The bacterial strains used in this study were derivatives of the parent strain UTI89. They were routinely cultured in Luria-Bertani (LB) broth (10 g bacto-tryptone, 5 g yeast extract, and 10 g NaCl/L) at 37°C, 200 rpm. Bacterial stock stored at −80°C was transferred into fresh LB medium and grown overnight before being used for persister experiments.

The deletion mutants of Hfq-dependent sRNAs were constructed using the λ Red recombination system as described by Datsenko and Wanner (2000). Primers used to amplify all knockout-DNA fragments and verify the correct constructs by polymerase chain reaction (PCR) are shown in Supplementary Tables 1, 2. To create double-deletion mutants, the chloramphenicol-resistance gene was removed from plasmid pCP20.

The arabinose-inducible plasmid pBAD202 was used to construct overexpression strains according to previous report (Ma et al., 2010). Primers (F: 5′-CATG CCATGGAAAAGCCAGCACCCGGC-3′ and R: 5′-CCCGGAATTCGCGATCAGGAAGACCCTCG-3′) used for the construction of the plasmid containing the RyhB gene were designed in this study. Genes were amplified with PCR primers, followed by digestion of both the PCR fragments and pBAD202 with the restriction enzymes NcoI and EcoRI (New England Biolabs, Ipswich, MA, USA) and ligation using the T4 DNA ligase (New England Biolabs). The new constructs along with the empty vector, pBAD202, were transformed into parent strain UTI89 for overexpression experiments. The deletion mutants and overexpression strains were verified by DNA sequencing. Arabinose (0.1%) was added to the cultures of overexpression strains to induce the conditional expression after bacterial culture for appropriate time.

Persister levels were determined by counting the number of colony forming units (CFUs) that grew on LB agar plates as previously described (Ma et al., 2010; Wang et al., 2015). The antibiotics levofloxacin (5 μg/mL), cefotaxime (128 μg/mL), gentamicin (30 μg/mL) were added directly to cultures at the exponential (about 3 h of cultivation, ~10⁸ CFU/mL) or stationary (10 h of cultivation, ~10⁹ CFU/mL) phase unless otherwise stated. Aliquots of the bacterial cultures exposed to antibiotics were incubated at 37°C at different time points and washed in phosphate-buffered saline before plating on LB plates in the absence of antibiotics to determine CFU count.

For heat shock, bacteria were placed in a water bath at 55°C for 3 h. For acid stress (pH 3.0) and hyperosmosis (NaCl, 4 M), cultures were washed twice with acid or hyperosmotic LB medium and resuspended in the same column of corresponding LB medium, respectively. Aliquots of the bacterial cultures exposed to various stresses were incubated at 37°C at different time points and washed in phosphate-buffered saline before plating on LB plates in the absence of antibiotics to determine CFU count.

Bacteria used for real-time PCR (RT-PCR) analysis were routinely cultured for 10 h in LB medium, followed by centrifugation at 5,000 rpm and 4°C to remove the supernatant. RNA Protect bacteria reagent (Qiagen, Hilden, Germany) was added to resuspended cells, and total RNA was isolated from cells using a bacterial RNA kit (Omega Bio-tek, Norcross, GA, USA) according to manufacturer protocol for the real-time PCR experiment. Total RNA was converted to cDNA using the PrimeScript TMRT reagent kit with gDNA Eraser (Takara, Shiga, Japan) according to manufacturer protocol and used as a template to perform real-time PCR on an Applied Biosystems 7500 real-time instrument (Applied Biosystems, Foster City, CA, USA). The primers for real-time PCR are listed in Supplementary Table 3. The 16S rRNA gene rrsB was used as the reference gene, and changes in expression were presented as the average of three biological replicates.

RNA extraction, quality assessment, sequencing, and analysis were performed by Shanghai Biotechnology Corporation (Shanghai, China) using method described previously (Xu et al., 2016). Briefly, total RNA was extracted using an RNeasy mini kit (Qiagen) according to manufacturer protocol, and the RNA integrity number was assessed using an Agilent Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA). Qualified total RNA was further purified using an RNA clean XP kit (Beckman Coulter, Brea, CA, USA) and a RNase-free DNase set (Qiagen). rRNA removal, fragmentation, synthesis of the second strand, adenylation of 3′ ends, adapter ligation, and amplification were prepared prior to sequencing using an Illumina HiSeq 2500 (Illumina, San Diego, CA, USA). Reads were assessed and quantified before mapping to genome of E. coli UTI89.

Intracellular ATP concentration was determined according to method described previously (Shan et al., 2017). At the appropriate time points, the BacTiter-Glo microbial cell viability assay kit (Promega) was used following the recommended protocol. Briefly, 100 μL BacTiter-Glo reagent and 100 μL cell-culture medium were mixed in 96-well plates on an orbital shaker and incubated for 5 min, followed by determination of luminescence using the SpectraMax Paradigm multi-mode detection platform (Applied Biosystems). The number of bacteria in the 100 μL cell-culture was also determined at the corresponding time. After that, the luminescence of a single cell was calculated.

Intracellular NAD⁺ and NADH concentrations were measured according to methods described previously (Vilcheze et al., 2005; Maeda et al., 2017). For NAD⁺ and NADH extraction, bacteria were cultured to the designed timepoints and harvested by centrifugation at 14,000 rpm at 4°C for 3 min. Using the NAD⁺/NADH Assay Kit (BioAssay Systems, Hayward, CA, USA), the extraction procedure and measurement were performed according to the manufacturer's instructions. The reaction was performed in flat-bottom 96-well plates. The NAD⁺/NADH concentration ratio was calculated based on the measured values of OD₅₆₅.

To explore the role of sRNA in persistence, we constructed the knockout-mutant strains of 20 known Hfq-binding sRNAs (RyhB, GcvB, MgrR, RybB, MicF, SgrS, RprA, DicF, SsrS, FnrS, GadY, DsrA, OmrB, ArcZ, RyeB, RydC, OmrA, MicA, MicC, and ChiX; Mandin and Gottesman, 2010; Kim et al., 2015) in E. coli UTI strain UTI89 and exposed the mutants to bactericidal antibiotic levofloxacin (5 μg/mL) in the stationary phase (cultured for 10 h, ~10⁹ CFU/mL). Among these mutants, the RyhB knockout mutant strain ΔryhB displayed a dramatic decrease in persister levels compared with the parent strain UTI89. Thus, RyhB was selected for further investigation below.

To better understand the role of RyhB in persistence, dynamic responses of the ryhB knockout mutant strain ΔryhB and the parent strain UTI89 upon exposure to antibiotics (levofloxacin, cefotaxime, gentamicin) were examined. In the stationary phase (cultured for 10 h, ~10⁹ CFU/mL), the ΔryhB mutant showed ~10⁵-fold lower persister level than the parent strain UTI89 upon treatment with levofloxacin (5 μg/mL) for 5 days (Figure 1A). The case was also applicable to gentamicin and cefotaxime. After gentamicin (30 μg/mL) or cefotaxime (128 μg/mL) treatment for 5 days, the ΔryhB mutant also had lower persister numbers than UTI89, which had a ~10⁴- and ~10⁵-fold decrease, respectively (Figures 1B,C).

Because persisters are heterogeneous, and the age of the bacterial culture can affect persister level (Li and Zhang, 2007; Luidalepp et al., 2011; Zhang, 2014), we determined whether RyhB influenced persister level in the exponential phase. Bacteria were cultured for about 3 h to reach the density of ~10⁸ CFU/mL before antibiotics were added. Following gentamicin (30 μg/mL) or cefotaxime (128 μg/mL) exposure, we found the survival of the ΔryhB mutant was decreased (~10²- and ~10³-fold, respectively) compared with the parent strain (Figure 1D). These results indicate that RyhB was indeed involved in persistence to a wide range of antibiotics in the exponential phase, as well as in the stationery phase.

Furthermore, the ΔryhB mutant displayed a similar defect in persister-formation capacity during the early stationary phase (cultured for 6 h, ~10⁹ CFU/mL) as well as late stationary phase (cultured for 24 h, ~10⁹ CFU/mL) (data not shown).

To test the effect of stress on the ΔryhB mutant, we subjected the ΔryhB mutant and the parent strain UTI89 to various stresses (hyperosmosis, acid, and heat). In the stationary phase (cultured for 10 h, ~10⁹ CFU/mL), we observed a ~10⁵-fold lower survival of the ΔryhB mutant than the parent strain UTI89 under hyperosmosis (NaCl, 4 M) exposure. In addition, the ΔryhB mutant was also more sensitive to other stresses including heat (55°C) and acid (pH 3.0), with a ~10³-fold lower survival to heat, and a ~10²-fold lower survival to acid, respectively (Figure 2A).

In addition, the effect of stress on the ΔryhB mutant could also be observed in the exponential phase (cultured for about 3 h, ~10⁸ CFU/mL). During hyperosmosis (NaCl, 4 M) or acid (pH 3.0) exposure, the ΔryhB mutant had ~10³-fold lower persister numbers than the parent strain UTI89 (Figure 2B).

To further determine if RyhB is involved in persister formation, we also overexpressed the RyhB in an inducible vector pBAD202-ryhB transformed into UTI89 and assessed their persister levels. We found a higher persister level in the RyhB overexpression strain than the parent strain carrying the empty vector pBAD202 at each antibiotic (levofloxacin, gentamicin, cefotaxime) and stress (hyperosmosis, heat, acid) we determined (Figure 3). As controls, we also determined the survival of all the strains in our experiments without any treatment, and found no difference at each timepoint (data not shown).

To determine whether Hfq was essential for the RyhB-specific persister phenotype, we transformed the RyhB-overexpression vector pBAD202-ryhB or the control empty vector, pBAD202 to an Hfq-knockout strain and subjected them to antibiotic exposure. The initial bacterial numbers were similar before antibiotics were added. The terminal numbers of the group without any treatment were similar too. Our results showed that the ryhB-overexpression strain was more tolerant to antibiotic challenge than the control strain in the absence of Hfq, with a ~10³-fold higher count of viable bacteria remaining following gentamicin (30 μg/mL), levofloxacin (5 μg/mL), or cefotaxime (128 μg/mL) exposure (Figure 4). These results suggest that RyhB-related bacterial persistence occurred independent of the Hfq protein. Additionally, we determined the susceptibilities of the Hfq-knockout strain to antibiotic challenge following exposure to gentamicin (30 μg/mL), levofloxacin (5 μg/mL), or cefotaxime (128 μg/mL). Compared with the parent strain, we observed defective persistence in the Hfq-knockout strain following exposure to each antibiotic (data not shown), which is consistent with a previous study (Kim and Wood, 2010).

The first persistence gene, hipA, was shown to depend upon relA-dependent ppGpp synthesis to mediate persistence (Korch et al., 2003; Maisonneuve et al., 2013). To investigate whether RyhB-related persistence is also dependent upon ppGpp, we constructed a relA-deletion mutant, ΔrelA, a RyhB mutant, ΔryhB, and a double-deletion mutant, ΔrelAΔryhB, and measured their persistence to different antibiotics and stresses. We observed that the number of surviving cells of the ΔrelAΔryhB mutant was significantly lower than either of the two single mutants. The number of viable ΔrelAΔryhB mutant cells decreased by ~10-fold following levofloxacin exposure (Figure 5A), ~10²-fold following cefotaxime exposure (Figure 5B), ~10⁴-fold during hyperosmotic stress (Figure 5C), and 5-fold upon exposure to acid pH (Figure 5D) as compared with either of the two single-mutant strains. These results indicate that the role of RyhB in persistence is independent of ppGpp, and that deletion of ppGpp further aggravated the persistence phenotype associated with ΔryhB mutant. Interestingly, there was no apparent difference between the ΔrelA mutant and the parent strain following antibiotic exposure, except that the cell number of the ΔrelA mutant decreased 4-fold relative to that of the parent strain under hyperosmostic stress.

To investigate whether this phenomenon is unique to the uropathogenic E. coli UTI89 strain, we constructed a knockout strain, ΔrelA in an E. coli K12 W3110 strain, and evaluated its effect on persistence. Our findings revealed that the W3110 ΔrelA strain exhibited apparent defects in persistence as compared with the parent strain W3110 (data not shown). Therefore, these results indicated that the persistence-related role of relA in uropathogenic E. coli UTI89 differed from that in the E. coli K12 strain.

We then determined ryhB-expression level in the UTI89 strain upon challenging stationary phase bacteria with gentamicin (30 μg/mL) or cefotaxime (128 μg/mL) for 1 day (Figure 5E). Upon antibiotic exposure, we observed higher ryhB-expression levels than that in the untreated strain. The ryhB-expression levels were 3.4- and 2.4-fold relative to that of the parent strain upon exposure to gentamicin and cefotaxime, respectively. These results indicated that RyhB may be induced to protect cells in response to antibiotic stress.

Since RyhB is regulated by the Fur repressor in the presence of iron, we wanted to address the relationship between Fur and RyhB on persistence. To test this, we constructed a single-mutant Δfur strain and a double-mutant ΔryhBΔfur strain. We found that the number of surviving cells of the Δfur mutant was more than 10³-fold lower than that in the parent strain upon exposure to cefotaxime (Figure 6A). Furthermore, we observed ~ 10²-fold lower viable cells following gentamicin (Figure 6B) or levofloxacin exposure (Figure 6C) and slightly lower levels (about 4-fold) following hyperosmosis (Figure 6D) in the Δfur mutant than the parent strain. These results were contrary to our hypothesis, suggesting that Fur is unable to repress the RyhB-mediated persistence.

Additionally, we found that the ΔryhBΔfur mutant had >10² times lower CFU than either that of the single-mutants ΔryhB or Δfur following exposure to cefotaxime or gentamicin, as well as hyperosmosis (Figures 6A,B,D). Furthermore, we observed ~10-fold lower cell survival following exposure to levofloxacin (Figure 6C) in the double mutant than that of either of the single mutants. These results indicated that the persister phenotype associated with the ΔryhBΔfur double mutant was more pronounced than that of either of the single ΔryhB or Δfur mutant, suggesting that RyhB exhibits a synergistic effect with Fur in persister formation.

Because our data indicated RyhB involvement in persistence, we wanted to investigate how RyhB is involved in persister formation. Therefore, we analyzed genes regulated by RyhB by comparing the ΔryhB mutant and the parent strain UTI89 through RNA-Seq analysis. We found that ~200 genes were upregulated by at least ≥2-fold in the ΔryhB mutant strain, whereas 130 genes were downregulated compared to the parent UTI89 strain. The upregulated genes were mainly comprised of genes belonging to cell motility, transport, toxin-antitoxin modules, DNA repair, transcription, and metabolism (Table 1).

Among elevated genes in the ΔryhB mutant strain, metabolic genes accounted for ~70% of the total genes according to Kyoto Encyclopedia of Genes and Genomes (KEGG) classification (Figure 7A). For this reason, we hypothesized that the ΔryhB mutant may lead to a metabolically hyperactive state. To test this hypothesis, we determined cellular ATP levels which represent the major energy currency of cells. If the strain is in a metabolically hyperactive state, the ATP level should be high. As expected, the ATP level of the ryhB-knockout strain was over 2-fold higher than that of the parent strain in both exponential and stationary phase cultures (Figure 7B).

To further confirm that RyhB deletion resulted in a metabolically hyperactive state, intracellular NAD⁺/NADH ratios were measured (Maeda et al., 2017). We found ΔryhB mutant showed reduced NAD⁺/NADH ratios in both exponential (2.76-fold decrease) and stationary phase (3.66-fold decrease) cultures (Table 2) when compared with the parent strain. These results indicate that the intracellular environment of the ΔryhB mutant was more reductive compared with the parent strain, suggesting the ΔryhB mutant caused a metabolically hyperactive state.

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.