Characterization and potential mechanisms of highly antibiotic tolerant VBNC Escherichia coli induced by low level chlorination

Escherichia coli is an important pathogenic indicator in drinking water. Viable but non-culturable (VBNC) E. coli induced by low level chlorination was found to have higher antibiotic tolerance. The emerging of VBNC bacteria in drinking water systems is posing challenges to the control of bio-safety. It is necessary to study the underlying mechanisms of VBNC state E. coli under actual residual chlorine condition of drinking water pipe network. In this study, we investigated the changes of morphology and gene expressions that might present such state. The results indicated that the size of VBNC E. coli was not remarkably changed or recovered culturability under favorable environmental conditions. Results from transcriptomic analysis revealed that the regulated genes related to fimbrial-like adhesin protein, putative periplasmic pilin chaperone, regulators of the transcriptional regulation, antibiotic resistance genes and stress-induced genes, rendering VBNC cells more tolerant to adverse environmental conditions. In total of 16 genes were significantly up-regulated under the VBNC state, including three genes encoding toxic protein (ygeG, ibsD, shoB), indicating that VBNC E. coil was still a threat to human. The work is of great relevance in the context of better understanding this poorly understood physiological state.

Bacteria are prime targets to remove from drinking water since waterborne pathogenic microorganisms can cause epidemic diseases. In many countries, chlorine at levels of 0.05-0.5 mg L −1 is generally applied to repress bacterial regrowth in drinking water distribution systems (DWDS) 1,2 so that the number of bacteria in tap water can meet the water quality standards. However, several reports have mentioned that some waterborne pathogens could be induced into a viable but nonculturable (VBNC) state by chlorine or chloramine disinfectants [3][4][5][6][7][8][9][10] . VBNC bacteria cannot be detected by the conventional plate counting method which leads to an underestimation of the potential microbiological risks. Thus, the pathogenic bacteria in the VBNC state may pose a threat to public health.
It has been reported that VBNC bacteria exhibit an extraordinary tolerance to adverse conditions. Escherichia coli in the VBNC state persisted in the presence of certain antibiotics, e.g., 200 μg mL −1 ampicillin (Amp) and 5 μg mL −1 ofloxacin (Ofx) (approximately 128 MIC and 64 MIC, respectively) 11 . In addition, Oliver et al. reported that Vibrio vulnificus could enter the VBNC state, protecting cells against a variety of potentially lethal environmental challenges (heat, oxidative stress, osmotic stress, pH fluctuations, ethanol, antibiotics and heavy metals) 12 . In our previous study, it was found that low doses of chlorine (0.5 mg L −1 ) disinfection effectively reduced the number of culturable E. coli and induced a VBNC state, after which the metabolic activity of the bacteria was reduced and persistence of E. coli in the presence of 9 typical antibiotics was enhanced 13 . Moreover, Moreno et al. also reported that low-level of residual chlorine in DWDS was capable of inducing VBNC state antibiotic-tolerant bacteria, which could be considered as an undetectable risk 8 . Chlorination, a vital step in controlling the biosafety of drinking water, was directly related to public and environmental health. Thus, it is necessary to thoroughly investigate the VBNC state induced by low concentrations of chlorine.
In addition to the compromise of physical structure, chlorine has also been reported to impact the metabolic processes, efflux pumps, and pathogenicity of bacteria by regulating gene expression. Therefore, it is important to comprehensively investigate the antibiotic tolerance of VBNC bacteria at the morphological and genetic levels. Generally, quantitative reverse transcription PCR (RT-qPCR) is applied to quantitatively determine the expression levels of known target genes in viable cells. It was reported that human pathogenic Vibrio spp. and Pseudomonas aeruginosa in the VBNC state could continuously express known toxin and virulence genes 14,15 . To survive under unfavorable conditions, major metabolic pathways (e.g., central carbohydrate metabolic pathways) could be altered to obtain energy and reduce the metabolic rate under the VBNC state 16 . Some studies focused on the gene expression of VBNC bacteria to explore their regulatory mechanisms [17][18][19] .
The regulation of gene expression involves complex and specific biological processes composed of a series of functionally related molecules 20 . Conventional investigation through individual genes is inadequate for discovering the complete regulation and cross talk of these genes. Nowadays, the high-throughput transcriptome sequencing (RNA-seq) technique has been applied in constructing frameworks of gene expression which has markedly accelerated the capacity of analyses 21 . Several comparative transcriptome analyses focused on some important environmental bacteria, including Pseudomonas syringae (Fmicb) 18 , Rhodococcus biphenylivorans 19 and Vibrio splendidus 22 . These studies highlighted a more comprehensive and data-intensive approach for better understanding of the mechanisms of bacterial stress responses. In DWDS, the inducer is chlorine, a chemical oxidant that may endow many different characteristics to the VBNC bacteria compared with physical inducers such as low temperature. Accordingly, it is worthwhile to analyze the adaptive responses of E. coli in the VBNC state induced by low-level chlorine via transcriptomic analysis.
In this study, as a continuation of the previous work, E. coli was selected as the test which is the indicator for pathogenic bacteria in the drinking water quality standards all over the world 23,24 . The bacteria were exposed to chlorine at the level comparable to residual chlorine in a DWDS to induce the VBNC state. The gene expression variation of the VBNC bacteria was assessed via transcriptome approaches. Then, the morphology was investigated by scanning electron microscope (SEM) and flow cytometry (FCM). It was expected that the special regulation behavior at low level chlorination circumstances could be revealed and deeper understanding of VBNC cells could be acquired.

Results and Discussion
Resuscitation of VBnc state. Chlorination is vital for controlling bacterial regrowth in DWDS. It usually takes several hours to days to transport water from a treatment plant to household taps. Such residence time is sufficient for the induction and resuscitation of bacteria into/from the VBNC state. To induce a VBNC state, we chose chlorine at 0.5 mg L −1 , the concentration generally applied in DWDS 1,2 . In our previous study, it demonstrated that 10 5-6 cells mL −1 of bacteria entered the VBNC state at 0.5 mg L −1 residual chlorine in 6 h ( Fig. S1 red column). Our result is consistent with Lee et al.; although no bacteria were observed in the chlorinated influent of the model system by heterotrophic plate count (HPC) method, a direct viable count (DVC) still presented observed growth in the range between 3-and 4-log cells mL −1 . Significantly high numbers of HPC and DVC were found both in the biofilm and effluent of the model system 25 .
VBNC microorganisms can be resuscitated under certain conditions. In this study, the resuscitation of VBNC E. coli (initial concentration of approximately 10 4 cells mL −1 ) was verified. Resuscitation experiments were performed with VBNC cells using liquid media. As shown in Fig. 1, the results of multiple experiments indicated that within the first 20 hours of culture, there was no colony formation on the plate. However, 1 hour later, the bacteria showed an "outbreak" increase of 3 orders of magnitude (from 0 to 1.4 × 10 3 CFUs mL −1 ). The same initial concentration of culturable E. coli was used to draw the growth curve as a control. We found that the growth curve of VBNC cells was significantly different from that of the conventional growth curve, suggesting the occurrence of resuscitation (Fig. S2). The theoretical growth rate of the cells in LB broth medium would be approximately 6 min (approximately 10 generations for one hour), which demonstrated that resuscitation occurred at a rate much www.nature.com/scientificreports www.nature.com/scientificreports/ faster than the typical doubling time of E. coli (which is approximately 30 min). In our results, approximately one in twelve (1/12) tubes displayed this recovery.
This recovery was not high for the absolute percentage of bacteria. However, the risk of resuscitation will be even more serious considering the large water supply in large modern cities. Currently in China, many cities have a water supply capacity of over 1 million tons/day. The probability of VBNC cells is approximately 10 1 to 10 3 cells/ mL in real drinking water. This means that residents in these cities will be exposed to up to 10 13-15 VBNC (10 1-3 cells/mL = 10 13-15 cells/million tons) cells per day, which might resuscitate under the appropriate conditions. characterization of VBnc E. coil. VBNC bacteria may present morphological changes such as cellular shrinkage and size reduction due to physical environmental stresses (e.g., low temperature) 26 . Unlike the VBNC state induced by cold temperature, the bacteria sterilized by low chlorine concentrations showed no changes in cell morphology. However, we observed a number of damages on the cell membrane (walls) of the chlorinated bacteria (red arrows), as shown Fig. 2(B), whereas we observed nearly intact cell membranes (walls) of the culturable cells (Fig. 2(A)). Jang et al. confirmed that bleach inhibits the biosynthesis of outer cell wall mycolic acids and induces oxidative damage in bactreria 27 . In the presence of 0.5 mg L −1 chlorine, some bacteria were relatively intact (marked by the green arrow in Fig. 2(B)), the cell membrane was not completely destroyed, and the bacteria in the VBNC state belonged to the group with intact membranes. The results of FCM corroborated the conclusion that some cells were SYTO 9 positive ( Fig. 2(C)). FCM analysis showed that the green fluorescence intensity of culturable and chlorinated cells (including VBNC cells) was greater than that of the negative control bacteria and the green fluorescence intensity of the VBNC cells was clearly weaker than that of the culturable cells. Furthermore, as shown in Fig. 2(C), the mean intensity of green fluorescence from the VBNC cells showed a trend of shifting to the left, and the red fluorescence intensity showed a reverse trend, revealing that the cell permeability of the VBNC bacteria increased. Moreover, we quantified the changes in cell membrane permeability under the VBNC state. Through the definition of the gate, 98.44% VBNC E. coli had enhanced permeability in the presence of selected chlorine concentrations (data not shown). These results were in accordance with the conclusions reached by Virto et al. 28 and Tatsumi et al. 29 who suggested that approximately a dose of 1 mg L −1 chlorine was linked to an increase in the membrane permeability of four types of bacteria including E. coli and neutrophils.
Global transcriptional profiles of E. coli in VBnc state and culturable state. Volcano Plot and venn diagram of VBNC versus culturable cells. Recently, several studies have focused on the induction conditions and properties of VBNC cells [30][31][32][33] , but few studies have investigated VBNC cells produced in municipal pipe networks. Their researches confirmed that VBNC cells existed in municipal pipe networks exhibited strong viability, and they could resuscitate 6,25 . Undoubtedly, more researches are needed to explore bacteria in a VBNC state, which may yield insight into strategies for the prevention and control of VBNC bacteria.
cDNA libraries derived from VBNC and culturable E. coli have been constructed. The libraries of VBNC cells and culturable cells had 27.97 million and 29.82 million reads respectively, and these reads were aligned with the E. coli K12 genome sequence (NCBI GenBank accession No. NC_000913). After incubation with a low concentration of chlorine, the expression of 362 genes of E. coli was found to be changed significantly (expressed in the log 2 (FoldChange) with base 2, hereinafter indicated as |log 2 (FoldChange)|) (|log 2 (FoldChange)| >1 and p < 0.05) ( Fig. S3(A)), of which 203 genes were upregulated (red plots) and 159 genes were downregulated (green plots). In particular, 3 downregulated and 22 upregulated genes with |log 2 (FoldChange)| >5 indicated that these genes had extremely significant changes after exposure to chlorine and needed special attention. Detailed information of the 362 genes with significant differences was provided in supporting information Table S3. These changes were categorized as follows: (1) biochemical process such as transport, establishment localization, localization, oxidation-reduction process, biological process; and (2) molecular function such as transporter activity, ATPase activity, GTPase activity, guanyl ribonucleotide binding, guanyl nucleotide binding, etc.
In the Venn diagram, the total expressed genes were plotted with two comparisons of transcriptional profiles ( Fig. S3(B)). Established on the standard of gene expression with FPKM >1, the sum of all the numbers (3900) in orange circle represented the expression level of genes in culturable cells, and the sum of all the numbers (3955) in green circle represented the expression level of genes in VBNC cells, suggesting that 3880 genes were in common between the two groups (overlapping yellow circle). Obviously, compared with the culturable state, the VBNC cells had some new genes that were activated or silenced to survive under environmental stresses.
Gene ontology (GO) and KEGG analysis. To ascribe gene functions to differentially expression genes (DEGs), a total of 3194 DEGs were assigned to 3 parts: biological process, molecular function and cellular component. To determine the Gene Ontology (GO) function of the up-or downregulated genes, GO enrichment analysis was performed on the 2048 upregulated and 1777 downregulated genes. The 30 most enriched GO terms are presented in Fig. 3. For the enriched GO terms of the 2048 upregulated genes, there were 1476, 467, 105 significantly enriched GO terms that corresponded to biological process, molecular function, and cellular component processes, respectively. It is worth mentioning that biological processes such as transportation, localization, and response to acid were significantly enriched. Furthermore, transporter activity and ATPase activity of the molecular component were also remarkedly enriched (P < 0.05), suggesting that the transport of some substance could be enhanced to excrete harmful substances in the VBNC state. The evidence indicated that VBNC cells were not just in a state of dormancy and still demonstrated highly expressed energy metabolism. The cellular component was relatively unchanged before or after treatment with a low concentration of chlorine, compared with the other 2 GO terms.
Additionally, the genes were mapped to the KEGG pathway. Among the mapped genes, the upregulated genes in the VBNC cells were mainly enriched in two pathways: ABC transporters and microbial metabolism in diverse environments. The enriched genes had lower Q values, indicating significant enrichment (Fig. 4). In Fig. 3(B), the GO enrichment of downregulated genes in the VBNC cells was presented by 2 groups: biological process and molecular function. Similar to GO terms identified for the upregulated genes, the cellular component category was unchanged before or after the low chlorine treatment. The top 2 enriched pathways were transmembrane transporter and RNA binding while no enrichment pathways were observed in downregulated GO.

functional analysis of the up/down-regulated genes in low chlorine-induced VBnc E. coli. A
VBNC state is a state of dormancy. In the VBNC state, many genes are downregulated to reduce energy consumption 4 . However, VBNC cells still have high level of gene expression, and the behavior of these genes has a greater contribution to the survival mechanisms and resuscitation of VBNC cells. The present study focused on highly expressed genes. Figure 5 revealed that highly expressed genes of VBNC E. coli were related to 6 functional groups, including ABC transporter, biotin, curli, biofilm, nitrate reductase and others. ABC transporters, including ATP-binding components of amino acid ABC transporters, ATP-dependent high affinity P-Type ATPase metal ion transporters, saccharides ABC transportors, etc., were upregulated from 3.15 to 6.39 log 2 (FoldChange), suggesting that some genes could be highly expressed in VBNC bacteria to survive 34,35 . With respect to the expression of other genes, some genes involved in biotin synthesis could encode proteins that produce some growth factors, increasing the survivability of bacteria. Enterobactin (also known as enterochelin) is a high-affinity www.nature.com/scientificreports www.nature.com/scientificreports/ siderophore that acquires iron for microbial systems. Pathogenic bacteria can steal iron from other living organisms using this mechanism, even though the concentration of iron is kept extremely low due to the toxicity of free iron 34,36,37 . The existence of such levels of biotin not only increases the pathogenicity of pathogenic bacteria (because of the involvement of metal ions in oxygen transport and gall storage, electron transfer, DNA synthesis and other physiological processes), but also sequesters the iron resources of other bacteria. Curli is composed of two subunits: CsgA and CsgB. CsgE prohibits the self-assembly of CsgA into amyloid fibers. CsgE acts directly on the secreted substrate CsgA to prevent premature subunit assembly 38 . The expression levels of the csgE gene were significantly higher in bacteria in the VBNC state (4.18 log 2 (FoldChange), P < 0.05) compared to the control bacteria, and the formation of pili was reduced, leading to the reduced motility of bacteria 39 . The cells that cannot swim could reduce energy consumption and are more suitable for dormancy. Additionally, biofilm formation would be regulated through the high expression of yjcH (the gene yjcG, is co-transcribed with the gene   www.nature.com/scientificreports www.nature.com/scientificreports/ Dormant bacteria improve their prospects of survival by reducing metabolic activity 4 . Thus, the downregulation of gene expression in dormant bacteria seems to be a common conclusion. In this study, we also listed genes that were extremely downregulated (log 2 (FoldChange) < −4) (Table S1). Eleven downregulated genes mainly encode hydrogenase 1 (HYD 1) (hyaA, B, C, D, E, F) and cold shock protein (cspA, B). Hydrogen metabolism in E. coli is tightly regulated by anaerobic conditions of growth produced. The genes hyaA, B, C, D, E, and F were downregulated, which could decrease the synthesis of active HYD1, and reduce energy metabolism s1 . CspA and CspB have been shown to be cold shock-inducible in E. coli. The low expression levels of CspA and B in VBNC cells suggested that bacteria were reduced in response to low temperature stress s2 . Other downregulated genes (yjiG, yhiD, appC) involved in ATP-dependent transport reduced the related biological process energy consumption s3, 4 . The results of downregulation gene expression were consistent with previous studies. extremely high expression genes of VBnc E. coli. For further analysis, we selected genes that were expressed at extremely high levels (log 2 (FoldChange) >5) ( Table 1). There were 16 upregulated genes, including 3 encoding toxic proteins (ygeG, ibsD, shoB), 5 encoding stress-induced proteins (spy, fade, narV, yeaG, ycgB), 3 encoding regulators (ebgC, ygcW, yaiV) and 3 encoding efflux systems (cusF, cusC, garP) ( Table 1). These genes could be classified into two groups. The genes of the first group were related to the transportation of substances (e.g., efflux pump-encoding genes and transcripts), which helped to lower the level of toxic or harmful substances in cells. The type III secretion system (T3SS) in several gram-negative bacteria, which contains the ygeG gene, an important virulence factor that subverts host cellular processes 45 . As a result, such "dormancy" cells still carry risks that threaten human health. The second group was involved in gene regulation (including transcriptional regulatory factors, stress-induced protein-encoding genes and virulence genes) which could be useful for regulating the metabolic levels. RpoS is the master regulator of the general stress response in E. coli. Rapidly growing cells contain little, if any, RpoS. In contrast, in order to response to many different stress conditions, cellular RpoS levels rapidly increased up to 20-fold. This regulation occurred at the levels of transcription and translation, and RpoS proteolysis occurred in various stress conditions (starvation, hyperosmotic shift, or pH downshift) that differentially affected these levels of control 46,47 . RpoS is a typical upregulated gene 47 which could lead to the inhibition of bacterial division and increase the survivability under high-level expression. In particular, the metal efflux system genes (cusC and F) 48,49 and the multidrug efflux protein gene (garP) 30 were extremely upregulated, which could increase metal-and antibiotic persistence.

conclusion
From the morphology and based on transcriptomic analysis, we investigated the regulation of gene expression of VBNC cells induced by a low chlorine level. E. coli cells could be induced into the VBNC state if water was treated with low concentrations (0.5 mg L −1 ) of chlorine and VBNC E. coli had the potential to resuscitate. Results of the SEM showed that: the size of VBNC bacteria was not markedly changed, and FCM examination showed that: VBNC cells' permeability was enhanced. Transcriptomic analysis suggested that the enhancement of the efflux system and stress-induced function contributed to the strong environmental tolerance of the VBNC cells and VBNC E. coli was still a threat to human health.
Methods preparation of strain. E. coli CMCC44103 15 was cultured with Luria-Bertani (LB) broth (Haibo, China) at 37 °C, 180 rpm until logarithmic phase (12 h). Cells were harvested (5000 rpm, 10 min) and re-suspended in sterilized buffered saline and then diluted to a final concentration of 10 7 -10 8 CFU mL −1 . The cell number was counted by plate counting and disposable counting chambers (CELL.VU, USA). The bacteria were divided into two parts, www.nature.com/scientificreports www.nature.com/scientificreports/ one of which was used to induce the VBNC state, and the other for control. The control was culturable E. coli with the same treatment except chlorination.
induction of VBnc E. coli. An initial disinfection concentration of 0.5 mg L −1 was made by addition of disinfectant stock solution. To maintain stable chlorine level, chlorine was added every 2 hours. The concentration of free chlorine was determined using the DPD method (Hach, USA). All cells were incubated at 25 °C, 180 rpm and sampled at 0 min, 15 min, 30 min, 1 h, 3 h, 6 h, 9 h, 12 h, 18 h and 24 h. Before sampling, 100 μL of 20% sodium thiosulfate (1:1 v/v in sterilized buffered saline) was added to inactivate residual chlorine.
In this study, culturable cells were identified by plate counting on NA agar. The number of viable cells was counted by flow cytometry. The number of VBNC cells was calculated as the difference between the numbers of viable and culturable cells. This is a classic characterization method that could count numbers of VBNC cells 4 . When cells were exposed to chlorine for 6 hours or longer, the numbers of VBNC and viable cells were equal because no colonies were formed on the plates. Cells were considered to have entered VBNC state (10 5-6 cells mL −1 ) and have been ready for the next preparation step in this study (Fig. S1 red column). The VBNC preparations were enriched about 50 times (final concentration of 10 7-8 cells mL −1 ) by centrifuging at 7800 rpm, 4 °C, 20 minutes and harvested for RNA extraction.

Resuscitation of VBnc cells.
Resuscitation experiments were conducted in LB broth medium with the prepared VBNC state E. coli. In each experiment, 4 mL of prepared VBNC state bacteria were separately dispensed into 50 mL centrifuge tubes. Each tube contained 36 mL sterilized LB broth medium and 4 mL bacterial culture, the ratio of VBNC E. coli to LB broth was 1:9. The initial concentration of VBNC cells was about 10 4 cells mL −1 . All of these tubes were cultured at 37 °C, 180 rpm. 1 mL of sample was collected hourly and coated on nutrient agar (NA) (Haibo, China) plate, and the resuscitation was evaluated by plate counting.
Analysis of viability and culturability. Cell viability was analyzed by flow cytometer (FCM) (Millipore Guava EasyCyte, USA). SYBR Green was used for total cells counting (Thermo Fisher Scientific, USA) 50 , whereas 5-cyano-2, 3-ditolyl tetrazolium (CTC) was used for viable cells counting (Thermo Fisher, USA) 51 . SYBR Green (The final concentration of 10,000 times dilution) was added into 0.2 mL suspensions in 1.5 mL tubes (20 min, room temperature). CTC was added to the same system at a final concentration of 2 mM (37 °C, 3 h). Samples were analyzed by FCM (Millipore Guava EasyCyte, USA) within 2 h after resuspension, and Forward scatter (FSC), side scatter (SSC), fluorescence signals SYBR Green and CTC using 525 nm and 620 nm band pass filters were measured in logarithmic amplification mode. 5,000 cells for each sample were collected. Dyeing operations were all conducted in the dark.
For culturability analysis, each sample was incubated on NA at 28 °C for 24 h according to the national standards 1 . All these experiments were performed in triplicate. Cells were considered to be in VBNC state when none of the colonies was demonstrated on a NA plate (1 mL of sample was coated).
RNA extraction and qualification. Previously prepared control (4.1) and VBNC (4.2) bacteria solutions (10 7-8 cells mL −1 ) were centrifuged at 12,000 × g for 1 min and re-suspended in sterilized buffered saline. The control in this paper was culturable E. coli with the same treatment except chlorination. RNA was extracted using EasyPure RNA Kit (Transgen Biotech, China). RNA purity was examined using the micro-spectrophotometer Nano-100 (Aosheng, China). Qubit (Life Technologies, USA) was used to measure the RNA concentration. The experiments were repeated three times under the same conditions. The RNA which was extracted from biological duplication was used for RNA sequencing. In addition, Pearson correlation between samples was checked to ensure accuracy of the results of RNA-seq. transcriptome sequencing. Library preparation and sequencing. Transcriptome sequencing was performed by the Novogene company (Novogene, China). Sequencing libraries were built using NEBNext ® Ultra ™ Directional RNA Library Prep Kit (NEB, USA) following the manufacturer's instructions. Briefly, the mRNA was fragmented and purified with poly-T oligo-attached magnetic beads (Beckman Coulter, USA). Then, first strand cDNA and second strand cDNA were synthesized. PCR was performed of universal PCR primers and Index (x) Primer. Finally, the Library preparations were sequenced on an Illumina Hiseq 2000 platform.
Gene expression level analysis. FPKM (Expected number of Fragments Per Kilobase of transcript sequence per Millions base pairs sequenced) method 52,53 was used to analyze the gene expression variations between culturable and VBNC cells. The reads numbers mapped to each gene was counted by HTSeq v0.6.1.
Differential expression analysis. DESeq R package was used to analyze differential expression of culturbale and VBNC state. After multiple testing correction, the adjusted P-value less than 0.05 by DESeq was selected as a threshold of differential expression 54 .
Gene Ontology (GO) enrichment analysis. GOseq R package was used to GO enrichment analysis of differential expressed genes. GO terms with adjusted P-value < 0.05 were assigned as significantly enriched by differential expressed genes 55 .
Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis. KEGG was used to understand functions of the biological system from molecular-level information. In this study, KOBAS software was adopted to analyze the statistical enrichment of differential expression genes in KEGG pathways 56 .