology (GO) identification developed at the GO Consortium (Ashburner et al., 2000).

Antimicrobial activity of 10× concentrated culture filtrates from bacterial strains grown in IMM and NB were assessed after 4 days of incubation in static conditions, using the well diffusion assay according to Licciardello et al. (2009). Two CLP-sensitive bioindicator strains R. pilimanae ATTC26432 and B. megaterium ITM100 previously grown in layers on top of agar potato dextrose agar plates (PDA, Oxoid, Milan, Italy) were used (Licciardello et al., 2009). The in zone for each antimicrobial compound tested was measured. All tests were carried out at least twice in triplicate.

Total exopolysaccharides (EPSs) were isolated from P. corrugata CFBP 5454 and derivative mutants grown in IMM at 28°C for 4 days. EPSs were also evaluated from WT strain grown on NB. After centrifugation at 16,300 × g for 20 min to remove cells, total EPSs were isolated according to Fett et al. (1996) with slight modifications (Licciardello et al., 2017). Three separate partially purified samples were prepared for each bacterial strain.

Data were analyzed by two-way ANOVA using IBM^® SPSS^® v20. Mean values were compared using the Student–Newman–Keuls test. Statistical significance was established at P ≤ 0.05 and P ≤ 0.001.

To investigate the regulatory functions of P. corrugata LuxR transcriptional regulators PcoR and RfiA, expression profiles from RNA-seq data were analyzed. The transcriptome of P. corrugata strain CFBP 5454 was compared to those of the mutant strains GL2 (pcoR mutant) and GLRFIA (rfiA mutant) (Licciardello et al., 2007, 2009) grown to the early stationary phase in IMM which facilitates CLP production (Scaloni et al., 2004; Licciardello et al., 2009, 2012). Libraries derived from single-stranded cDNAs were sequenced and mapped against P. corrugata CFBP 5454 reference genome (ATKI01000000). Genes with increased or decreased expression in the WT strain compared to the mutant strains were considered to be positively or negatively regulated by PcoR or RfiA.

With a false-discovery rate (FDR) correction of 5%, 152 genes (46 increased and 106 decreased) differed significantly in the pcoR mutant, and 130 genes (52 increased and 78 decreased) differed significantly in the rfiA mutant compared to the parent strain CFBP 5454 (Figure 1A). Overall, the expression of 92 genes, which represent 3% of the annotated genes in the CFBP 5454 draft genome, differed significantly in both pcoR and rfiA mutants (Supplementary Files 2, 3). The remaining 60 (out of 152) and 38 (out of 130) genes were independently regulated either by PcoR or RfiA, respectively (Figure 1A) (Supplementary Files 4, 5). The Supplementary Files contain a thorough analysis of the transcripts and their predicted functions found to be associated with the role of PcoR and RfiA (Supplementary Files 2–5). In order to assemble a catalog of functions strongly linked to these transcriptional regulators, differentially expressed genes for both mutants were grouped based on their GO utilizing GO Consortium². Genes were grouped into 14 functional categories on the basis of PseudoCAP and were plotted with respect to down-regulation and up-regulation (Figure 1B).

The largest group consisted of enzymes associated with transport systems, 34 of which were differentially expressed in the WT strain compared with expression in pcoR mutant and 27 with rfiA mutant. The second largest group were the genes involved in redox and oxidative stress, most of which were down-regulated in both pcoR (17 genes) and rfiA (11 genes) mutants. Genes predicted to be related with alginic acid biosynthesis (12 genes) and secondary metabolite production (11 genes) were well represented among the over-expressed genes in the WT strains in comparison to mutants, thus revealing the predominantly positive control of both PcoR and RfiA in the biosynthesis of these molecules. Transcriptional regulator genes account for a significant number of transcripts affected by pcoR and rfiA mutations, including up- (8 genes) and down- (11 genes) regulated genes that show a wide-ranging control through a cascade of other regulators. Other gene categories affected are involved in carbohydrate metabolic processes, fatty acids, amino acids, and purine and pyrimidine metabolisms.

RNA-seq analysis showed that among the transcripts differentially expressed in both pcoR and rfiA mutants, there are 21 genes putatively involved in CLP production, as ascertained by homology BLAST analysis. The genome of P. corrugata CFBP 5454 was assembled into 156 contigs and NRP genes were located in at least 10 different contigs (Licciardello et al., 2014; Trantas et al., 2015). Therefore, for the in-silico reconstruction of P. corrugata secondary metabolite clusters, we used the annotated sequence of strain LMG 2172^T (also known as BS3649, Genbank accession NZ_LT629798.1) which shared an average nucleotide identity (ANI) of 99.53% with strain CFBP 5454. Using combined AntiSMASH and BLAST analyses, we found that the differentially expressed genes were located in a large HSL-NRPS cluster accounting for approximately 3.4% of the LMG 2172^T genome (Figure 2A, Table 2 and Supplementary File 6).

This large cluster includes six NRPS genes most of which were putatively attributed to two closed CLP biosynthetic clusters for the synthesis of corpeptins, 22 amino acid lipopeptides, and the nonapeptide cormycin A. A similar topology was observed for nunapeptins and nunamycin in P. fluorescens In5 (Michelsen et al., 2015; Hennessy et al., 2017) and for thanapeptins, and thanamycin of Pseudomonas sp. SH-C52 (Mendes et al., 2011; Van Der Voort et al., 2015). In addition, in close proximity to the cormycin cluster, a biosynthetic cluster of five genes was identified, which was highly homologous to the CLP brabantamide cluster described in Pseudomonas sp. SH-C52 (Schmidt et al., 2014; Van Der Voort et al., 2015). This biosynthetic cluster is also present in P. fluorescens In5 (Hennessy et al., 2017). BLAST analysis also revealed that 3 of the 21 genes differentially expressed were within the same open-reading frame (ppsE_1; grsB_1; grsB_2). These include all the three putative corpeptin NRPS genes (tycB, dhbF_4, ppsE_1; Table 2) and the two downstream genes coding for an ABC transporter system (macA, macB2; Table 2 and Figure 2A). The genes ppsE_1, macA, and macB have been demonstrated to be part of the same transcriptional unit known as crpCDE (Strano et al., 2015). Insertional mutants in crpC and crpD were no longer able to produce corpeptins, but still produced cormycin (Strano et al., 2015). CrpC was significantly upregulated in the WT strain compared to the pcoR and rfiA mutants by 5.78 and 5.79 Log-fold changes (LogFC), respectively (Table 2 and Supplementary Files 2, 3).

None of the putative cormycin NRPS genes were detected among the differentially expressed genes. However, two genes coding for an ABC transporter system and a gene annotated as syrD_2 were over-expressed in the WT CFBP 5454 strain in comparison with pcoR- and rfiA-mutant strains (5.28- and 5.51-fold, respectively). In addition, among the differentially expressed genes, we identified the yeaM gene coding for an AraC family transcriptional regulator, in proximity of the putative cormycin NRPS genes, overexpressed 3.05- and 2.93-fold in WT compared to the pcoR- and rfiA-mutant strains, respectively. Four out five genes of the putative brabantamide biosynthesis cluster were differentially expressed. Although the production of this metabolite has not yet been described in P. corrugata, it could be argued that previous experimental conditions prevented it from being detected (Emanuele et al., 1998; Scaloni et al., 2004).

Cell-free culture filtrates of the pcoI- and rfiA-mutant strains grown on IMM medium for all RNA extractions didn’t show antimicrobial activity against the two CLP bioindicators, the yeast R. pilimanae ATCC26423 and the Gram-positive bacterium B. megaterium ITM100 (Figure 3A) as opposed to the parent strain. The antagonistic activity was complemented at the same levels as those of the CFBP 5454 strain by expression in trans of the pcoR and rfiA genes into the respective mutant strains. In addition, the expression in trans of rfiA was sufficient to restore the antagonistic activity of the culture filtrate of the pcoR mutant (Figure 3A).

Figure 3C shows the antimicrobial activity of the living cells of strain CFBP 5454 and also the derivative mutants. It is worth noting that living cells of the two regulatory mutants still demonstrate antimicrobial activity against P. digitatum although to a different extent.

RNA-seq analysis revealed that 12 genes putatively involved in the EPS alginate biosynthesis were upregulated in strain CFBP 5454 compared to both luxR derivative mutants with LogFC ranging from 4.09 (algD) to 1.93 (algG) in comparison with the pcoR-mutant strain, and from 4.09 (algD) to 2 (algG) to the rfiA-mutant strain (Table 2 and Supplementary Files 2, 3).

In the P. corrugata CFBP 5454 genome, these genes were located in an 18 kb region constituting most of the core structural/biosynthetic cluster (contig38, Figure 2B) except for algC which was located elsewhere (contig86) and is not differentially expressed. This cluster encodes the biosynthetic enzymes and membrane-associated polymerization, modifications, and exports proteins necessary for the alginate production. The order and arrangement of the alginate structural gene cluster in P. corrugata is similar to those already described for P. aeruginosa and P. syringae (Fialho et al., 1990; Shankar et al., 1995; Peñaloza-Vázquez et al., 1997). The expression of other genes implicated in alginate regulation and switching phenotype and dispersed in other parts of the genome were not altered in these mutants (data not shown). No other EPS clusters in Pseudomonas strains, i.e., the pel and psl clusters described in P. aeruginosa (Franklin et al., 2011), the epm cluster, responsible for the production of an alginate-like EPS in P. alkyphenolia (Lee et al., 2014) or levane, were detected in the P. corrugata genome, as assessed by BLAST analysis. Total EPSs were isolated after isopropanol precipitation from the supernatant of the pcoR and rfiA mutant and complemented strains growing in IMM and compared to the P. corrugata parent strain CFBP 5454 (Figure 3D). An approximate 10-fold reduction of EPS yield was recorded in both mutants. The production of EPS was almost restored after complementation of the pcoR and rfiA (P ≤ 0.01) (Figure 3D). As P. corrugata has been demonstrated by Fett et al. (1996) to produce alginate as polymannuronic acid and not levan as EPSs, PcoR, and RfiA would seem to play a role in alginate production regulation.

In addition to the genes described above, there are eight genes whose expression was significantly modified in both the pcoR and rfiA mutants. Only two of them, oprM_3 and bepE_1, are significantly down-expressed (LogFC ≥-2 and P-value ≤ 0.005) in the WT in comparison with mutant strains and are predicted to codify for multidrugs efflux systems. The ompM_3 gene encodes a putative outer membrane protein and bepE_1 an efflux pump membrane transporter (Table 2). Although they are located adjacently in the same genomic region, there was no evidence of their function. No gene homologs were found in the genome of the LMG 2172^T. Three adjacently located genes (DIT1_1, gph_2, and DIT1_2) were over-expressed in the WT and showed more than a three-LogFC in transcript abundances compared to both pcoR and rfiA mutants. Analysis of the DIT1_2 putative protein revealed the presence of a DIT1_PvcA superfamily conserved domain, common to pyoverdine/dityrosine biosynthesis proteins. Blastx analysis showed a 40% homology with PvcA protein of P. aeruginosa, involved in the biosynthesis of the paerucumarin, a new metabolite described as an isonitrile functionalized cumarin (Clarke-Pearson and Brady, 2008). DIT1_1 differed from DIT1_2 in terms of an additional conserved domain belonging to the CAS-like superfamily, responsible for clavaminic acid biosynthesis. The gph_2 gene encodes a putative phosphoglycolate phosphatase. Among the most differentially expressed genes (LogFC > 4 and P-value ≤ 0.005), rhbA_1 also needs mentioning. This gene is a putative diaminobutyrate-2-oxoglutarate aminotransferase located in the ornicorrugatin gene cluster of P. fluorescens SBW25 and in histicorrugatin of P. thivervalensis (Cheng et al., 2013; Matthijs et al., 2016).

A total of 60 and 38 differentially expressed genes were identified in either the pcoR- or rfiA-mutant strains, respectively, mainly associated with transport systems, transcriptional regulation, and redox and oxidative stress. Of these, six transcriptional regulators were over-expressed in the WT in comparison to the pcoR mutant, although at low LogFC (0.77–1.10). Three of them belong to the HTH transcriptional regulator family, whose role to the best of our knowledge has not been investigated. Compared to the rfiA mutant in the WT strain, only one HTH regulator was over-expressed and two were down-expressed (LogFC 1) (Supplementary File 4). We observed that most of these genes had a very low LogFC < 1; thus, we decided to focus only genes with a minimum two LogFC. We found a strong overexpression of the traI gene coding for the AHL synthase (pcoI by Licciardello et al., 2007) in the WT compared to pcoR mutant. Without PcoR, the AHL-QS would not be able to work, since it is strictly dependent on the PcoR–AHL complex, and pcoI is only expressed at the basal level. Three genes involved in copper metabolism are among the most over-expressed in the WT compared to the pcoR mutant (Supplementary File 4). The Cyp4d2 gene, which codes for a cytochrome P450 involved in redox and oxidative stress, was differentially expressed only in the pcoR mutant. It is down-regulated in the WT with a LogFC of -2.4 (Supplementary Files 4, 5).

Thirteen genes among those co-regulated by PcoR and RfiA, putatively involved in biosynthesis secondary metabolites (six genes) and transport (seven genes), and three genes putatively responsible of alginate biosynthesis, were selected to validate RNA-seq results. qPCR was carried out with gene-specific primers (listed in Supplementary File 1) and the gene expression of WT versus the mutant strains was analyzed. Although there was a difference in the fold change estimated by the two methods (RNA-seq and qPCR), the expression pattern was the same (Figure 4A). A close correlation (Pearson’s R² = 0.796) was observed between LogFC measured by RNA-seq and qPCR.

The data confirmed the positive regulation of PcoR and RfiA of all the selected genes, and the negative regulation of bepE_1 and opmR_3, which were down-regulated in the WT compared to both mutants, in accordance with RNA-seq data.

Since the RNA-seq experiment relied on conditions known to stimulate CLP production, we investigated the expression of NRP and alginate genes in the P. corrugata CFBP 5454 strain grown in complex undefined medium (NB) and in planta. The results demonstrated that genes involved in NRP biosynthesis and transport were activated two to sixfold more in minimal medium compared to NB, and two to five in planta compared to NB medium (Figure 5). Thus, cell culture filtrates of CFBP 5454 grown on NB showed very little or no activity against B. megaterium and R. piliminae, respectively (Figure 3B). AlgG gene in P. corrugata CFBP 5454 was upregulated both in NB and in planta compared to IMM. Total EPS production was higher in NB (114 ± 14 mg/100 ml) compared to IMM (82 ± 14 mg/100 ml), showing that rich medium provides better conditions for EPS production (Figure 3D).