The effect of two ribonucleases on the production of Shiga toxin and stx-bearing bacteriophages in Enterohaemorrhagic Escherichia coli

Enterohaemorrhagic Escherichia coli (EHEC) comprise a group of intestinal pathogens responsible for a range of illnesses, including kidney failure and neurological compromise. EHEC produce critical virulence factors, Shiga toxin (Stx) 1 or 2, and the synthesis of Stx2 is associated with worse disease manifestations. Infected patients only receive supportive treatment because some conventional antibiotics enable toxin production. Shiga toxin 2 genes (stx2) are carried in λ-like bacteriophages (stx2-phages) inserted into the EHEC genome as prophages. Factors that cause DNA damage induce the lytic cycle of stx2-phages, leading to Stx2 production. The phage Q protein is critical for transcription antitermination of stx2 and phage lytic genes. This study reports that deficiency of two endoribonucleases (RNases), E and G, significantly delayed cell lysis and impaired production of both Stx2 and stx2-phages, unlike deficiency of either enzyme alone. Moreover, scarcity of both enzymes reduced the concentrations of Q and stx2 transcripts and slowed cell growth.


Results
Growth profile of EHEC with deficiency in RNase E, RNase G, or both enzymes in MMC-treated cultures. The RNase G-encoding gene (rng) was deleted in the EHEC strains TEA028 (parental) and TEA028rne to generate the strains TEA028-Δrng and TEA028-rne-Δrng, respectively (Supplementary Table S1). RNase E is IPTG-inducible in TEA028-rne, producing normal levels of RNase E in medium supplemented with 100 µM IPTG, but low levels of RNase E at or below 1 µM IPTG 49,50 . Deletion of the rng gene in TEA028-rne-Δrng did not change the inducible expression pattern of RNase E ( Supplementary Fig. S1).
To induce the stx2-phage lytic cycle, cultures of the TEA028 parental strain and its RNase E and G derivatives were treated with subinhibitory concentrations of MMC (1 µg/mL). The strains were grown to optical density at 600 nm (OD 600 ) of 0.30-0.35 (time 0), and then MMC was added to an aliquot of the cultures. Thereafter, the turbidity of the cultures was determined at various time intervals (Supplementary Fig. S2).
The difference of turbidity readings between MMC-treated and non-treated cultures is a measure of cell lysis progression after MMC addition due to activation of the phage lytic cycle. As reported previously 50 , RNase E deficiency resulted in slower rate of lysis (Fig. 1, compare TEA028 cells vs. TEA028-rne cells at 0.1 µM IPTG). The absence of RNase G in RNase E-deficient cells (TEA028-rne-Δrng at 0.1 µM IPTG) provoked a substantial delay in cell lysis; in contrast, the absence of RNase G alone (TEA028-Δrng) had no effect. Complementation of the rng deletion with the rng gene expressed from a plasmid in the TEA028-rne-Δrng (prng) strain resulted in a similar rate of cell lysis to TEA028-rne.
Supplementation with 100 µM IPTG did not affect significantly the response to MMC treatment of the strains TEA028 or TEA028-Δrng ( Supplementary Fig. S3). Likewise, the strains TEA028-rne at 100 µM IPTG (normal Figure 1. Growth of EHEC TEA028 and its RNases E and G derivative strains in cultures treated or nontreated with mitomycin C (MMC). Cultures of TEA028 (parental strain) and its RNases E and G derivatives were grown in Luria Bertani medium to optical density at 600 nm (OD 600 ) of 0. 30-0.35 at which point the cultures were split and an aliquot was treated with MMC (1 µg/mL) to induce the stx2-phage lytic cycle. Thereafter, samples were collected at various time points to measure OD 600 . The strains TEA028-rne and TEA028-rne-Δrng underproduce RNase E when the medium is supplemented with low levels of isopropyl β-d-1-thiogalactopyranoside (IPTG), as indicated. For each time point, the OD 600 of MMC-treated cultures was subtracted from the OD 600 of non-treated aliquots (ΔOD 600 ). Means and standard errors or at least 4 biological replicates are graphed. Lack of error bars indicates that the standard error was smaller than the plot symbol.  Fig. S4) in agreement with our previous results 50 .
The delayed kinetics of cell lysis after induction of the phage lytic cycle suggested that deficiency of RNases E and G could impair the production of stx2-phages, Stx2, or both.
Production of stx-phages in EHEC with deficiency in RNase E, RNase G, or both enzymes in MMC-treated cultures. Phage yields were determined in culture supernatants of TEA028 and its RNases E and G derivatives after MMC-treatment for 6 h. After treatment with MMC, EHEC TEA028 produces stx2phages, which were detected by plaque hybridization under the same experimental conditions in a previous study 50 ; however, plaques of stx1-phages were not detected. RNase E deficiency (TEA028-rne at 0.1 µM IPTG) impaired the production of infectious phages in agreement with our previous findings ( Fig. 2) while TEA028-rne at 100 µM IPTG (normal RNase E levels) produced similar phage yields to the control ( Supplementary Fig. S4). Deficiency of both RNase E and RNase G (TEA028-rne-Δrng at 0.1 µM IPTG) reduced phage yields even further, whereas the absence of only RNase G caused no effect. Complementation of the rng deletion partially restored the production of phages (TEA028-rne-Δrng (prng) vs. TEA028-rne-Δrng at 0.1 µM IPTG).
Production of toxin in EHEC with deficiency in RNase E, RNase G, or both enzymes in MMC-treated cultures. Next, Stx2 production was measured in extracts prepared from MMC-treated cultures of TEA028 and its RNase E and RNase G derivatives. Specifically, the EHEC strain TEA028 and its derivatives produce Stx2a subtype. At 6 h after MMC addition, lysates from the strains lacking RNase G (TEA028-Δrng) or underproducing RNase E (TEA028-rne at 0.1 µM IPTG) had toxin levels that were not significantly different from TEA028 (control) (Fig. 3). Similarly, strain TEA028-rne at 100 µM IPTG (normal RNase E levels) produces toxin as the control ( Supplementary Fig. S4, and Ref. 49 ). Deficiency of RNase E and G (TEA028-rne-Δrng at 0.1 µM IPTG) resulted in 245 ± 36 ng/mg of protein (mean ± standard error) contrasting with the production of 4,677 ± 173 ng/mg of protein (mean ± standard error) by TEA028. Because of the slower kinetics of growth and lysis in TEA028-rne-Δrng at 0.1 µM IPTG, Stx2 was also measured in cell extracts at later time points (10 and 24 h) after MMC addition. Toxin production by TEA028 was not measured beyond the 6-h period because at this point cells are mostly lysed, and thus, a small increment or no increment in Stx2 accumulation would be expected at later time points. In cultures of TEA028-rne-Δrng at 0.1 µM IPTG, toxin accumulation increased 1.7-fold at 10 h and 2.2-fold at 24 h when compared with the 6-h time point. However, there was still about eightfold difference in the mean of Stx2 levels between cell lysates of the control (TEA028) at 6 h and cell lysates of TEA028-rne-Δrng at 0.1 µM IPTG at 24 h. For comparison, Stx2 concentrations under RNase E scarcity alone (TEA028-rne at 0.1 µM IPTG) were still high at 10 h. Complementation of the rng deletion in the TEA028-rne-Δrng (prng) strain, partially restored Stx2 production at 6, 10, and 24 h after MMC addition. Partial complemen- www.nature.com/scientificreports/ tation in the production of toxin and phages may be caused by higher than normal physiological levels of RNase G produced in the complemented strain. To examine this possibility, RNase G concentrations were analyzed by Western blotting in TEA028, its rng deletion mutant derivatives, and in an E. coli laboratory strain overexpressing RNase G as a positive control. Supplementary Fig. S5 shows RNase G was detected as a prominent band of the expected molecular weight in the positive control. However, RNase G was undetectable in the parental strain TEA028, but it was detected in the complemented strain. Attempts to improve the sensitivity of detection in TEA028 failed. Therefore, RNase G in EHEC is at low physiological levels in accordance with previous data in nonpathogenic E. coli 51 . As a strategy to reduce RNase G concentrations in the complemented strain, the ATG start codon of the rng gene was mutated to GTG. In E. coli, the GTG codon reduces the efficiency of translation initiation and is a more infrequent start codon than ATG 52 . Complementation with the GTG-mutated rng gene resulted in higher toxin yields than complementation with the wild-type variant ( Supplementary Fig. S6), suggesting that higher-than-normal levels of RNase G in the TEA028-rne-Δrng (prng) strain caused the partial complementation result.
Quantification of mRNA levels of the stx2, recA, and Q mRNAs in EHEC and its RNases E and G derivatives. Reverse transcription real-time PCR (RT-qPCR) was used to examine whether the reduction in Stx2 levels under RNases E and G deficiency could be explained by a concomitant reduction in stx2 mRNA levels. In addition, recA and Q transcript levels were also examined since expression of these genes is essential for initiation of the stx2 phage lytic cycle and Stx2 production when DNA damage occurs. Cultures of each strain were grown to OD 600 of 0.3-0.35 (time 0) at which point MMC (1 µg/mL) was added; thereafter, aliquots were collected at various time points for total RNA extraction and RT-qPCR analysis. For each gene, fold change was calculated as the cDNA present at each time point with respect to the cDNA at time 0 (described in "Methods" section).
In the TEA028 strain, the stx2B transcript reached high levels 1 h after the addition of MMC (Fig. 4a). Similar stx2B mRNA kinetics were observed in strains TEA028-rne at 100 µM IPTG (Supplementary Fig. S7b) and TEA028-Δrng ( Supplementary Fig. S7c). Under RNase E deficiency, high transcript levels were observed at 2.5 h and later time intervals, after an initial delay (Fig. 4d). In contrast, deficiency of both RNases reduced significantly the levels of stx2 transcripts, with 50-fold decrease at 2.5 h when compared with the control (TEA028) (Fig. 4a, g).
The recA transcripts showed similar kinetics in the TEA028, TEA028-rne at 100 µM IPTG, and TEA028-Δrng strains, with high levels at 30 min and falling afterwards to a fold change below 1 in some experiments (Supplementary Fig. S7a, S7b, and S7c). Depletion of RNase E or of RNases E and G did not reduce recA transcript levels, and fold changes were above 1 at all time points in most of the experiments (Fig. 4e,h). In particular, recA transcript levels stayed significantly elevated at three time points under RNases E and G scarcity when compared with the control (TEA028) (Fig. 4b,h).
The kinetics of Q transcript levels were similar in TEA028, TEA028-rne at 100 µM IPTG, and TEA028-Δrng ( Supplementary Fig. S7a, S7b, and S7c), resembling the expression pattern of stx2 mRNA. Under RNase E scarcity, there was a slight delay in the Q mRNA peak, but the levels were as high as in the parental strain at later time points (Fig. 4f). Under RNases E and G deficiency, the Q transcript levels were significantly lower at all time points with about 42-fold reduction at 2.5 h when compared with TEA028 (Fig. 4c,i). These findings indicate www.nature.com/scientificreports/ that RNases E and G scarcity significantly impaired Q expression, which could have contributed to reduction in stx2 transcription and Stx2 production.

Determination generation times in EHEC and its RNases E and G derivative strains.
In addition to the reduction in Q transcript levels, the impairment of Stx2 production under RNases E and G deficiency may be an indirect consequence of increased generation times from the burden of slowed RNA metabolism. In support of this hypothesis, the generation time was significantly increased in TEA028-rne-Δrng at 0.1 µM IPTG when compared with the control or deficiency of RNase E alone ( Supplementary Fig. S8).

Discussion
This study establishes that depletion of RNases E and G significantly impair the production of both Stx2 and stx2phages in EHEC. In contrast, the absence of RNase G does not impair the production of either toxin or phages, and deficiency of RNase E impairs phage production only (this study and Ref. 50 ). The inducing agent used in this www.nature.com/scientificreports/ study, MMC, strongly triggers the cellular response to DNA damage. Thus, cells deficient in both RNases still undergo the initiation of the phage lytic cycle with concomitant toxin production although at a slower rate and resulting in lower Stx2 yields even at 24 h after MMC addition. RecA and Q proteins are critical for the initiation of the phage lytic cycle and toxin production. The reduced stx2 transcript levels under RNase E and G deficiency cannot be explained by a change in recA transcript levels, which remained higher than in the control. In contrast, RNase E and G scarcity caused a drastic reduction in transcript levels of the Q antiterminator, which is required for transcription to continue to stx2 genes. This result can explain, at least partially, the observed decrease in stx2 transcript concentrations. In addition, the impaired expression of Q could have affected the transcription of late genes required for the assembly of phage particles. The reduction of Q transcript levels could be the result of impaired Q gene transcription or transcript stability. If the latter scenario is correct, RNases E/G effect on Q mRNA lifetimes is probably indirect because they are required for Q transcripts to reach high levels. Deficiency of both RNases likely creates metabolic burden by affecting the stability and processing of many RNAs. As a consequence, the growth rate is decreased, probably having an indirect effect on the production and release of toxin and phages. The mechanisms underlying these effects are probably many and complex since depletion of RNase E alone changes the half-life of thousands of transcripts in nonpathogenic E. coli 53 . In EHEC, direct RNase E or G degradation or processing of transcripts encoding virulence factors, such as Shiga toxin, has been scarcely explored 37,38,54 .
The importance of the production of actual phage particles in disease progression is debatable. Some studies indicate that infection of commensal bacteria by stx phages may amplify the production of toxin 55,56 , potentially worsening disease outcomes. However, subsequent work with a microbiome-replete murine model of EHEC infections indicates that the actual production of phage particles is negligible for disease 57 . Nevertheless, potential transmission of stx genes to naive bacterial hosts is problematic, as exemplified by the novel and highly virulent Stx2-producing E. coli serotype O104:H4 that caused an outbreak in Europe in 2011 58 . Therefore, the reduction of both toxin and infective phage particles produced under RNases E and G scarcity is important.
Despite remarkable advances in the understanding of EHEC pathogenesis and its virulence factors, specific therapeutics against EHEC infections are still lacking. The neutralization of Stx by a binding agent (Synsorb P) was ineffective in a multicenter randomized controlled clinical trial, and the authors concluded that success of similar strategies was doubtful 59 . Although humanized monoclonal antibodies against Stx1 or Stx2 showed efficacy at controlling fatal complications in animal models [60][61][62][63] , clinical trials have not advanced beyond the determination of safety and pharmacokinetic parameters [63][64][65] . In the United States, antibiotics are contraindicated for treatment of Stx-producing E. coli infections 28 .This topic remains controversial since there are some conflicting data from in vitro, animal model and clinical studies (reviewed in 66 ). Nevertheless, it is clear that the effect of antibiotics on Stx production depends on the antimicrobial compound and the particular Stx-producing E. coli strain. Given the complexity of the problem, the therapeutic potential of the findings reported here is unknown, at this moment. Inhibitors of purified RNase E/G have been isolated 67,68 ; however, there are currently no reported inhibitors of those enzymes that can be tested in vivo, i.e., on the EHEC capacity to produce toxin.
Oligonucleotides. The oligonucleotide sequences used in this study are described in Supplementary   Table S2.

Construction of TEA028-Δrng and TEA028-rne-Δrng strains.
The rng gene from strainsTEA028 and TEA028-rne was deleted following the gene-doctoring procedure 69 . The kanamycin (kan) resistance gene was amplified with primers f-EcoRI-rng and r-XhoI-rng (Supplementary Table S2) using plasmid pDOC-K as the template. The PCR fragment was cloned between the EcoRI and XhoI sites of plasmid pDOC-C (or pDOC-C-Gen [see below]), which carries an amp resistance marker, and the sacB gene that confers resistance to sucrose. The resulting plasmid was then introduced into the strain TEA028. Next, TEA028 was transformed with the plasmid pACBSCE, which carries the genes encoding for the λ-Red proteins and the gene encoding for the restriction enzyme I-SceI. Cultures were grown for 2 h at 37 °C, and then the cells were pelleted and resuspended in LB supplemented with 0.5% L-arabinose to induce the expression of λ-Red proteins. Appropriate dilutions were plated to select for kan resistant and sucrose insensitive recombinants. The colonies were also checked for loss of the donor plasmids. Next, the recombinants were screened with PCR primers CC1 and CC2 and with complementary primers to the rng gene flanking region as described by Lee et al. 69 . To eliminate the kan resistance marker inserted into the rng gene, the Flp recombinase was produced from the pCP20 plasmid (E. coli Genetic Stock Center), which was then eliminated by incubation at the restrictive temperature. To delete the rng gene in the TEA028-rne strain, which is amp resistant, the pDOC-C plasmid was modified by introducing a gentamicin resistance gene (aacC1). The aacC1 gene was amplified with PCR primers f-SphI-gen and r-SphI-gen  Table S2) from plasmid pBAMD1-6 (Addgene). The PCR fragment was cloned into the SphI restriction site of plasmid pDOC-C to generate plasmid pDOC-C-Gen. To delete the rng gene from the TEA028rne strain, the same procedure described above was followed, except the LB medium was supplemented with 100 µM IPTG to induce production of RNase E. After the rng gene was deleted, we introduced into the TEArne-Δrng strain the plasmid pLacI Q which carries the lacI Q gene or the plasmid prng which carries the lacI Q gene and the rng gene. The LB plates were supplemented with antibiotics as described by Lee et al. 69 and gentamicin at 10 µg/mL, when appropriate.
Construction of plasmid prng and site-directed mutagenesis. The rng gene was amplified using genomic DNA as template and primers f-XbaI-rng and r-HindIII-term-rng. The M13 transcriptional terminator, which has high termination efficiency, was added at the end of the primer r-HindIII-term-rng 70  RNA isolation and mRNA quantification. Culture samples (3, 5, or 6 mL) were collected at 4 °C and added in a 10:1 ratio to a mixture of 95% ethanol and 5% saturated phenol. After centrifugation, the cell pellets were stored at − 70 °C. To extract total RNA, the cell pellets were resuspended in 300 µL of lysis buffer (10 mM Tris-Cl, pH 8.0, 1 mM EDTA, 0.5 mg/mL lysozyme), 20 µL of 20% SDS, and 20 µL of water. The lysate was incubated at 64 °C for 2 min, and then 1.2 mL of TRI Reagent solution (Invitrogen) was added. Next, total RNA was isolated with the Direct-zol RNA Miniprep kit (Zymo Research). On-column DNA digestion was performed during the RNA extraction following the manufacturer's directions. RNA was eluted in water, and the concentration determined with a NanoDrop 2000c spectrophotometer. To digest remaining DNA contamination, an aliquot of each RNA sample was treated with TURBO DNase (Ambion) in a reaction mixture containing SUPERase•In RNase Inhibitor (Thermo Scientific). The total RNA was then repurified using saturated phenol and chloroform following standard procedures and stored at − 70 °C. The RNA integrity was checked by agarose gel electrophoresis, and potential DNA contamination was detected by qPCR amplification with primers f-qPCR-stx2B and r-qPCR-stx2B (Supplementary Table S2). cDNA was synthesized with the iScript cDNA Synthesis Kit (Bio-Rad) following the manufacturer's instructions. The stx2B, recA, Q, and rrsH genes were amplified by qPCR from appropriate dilutions of cDNA using the primers described in Supplementary Table S2. The reactions (10 µL) were performed in a CFX Connect real-time detection system (Bio-Rad) using the iTaq Universal SYBR Green Supermix (Bio-Rad). The cycling protocol was 95 °C for 3 min and then 40 cycles at 95 °C for 5 s and 60 °C for 30 s. A melting curve analysis was completed afterwards to confirm the presence of only one product of amplification and the absence of primer dimers. The initial concentration of molecules (N 0 ) was calculated using the open-source software LinReg 71,72 . Then, the N 0 of stx2B, recA, or Q cDNA was normalized to the N 0 of the rrsH cDNA for each experiment. For each gene, the fold change was calculated as the ratio of the normalized N 0 at each time point after MMC addition to the normalized N 0 before the MMC addition (time 0). Fold changes were log-transformed before statistical analysis.

Statistical analysis.
The results of this study are derived from at least three biological replicates. The statistical analysis was performed with the software GraphPad Prism 7.0. Groups were compared using one-way or two-way analysis of variance and Dunnett's or Sidak's multiple comparison test, or a t-test when two groups were compared. Measurements of toxin, protein concentrations, and qPCR were performed in three technical replicates.

Data availability
Data are available from the author upon request.