Bacterial strains, plasmids, and phage used in this study are listed in Table 1. S39006 strains were grown at 30°C and Escherichia coli strains were grown at 37°C in sealed plastic universals containing Lysogeny Broth (LB; 5 g l⁻¹ yeast extract, 10 g l⁻¹ tryptone, and 5 g l⁻¹ NaCl) or on solid LB agar plate (LBA) supplemented with 1.5% (w/v) agar. Bacterial growth was measured in a Unicam Heλios spectrophotometer and expressed as OD₆₀₀. Where required, antibiotics were added at the following final concentrations: kanamycin (Kn), 50 μg ml⁻¹, ampicillin (Ap), 100 μg ml⁻¹, and chloramphenicol (Cm), 25 μg ml⁻¹.

To study growth under aeration conditions, 25 ml LB in 250 ml Erlenmeyer flasks were inoculated with an overnight culture of the test strain to an OD₆₀₀ of 0.05. The culture was incubated at 30°C with shaking at 215 rpm. Sampling was carried out every 2 h. Similar procedures were used to study growth under microaerophilic conditions except that 25 ml mineral oil was overlaid on 25 ml LB and the flask was shaken at 80 rpm. For flotation assays, experiments were carried out as described in Tashiro et al. (2016). For bacterial spot tests, cultures were normalized to OD₆₀₀ of 1 and 10 μl was spotted on LBA. Transduction was carried out as describe by Evans et al. (2010) using φOT8. Transductants were selected on LBA supplemented with Kn.

Transposon mutagenesis of S39006 was carried out as described previously (Monson et al., 2015). Briefly, the plasmid pKRCPN1 was delivered to the recipient by conjugation with E. coli β2163. A mating patch containing a ratio of 3:1 S39006 or LacA ΔpigC and E. coli β2163 was prepared and incubated overnight on LBA supplemented with 300 μM diaminopimelic acid (DAPA). Samples from the patch were serially diluted and plated on LBA plates containing Kn. The plates were incubated at 30°C for 48 h. Transconjugants were screened visually for their colony appearance. S39006 colonies are normally opaque but become translucent when they lose their ability to make GVs (Ramsay et al., 2011). The insertion site of the transposon in translucent mutants was then identified by random priming PCR (RP-PCR) analysis (Jacobs et al., 2003) followed by DNA sequencing across the transposon junction using oligos MAMV1-KRCN1 and MAMV2-KRCN1 (Table 2).

β-glucuronidase (β-glu) activity was measured as described by Ramsay et al. (2011). A 100 μl sample of culture was taken at each time point and frozen at −80°C. The sample was then thawed at room temperature. Phosphate-buffered saline (100 μl) containing 400 μg ml⁻¹ lysozyme, 250 μg ml⁻¹ 4′-methylumbelliferyl-β-D-glucuronide (MUG) was added to 10 μl of sample. The fluorescence emitted by the samples was then immediately monitored (excitation 360 nm, emission 450 nm, cut-off 435 nm, eight reads per well, measured every 30 s for 30 min at 37°C) using a Gemini XPS plate reader and expressed as RFU OD₆₀₀⁻¹. β-galactosidase activity was determined using similar procedure except that MUG was replaced by 4′-methylumbelliferyl-β-D-galactoside.

Phenotypic assays for siderophore, cellulase, pectate lyase, BHL, prodigiosin, and carbapenem production, plus swarming and swimming assays, were performed as described previously (Schwyn and Neilands, 1987; Slater et al., 2003; Williamson et al., 2008; Poulter et al., 2010). In all phenotypic plate assays, overnight cultures of the test strains were adjusted to an OD₆₀₀ of 1.0, 10 μl (or 5 μl for swimming and swarming assays) spotted on appropriate agar plates and incubated at 30°C for 48 h. Indicator plates for BHL detection were prepared by adding 100 μl of Serratia biosensor strain SP19 with 3 ml of molten 0.75% agar. Detection of BHL was indicated by the production of a red halo by the biosensor strain around the test colonies (Poulter et al., 2010). Indicator plates for carbapenem production were made using 0.75% top agar lawns seeded with E. coli strain ESS. The production of antibiotic was indicated by the formation of inhibition zones around the test colonies (Slater et al., 2003). Phenotypic assay plates for cellulase production were developed by flooding with 0.2% (w/v) of congo red for 20 min, bleached with 1 M NaCl and finally stained with 1 M HCl for 5 min. For assessing pectate lyase activity, the agar plate was flooded with 7.5% (w/v) copper acetate for 1–2 h. Enzyme activity was indicated by halos formed around the test strain. Swimming and swarming plates were assessed visually. Swarming behavior varies between plates and therefore images of plates shown are representative of those observed. Comparisons of swarming behavior were always between strains swarming on the same plate to avoid plate-to-plate variation.

Phase contrast microscopy (PCM) images were obtained from wet mounts of bacterial samples from a colony or liquid culture, using an Olympus BX-51 with a 100X oil-immersion lens. For transmission electron microscopy (TEM), a carbon-coated glow discharge grid was treated with 0.01% poly-L-lysine (2 min) and the bacterial sample was attached to the grid for 10 min. The grids were washed twice with dH₂O and stained with 2% phosphotungstic acid (pH 7.0) for 5 min. The grids were viewed using a FEI Tecnai G2 TEM in the Cambridge University Advanced Imaging Facility.

To construct pBAD-rbsR for complementation assays, the rbsR gene was first amplified using oligonucleotide pair, oCML24 and oCML37. Oligonucleotides used for PCR amplification were purchased from Sigma Aldrich and are listed in Table 2. The PCR product and plasmid pBAD30 were digested with KpnI (NEB) and HindIII (NEB) at 37°C for 2 h. The digested PCR product was subsequently ligated into compatibly digested pBAD30 using T4 DNA ligase (NEB) according to the manufacturer's instructions. Plasmid pQE80-rbsR was constructed using oligonucleotides oCML24 and oCML36 to amplify the rbsR open reading frame (ORF). The amplified fragment was digested with KpnI (NEB) and HindIII (NEB) and ligated with compatibility digested pQE80-oriT. For the construction of pBAD-rbsK, oligonucleotides oREM726 and oREM727 were used to amplify the rbsK ORF and the subsequent PCR product was digested with KpnI and HindIII. This was ligated with compatibly digested pBAD33 to form pBAD-rbsK. All plasmids were subjected to sequencing (GATC Biotech) to confirm the sequence was correct.

The pQE80-rbsR plasmid was transferred into S39006 by conjugation using E. coli β2163 and RbsR production was induced by addition of 1 mM IPTG to the bacterial culture. The RbsR protein was purified using Ni-NTA agarose according to the manufacturer's instructions (Qiagen, Germany). The identity and nature of the purified protein was checked by SDS-PAGE and Western blot analysis. Protein concentration was measured by DC protein assay (Biorad) according to the manufacturer's instructions.

DNA probes for EMSA analyses were prepared according to the LUEGO method. The universal “third oligonucleotides” (LUEGO), which was fluorescently labeled at 5′ and 3′ ends, was used to generate many different probes (Jullien and Herman, 2011). To perform EMSA, oligonucleotide mixtures (5:5:1 LUEGO:short:long) were prepared and annealed in a thermocycler by the following profile: 2 mins at 95°C, cooled down to 70°C at 5°C s⁻¹ followed by slow cooling at 0.02°C min⁻¹ to 18°C. The mixture was then diluted to 5 nM and mixed with the indicated concentrations of the RbsR protein. This protein-probe mixture was added to the EMSA reaction buffer (10% glycerol, 10 mM Tris pH 8.0, 1 mM DTT, 10 mM KCl, 0.5 mM EDTA, and 20 μg m⁻¹ BSA) to a total volume of 20 μl. The reaction was incubated on ice for 30 min then 10 μl of the sample were loaded onto a 7.5% native acrylamide 1X TGE (25 mM Tris pH 8.0, 192 mM Glycine, and 2 mM EDTA) gel and subjected to 100 V at 4°C. The gel was scanned using a Typhoon 9400 scanner with the following parameters: Acquisition mode: fluorescence, focal plane: +3 mm, emission filter: 670 BP30 Cy5, PMT: 600, laser: red (635 nm). Where indicated, a non-specific probe, made using oligonucleotides oMC100 and oMC101, was used. Together these oligos form a binding site consisting of 10 repeats of the nucleotides CAT (Table 2).

Potato rotting assays were performed as described by Fineran et al. (2007) to assess the virulence of the rbsR mutant in plants. Potatoes were inoculated with 1 × 10⁶ cfu of S39006 or the rbsR mutant. After 5 days incubation at 30°C, rotted tissue was weighed and compared. To assess colony counts, 0.1 gram of rotten potato tissue (or a normalized amount where the amount recovered was low) was serially diluted in LB. Serial dilutions were plated out onto LB and colony counts assessed. Statistical analyses were performed using a two-tailed t-test; differences were significant if the P-value < 0.05.

The rbsR sequence was compared to those available in Genbank (Benson et al., 2013) using Basic Local Alignment Search Tool (BLAST) (Altschul et al., 1990). Conserved protein domains were analyzed using pfam (Bateman et al., 2004) and Conserved Domain Database (CCD) (Marchler-Bauer et al., 2014). The binding motifs of RbsR were predicted using the MEME suite (Bailey et al., 2009). The amino acid sequences of rbsR from Serratia and closely related strains were used to construct the phylogenetic tree using Molecular Evolutionary Genetic Analysis (MEGA) version 7.0 (Kumar et al., 2016). The evolution history was inferred by the Maximum Likelihood method based on the JTT matrix-based model (Jones et al., 1992). All positions containing gaps and missing data were eliminated. Bootstrap trials were replicated 1,000 times to estimate confidence values of the phylogenetic tree.

To identify new regulators of GV production, plasmid pKRCPN1 was first used to randomly mutagenize S39006. The plasmid contains a Tn5 transposon derivative containing a promoterless lacZ gene and a kanamycin resistance cassette (mini-Tn5KnlacZ1) (Monson et al., 2015). S39006 colonies are normally opaque because of light-refracting GVs in the bacterial cells. The bacterial colonies become translucent when GV production is inhibited and so presumptive GV-defective mutants could be identified based on their translucent colonial morphology on agar plates. However, visual screening for GV-deficient mutants in a strain making the red pigment prodigiosin proved problematic as definition of the translucent phenotype was sometimes ambiguous. To overcome this, S39006 LacA ΔpigC, a non-pigmented mutant carrying a mutation in pigC, was used for further transposon mutant screens.

A comprehensive mutant library containing 71,000 insertion mutants was generated by random transposon mutagenesis. From 67 independent conjugations, 311 putative GV-defective mutants were obtained. The transposon insertion sites in the mutants were determined by RP-PCR. Three of these GV-defective mutants (CML25, CML26, and RA119) had a transposon inserted in a putative LacI family transcription regulator gene (or508 or rbsR) and two (CML33, RA79) had an insertion in the ribokinase gene (rbsK) of the ribose operon (Figure 1B) that is responsible for ribose transport and utilization in some bacteria. In addition to mutations in rbsR and rbsK, the screen also identified transposon insertions in two previously defined regulators of gas vesicle biosynthesis in S39006; namely pigX and smaI (Ramsay et al., 2011). Transposons insertions were also identified in gvpA1 and gvpN—two genes that lie within the gas vesicle biosynthetic cluster and which we showed previously to be essential for robust gas vesicle formation (Tashiro et al., 2016).

The ORF or508 (locus of transposon inserted in mutant strain CML26) of S39006 is predicted to encode the LacI family transcriptional regulator of the ribose operon. The amino acid sequence of or508 exhibited high similarity to several putative LacI family transcriptional regulators from other enterobacteria, including Dickeya dadantii, D. dianthicola, D. solani, D. zeae, D. chrysanthemi, and Pectobacterium atrosepticum (Supplementary Table 1). The or508 gene was designated as rbsR based on similarity of the corresponding gene product with the E. coli version (Mauzy and Hermodson, 1992). The RbsR protein contains two domains, a N-terminal helix-turn-helix (HTH) DNA binding domain of the LacI family (Pfam PF00356) and a C-terminal ligand-binding domain which is very similar to the sugar-binding domain of an ABC transporter (Pfam PF13377) (Figure 1C). The presence of domains predicted DNA or sugar binding domains therefore suggest that RbsR is a transcriptional factor involved in carbohydrate metabolism (Pérez-Rueda and Collado-Vides, 2000).

Bioinformatic analysis revealed that the genes contiguous with the putative rbsR are orthologous with other genes (rbsD, rbsA, rbsC, rbsB, and rbsK) in the ribose operon (Fineran et al., 2013). The predicted rbsDACBKR gene cluster organization in S39006 is the same as in other enterobacteria such as E. coli, Serratia marcescens Db11, D. dadantii, P. atrosepticum, Erwinia amylovora, Yersinia enterocolitica, and Klebsiella pneumoniae (Supplementary Figure 1). In E. coli, the rbsD gene encodes a ribose mutarotase which converts the β-form of D-ribose into the α-furan form (Ryu et al., 2004); the rbsABC genes encode the ABC transporter (in which rbsA encodes an ATPase subunit, rbsB encodes a periplasmic binding protein and rbsC encodes a membrane permease; Park et al., 1999); and the rbsK gene encodes the ribokinase which phosphorylates D-ribose to D-ribose-5-phosphate. The rbsR gene, encoding the repressor of the rbsDACBKR operon, is the terminal gene of the ribose operon (Nentwich et al., 2009).

The phylogenetic relationships of the RbsR proteins in S39006 and 12 taxonomically related strains were analyzed. The phylogenetic tree shows two major clades (Supplementary Figure 2). The RbsR protein in S39006 is monophyletic with the LacI type family transcriptional regulator in D. dadantii and closely related to RbsR of P. atrosepticum. The RbsR protein is more distantly related to LacI type family transcriptional regulators in other enterobacteria such as Y. enterocolitica, E. amylovora, K. pneumonia, and E. coli.

The rbsR transposon insertion in S39006 LacA ΔpigC was transduced into the wild type S39006 strain to first confirm that the phenotypes observed were due entirely to the transposon. In contrast to wild type S39006 (which is opaque and red) the rbsR mutant (transductant) was translucent and non-pigmented on agar plates (Figure 1D). PCM showed no phase-bright structures in the mutant bacterial cells. The phase-bright gas “vacuoles” seen in a wild type strain are aggregates of GVs that appear as bright refractile structures under PCM. GVs were not visible by transmission electron microscopy (TEM) in the rbsR mutant (Figure 1E). The absence of GVs was further confirmed using flotation assays. Cultures of wild type S39006 remained buoyant after 48 h, whereas the rbsR mutant gradually settled to the bottom of the static liquid culture. Together, these results confirmed that a mutation in rbsR repressed GV formation.

The rbsR mutation was complemented by a plasmid carrying the wild type allele (pBAD-rbsR) and, on induction with arabinose, production of GVs was restored at the indicated arabinose concentrations (Figure 1F). Complemented strains were also restored for flotation and prodigiosin production (Figure 1F, see Supplementary Figure 3D for complementation of additional phenotypes). These results confirmed that the highly pleiotropic phenotypic changes seen in an rbsR insertion mutant were fully attributable to the transposon insertion. The rbsR gene was also overexpressed in a wild type background but no obvious effects were observed on GV production (data not shown).

To investigate regulation by RbsR at the transcriptional level, strains carrying either a chromosomal gvpA1::uidA or gvrA::uidA reporter gene fusion were employed. The gvpA1 and gvrA genes are the first genes of the two operons of the GV gene locus (Ramsay et al., 2011; Tashiro et al., 2016). The expression of gvpA1 or gvrA in the wild type and the rsbR mutant background was monitored throughout growth using a β-glu assay (a proxy for changes in transcriptional activity from the corresponding promoters). The β-glu activity expressed from gvpA1::uidA or gvrA::uidA transcriptional fusions increased between hours eight and 14 in both strains, during stationary phase. However, transcription from the gvpA1 or gvrA promoters in the rbsR mutant background was significantly reduced (3–5 fold lower, see ANOVA statistics in Figure Legend) compared to that in S39006 (Figures 2A,B). Similarly, under microaerophilic conditions, transcription of both gvrA and gvpA1 was greatly reduced in an rbsR mutant (Figures 2C,D). We also noted a minor growth defect in the rbsR mutant when compared to S39006, but this did not affect the data analysis as the β-glu activity was normalized to OD₆₀₀.

The rbsK mutant also showed the GV-negative phenotype—as seen in the rbsR mutant (Supplementary Figure 3A). This was not unexpected because the transposon, containing a transcriptional terminator (Monson et al., 2015), had been inserted within the rbsK gene and so should be polar on the downstream rbsR gene. To confirm this predicted polarity, we attempted to complement a mutation in rbsK with either one of the plasmids pBAD-rbsR or pBAD-rbsK. With induction, the plasmid expressing RbsR restored the ability of the rbsK mutant to produce GVs and prodigiosin (Supplementary Figure 3A). Further, gene expression from an rbsK::lacZ fusion was measured and the expression of rbsK was significantly reduced when plasmid-encoded RbsR was induced. (Supplementary Figure 3B). In contrast, the plasmid expressing RbsK was not capable of complementing GV formation or prodigiosin production in a rbsK mutant (Supplementary Figure 3C). These results are consistent with RbsR being a negative regulator of the ribose operon in this strain.

S39006 produces bioactive secondary metabolites (a carbapenem and prodigiosin), and PCWDEs, such as cellulase and pectate lyase (Coulthurst et al., 2005; Williamson et al., 2006; Fineran et al., 2007). In addition, S39006 is capable of swimming and swarming (Williamson et al., 2008). Most of these physiological characteristics are under QS control via production of BHL (Thomson et al., 2000; Williamson et al., 2008). A previous study demonstrated that QS also regulated the synthesis of GVs (Ramsay et al., 2011; Ramsay and Salmond, 2012). As the rbsR mutation regulated GV and prodigiosin production, we postulated it might also regulate other secondary metabolites and modes of motility. Production of the carbapenem and prodigiosin antibiotics was monitored throughout growth (Figures 3A,B). Both secondary metabolites were impacted: carbapenem antibiotic production was abolished (ANOVA results comparing the wild type and rbsR: F = 2614.62 > F_crit = 4.20, p-value 3.38^*10⁻²⁹) and prodigiosin production was significantly reduced in the rbsR mutant compared to wild type S39006 (ANOVA results comparing the two strains: F = 135.23 > F_crit = 4.20, p-value 3.11^*10⁻¹²). Furthermore, in an rbsR mutant, activity of a carA::lacZ fusion (the first gene in the carbapenem biosynthetic operon) was restored to wild type levels upon expression of RbsR from a plasmid (data not shown).To avoid suicide through carbapenem production, S39006 possesses a carbapenem intrinsic resistance mechanism (Coulthurst et al., 2005). Despite being defective in carbapenem production, the rbsR mutant was still resistant to the carbapenem antibiotic as indicated by the absence of a halo of inhibition in resistance assays (Figures 3C,D).

The secretion of cellulase and pectate lyase has been found to contribute to plant pathogenicity of S39006 (Fineran et al., 2007). Similarly, siderophores are produced by many bacteria in an attempt to acquire iron and this can be an important trait for plant virulence (Neilands, 1995). In other plant pathogens, siderophore production has also been found to be under the control of QS (Monson et al., 2013). In the present study, production of both cellulase and the siderophore of S39006 was reduced in the rsbR mutant (Figures 3E,F) though no significant changes in production of the QS signaling molecule, BHL, or pectate lyase production were observed (Supplementary Figure 4). Swimming and swarming abilities can also play roles in S39006 virulence (Williamson et al., 2008; Wilf et al., 2013) and the rbsR mutant also exhibited increased swimming and swarming motility (Figures 3G,H). We were also able to complement the effects of an rbsR mutation on swimming motility, swarming motility, siderophore production and cellulase production by expression of rbsR from a plasmid (Supplementary Figure 3D). These observations further highlighted the extent of the pleiotropy caused by the rbsR mutation in S39006.

Bioinformatic analysis revealed that the predicted binding motif of RbsR is conserved among different bacterial lineages (Laikova et al., 2001). The potential DNA binding site of RbsR in S39006 was analyzed using the MEME suite (Bailey et al., 2009). RbsR was predicted to bind to AAACGTTT and this binding motif was found in the upstream region of 18 genes, including rbsD and gvrA, in S39006 (Figure 4; Supplementary Table 2). The motif was also found upstream of a wide range of genes including those encoding transporters, transcriptional regulators, ATPase, transferase and biosynthetic genes (Supplementary Figure 5; Supplementary Table 2).

To assay the predicted binding targets of RbsR, EMSA was carried out using the purified RbsR protein. The N-terminal His-tagged RbsR protein was produced. Using the LUEGO method, a gel shift assay with increasing amount of RbsR protein demonstrated that the mobility of the rbsD probe was reduced and a complete band-shift was achieved with 14 nM of RbsR (relative ratio protein to DNA of 2.8:1) (Figure 5A). However, there were no discernable changes in the mobility of the gvrA probe at the concentrations used (Figure 5B). The EMSA probe was outcompeted by excess (20x) unlabeled probe but not by a non-specific probe, indicating that the binding was specific. Furthermore, in the presence of 1% ribose, RbsR failed to bind to the rbsD or gvrA probe (Supplementary Figure 5). These results suggest that, under these conditions, ribose is an effector of the RbsR protein.

To assess any impacts on plant virulence, potato rotting assays of S39006 and the rbsR mutant were performed. Potato tubers were inoculated with 1 × 10⁶ cfu of bacterial cells and incubated at 30°C for 5 days. The rotted tissue produced by the rbsR mutant was significantly lower than that produced by the wild type (Figure 6). This observation is consistent with the cellulase plate assay which showed significant reduction in elaboration of this virulence determinant (Figure 3C). However, the bacterial viable cell count revealed that the mutant exhibited impaired growth in potato tubers. For wild type cells, the number of colony forming units per tenth of a gram of rot was always greater than 10⁸. In contrast, for the rbsR mutant, colony numbers never exceeded 10⁴ for the same weight of inoculated potato rot.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer GJ and handling Editor declared their shared affiliation, and the handling Editor states that the process nevertheless met the standards of a fair and objective review.