Growth inhibition and experimental evolution in succinate minimal medium

Enteric bacteria possess the catabolic enzymes and efficient uptake systems required to grow with C₄-dicarboxylates as sole C-source [2,28]. However, some environmental and clinical isolates of Escherichia coli and Salmonella have surprisingly slow growth rates in succinate-containing minimal media [29–31]. To investigate this phenomenon in pathogenic and non-pathogenic enteric bacteria, we assessed the growth of four bacterial species on agar plates containing succinate as a sole C-source (M9+Succ). We studied growth for up to 96 hours, and used a variety of laboratory strains with an emphasis on Salmonella (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Inhibition of growth of most Salmonella serovars and K. pneumoniae on succinate minimal medium.

The indicated strains of E. coli, Salmonella, Citrobacter and Klebsiella were spread on M9+Succ agar plates and incubated at 37°C. Photographs of bacterial growth were taken every 24 hours for 2 days or 4 days. For Salmonella enterica isolates, the serovar and the clade A/B status [35,36] are indicated by the colour of the picture frame. Experiments were carried out as biological triplicates, and a representative picture is shown for each strain.

https://doi.org/10.1371/journal.pgen.1011142.g001

For S. enterica serovars Typhimurium and Enteritidis, we tested the growth of the well-characterised S. Typhimurium strains LT2, 4/74 and 14028 and of the S. Enteritidis strain P125109. In addition, we assessed the growth of multidrug resistant S. Typhimurium ST313 strain D23580 and S. Enteritidis strain D7795, two representative strains that cause invasive non-typhoidal Salmonella disease in Africa [32–34]. We included the reptile-associated Salmonella serovars Soahanina, Hadar, Newport and Infantis which belong to the two metabolically-distinct clades A and B of S. enterica [35,36]. We and others have previously generated genome sequences of most of the Salmonella strains that were used [34,35,37]. We also tested a multidrug resistant strain of Klebsiella pneumoniae (strain KP52.145) [38], a Citrobacter rodentium strain (ATCC51459) [39] and the classic laboratory strain E. coli K-12 strain MG1655.

After 24 hours of incubation at 37°C, the only strains that displayed substantial growth with succinate as a C-source were S. Typhimurium LT2, E. coli MG1655 and C. rodentium ATCC51459. The growth of the S. Typhimurium laboratory strain LT2 upon succinate at 24h is the result of the mutated variant of rpoS carried by this strain [40], which explains its faster growth [16]. We found that growth on succinate is a serovar-dependent phenotype. Of the six Salmonella serovars that were tested, only S. Newport displayed growth after 2 days. Subsequently, following 2–3 days of incubation, large isolated colonies were observed within the bacterial lawns of the K. pneumoniae strain and the other Salmonella isolates.

Previous experiments in liquid minimal medium containing succinate as sole C-source showed that S. Typhimurium exhibited a particularly long lag phase [16,23]. This extended lag time could reflect a particularly slow metabolic remodelling, preparing Salmonella for the exponential phase in the presence of succinate. Alternatively, robust inhibition of succinate assimilation might be occurring under these conditions, preventing growth until spontaneous fast-growing mutants (hereafter referred as Succ⁺ mutants) have emerged.

To test these hypotheses, we used an experimental evolution approach to generate Succ⁺ mutants. We assessed the growth of the well-characterised S. Typhimurium strain 4/74 (henceforth referred to as 4/74 or Salmonella) in liquid M9+Succ media inoculated with a stationary phase culture made in rich medium (LB). The four independent 4/74 cultures (cultures I-IV) exhibited the reported 30–35 hour lag time at 37°C [16,23] (Fig 2A). We collected the Salmonella that eventually reached stationary phase and cultured the bacteria in LB for two passages before re-inoculation in M9+Succ. For all the succinate-evolved cultures, the lag time in M9+Succ was shortened to 4–5 hours, and stationary phase was reached after approximately 14 hours (Fig 2B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Experimental evolution of S. Typhimurium in succinate minimal medium.

(A) Growth of S. Typhimurium 4/74 displays an extended lag time in M9+Succ medium. The growth curves of four independent cultures (I-IV) of 4/74 WT in M9+Succ are presented. Cultures were inoculated with bacteria grown beforehand to stationary phase in LB. the succinate evolved bacteria were harvested in ESP (50 μL of culture) and were grown twice in LB prior to re-inoculation in M9+Succ. (B) The succinate-evolved bacteria grow fast in M9+Succ in comparison with the WT strain. Growth curves of the succinate evolved cultures I-IV from (A) in M9+Succ are presented: the 4/74 WT and ΔrpoS (JH3674) strains were included as controls. (C) Succinate fast growing (Succ⁺) mutants were detected in liquid M9+Succ 4/74 cultures. Cultures from (A) were spread on LB plates at the indicated growth phase and growth with succinate was assessed for 15 isolated colonies per replicate on M9+Succ agar plates after 48 hours of incubation. The graph shows the proportion (%) of Succ⁺ clones. ND = not detected. (D) Succ⁺ spontaneous mutants emerge from 4/74 WT bacterial lawns on M9+Succ agar plates. 4/74 WT cultures (~10⁷ CFU) were spread on a M9+Succ agar plates and the picture of a representative plate was taken after 3 days of incubation at 37°C. For the growth curves (A&B), bacteria were grown at 37°C with aeration in 25 ml of M9+Succ (in 250 ml conical flasks) with an initial inoculum of ~10⁷ CFU/mL (OD₆₀₀ = 0.01). Growth phases are indicated in (A&C): Lag phase (Lag); Early exponential phase (EEP); Mid-exponential phase (MEP); Early stationary phase (ESP).

https://doi.org/10.1371/journal.pgen.1011142.g002

To investigate heritability of the succinate growth phenotype, the initial M9+Succ cultures (Fig 2A) were spread on LB agar plates at different stages of growth, and isolated colonies were tested on M9+Succ plates. Of 60 colonies obtained from lag phase, none grew faster than the wild type (Fig 2C). However, fast growing Succ⁺ mutants harvested from early exponential, mid-exponential and early stationary phase culture were detected at a frequency of 20%, 78% and 90%, respectively (Fig 2C). When ~10⁷ Colony Forming Units (CFU) of 4/74 wild-type (WT) were spread on M9+Succ plates, between 100 and 1000 Succ⁺ colonies grew in the bacterial lawn after 3 days of incubation (Fig 2D).

Collectively, these results indicated that Salmonella growth upon succinate was consistently inhibited in our experimental setup. The eventual initiation of exponential phase did not result from an orchestrated metabolic switch, but reflected the emergence of spontaneous mutants that efficiently utilised succinate, and proliferated to outcompete the WT bacteria. This Succ⁺ phenotype remained stable after two passages on LB medium, indicating that the trait was not a phase-variable phenomenon caused by epigenetic mechanisms, as has been observed for other reversible phenotypes in Salmonella [41,42]. Moreover, our data suggest that other pathogenic bacteria, such as Klebsiella also suppress succinate assimilation (Fig 1).

Here, we define “succinate utilisation” as the ability of Salmonella to grow with succinate as a sole carbon and energy source. We selected S. Typhimurium strain 4/74 for further study of the suppression of succinate utilisation because it is the prototrophic parent of strain SL1344, which has been used for a plethora of regulatory and infection studies in the past [43]. We aimed to identify novel genetic determinants involved in the control of the uptake and catabolism of this infection-relevant C-source.

Identification of novel mutations that abolish inhibition of succinate utilisation under aerobic conditions

To identify mutations that ablate the inhibition of succinate utilisation, we used three complementary unbiased approaches. We first screened a collection of published S. Typhimurium 4/74 mutants [44] that lacked key regulatory proteins (S1A Fig), and focused on mutations that both promoted growth on succinate agar plates and in liquid medium with aeration. This screen revealed that mutants lacking the RNA chaperone Hfq (Δhfq) and the polynucleotide phosphorylase (PNPase, mutant Δpnp) had a Succ⁺ phenotype. Complementation with low-copy plasmids carrying hfq⁺ or pnp⁺, restored the Succ^- WT phenotype in the corresponding mutant (S3C and S3D Fig).

To explore metabolic suppression in more depth, we used global Tn5 transposon mutagenesis to generate insertions that promoted growth on succinate (Methods). RpoS has a key role in the inhibition of succinate utilisation [29–31] and rpoS inactivation did cause the drastic shortening of the lag time of 4/74 in M9+Succ, similarly to the succinate evolved cultures (Figs 2B, S1B, and S3B). Therefore, we developed a strategy to avoid the selection of Succ⁺ rpoS mutants by constructing a strain that carried two chromosomal copies of rpoS (4/74 rpoS^2X; S1B and S1C Fig). Following Tn5 mutagenesis of the rpoS^2X strain, individual Succ⁺ Tn5 mutants were isolated. The Tn5 insertions were P22-transduced into 4/74 WT and the Succ⁺ phenotype of the transductants was confirmed (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Fifteen mutations that stimulated growth of S. Typhimurium upon succinate.

https://doi.org/10.1371/journal.pgen.1011142.t001

In parallel, we isolated spontaneous Succ⁺ mutants from either M9+Succ agar plates (Fig 2D) or from liquid cultures (Methods). We first verified the RpoS positive (rpoS⁺) status of each spontaneous Succ⁺ mutant (Methods), and then used whole genome-sequencing to identify relevant nucleotide changes, which were associated with seven genes (Table 1).

These complementary genetic screens identified Succ⁺ mutants that carried Tn5 insertions in iraP, cspC, rbsR and fliD or in the 5’ untranslated region (5’-UTR) of the yobJ-cspC operon (Table 1). In addition, a nonsense spontaneous mutation in iscR (yfhP) and a spontaneous in-frame insertion of 4 codons in rbsR were identified in spontaneous Succ⁺ mutants (Table 1). To independently confirm the function of these genes, λ red recombination was used to generate ΔiraP, ΔcspC, ΔrbsR and ΔiscR deletion mutants. Each of the four mutants had the Succ⁺ phenotype (Fig 3). We confirmed that the corresponding WT proteins could inhibit succinate utilisation by plasmid-borne complementation experiments (S3E–S3H Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Growth phenotype of 5 genome-edited Succ⁺ mutants and 8 regulatory mutants on solid succinate minimal medium.

S. Typhimurium strains 4/74 WT, ΔrpoS (JH3674), Δhfq (JH3584), ΔiraP (SNW188), ΔfliST (SNW288), ΔcspC (SNW292), ΔiscR (SNW184), ΔrbsR (SNW294), Δpnp (JH3649), dctA^mut1 (SNW160), dctA^mut2 (SNW315), oxyR^mut (SNW318), rrsA^mut (SNW336), rrsH^mut (SNW314), were spread on M9+Succ agar plates and incubated for 48 hours at 37°C. Experiments were carried out with biological triplicates and a representative picture is shown for each strain.

https://doi.org/10.1371/journal.pgen.1011142.g003

We found that inactivation of fliD alone did not cause a Succ⁺ phenotype. The fliD gene is co-transcribed with the downstream fliS and fliT genes [45]. Tn5 insertions may have polar effects on the expression of surrounding genes [46], raising the possibility that the fliD::Tn5 insertions (S2A Fig) modulated expression of fliS or fliT. Our genetic dissection of the fliDST operon revealed that inactivation of either fliS or fliT caused the Succ⁺ phenotype, suggesting that the two genes contribute to the inhibition of succinate utilisation (S2B Fig). The Succ⁺ phenotype of the ΔfliST mutant was confirmed, and complementation of the double mutation restored the WT Succ^- phenotype (S3I Fig).

During the complementation experiments, we observed that the presence of chloramphenicol (Cm) mildly stimulated the growth of the WT strain (S3J Fig), an observation that we have investigated further (S2 Text).

Certain spontaneous mutations that stimulated growth with succinate did not reflect a typical loss-of-function scenario. For example, one of the spontaneous Succ⁺ mutants had an additional CTA codon resulting in an extra leucine between residues 239 and 240 of the transcriptional regulator OxyR (mutant oxyR^mut) (Table 1). We also identified two Succ⁺ mutants with a single nucleotide polymorphism (SNP) located in the 5’-UTR of dctA (mutants dctA^mut1 and dctA^mut2, Table 1), that encodes for the aerobic succinate transporter DctA [47]. Finally, two Succ⁺ mutants carried a SNP in the anti-Shine-Dalgarno sequence of the rrsA and rrsH genes that encode two 16S ribosomal RNAs; mutants rrsA^mut and rrsH^mut (Table 1).

The function of the mutations associated with genes oxyR, dctA, rrsA and rrsH was confirmed by scarless genomic editing to generate exactly the same nucleotide changes in the WT background (Methods). All these reconstructed mutations caused the Succ⁺ phenotype (Fig 3).

In summary, we identified novel mutations that promote Salmonella growth upon succinate and shortened the lag time in comparison with the WT (S3 and S4 Figs). These included mutations that involved the succinate transporter DctA, the transcriptional regulators (RbsR, IscR, OxyR), RNA binding proteins (PNPase, Hfq and CspC), flagellar protein chaperones (FliS and FliT) and in ribosomal RNAs (RrsA and RrsH) (Fig 3). The eleven novel Succ⁺ mutations also promoted Salmonella growth upon fumarate or malate, suggesting that the regulatory systems play a general role in the de-inhibition of C₄-dicarboxylate utilisation (S1D Fig).

The OxyS sRNA stimulates growth upon succinate by repressing expression of CspC

The spontaneous Succ⁺ mutants included an oxyR variant (oxyR^mut) that encoded an extra leucine residue in the C-terminus domain of the OxyR transcriptional regulator (Table 1 and Fig 4A). OxyR senses oxidative stress and is activated by disulfide bond formation in the presence of reactive oxygen species [48,49]. In E. coli, the OxyR regulon includes about 40 genes, mainly associated with oxidative stress resistance [50,51]. In addition, the oxidized form of OxyR triggers the transcription of OxyS, an Hfq-binding small regulatory RNA (sRNA) [52–54]. Previous studies reported the isolation of constitutively-active OxyR variants that carried mutations in the same region as the extra leucine residue of the oxyR^mut variant [55–57].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. The Succ⁺ oxyR^mut mutant expresses OxyS constitutively and stimulates Salmonella growth with succinate in an Hfq-dependent manner.

(A) Schematic representation of the oxyRS locus. The Succ⁺ mutation oxyR^mut has an additional CTA codon encoding for an extra leucine in oxyR. (B) OxyS is constitutively expressed in the oxyR^mut strain. Northern blot detection revealed OxyS expression in strains 4/74 WT, oxyR^mut (SNW318) and ΔoxyR (SNW320) grown in LB and exposed or not to 2 mM H₂O₂ for 30 min (Methods). The detection of the 5S rRNA was used as a loading control. (C) OxyS stimulates Salmonella growth with succinate. Growth curves showed the fast growth of the oxyR^mut strains in comparison with the WT and the oxyR^mut ΔoxyS (SNW340) strains. Growth is restored in the complemented strain oxyR^mut ΔoxyS oxyS^chr+ (SNW362), carrying a chromosomal oxyS copy. (D) The plasmid-borne expression of OxyS stimulates the growth of the WT but not of the ΔoxyR mutant. Growth was assessed for strains 4/74 WT and ΔoxyR (SNW320), carrying either the empty (vector, pP_L) or the oxyS expressing (oxyS++, pNAW255) plasmids, schematised at the top of the figure. The bent arrows represent the constitutive promoter P_{L lacO-1} of the Ap^R pP_L plasmid, carrying the oriColE1 replicon. (E) Hfq inactivation suppresses the fast growth of the oxyR^mut mutant. Growth was assessed for strains 4/74 WT, Δhfq (JH3584), oxyR^mut (SNW318) and oxyR^mut Δhfq (SNW663). The medium used is indicated for each experiment. Growth curves were performed in 96-well plates, as specified in Methods. The details about the construction of strains SNW320, SNW340 and SNW362 are depicted in the S7 Fig.

https://doi.org/10.1371/journal.pgen.1011142.g004

Consequently, we investigated whether the oxyR^mut allele drove constitutive transcription of the OxyS sRNA. Northern blot analysis revealed that in the oxyR^mut strain, OxyS was strongly expressed both in the absence and in the presence of hydrogen peroxide, indicating that the OxyR^mut protein is constitutively active (Fig 4B). We then determined whether constitutive expression of OxyS was responsible for the Succ⁺ phenotype of the oxyR^mut mutant. The deletion of oxyS in the oxyR^mut strain (oxyR^mut ΔoxyS) totally abolished the Succ⁺ phenotype (Fig 4C). A complementation experiment was carried out by re-introducing a single copy of oxyS and its native promoter (oxyS^chr+, S7 Fig) into the chromosome of the oxyR^mut ΔoxyS strain (Methods). This chromosomal complementation restored the fast growth of the oxyR^mut ΔoxyS mutant in M9+Succ (Fig 4C). Furthermore, the plasmid-borne expression of OxyS boosted the growth of 4/74 WT in M9+Succ, confirming that high level expression of the OxyS sRNA stimulated growth with succinate (Fig 4D). The same plasmid did not stimulate the growth of the ΔoxyR strain indicating that other OxyR-dependent genes are required to grow under this condition.

We previously showed that Hfq inactivation boosted succinate utilisation, but in the oxyR^mut genetic background the same Hfq inactivation dramatically reduced growth and extended the duration of lag time in M9+Succ (Fig 4E). Collectively, our findings show that the OxyS sRNA orchestrates the de-inhibition of succinate utilisation in concert with Hfq.

In E. coli, OxyS acts as an indirect repressor of RpoS expression, likely via the titration of Hfq [53]. OxyS also represses the expression of the yobF-cspC operon, probably by base-pairing near the SD motif of the yobF 5’-UTR [58,59]. Because RpoS, Hfq and CspC repress succinate utilisation (Fig 3), we tested the effects of the plasmid-borne overexpression of rpoS⁺, hfq⁺ or cspC⁺ on the growth of the oxyR^mut strain. The overexpression of Hfq and RpoS slightly increased the lag time of oxyR^mut strain, while the plasmid-borne expression of CspC totally abolished the Succ⁺ phenotype in this genetic background (Fig 5A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. The OxyS sRNA stimulates Salmonella growth with succinate by repressing expression of CspC.

(A) Overexpression of cspC inhibits the growth of the oxyR^mut strain (SNW318), while hfq and rpoS overexpression have only a mild effect. The oxyR^mut strain was carrying either the empty (vector, pNAW125), the prpoS (pNAW95), the phfq (pNAW45) or the pcspC (pNAW92) plasmids, schematised at the top of the figure. These low-copy Cm^R plasmids are carrying the ori101* replicon and the genes of interest are under the control of the strong constitutive promoter P_{L tetO-1} (bent arrows). As a control, the growth of 4/74 WT carrying the empty plasmid (vector) was also assessed. (B) Strategy to test whether the Salmonella sRNA OxyS represses the expression of the yobF-cspC mRNA at the post-transcriptional level, as previously reported in E. coli [58]. The plasmid-borne translational fusion yobF::_sfgfp (pNAW258) is depicted and was constructed as described in Methods and in Corcoran et al., 2012 [99]. This fusion is under the control of the constitutive promoter P_{L tetO-1}.SD = Shine-Dalgarno. (C) The yobF::_sfgfp activity is reduced in the oxyR^mut strain, that expresses constitutively OxyS: fluorescence GFP intensities were measured in strains 4/74 WT and oxyR^mut carrying yobF::_sfgfp. (D) Prediction of the RNA secondary structure of Salmonella OxyS and of yobF 5’-UTR, using Mfold [119]. The putative kissing complex between the two RNA molecules was predicted with IntaRNA [120] and the corresponding nucleotides are indicated in magenta for yobF 5’-UTR or in blue for OxyS. The mutation oxyS^GG is indicated and the corresponding nucleotides are framed. (E) The plasmid-borne overexpression of OxyS represses the yobF::_sfgfp activity and the oxyS^GG mutation attenuates this repression: fluorescence GFP intensities were measured in the ΔoxyS strain (SNW338) carrying yobF::_sfgfp and the empty (empty vector, pP_L), the pP_L-oxyS (oxyS^WT++, pNAW255) or the pP_L-oxyS^GG (oxyS^GG++, pNAW259) plasmids. (F) The oxyS^GG mutation reduces the growth of the oxyR^mut strain: strains 4/74 WT, oxyR^mut and the oxyR^mut mutant carrying the oxyS^GG mutation (SNW670) were grown in M9+Succ. (G) Effects of rpoS overexpression on the growth of the ΔcspC mutant: the strains 4/74 WT and ΔcspC (SNW292), carrying either the empty plasmid (vector, pNAW125) or the prpoS (pNAW95) plasmid were grown in M9+Succ. For A, F and G, growth curves were carried out in the indicated medium with 6 replicates grown in 96-well plates. For C & E, strains were grown to OD₆₀₀ ~ 2 in LB, supplemented with the appropriate antibiotic(s). GFP fluorescence intensities were measured, as specified in Methods. The graphs represent the relative fluorescence intensities (%), in comparison with the indicated reference strain (100% of intensity). The same strains carrying GFP fusions were grown on LB agar plates and pictures were taken under blue light exposure. The data are presented as the average of biological triplicates ± standard deviation and the statistical significance is indicated, as specified in Methods.

https://doi.org/10.1371/journal.pgen.1011142.g005

To confirm that the OxyS-driven repression of the yobF-cspC was conserved in Salmonella we used a plasmid-borne yobF::_sfgfp translational reporter (Fig 5B). In comparison with the WT, the yobF::_sfgfp activity was significantly lower in the oxyR^mut strain (~2-fold repression), confirming that OxyS represses the expression of the yobF-cspC operon in Salmonella (Fig 5C). Bioinformatic analyses identified putative secondary structures and a potential base-pairing interaction between OxyS and the yobF 5’-UTR (Fig 5D), which was predicted to be an 12 nucleotide-long kissing complex between OxyS and the yobF 5’-UTR, consistent with the proposed interaction in E. coli [59] (Fig 5D).

To assess the role of the kissing complex experimentally, we generated a mutated version of OxyS with a CC➔GG mutation in the loop of the first RNA hairpin (allele oxyS^GG, Fig 5D). This mutation was introduced into the chromosome of the oxyR^mut strain (strain oxyR^mut oxyS^GG) and the oxyS^GG gene was cloned into the pP_L expression vector [60]. To test the impact of sRNAs OxyS and OxyS^GG upon yopF-cspC operon expression, the empty pP_L vector, the pP_L-oxyS or the pP_L-OxyS^GG plasmids were transferred into the oxyS null strain carrying the yobF::_sfgfp fusion and the GFP signal was measured. In this genetic background, oxyS expression (oxyS++) reduced the GFP fluorescence intensity by ~3-fold, but only by ~1.5-fold in the strains expressing oxyS^GG (oxyS^GG++) (Fig 5E). Consistent with the attenuated repression of yobF in the presence of OxyS^GG, the oxyR^mut oxyS^GG strain had a longer lag time than the oxyR^mut mutant, confirming that the mutated region of OxyS is involved in the de-inhibition of succinate utilisation (Fig 5F).

In E. coli, CspC stabilises rpoS mRNA and increases the cellular level of RpoS [61,62]. To investigate whether the Succ⁺ phenotype of the CspC null mutant was caused by changes in RpoS expression, we tested the effect of RpoS overexpression in the ΔcspC mutant (Fig 5G). The plasmid-encoded overexpression of RpoS only marginally extended lag time in the ΔcspC mutant, indicating that repression of succinate utilisation by CspC is RpoS-independent. A recent study corroborated this observation, as the CspC-mediated activation of RpoS did not occur in Salmonella [63].

Collectively, our results indicate that the OxyS sRNA is a key determinant in the de-inhibition of succinate utilisation by Salmonella. Despite the fact that OxyS can regulate RpoS expression levels, we propose that OxyS stimulates the Succ⁺ phenotype by repressing the expression of CspC via base-pairing in the vicinity of the yobF SD motif. CspC is an RNA binding protein, belonging to the cold shock protein family [64]. CspC and its paralog CspE often have redundant functions, being involved in biofilm formation, motility, stress resistance and virulence modulation in S. Typhimurium [61,63]. It remains unclear how the OxyS-driven inhibition of CspC expression impacts upon the catabolism of succinate. One possibility is that CspC directly represses succinate catabolic genes. In line with this hypothesis, a transcriptomic study in S. Typhimurium, revealed that the sdhC, sdhD and sdhA genes are moderately up-regulated in a ΔcspEC mutant [63].

Our study also revealed that two other RNA binding proteins, Hfq and PNPase in concert with their cognate sRNAs also contribute to the repression of Salmonella growth upon succinate (S1 Text and S5 Fig). We discovered that Hfq played a dual role on the regulation of succinate utilisation by stimulating succinate utilisation in concert with OxyS, whilst maintaining succinate growth inhibition, probably via the stimulation of rpoS expression [65].

In summary, these results demonstrate the crucial role of RNA-binding proteins and their associated sRNAs in the fine-tuning of Salmonella carbon metabolism.

The iron-sulphur cluster regulator IscR inhibits growth upon succinate by repressing DctA expression

In E. coli, the C₄-dicarboxylate transporter DctA mediates succinate uptake under aerobic conditions [47]. DctA is also a C₄-dicarboxylate co-sensor that modulates the expression of several genes, including dctA itself, in concert with the two-component system DcuR/S [66]. The transcription of dctA is controlled by catabolic repression and putative CRP binding sites have been identified in the dctA promoter region [67,68] and are conserved in Salmonella (Fig 6A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. IscR represses the expression of DctA and inhibits Salmonella growth upon succinate.

(A) Detailed schematic representation of the dctA promoter (P_dctA) region in S. Typhimurium (S. Tm) 4/74 and in E. coli MG1655. Conserved nucleotides in 4/74 and MG1655 are highlighted in light gray. The promoter -35 and -10 boxes and the transcription start site (TSS, numbering +1) are indicated according to Davies et al., 1999 [67]. The DcuR binding site [121] and the putative CRP binding site [67,68] are indicated. A putative IscR binding site is depicted and the mutations identified in the Succ⁺ mutants dctA^mut1 and dctA^mut2 are indicated in red. In addition, the corresponding region of E. coli B strain REL606 is depicted: in this strain, a G➔A SNP (in red) causing dctA up-regulation and the stimulation of succinate utilisation was previously described [75]. SD = Shine-Dalgarno motif. Promoters (P) are represented by bent arrows. The symbol “*” denotes that the binding sites were not experimentally demonstrated. (B) The stimulation of DctA expression and the dctA^mut1 and dctA^mut2 mutations boost Salmonella growth with succinate. The AHT-inducible strain tetR-P_tetA-dctA (SNW133) is depicted: the tetR repressor gene, the P_tetR and the P_tetA promoters are indicated. The residual FLP recognition target site sequence is denoted by”frt”. In the absence of AHT no growth was detected for both WT and tetR-P_tetA-dctA strains, while AHT addition stimulated the growth of the tetR-P_tetA-dctA strain. Similarly, the dctA^mut1 (SNW160) and dctA^mut2 (SNW315) strains displayed a fast growth in M9+Succ. (C & D) The SNP mutations dctA^mut1 and dctA^mut2 and the inactivation of IscR stimulate the expression of DctA. The chromosomal transcriptional/translation dctA::_sfgfp fusion is depicted: _sfgfp encodes for the superfolder GFP fused in frame to DctA C-term. The GFP fluorescence intensity was measured in strain WT dctA::_sfgfp (SNW296), dctA^mut1 dctA::_sfgfp (SNW310), dctA^mut2 dctA::_sfgfp (SNW316), ΔiscR dctA::_sfgfp (SNW329) and ΔrpoS dctA::_sfgfp (SNW313).(E) Both Apo- and holo- forms of IscR repress Salmonella growth with succinate. The growth was assessed for strains 4/74 WT and ΔiscR (SNW184) carrying either the empty plasmid (vector, pXG10-SF), the piscR (pNAW96) or the piscR^3CA (pNAW97) plasmids. The piscR^3CA expressed the IscR^3CA variant that prevents the binding of an iron-sulphur cluster, maintaining IscR in its apo-form (see text for details). Growth curves (B & E) were carried out in the indicated medium with 6 replicates grown in 96-well plates. For C & D, strains carrying the dctA::_sfgfp fusion were grown in M9+Gly+Succ to OD₆₀₀ 0.5–1 and GFP fluorescence intensity was measured by fluorescence microscopy (C) or by flow cytometry (D), as specified in Methods. The graph (D) represents the relative fluorescence intensities of each strain (%), in comparison with the WT strain carrying dctA::_sfgfp (SNW296, 100% of intensity). The data are presented as the average of biological triplicates ± standard deviation and the statistical significance is indicated, as specified in Methods. NS, not significant.

https://doi.org/10.1371/journal.pgen.1011142.g006

To determine whether DctA was required for Salmonella growth under our experimental conditions, we constructed a chromosomal inducible dctA construct, by replacing the dctA promoter with tetR (encoding the TetR repressor) and the tetA promoter (strain tetR-P_tetA-dctA, Fig 6B). In the absence of anhydrotetracycline (AHT) inducer, the tetR-P_tetA-dctA strain did not grow at all in M9+Succ. However, upon addition of AHT, the tetR-P_tetA-dctA strain displayed a fast growth phenotype that was not observed with the WT (Fig 6B). A long incubation (>60 hours) confirmed that dctA was essential for Salmonella growth and the emergence of the Succ⁺ mutants, ruling out a role for other C₄-dicarboxylate transporters under the conditions tested (S8 Fig). We conclude that Salmonella requires the DctA transporter to grow on succinate and that dctA expression is likely to be repressed in the WT, as previously hypothesised by Hersch and co-workers [16].

The two spontaneous Succ⁺ mutants dctA^mut1 and dctA^mut2 (Table 1) carry SNPs in the 5’-UTR of dctA (Fig 6A), and both promoted growth with succinate (Fig 6B). We reasoned that the Succ⁺ mutations could de-repress dctA expression, which we examined with a chromosomal dctA::_sfgfp transcriptional/translational reporter fusion in the Succ⁺ mutant backgrounds (Methods and Fig 6C). To allow homogenous growth for all the strains (including the Succ^- 4/74 WT), bacteria were grown in M9 supplemented with glycerol (40 mM) as the main C-source and the addition of 10 mM succinate, to stimulate expression of succinate-induced genes [69]. The GFP fluorescence intensity of single bacteria was measured for each strain by flow cytometry. In comparison with the low levels of GFP fluorescence seen in the 4/74 WT background, higher GFP levels were only observed in the presence of the dctA^mut1, dctA^mut2 and ΔiscR mutations (S9A–S9N Fig). The regulation was confirmed by fluorescence microscopy (Fig 6C) and flow cytometry, using biological triplicates (Fig 6D).

A recent report proposed that RpoS indirectly represses dctA in Salmonella [16], but such an increase of the dctA::_sfgfp fusion activity was not observed in the ΔrpoS strain under our experimental conditions (Figs 6D and S9P). We found that the dctA::_sfgfp-tagged ΔrpoS, dctA^mut1, dctA^mut2 and ΔiscR mutants grew much faster in M9+Succ than the isogenic WT strain (S9O Fig), with the same growth rate we observed previously with the corresponding untagged mutants (Figs 6B,6E, and S1B). This indicated that the C-terminal addition of sfGFP did not impede the function of DctA as a succinate transporter or co-sensor.

The up-regulation of dctA in the Salmonella IscR null mutant was consistent with the IscR-driven repression of dctA proposed in E. coli [70]. In most Gram-negative bacteria, the dual regulator IscR controls the transcription of the iron-sulphur (Fe-S) cluster biosynthesis operon iscRSUA and the Fe-S cluster assembly genes sufABCDSE [71,72]. The apoprotein form of IscR (apo-IscR) is matured by the Isc system into its [Fe₂-S₂]-containing holo-form. The resulting holo-IscR represses the expression of several genes including the isc operon. Under iron-limitation and in the presence of reactive oxygen species, the IscR apo-form predominates and stimulates the expression of the suf operon, in concert with OxyR [71,73]. IscR binds to two classes of DNA motifs: Type 1 motifs are only bound by holo-IscR, while Type 2 motifs are recognised by both holo- and apo- forms [70,74].

Analysis of the promoter region of the Salmonella dctA revealed the presence of a putative Type 2 IscR-binding site (ATAACCTTACAAGACCTGTGGTTTTT) [74] located 10 bp downstream of the transcription start site of dctA (Fig 6A). Both the Succ⁺ mutants dctA^mut1, dctA^mut2 carried SNPs within this DNA motif. This motif is also conserved in E. coli MG1655 (Fig 6A) and a similar SNP, that stimulated dctA transcription and succinate utilisation, has previously been identified in E. coli B [75].

To investigate which of the apo/holo-forms of IscR was repressing succinate utilisation in Salmonella, we constructed a plasmid expressing an IscR variant carrying three Cys➔Ala substitutions (IscR^3CA,Cys_92,98,104➔Ala_92,98,104) that prevent the binding of [Fe₂-S₂] to IscR, and maintain the apo-form of the protein [76,77]. The plasmid-borne expression of both IscR and IscR^3CA complemented the ΔiscR deletion and suppressed the Succ⁺ phenotype (Fig 6E), indicating that both apo- and holo- IscR repress succinate utilisation.

Collectively, these results demonstrated that IscR plays a critical role in the repression of dctA and in the inhibition of succinate utilisation. The apo-IscR represses succinate utilisation, suggesting that IscR represses dctA expression by binding the putative Type 2 DNA motif identified downstream of the dctA promoter. The finding of dctA-stimulating SNPs in this motif supports our hypothesis, and it is possible that the binding of IscR downstream of the promoter acts as a “roadblock” that inhibits dctA transcription, as proposed for the repression of mgtC by PhoP [78]. However, the repression of dctA by IscR could also be indirect with the dctA^mut1 and dctA^mut2 mutations stimulating dctA transcription by another mechanism. Further study is required to understand the IscR-driven repression of dctA at the mechanistic level.

Our data show that the de-repression of dctA expression is sufficient to stimulate Salmonella growth with succinate, consistent with a previous report [16]. However, dctA de-repression was only observed in the dctA^mut1, dctA^mut2 and ΔiscR mutants, raising the question of whether DctA-driven succinate uptake is the key limiting factor for rapid growth of Salmonella in M9+Succ media. In E. coli, the presence of fumarate stimulates the expression of sdhC and other TCA cycle genes in a DcuR-dependent manner [18]. Because DctA can act as a co-sensor of the DcuS/R transduction cascade [66], it is possible that the increase of DctA abundance impacts upon the expression of the sdh operon, stimulating growth upon succinate.

Succinate utilisation is inhibited by RbsR and FliST via RpoS

Factors that modulate RpoS expression, stability or activity are likely to control succinate utilisation in Salmonella. For example, the inactivation of the anti-adapter IraP stimulates succinate utilisation by increasing RssB-facilitated proteolysis of RpoS by the ClpXP protease [16,79,80]. Our genetic screens showed that succinate utilisation was stimulated by the absence of IraP, CspC, RbsR and FliST and by increased expression of OxyS.

To assess RpoS levels in the corresponding Succ⁺ mutants, we used Western blot detection (S10A Fig). The levels of RpoS protein in the ΔcspC, ΔfliST, oxyR^mut mutants were similar to the WT strain, while lower levels of RpoS were observed in the ΔiraP and ΔrbsR mutants. We confirmed the role of RbsR in RpoS activation by complementing the ΔrbsR mutation with the prbsR plasmid, revealing that inactivation of RbsR reduced RpoS abundance in exponential and early stationary phases, but not in stationary phase (Figs 7A and S10B). During the characterisation of the Succ⁺ mutants, we noticed that the RbsR null strain was impaired in its capacity to form red, dry and rough colonies (RDAR), another RpoS-dependent phenotype of Salmonella [81]. The RDAR morphotype was restored in the ΔrbsR mutant by complementation with the plasmid prbsR (Fig 7B). We observed that the plasmid-borne overexpression of RpoS in the RbsR null mutant totally abolished the Succ⁺ phenotype, indicating that RbsR represses succinate utilisation via the activation of RpoS (Fig 7C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. RbsR and FliST inhibit Salmonella growth upon succinate via RpoS.

(A) RbsR inactivation reduces the cellular level of RpoS in exponential and early stationary phases. Strains 4/74 WT and ΔrbsR (SNW294) carrying the empty (vector, pXG10-SF) or the prbsR (pNAW93) plasmids were grown in LB (without Cm) to late exponential (OD₆₀₀~1), early stationary (OD₆₀₀~2) and stationary phase (OD₆₀₀~4). The cellular levels of RpoS and DnaK (loading control) were assessed by Western blotting (Methods). As negative control, the ΔrpoS mutant (JH3674) carrying pXG10-SF was included. The experiment presented is representative of three independent experiments and two replicates are presented in S10B Fig. (B) RbsR inactivation reduces red, dry and rough (RDAR) morphotype, another RpoS-dependent phenotype. The RDAR phenotypic assays were carried out as specified in Methods with strain 4/74 WT, ΔrbsR and ΔrpoS carrying the indicated plasmids (vector = pXG10-SF). At least three independent experiments were performed, and representative RDAR colony pictures are presented. The plasmid-borne rpoS overexpression suppresses the Succ⁺ phenotype of the ΔrbsR (C) and ΔfliST (D) mutants. Growth was assessed in the indicated medium in a 96-well plate for strains ΔrbsR (SNW294) or ΔfliST (SNW288) carrying either the empty plasmid (vector, pNAW125) or the prpoS plasmid (pNAW95) and strains 4/74 WT carrying the empty plasmid.

https://doi.org/10.1371/journal.pgen.1011142.g007

RbsR is a LacI-type transcriptional regulator that inhibits the transcription of the ribose utilisation operon (rbsDACBKR) in the absence of ribose [82]. To investigate whether RbsR stimulated rpoS transcription, we used a chromosomal transcriptional GFP reporter fusion, with a gfp⁺ gene (including its SD sequence) inserted downstream of the main transcription start site of the rpoS locus (S10C Fig). Similar GFP levels were observed in the WT and the ΔrbsR strains grown to either exponential, early stationary or stationary phase (S10D Fig).

Taken together, our findings show that RbsR acts as a pleiotropic regulator in Salmonella, controlling growth with succinate and the RDAR morphotype, via the positive regulation of rpoS. RbsR does not directly stimulate rpoS promoter activity and we propose an indirect RbsR-driven activation of RpoS at the post-transcriptional or the post-translational level. In line with this hypothesis, it was recently observed that RpoS is repressed at the post-transcriptional level, when the RbsR-dependent rbsD gene is over-expressed in E. coli [83].

Our work also revealed that the two flagellar chaperones FliS [84] and FliT [85] control succinate utilisation (Figs 3 and S2B), suggesting a link between the control of the flagellar machinery and Salmonella central carbon metabolism. In the ΔfliST mutant, RpoS overexpression totally suppressed the Succ⁺ phenotype (Fig 7D) indicating that the regulation is RpoS-dependent. However, Western blots did not show reduced levels of RpoS in the FliST null mutant (S10A Fig). It remains unclear how these protein chaperones inhibit succinate utilisation, and whether FliS and FliT are capable of stimulating the expression or the activity of RpoS.

Perspective

Here, we have revealed that in addition to the RpoS-driven inhibition of succinate utilisation CspC and IscR systems control the utilisation of this infection-relevant C-source. The fact that mutations in the anti-Shine-Dalgarno sequence of two 16S ribosomal RNAs or the presence of ribosome-inhibiting antibiotic chloramphenicol boosted growth with succinate added extra complexity (S2 Text and S11 and S12 Figs). More work will be required to fully unravel the genetic mechanisms by which these redundant regulatory systems impact upon succinate catabolism and/or succinate uptake genes.

These distinct regulatory mechanisms are likely to adjust Salmonella metabolism during the journey of the pathogen through the host; from the colonisation of the gastrointestinal tract to intra-macrophage replication. We showed that the sRNA OxyS antagonises CspC-dependent inhibition. Because OxyS is induced by oxidative stress, our findings raise the possibility that the reactive oxygen species produced in the inflamed gut [86,87] can stimulate Salmonella growth upon microbiota-derived succinate in this niche [3,5]. Despite, the abundance of succinate within infected macrophages [12], the intracellular proliferation of Salmonella does not require succinate catabolic genes [15]. The high levels of intra-macrophage expression of the iscR [88] and rpoS [88,89] lead us to propose that the DctA-driven uptake and catabolism of succinate are repressed in the intra-macrophage niche.

The crucial open question that remains is why Salmonella and other enterobacteriaceae display an aversion to utilising succinate for growth. Because succinate is an important infection-relevant C-source, we hypothesise that the distinct regulatory mechanisms we have identified act in response to various environmental stimuli to fine-tune Salmonella succinate metabolism during the journey of the pathogen through the host. We propose a model that explains how the overlapping regulatory systems allow Salmonella to utilise succinate “at the right place and at the right time” during infection (Fig 8).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. A model depicting the modulation of Salmonella succinate utilisation by multiple environmental stimuli under aerobic conditions.

(A) RpoS expression stimulation by Hfq, PNPase, sRNAs and RbsR inhibit Salmonella succinate utilisation. The sRNAs ArcZ, DsrA and RprA stimulate rpoS mRNA transcription elongation [122] and translation [123] in concert with Hfq. RpoS represses the transcription of several genes of the TCA cycle, including the sdh operon [24,25], inhibiting succinate catabolism and Salmonella growth with this C-source. PNPase presumably represses succinate utilisation indirectly by its role in the stabilization of several Hfq-associated sRNAs [124] and in the translational activation of rpoS [125]. The ribose sensor RbsR stimulates the expression of RpoS, presumably at the post-transcriptional level (Fig 7), repressing growth upon succinate. (B) The sRNAs RyhB-1, Spf and RybB repress sdh mRNA translation [126,127] in concert with Hfq and attenuate succinate dehydrogenase (SDH) synthesis, inhibiting Salmonella growth upon succinate. In addition, PNPase may promote the degradation of the sdhCD mRNA [128]. (C) Under aerobic conditions, succinate is mainly imported by DctA [47]. DctA interacts with the DcuS protein and acts as a co-sensor of C₄-dicarboxylates [66,129]. In the presences of succinate, DcuS/DctA activates the response regulator DcuR, that stimulates the expression of several genes, including dctA [121]. De novo synthetised DctA accumulates and increases the uptake of succinate and DctA accumulation may also stimulate the transcription of succinate utilisation genes in concert with DcuS/R [18]. However, dctA expression is robustly repressed by the both halo- and apo-forms of the iron-sulphur cluster regulator IscR (Fig 6), which is up-regulated under iron limiting conditions [111]. Therefore, IscR plays a pivotal role in the succinate utilisation repression and blocks Salmonella growth with this C-source (D) OxyR is stimulated by oxidative stress and stimulates the expression of the sRNA OxyS. OxyS stimulates Salmonella succinate utilisation by repressing the small RNA binding protein CspC. CspC represses succinate utilisation by a still unknown mechanism.

https://doi.org/10.1371/journal.pgen.1011142.g008

Bacterial strains and growth conditions

Precise details of all the chemicals, reagents, DNA oligonucleotides (primers), plasmids and bacterial strains used in this study are listed in S1 Table. The Salmonella mutant strains were all derivatives of Salmonella enterica serovar Typhimurium strain 4/74 [90]. Strain 4/74 is now available from the UK National Collection of Type Cultures (https://www.culturecollections.org.uk/products/bacteria/index.jsp) as NCTC 14672. All the nucleotide coordinates given for 4/74-derived strains correspond to the published genome: GenBank CP002487.1 [37]. Escherichia coli strains Top10 (Invitrogen) and S17-1 λpir [91] were used as hosts for the cloning procedures.

Unless otherwise specified, bacteria were grown at 37°C with aeration (orbital shaking 220 rpm) in Lennox Broth (LB: 10 g/L BD Tryptone, 5 g/L BD Yeast Extract, 5 g/L NaCl), LBO (10 g/L BD Tryptone, 5 g/L BD Yeast Extract) or in M9 minimal medium [92], prepared with M9 Salts, 2 mM MgSO₄, 0.1 mM CaCl₂ and 40 mM sodium succinate dibasic hexahydrate (succinate) or 40 mM glycerol + 10 mM Succinate, as sole C-sources (henceforth, media M9+Succ and M9+Gly+Succ). Agar plates were prepared with the same media, solidified with 1.5% BD Bacto agar. When required, 10–100 μM of FeCl₃ were added to the M9 media.

To seed the M9-derived media of all the experiments, stationary phase pre-cultures were prepared by inoculating isolated colonies into 5 mL LB (in 30 mL Universal glass tubes) and the cultures were incubated for 6–20 hours at 37°C with aeration. Bacteria were harvested by centrifugation, washed once, and Optical Density at 600 nm (OD₆₀₀) was adjusted to 1 with the minimal medium used for the cultures, or with 1 X Phosphate-Buffered Saline (PBS) to generate a standardised inoculum. Subsequently, bacteria were grown aerobically in conical flasks (topped with aluminium foil) or in Greiner 50 mL plastic tubes (with lids slightly open to allow gaseous exchange). Washed bacteria from the pre-cultures were inoculated as a 1:100 dilution to give a starting OD₆₀₀ of 0.01 (~10⁷ CFU/mL), in a final medium volume corresponding to 10% of the flask/tube capacity, to ensure optimal oxygenation by shaking.

For growth curves in 96-well microplates (Greiner #655180), bacteria grown beforehand for ~6 hours in LB, were washed with the minimal medium used for the cultures, or with PBS, and inoculated to give a starting OD₆₀₀ of 0.01 in 200 μL of medium per well. The microplates were incubated at 37°C with orbital shaking (500 rpm), in a FLUOstar Omega plate reader (BMG Labtech), and the OD₆₀₀ was monitored every 15–30 min, using the appropriate growth medium as blank.

When required, antibiotics were added as follows: 50 μg/mL kanamycin monosulfate (Km), 100 μg/mL Ampicillin sodium (Ap), 25 μg/mL tetracycline hydrochloride (Tc), 20 μg/mL gentamicin sulfate (Gm) and 25 μg/mL chloramphenicol (Cm).

For strains carrying the tetR-P_tetA module, the P_tetA promoter was induced by adding 500 ng/mL of anhydrotetracycline hydrochloride (AHT, from a 1 mg/mL stock solubilised in methanol). The same volume of methanol was added to the mock-induced cultures. To stimulate expression of genes controlled by the P_BAD promoter (e.g. in plasmid pWRG99), 0.2% L-(+)-arabinose was added to the culture. For the strains carrying the plasmid pSW-2 [93], the P_m promoter was induced by adding 1 mM m-toluate (500 mM stock titrated with NaOH to pH 8.0).

Bacterial transformation and Tn5 mutagenesis

Chemically-competent E. coli were prepared and transformed as previously published [94]. Electrocompetent cells were prepared with Salmonella cultures grown in salt free LBO medium and were electroporated, as described previously [95]. After recovery in LB at 37°C (30°C for temperature-sensitive plasmids) transformation reactions were spread on selective LB agar plates and transformants were obtained after incubation at 30°C or 37°C.

For the Tn5 transposon mutagenesis, ultra-competent Salmonella were prepared from LBO cultures grown at 45°C, as previously reported [95,96]. The λpir-dependent plasmid pRL27 [97], that encodes the Tn5 transposase gene (tnp) and the mini Tn5-oriR6K-Km^R transposon (Tn5), was used to generate the Salmonella Tn5 libraries, as follows: 50 μL of ultra-competent Salmonella were electroporated with 500 ng of the non-replicating Tn5 delivery plasmid pRL27. After 1 hour recovery in LB the transformation reactions containing the Tn5-carrying Salmonella were washed in PBS or minimal media and 1% of transformations was spread on LB Km plates to estimate the size of the resulting Tn5 library. The remainder of the Tn5 libraries were stored for further experiments.

Cloning procedures

Enzymes, buffer and kits used are listed in S1 Table. DNA manipulations were carried out according to standard protocols [92]. DNA fragment were purified from enzymatic reactions or from agarose gel using the Bioline ISOLATE II PCR and Gel Kit. Plasmids were extracted with the Bioline ISOLATE II Plasmid Mini Kit. Genomic DNA (gDNA) was isolated from 0.5–1 mL of stationary phase cultures with the Zymo Quick DNA Universal Kit.

For PCR, DNA was amplified with Phusion High Fidelity DNA polymerase, template DNA and 0.5 μM primers, in the presence of 3% Dimethyl Sulfoxide and 1 M betaine. For plasmid/strain verifications by Sanger sequencing, PCR reactions were carried out from bacterial colonies with MyTaq Red Mix 2 X and PCR fragment were Sanger sequenced with the appropriate primers (Lightrun service, Eurofins Genomics).

DNA digestion/ligation procedures or the restriction-free “PCR cloning” technique [98] were used to insert DNA fragments into the plasmids pEMG [93], pJV300 (pP_L) [60] and pXG10-SF [99]. E. coli Top10 was used as host for the construction of pXG10-SF and pP_L derived plasmids, while S17-1 λpir was used for pEMG derivatives.

The construction of each plasmid is detailed in S1 Table. For complementation experiments, the genes of interest (including their native ribosome binding site) were PCR-amplified from 4/74 gDNA and were cloned between the NsiI and XbaI sites of the low copy plasmid pXG10-SF, resulting in plasmids phfq (pNAW45), prpoS (pNAW95), piraP (pNAW98), pfliST (pNAW94), prbsR (pNAW93), piscR (pNAW96) and pcspC (pNAW92). For the construction of the piscR^3CA plasmid (pNAW97, carrying three Cys➔Ala substitution at positions 92, 98 and 104 in IscR), two fragments carrying the appropriate mutations were first amplified with primer pairs NW_461/NW_467 and NW_462/NW_466 and the resulting amplicons were fused by overlap extension PCR [100] before insertion into pXG10-SF. In all these plasmids, the genes of interest were under the control of the strong constitutive promoter P_{L tetO-1} [101]. For the construction of ppnp (pNAW256), the pnp gene and its promoter region (including sraG) were amplified and inserted between the XhoI and XbaI sites of pXG10-SF. In the resulting plasmid, pnp expression was controlled by its native promoter. For the construction of poxyS (pP_L-oxyS, pNAW255), oxyS was PCR amplified and inserted by PCR cloning downstream of constitutive P_{L lacO-1} promoter of the pP_L vector, as described earlier [102]. The plasmid-borne translational fusion yobF::_sfgfp (pyobF::_sfgfp, pNAW258) was constructed by cloning the 5’-UTR of yobF and its 31 first codons in frame with the _sfgfp gene of pXG10-SF, as previously described [99].

Plasmid site-directed mutagenesis [103] was used to introduce the CC➔GG mutation in the pP_L-oxyS: complementary primers NW_1022 and NW_1023, carrying the mutations were annealed and elongated by PCR for 10 cycles. Fifty nanograms of plasmid poxyS (pP_L-oxyS, pNAW255) were added to the reaction and the PCR reaction was resumed for 25 cycles. After DpnI treatment and transformation in E. coli Top10, the mutated plasmids pP_L-oxyS^GG (pNAW259) was obtained.

For the construction of the empty vector pNAW125 (pXG10-SF lacking the lacZ¹⁸⁶::_sfgfp fragment), a 3.5 kb fragment was PCR-amplified from pXG10-SF with the primer pair NW_348/NW_565, the resulting fragment was digested with NsiI and self-ligated.

All the pXG10-SF, pEMG and pP_L derived plasmids were verified by Sanger sequencing: primers pXG10_R2 and pZE-CAT were used for plasmid pXG10-SF, primers M13_-40_long and M13_Rev_long for plasmid pEMG and primer pZE-A for plasmid pP_L.

Genome editing techniques

The λ red recombination methodology was used to insert or delete genes in the Salmonella chromosome, using the heat-inducible λ red plasmid pSIM5-tet [104]. Salmonella carrying the pSIM5-tet were grown in LBO at 30°C and electrocompetent bacteria were prepared after a 15 minute heat shock at 42°C, as previously described [95,105]. PCR fragments carrying a resistance gene were PCR-amplified from the template plasmids pKD4, pKD3, pNAW52, pNAW55 or pNAW62 [95,106]. Electrocompetent bacteria (50 μL) were transformed with 500–3000 ng of the PCR fragments, and recombinants were selected on LB agar plates that contained the appropriate antibiotics.

The deletions/insertions linked to a selective marker and the Tn5 insertions were transduced to S. Typhimurium strains using the P22 HT 105/1 int-201 (P22 HT) transducing phage, as previously reported [107,108].

When required, antibiotic resistance cassettes flanked by FLP recognition target sites (frt), were removed from the Salmonella chromosome with the FLP recombinase-expressing plasmid pCP20 [109]. Subsequently, the temperature sensitive plasmid pCP20 plasmid was eliminated by a single passage at 42°C.

The construction of each strain is detailed in S1 Table. The marked mutations ΔrpoS::aph, Δhfq::aph, ΔrybB::aph, Δpnp::cat (pnp-539) and ΔarcA::aph were transduced from published 4/74 and SL1344 derivative strains. For the transduction of the Δhfq::aph mutation, the donor strain JH3584 was first complemented with the plasmid phfq (pNAW45), because Hfq is required for P22 transduction [60].

The 4/74 tetR-P_tetA-dctA strain was constructed according to the principle described by Schulte and colleagues [110]. The promoter and the 5’-UTR of dctA (coordinates 3812883–3812972) were replaced by a frt-aph-frt-tetR-P_tetA module from plasmid pNAW55. In the resulting strain, dctA has the tetA ribosome binding site and is controlled by the AHT-inducible P_tetA promoter.

To measure rpoS expression, a chromosomal transcriptional rpoS-gfp⁺ fusion was constructed as follows: the gfp⁺-frt-aph-frt module of pNAW52 (including the gfp⁺ SD) was inserted after the main rpoS transcription start site (TSS), previously mapped at the coordinate 3089613 on the 4/74 chromosome [111].

To measure dctA expression, a chromosomal transcriptional/translational dctA::_sfgfp fusion was constructed: the flexible amino acid linker GSAGSAAGSGEF and the sequence encoding for superfolder GFP (_sfgfp) were fused to the dctA C-term, using a _sfgfp-frt-aph-frt module amplified from pNAW62 [95].

For complementation in oxyS and rpoS null mutants, a single copy of these genes was inserted in a non-transcribed chromosomal locus in two steps: first, an antibiotic cassette was inserted upstream of the oxyS or rpoS promoters, coordinates 4364521 and 3089777, respectively. Then, the aph-oxyS and the cat-rpoS modules were PCR-amplified from the resulting strains and inserted into the non-transcribed pseudogene STM474_1565 (between coordinates 1585970–1586170 on the 4/74 chromosome) [95]. The STM474_1565 gene is also known as STM1553 or SL1344_1483.

For scarless transfer of rrsA^mut and rrsH^mut mutations located in the 16S rRNA genes, the two step λ red recombination-based methodology described by Blank and colleagues was used [112]. The rrsA and rrsH genes (including their promoters) were replaced by a I-SceI-aph module, amplified from plasmid pKD4-I-SceI [107]. The mutations were transduced in 4/74 WT, yielding the strains ΔrrsA::(I-SceI-aph) and ΔrrsA::(I-SceI-aph). The two strains were transformed with the λ red plasmid pWRG99, expressing the I-SceI nuclease in the presence of AHT [112]. Electro-competent cells were prepared with the two Salmonella+pWRG99 strains in the presence of arabinose, as described above. Competent bacteria were electroporated with 2 μg of the rrsA^mut or the rrsH^mut fragments, obtained by PCR from spontaneous mutants SNW245 (rrsA^mut) and SNW246 (rrsH^mut). Replacement of the I-SceI-aph module by the corresponding rrs mutated PCR fragment was selected on LB agar supplemented with Ap and AHT at 30°C. The rrsA^mut or rrsH^mut insertions were confirmed by PCR and Sanger sequencing and the temperature-sensitive pWRG99 plasmid was eliminated by a passage at 42°C.

For other scarless genome editing procedures, the pEMG-based allele exchange technique [93] was used as previously described [107,113]. For the transfer of the dctA^mut1, dctA^mut2 and oxyR^mut mutations, fragments encompassing the mutations and the ~500 bp flanking regions were PCR-amplified from the corresponding spontaneous Succ⁺ mutants. For the construction of the oxyS^GG mutant, two fragments flanking the mutations were PCR amplified with primer pairs NW_1021/NW_1022 and NW_1023/NW_1024. The two amplicons carrying the CC➔GG mutation on one of their extremities were fused by overlap extension PCR. All the PCR fragments carrying the mutations were inserted by digestion/ligation between the EcoRI and BamHI sites of the suicide plasmid pEMG and E. coli S17-1 λpir was transformed with the resulting ligation reactions. The resulting suicide plasmids were mobilised into the recipient Salmonella by conjugation and recombinants were selected on M9-glucose agar plates containing Km. Merodiploid resolution was carried out with the pSW-2 plasmid, as previously described [107]. The relevant mutations were confirmed in Km^S candidates by PCR and Sanger sequencing. Finally, the unstable pSW-2 plasmid was eliminated by 2–3 passages on LB.

Experimental evolution to select Succ⁺ mutants

A 4/74 strain with an additional rpoS copy inserted in the STM474_1565 pseudogene (strain rpoS^2X, SNW226, Cm^R) was transformed with pRL27 and ten libraries of approximately 10,000 Tn5 mutants were grown aerobically in 25 mL of M9+Succ containing Cm and Km in ten 250 mL conical flasks. After 48 hours incubation at 37°C, the cultures were spread on LB + Km plates and isolated colonies were passaged twice on LB agar. Growth on M9+Succ plates was assessed for >10 isolates per library. The Tn5 insertions from fast-growing colonies (Succ⁺ phenotype) were P22-transduced into 4/74 WT, and the Succ⁺ status of the transductants was verified.

For the isolation of Succ⁺ spontaneous 4/74 mutants, bacteria obtained from stationary phase LB cultures were washed with PBS and the OD₆₀₀ was adjusted to 0.1 (~10⁸ CFU/mL). Approximately 10⁷ CFU (100 μL) were spread on M9+Succ agar and the plates were incubated at 37°C, until Succ⁺ large colonies were visible (3–4 days). Alternatively, spontaneous mutants were obtained from liquid M9+Succ cultures (25 mL), inoculated with ~100 Salmonella. The cultures were grown aerobically at 37°C, until substantial growth was observed (typically after 3 days incubation) and the cultures were spread on LB agar plates. All the presumed Succ⁺ spontaneous mutants were passaged twice on LB plates before confirming the Succ⁺ phenotype on M9+Succ agar plates. The rpoS positive status (rpoS⁺) of each mutant was tested by phenotypic assays (see below) and confirmed by PCR and Sanger sequencing, using primers NW_403, NW_252 and NW_252.

The genomes of a collection of rpoS⁺ Succ⁺ mutants were sequenced by the Illumina whole genome sequencing service of MicrobesNG (Birmingham, UK). Mutations were identified using the VarCap workflow [114] available on Galaxy (http://galaxy.csb.univie.ac.at:8080), using the published 4/74 genome as reference. The identified mutations were confirmed by PCR and Sanger sequencing, and were transferred into 4/74 WT using the two scarless genome editing techniques described above. After transfer, the Succ⁺ phenotype of all genome-edited mutants was confirmed.

Phenotypic characterisation of the Succ⁺ mutants

Phenotypes linked to the rpoS status were tested for each of the Succ⁺ Tn5 or Succ⁺ spontaneous mutants. The RpoS-dependent catalase activity was assessed with hydrogen peroxide directly on colonies or with stationary phase LB cultures, as described earlier [115,116]. The RpoS-dependent RDAR (red, dry and rough) morphotype [81] was tested by adding 2 μL of a stationary phase LB cultures on LBO agar plates containing 40 μg/mL of Congo Red. The RDAR morphotype was observed after at least 3 days of incubation at room temperature.

Mapping of Tn5 insertion sites

The Tn5 insertion sites of the Succ⁺ mutants were mapped by an arbitrary PCR approach [117]. For each Tn5 mutant, the arbitrary PCRs were carried out directly from colonies with the primer pair NW_319/NW_320 (0.5 μM each) and MyTaq Red Mix 2 X in a final volume of 20 μL. The arbitrary PCR conditions were: 95°C 120 sec; 6 X [95°C 15 sec 30°C; 30 sec; 72°C 90 sec]; 30 X [95°C 15 sec; 50°C 30 sec; 72°C 90 sec]; 72°C 300 sec; 4°C. The amplicons were purified on column and eluted in 20 μL of water. For the nested PCRs, 2 μL of the arbitrary PCR products were used as template and mixed with primer pair NW_318/NW_321 (0.5 μM each) and MyTaq Red Mix 2 X in a final volume of 40 μL. The second PCR conditions were: 95°C 120 sec; 30 X [95°C 15 sec; 50°C 30 sec; 72°C 90 sec]; 72°C 300 sec. The PCR products were separated by electrophoresis on a 1% agarose gel containing Midori Green for DNA UV-visualization. The most prominent DNA bands were excised, the DNA was purified and Sanger-sequenced with primer NW_318. Insertions were mapped by BLAST, using the 4/74 genome as reference.

Quantification of GFP fluorescence intensity

Strains carrying the chromosomal fusion rpoS-gfp⁺ or the plasmid-borne yobF::_sfgfp translational fusion were grown in the indicated conditions in biological triplicates and bacteria were harvested by centrifugation and re-suspended in the same volume of PBS. The GFP signal was measured with a FLUOStar Omega plate reader (BMG Labtech) with 200 μL of bacterial suspension per well in black microplates (Greiner #655090). For each strain, the PBS background fluorescence was subtracted from the GFP signal (in arbitrary unit [a.u.]). The fluorescence values were divided by the OD₆₀₀ of the cell suspensions. The fluorescence background of a WT unlabeled strain (carrying the empty plasmid pNAW125, when required) was measured similarly, and was subtracted from the fluorescence signal of the GFP-labelled strain. For each strain the GFP fluorescence intensity is represented as absolute values (GFP fluorescence intensity/OD₆₀₀ [a.u.]) or as a relative GFP fluorescence intensity (in %).

To measure the activity of the dctA::_sfgfp chromosomal fusion, bacteria were grown in M9+Gly+Succ to OD₆₀₀ 0.5–1. Cells were harvested and re-suspended in PBS, prior to fixation with 4% paraformaldehyde and washes with PBS [118]. To measure the GFP fluorescence intensity in all the strains carrying dctA::_sfgfp, the IntelliCyt iQue Screener PLUS (Sartorius) was used. To quantify the dctA::_sfgfp activity more precisely in the WT, dctA^mut1, dctA^mut2, ΔiscR and ΔrpoS genetic backgrounds, bacteria were grown in biological triplicates and were fixed with paraformaldehyde. The FITC-H GFP fluorescence intensity (median of the population) was measured using a FACSCanto II flow cytometer (BD Biosciences). The fluorescence background of a WT unlabelled strain was subtracted from the fluorescence intensity of each dctA::_sfgfp carrying strain. The data are represented as the GFP fluorescence intensity of each mutant, relative to the intensity of the WT isogenic strain (%). All flow cytometry data were analysed using the FlowJo software (BD Biosciences).

For fluorescence microscopy, the dctA::_sfgfp-carrying strains were grown in M9+Succ or M9+Gly+Succ and bacteria were immobilised in PBS solidified with 0.75% low melting point agarose. Pictures were taken with the EVOS FL cell imaging system (Thermo Fisher), as previously described [102].

RpoS detection by Western blotting

The strains of interest were grown in the indicated condition, and bacteria (~10⁹ CFU, estimated by OD₆₀₀) were pelleted by centrifugation and stored at -80°C. Bacteria were re-suspended in 100 μL of PBS and 10 μL of the cell suspensions were mixed with 990 μL of PBS to measure the OD₆₀₀ 1/100 of each suspension. Bacteria were lysed by adding 100 μL Laemmli Buffer 2 X [120 mM Tris-HCl pH 6.8, 4% (wt/vol) SDS, 20% (vol/vol) glycerol, Bromophenol blue 0.02% (wt/vol)] and 10 μL β-mercaptoethanol (5% vol/vol final). The lysates were boiled for 15 min, chilled on ice for 1 min and spun down for 5 min at 4°C (14,000 rpm). Bacterial extracts were separated by SDS polyacrylamide gel electrophoresis and the proteins RpoS and DNaK were detected by Western Blotting, as described earlier [105]. The volume of protein extract loaded (~10 μL) on the SDS 10% polyacrylamide gel was normalised by the OD₆₀₀ of the bacterial suspensions prior to lysis.

Catalogue numbers of all antibodies are listed in S1 Table. The primary antibodies, Anti-E. coli RNA Sigma S Antibody (diluted 1:5,000) and anti-DnaK mAb 8E2/2 (diluted 1:10,000), were used for the detection of RpoS and DnaK (loading control), respectively. For detection, the secondary antibody Goat anti-mouse IgG (H + L)-HRP (diluted 1:2,500) and the Pierce ECL Western blotting substrate were used. The chemiluminescent reaction was detected with the ImageQuant LAS 4000 imaging system (GE Healthcare Life Sciences).

OxyS detection by Northern blotting

The WT, ΔoxyR and oxyR^mut strains were grown in 25 mL of LB to OD₆₀₀ = 1 and the cultures were split in two 10 mL subcultures. For each strain, hydrogen peroxide (H₂O₂, 2 mM) was added to one of the subcultures. After 30 min at 37°C, cellular RNA transcription and degradation processes were stopped by adding 4 mL of ice-cold STOP solution (95% ethanol + 5% acid phenol) to the 10 mL cultures. After a 30 min incubation on ice, bacteria were pelleted by centrifugation and total RNA was extracted with Trizol, as described previously [102].

Probe synthesis and OxyS sRNA detection by Digoxigenin (DIG)-based Northern blotting were carried out with the DIG Northern Starter Kit, according to the DIG Application Manual for Filter Hybridization (Roche) and a previous study [102]. Briefly, heat-denatured RNA (2.5 μg) was separated on an 8.3 M urea, 7% polyacrylamide gel in TBE 1X. RNA was transferred to a positively charged nylon membrane with the Bio-Rad Semi Dry transfer system (#170–3940). RNA was UV-crosslinked to the membrane before hybridization with the DIG-anti-OxyS probe in DIG Easy Hyb buffer at 68°C for 20 hours. The membrane was washed and the OxyS transcripts were detected using the Anti-Digoxigenin antibody and the CDP-Star substrate. Finally, the chemiluminescent reaction was visualised using the ImageQuant LAS 4000 imager. After OxyS detection, the membrane was stripped and re-probed with the DIG-anti-5S probe to detect the 5S ribosomal RNA, used as a loading control. The ssRNA DIG-labelled probes DIG-anti-OxyS and DIG-anti-5S were synthesised with the T7 polymerase and DNA templates obtained by PCR with template 4/74 gDNA and primer pairs NW_485/NW_485 and DH58/DH59, respectively.

Quantification and statistical analysis

Numerical data were plotted and analysed using Microsoft Excel (version 16.46) and GraphPad Prism 9.3.1. Data are presented as the mean of three to six biological replicates ± standard deviation, as indicated in the figures. The unpaired t-test was used to compare the groups and statistical significance is indicated on the figures. P values (two-tail) are reported using the following criteria: 0.0001 to 0.001 = ***, 0.001 to 0.01 = **, 0.01 to 0.05 = *, ≥ 0.05 = NS