Viable but Nonculturable and Persister Cells Coexist Stochastically and Are Induced by Human Serum

Abstract

Dormancy holds a vital role in the ecological dynamics of microorganisms. Specifically, entry into dormancy allows cells to withstand times of stress while maintaining the potential for reentry into an active existence. The viable but nonculturable (VBNC) state and antibiotic persistence are two well-recognized conditions of dormancy demonstrated to contribute to bacterial stress tolerance and, as a consequence, yield populations that are tolerant to high-dose antibiotics. Aside from this commonality, more evidence is being presented that indicates the relatedness of these two states. Here, we demonstrate that VBNC cells are present during persister isolation experiments, further indicating that these cells coexist and are induced by the same conditions. Interestingly, we reveal that VBNC cells can exist stochastically in unstressed growing cultures, a finding that is characteristic of persisters. Furthermore, human serum induces the formation of both VBNC cells and persisters, a finding not previously described for either dormancy state. Lastly, we describe the role of toxin-antitoxin systems (TAS) in the induction of the VBNC state and report that these TAS, which are classically implicated in persister cell formation, are also induced during incubation in human serum. This study provides evidence for the recently proposed “dormancy continuum hypothesis” and substantiates the physical and molecular relatedness of VBNC and persister cells in a standardized model organism. Notably, these results provide new evidence for the clinical significance of VBNC and persister cells.

INTRODUCTION

As inhabitants of a dynamic biosphere, bacteria are constantly challenged with potentially harmful environmental uncertainty. To defy such perpetual instability, many microorganisms maintain subpopulations with the capability to enter a temporary state of dormancy during which cells exhibit reduced growth rates and metabolic demand (1). When the environment becomes permissive, dormant cells can resuscitate and subsequently regain growth (2, 3). The evolutionary role of the maintenance of such population heterogeneity is analogous to a bet-hedging strategy in which cells of various phenotypes arise and increase the chance of survival in a fluctuating milieu (1). Importantly, dormancy that allows bacteria to oppose environmental stress can also render them tolerant to antibiotics (4–6), highlighting the clinical relevance of this physiological state.

Currently, two well-defined dormancy states have been described in nonsporulating bacteria: the viable but nonculturable state (VBNC) and antibiotic persistence (2, 7). Persister cells are described as slow or nongrowing subpopulations present within a growing culture that are consequently able to withstand multiple types of antibiotics (8). As opposed to antibiotic-resistant cells, persister cells are thought to be genetically identical to the nonpersister cells but exhibit a drug-tolerant phenotype (9). The persister phenotype has been shown to exist stochastically within growing cultures (10) but can also be induced by stressful environments such as starvation, oxidative stress, DNA damage, stressful pH, and antibiotics (11–16). Therefore, persister cells are of medical relevance due to the potential to cause recurrent bacterial infections such as those exhibited by Mycobacterium tuberculosis, Salmonella, pathogenic Escherichia coli, and Staphylococcus aureus (17). This claim was supported when high-persistence mutants were isolated from cystic fibrosis patients receiving repeated antibiotic therapy (18).

At least 85 species of bacteria have been found to enter a mode of dormancy referred to as the viable but nonculturable (VBNC) state (2). This state has alternatively been referred to by others as conditionally viable environmental cells (CVEC) (19), active but nonculturable cells (ABNC) (20), and dormant cells (21). These cells are reported to be viable due to their intact cell membranes, low-level metabolic activity, and continued gene expression (6, 22). However, they are nondividing and, unlike persisters, are unable to immediately regain the ability to divide when plated on routine laboratory medium (2). The VBNC state is considered to be an effective survival strategy for the bacterium as it allows cells to endure adverse environmental conditions and to resuscitate to a replicative form when environmental conditions improve. Indeed, it has been shown that the VBNC state is induced by a variety of environmentally relevant stressors such as starvation, hypoxia, stressful temperature, salinity, and pH (23–26). Furthermore, while in the VBNC state, cells have been shown to tolerate typically fatal stressors, including high-dose antibiotics (4). Importantly, VBNC cells are able to resuscitate in vivo and regain their virulence (27–29). These findings are further corroborated by a study showing that E. coli VBNC cells were present in a mouse urinary tract infection model after antibiotic treatment and were able to resuscitate when antibiotic treatment was stopped (30). Furthermore, a study by Colwell and colleagues demonstrated that VBNC Vibrio cholerae O1 was converted to a culturable state during passage through human participants (31). The clinical relevance of VBNC cells is further supported by a Lleo et al. finding that 14 to 27% of infections in which organisms could not be cultured on clinical laboratory medium were, in fact, positive for pathogenic organisms, as determined using PCR-based detection (32). These findings strongly suggest that VBNC cells are a cause of antibiotic failure and recurrent infections, thereby posing a significant clinical and public health risk.

There is compelling evidence that persisters and VBNC cells are related; however, this relationship is rarely discussed in the literature. Their association was described in detail in a recent review, where it was proposed that VBNC cells and persisters are part of a “dormancy continuum,” in which they share similar mechanisms but are found in different physiological positions on the dormancy range (33). The model is based on a molecular mechanism of stochastic persister formation, which is based on the action of toxin-antitoxin systems (TAS) (7). TAS are typified as two-gene operons encoding a protein toxin and cognate antitoxin (7). When cytosolic concentrations of the two components are proportional, coupling with the antitoxin neutralizes the toxin; any disturbance that decreases complexation liberates the toxin. Some of these toxins go on to inhibit translation and eventually cause growth inhibition (7). It is tempting to speculate that variable levels of free toxin in individual cells drive the production of a heterogeneous population that harbors actively growing cells, persisters, and VBNC cells.

Experimentally, we discriminate VBNC cells from persisters based on their ability/inability to grow on routine medium after the inducing stress is removed (33). Whereas persisters regain growth on nutrient medium shortly after the removal of stress (antibiotics), VBNC cells are unable to do so unless they are given time and/or the proper conditions to resuscitate (2). Therefore, cells that are able to grow on solid medium after antibiotic treatment are defined as persisters, while cells that require up to 24 h of resuscitation treatment before they are able to grow on medium are defined as VBNC cells. The purpose of this study was to begin addressing questions previously set forth in our recent opinion article describing the “dormancy continuum hypothesis” (30). The work described herein aimed to address three of these outstanding questions: (i) Do persisters become VBNC cells upon prolonged exposure to stress? (ii) In a characteristic persister isolation experiment, is there a VBNC population that is normally not detected by culture-based methods? (iii) Can evidence for overlapping VBNC and persister mechanisms be observed in a standardized model? Overall, our goal was to test this hypothesis in a model organism that is capable of easily forming both persister and VBNC cells to further elaborate on the relationship between these two dormancy states. Specifically, we sought to determine if persister cells are primed for entry into the VBNC state and to test for the presence of preexisting VBNC cells within actively growing and persister populations. Furthermore, since TAS are important in the modulation of persistence, we tested whether VBNC cells also exhibit increased expression of TAS genes. We further evaluated whether a clinically relevant stressful condition (human serum) could induce persistence and/or the VBNC state and whether this induction occurred through TAS regulation.

MATERIALS AND METHODS

Strains and growth media.

Strains used in this study were Vibrio vulnificus CMCP6 and Escherichia coli K-12. These were stored at −80°C in Bacto Luria broth (LB) (BD, Franklin Lakes, NJ) containing 20% glycerol. All strains were grown in Bacto heart infusion (HI) broth (BD, Franklin Lakes, NJ) for 24 h at 30°C with aeration.

Persister cell isolation.

Persister cells were isolated by growing V. vulnificus or E. coli in HI broth to log phase (optical density at 610 nm [OD₆₁₀] of 0.15 to 0.25) and then treating these cultures with 100 μg/ml ampicillin for 4 h at 30°C with aeration. Subsequently, antibiotic-treated cultures of V. vulnificus or E. coli were washed four times to remove antibiotics, and the number of surviving culturable cells was determined using a standard plate count method (serial dilutions into phosphate-buffered saline [PBS] and plating on HI or MacConkey agar, respectively). To allow for the potential resuscitation of VBNC cells within these same cultures, cells were washed four times and resuspended in either 1/2 artificial seawater (ASW) for V. vulnificus or 0.85% NaCl for E. coli and incubated for an additional 24 h at 20°C before standard plate counts were performed to assess culturability.

VBNC cell isolation.

To isolate V. vulnificus VBNC cells, log-phase cultures were washed twice using 1/2 ASW to remove nutrients before being diluted 1:100 (vol/vol) into 1/2 ASW. These cultures were statically incubated at 4°C and quantified daily until cells were no longer culturable on HI agar (<10 CFU/ml detectable). Experiments in which persister cells were allowed to enter the VBNC state were performed by isolating persisters as described above, washing these cells four times to remove nutrients and antibiotic, resuspending cells in 1/2 ASW (without diluting cells), and then placing cultures at 4°C until persisters were no longer culturable on HI agar (<10 CFU/ml detectable). Cells were resuscitated by incubating cultures at 20°C for 24 h.

Determination of viability.

To detect cells while in the VBNC state, viability was determined using a BacLight Live/Dead bacterial viability assay (Molecular Probes, Inc., Eugene, OR) according to the manufacturer's instructions. This assay is used to distinguish between viable cells with intact cell membranes and dead/dying cells with damaged cell membranes. Viable cells exhibit green fluorescence due to staining with SYTO 9, whereas dead cells exhibit red fluorescence due to staining with propidium iodide. Viable cells were quantified by calculating the average number of green fluorescent cells in 30 fields.

Serum studies.

Pooled human serum (MP Biomedicals, USA) was used in all experiments. To test whether human serum induces persistence and/or the VBNC state, log-phase cells were washed twice and resuspended in 2 ml of active or heat-inactivated (56°C for 30 min) human serum. These suspensions were incubated at 37°C for 1 h before culturability was quantified. While cells were maintained in human serum, ampicillin was added at a final concentration of 100 μg/ml, and cultures were incubated at 37°C for another 4 h. Cells were then washed four times to remove antibiotics and human serum, resuspended in 1/2 ASW, and either plated immediately (to quantify the persister fraction) or incubated at 20°C for 24 h (to quantify the VBNC fraction). Cells were quantified using a standard plate count method at each step during the experiment (log phase, after 1 h in serum, after 4 h in serum with ampicillin, and after resuscitation for 24 h).

Gene expression studies. (i) Treatment conditions for gene expression in human serum.

V. vulnificus was grown in HI broth for 24 h at 30°C with aeration and subsequently diluted 1:100 (vol/vol) in HI broth and grown to log phase (OD₆₁₀ of 0.15 to 0.25). Cells were subsequently washed twice in phosphate-buffered saline (PBS), resuspended in 2 ml of active or heat-inactivated pooled human serum, and incubated at 37°C for 4 h. Serum-treated cells and an aliquot of log-phase cells were treated with RNAprotect (Qiagen), according to the manufacturer's instructions, to preserve samples for subsequent RNA extraction.

(ii) Treatment conditions for gene expression in the VBNC state.

VBNC cultures of the V. vulnificus strain were prepared as described in the “VBNC cell isolation” above. VBNC cells, cells exposed to VBNC-inducing conditions for 60 min, and log-phase cells were collected and treated with RNAprotect (Qiagen) according to the manufacturer's instructions to preserve samples for subsequent RNA extraction.

(iii) RNA extraction and purification.

Cell pellets treated with RNAprotect were resuspended in Tris-EDTA (TE) buffer with 10 mg/ml lysozyme and incubated at room temperature on a vortex at medium speed for 30 min. Lysed cells were then subjected to RNA extraction using an RNeasy Minikit (Qiagen) according to the manufacturer's instructions, in conjunction with Qiagen's on-column DNase I treatment. Total RNA was treated with a second DNA digest using Turbo DNA-free (Ambion) according to a rigorous DNase treatment protocol. RNA quality and quantity were assessed using a NanoDrop spectrophotometer (Thermo), and all RNA samples were found to have a 260/280-nm absorbance ratio of ≥1.7. RNA samples were stored at −80°C.

Endpoint PCR was performed on RNA samples to confirm the complete digestion of DNA. PCR was performed using Promega GoTaq DNA polymerase, 5× Green GoTaq reaction buffer, 10 mM deoxynucleoside triphosphate (dNTP) mix, and primers for vvhA (species-specific gene). Cycling parameters were performed according to the manufacturer's recommendations (Promega), with an annealing temperature of 53.1°C and 40 cycles of amplification. Amplification of the vvhA gene was indicative of DNA contamination, in which case the RNA was not used for downstream processes.

(iv) Primer design for qRT-PCR.

V. vulnificus CMCP6 chromosomal loci identified as being homologous to relE and hipA genes by the Toxin-Antitoxin Database (TADB) website (34) were chosen as amplification targets. Working gene names were assigned to locus tags as shown in Table 1. Primers were designed based on NCBI's publicly available sequence data for CMCP6 and were designed to be suitable for quantitative reverse transcription-PCR (qRT-PCR). Optimal primer quality and fidelity were assessed using IDT OligoAnalyzer, version 3.1, software, and PCR efficiency of each primer pair was preliminarily evaluated using an in silico PCR efficiency estimation tool (35). Primers were purchased from Sigma-Aldrich and validated using endpoint PCR to ensure specific amplification of the DNA targets of interest. Relevant information regarding primers used in this study is listed in Table 1.

TABLE 1.

Primers used in this study

Gene namea	Locus tag	Primer sequence (5′–3′)b	Product size (bp)	Primer efficiency
hipA1	VV1_0401	F, AGAGGCTCAAGAAGCACCATC	202	1.841244
		R, CAGATTCCATGCCAAATCTCTC
hipA2	VV1_2047	F, CCTATCTCGCTTTCTATGCCTCTC	215	2.035363
		R, TCTACATCAAAGCCCTCTGGAAC
hipA3	VV1_2190	F, CAAATGGTCTCTCGCACCAG	210	1.955265
		R, CGCCAAGTTGATACGTGCTC
relE1-relE2	VV1_2410 VV1_2433	F, CGCTATCCGACCTCAATGAT	119	2.100742
		R, CGATTCAGGGAAAGCCTCTA
relE3	VV1_2547	F, TAACCCAGTGGCAGCAGAAAG	139	1.947688
		R, TCGGTAGTGGCCAAAAAGG
relE4	VV2_1493	F, TGACTTCAACCCTGATGTGC	194	1.921710
		R, GGTGGAGGACACGCATAACT
relE5	VV1_0110	F, CTCAATGTGTAGTGGTGGAAG	197	1.893049
		R, GTAATTCTCATGAGGCCCAAC
rplT	VV1_2400	F, GCTCGTGCACGTCATAAGAA	124	1.878874
		R, GACGGTCACGGTAAGCGTAT
gyrB	VV1_0996	F, CTGAAGGGTCTGGATGCGG	65	2.032692
		R, GTGCCATCATCTGTGTCCCC

^aOne copy of each gene was detected. relE1-relE2 represent two loci that are 100% homologous.

^bF, forward; R, reverse.

(v) Determining primer efficiencies.

The primer efficiency of each primer set was calculated by generating standard curves in order to take into account potential differences in PCR efficiencies between primer sets. Amplicons for all target genes were PCR amplified from genomic DNA using the PCR protocol detailed in the “RNA extraction and purification” section above. Products were agarose gel extracted and purified using a Wizard SV Gel and PCR Clean-Up System (Promega). DNA concentration was determined using a NanoDrop instrument (Thermo), and serial dilutions in nuclease-free water were prepared with a range of 10⁻⁵ to 10⁻¹⁰ of the initial concentrations. The C_T value for each dilution was measured in triplicate wells using an Applied Biosystems 7500 real-time PCR system and PerfeCTa SYBR Green FastMix, Low ROX (6-carboxy-X-rhodamine) (Quanta Biosciences), reagents. PCR amplification efficiency was analyzed by generating a standard curve and evaluating the slope. From this slope, an efficiency value was calculated using Pffafl's method (36) and incorporated into the final calculation for relative expression.

(vi) Relative qRT-PCR.

Gene expression levels of toxin genes were examined in human serum relative to log phase cells or in VBNC cells relative to culturable log-phase cells, using relative quantitative reverse transcription-PCR (qRT-PCR) as previously described (37). Total RNA (1 μg) was reverse transcribed using qScript cDNA SuperMix (Quanta Biosciences), and 50 ng of cDNA template was carried over for quantitative PCR (qPCR). qRT-PCR was performed on three technical replicates for each sample using an Applied Biosystems 7500 real-time PCR system and PerfeCTa SYBR Green FastMix, Low ROX (Quanta Biosciences), reagents. Negative controls and no-RT controls were employed to rule out the influence of DNA contaminants and residual genomic DNA, respectively. Expression levels of each gene were normalized by using an endogenous control gene to correct for sampling errors. DNA gyrase subunit B (gyrB) was used as the control for serum studies, whereas the 50S ribosomal protein L20 (rplT) was used as the control gene for VBNC studies. Each control gene was determined to be normally expressed across conditions prior to subsequent data analysis. Fold changes in expression levels were measured using the Pfaffl equation (36), taking into account differences in PCR efficiencies between primer sets.

Statistical analyses. (i) Culture-based methods.

Statistical significance was determined using an unpaired Student's t test or one-way analysis of variance (ANOVA), followed by Bonferroni's post hoc test for multiple comparisons (as indicated in the figure legends). Data were analyzed using GraphPad Prism (version 6.0; GraphPad Software, Inc.).

(ii) Gene expression studies.

Significant differences in gene expression levels were assessed using either one-way analysis of variance (ANOVA) followed by Bonferroni's post hoc test for multiple comparisons or Kruskal-Wallis ANOVA followed by Dunn's post hoc test (in the cases where data were nonparametric). Significance was determined by using a 95% confidence interval. Gene expression data were analyzed by using GraphPad Prism (version 6.0; GraphPad Software, Inc.).

RESULTS AND DISCUSSION

Do persister cells efficiently enter the VBNC state?

We have previously proposed that VBNC cells and persisters may be part of a dormancy continuum in which active cells stochastically, or during environmental stress, give rise to persister cells which may be transitory cells that lead to the formation of VBNC cells (33). To test this hypothesis we asked whether persister cells are more efficient at entering the VBNC state than a log-phase culture. We hypothesized that persisters would be more efficient at entering the VBNC state than cells of a log-phase culture. V. vulnificus log-phase cells enter the VBNC state in approximately 7 to 10 days at 4°C in 1/2 ASW and can typically resuscitate within 7 to 9 h after the temperature is increased to 20°C (38). Interestingly, in V. vulnificus only about 1 to 10% of the population resuscitates, while the rest of the population is presumed dead. For a pure persister population, we hypothesized that a higher proportion of cells would be able to enter the VBNC state and resuscitate since they are already in a state of slow or no growth. To test this, we isolated persister cells using antibiotic treatment, removed the antibiotic, and incubated these cells at 4°C to subsequently induce the VBNC state in this persister population (as is typically done to induce VBNC in V. vulnificus). Interestingly, the persister populations were able to enter the VBNC state faster (4 to 5 days) than log-phase cells (7 to 10 days). This may indicate that persister cells are able to enter the VBNC state more efficiently than actively growing cells, which may be in part due to stress caused by the preceding antibiotic treatment, which primes the cells to become VBNC quickly. It is important to note that simply the lower number of cells present after antibiotic treatment (i.e., persisters) may also result in a more rapid entry into the VBNC state.

To our surprise, upon resuscitation there were approximately 100 times more (P = 0.0254) culturable cells than the number of persister cells that were initially allowed to enter the VBNC state (Fig. 1). Due to the absence of nutrients during VBNC experiments, it is highly improbable that this magnitude of resuscitation is a result of cellular regrowth. Hence, VBNC cells must have already been present before and/or after antibiotic treatment but were, by definition, not detectable using culture-based methods during persister quantification. This unexpected yet intriguing finding could be explained by one of two phenomena. It is possible that VBNC cells along with the persisters survived the antibiotic treatment and the 4°C treatment and that both populations resuscitated after the temperature upshift. Alternatively, it may be that the persister population did not survive the 4°C treatment and that the resuscitated population arose purely from preexisting VBNC cells. Unfortunately, this experiment does not allow us to distinguish between these two possibilities, and for this reason it is difficult to address our original question of whether persisters are more efficient at entering the VBNC state. However, this study unearthed an interesting question regarding VBNC dynamics and their preexistence in growing cultures. Future studies using single-cell tracking with specific genetic markers could allow one to better observe whether persisters are able physiologically to transition to the VBNC state.

FIG 1.

Induction of the VBNC state from persister cells. Log-phase cells of V. vulnificus were treated with ampicillin to isolate persister cells, which were then washed and incubated at 4°C to induce the viable but nonculturable state. Once cells were determined to be nonculturable (third bar; arrow indicates that cells were below level of detection), cultures were incubated at 20°C for 24 h to allow for resuscitation. Error bars represent standard errors of three biological replicates. Different letters indicate statistically significant differences.

Are VBNC cells present during persister isolation?

To further determine if VBNC cells are present after antibiotic treatment, a culture of persisters (selected for by antibiotic treatment) were cleansed of antibiotics and nutrients and subsequently given the opportunity to resuscitate at 20°C for 24 h. Once again, more cells were culturable after a 24-h resuscitation period, further supporting the idea that VBNC cells were present during the persister isolation experiment and resuscitated after the antibiotics were removed (Fig. 2). There was no significant difference in culturability between cultures that were resuscitated for 24 h directly after isolation of persisters and cultures containing persisters that were additionally induced into the VBNC state at 4°C before resuscitation (Fig. 2). These results indicate that VBNC cells are present along with persisters after antibiotic treatment and that VBNC cells can resuscitate when the antibiotics are removed.

FIG 2.

VBNC cells resuscitate after antibiotic treatment. Log-phase cells were treated with ampicillin to isolate persister cells, which were then washed and either incubated at 20°C for 24 h to resuscitate VBNC cells (third bar) or incubated at 4°C to induce the VBNC state in persister cells prior to resuscitation (fourth bar). Error bars represent standard errors of three biological replicates. Different letters indicate statistically significant differences.

To further validate these findings, we employed fluorescence microscopy using a BacLight Live/Dead viability staining protocol to quantify viable cells. It is important to note that this kit does not directly detect cellular viability or the VBNC state itself; however, it detects cell membrane integrity, a marker of cellular viability. Culturability and viability were quantified in parallel during log phase, after persister isolation (after antibiotic treatment), and postresuscitation (antibiotics and nutrients removed and cells incubated at 20°C for 24 h). Surprisingly, we discovered that even during log phase, there were more viable cells than there were culturable cells in V. vulnificus, indicating that there are VBNC cells present stochastically during log-phase growth (Fig. 3A). These findings are consistent with the idea that both VBNC and persister cells are part of a heterogeneous population present even during optimal conditions as part of a bet-hedging strategy to endure future environmental challenges (33). As expected, after persister isolation, there was a much larger viable population than culturable (persister) population (Fig. 3A). The difference between the amount of culturable persisters and the amount of viable cells is, by definition, equal to the amount of VBNC cells. Whereas there was a 99.6% decline in culturability after antibiotic treatment, there was only a 61.3% reduction in viability. This indicates that the antibiotics are either inducing the VBNC state and/or that the VBNC cells are present prior to antibiotic treatment, are able to withstand antibiotic treatment, and resuscitate when the antibiotic challenge is removed (as indicated in Fig. 3).

FIG 3.

VBNC cells are present stochastically and coexist with persisters. V. vulnificus (A) and E. coli (B) persister cells were isolated from log-phase cells and then washed and allowed to resuscitate for 24 h at 20°C. Culturability was assessed on HI agar (as log CFU/ml) while viability (number of cells staining green) was assessed using a Live/Dead viability staining kit. Error bars represent standard errors of three biological replicates. Asterisks represent statistically significant differences between corresponding values for live and culturable cells (P < 0.05, by one-way ANOVA).

To determine if this finding was unique to V. vulnificus, we replicated the experiment using E. coli. Indeed, similar results were obtained for E. coli cells subjected to antibiotic treatment (Fig. 3B). However, there was not a statistically significant difference in viability and culturability in log-phase cells, indicating that E. coli does not maintain a large preexisting VBNC population. However, even in E. coli there was a large population that resuscitated after the antibiotic was removed, demonstrating that the VBNC state was likely induced by the antibiotic treatment.

An interesting finding was that the majority (81%) of a log-phase growing culture of V. vulnificus was nonculturable (i.e., in the VBNC state). The fact that VBNC cells are present stochastically is intuitive, and such a bet-hedging process has clear benefits. However, the finding that the majority of the population was in this state was very unexpected. One explanation may be that during growth, VBNC cells are accumulating. Cells that were once active are either dying or becoming dormant (VBNC/persister). Furthermore, since the environment during exponential growth is not stressful, these cells are more likely to enter a dormancy state rather than die. As logarithmic growth progresses, fit cells continue to divide, while more VBNC cells continue to accumulate. When the culture approaches zero population growth (i.e., stationary phase), VBNC and dead cells are being produced at the same rate at which new cells are formed. Furthermore, it is possible that VBNC cells that accumulate in stationary-phase cultures (overnight cultures) are carried over into new log-phase cultures. These two processes together can potentially rationalize the finding that there are more VBNC cells present during log-phase growth than actively growing culturable cells.

Overall, these findings not only indicate that VBNC cells coexist with persisters but also illuminate the shortcomings of current antibiotics and highlight the potential clinical role of VBNC cells in recalcitrant infections. Since we show that VBNC cells can represent a larger portion of the population than persisters, they may be an even more significant reservoir of multidrug-tolerant cells, which upon proper stimulation, cessation of treatment, or immune suppression can resuscitate and once again cause infection.

This study also underscores the limitations of using culturability-based methods as a proxy for viability. It is becoming more apparent that this approach overlooks a very large portion of clinically relevant dormant cells, and some investigators have begun to implement methods to evaluate the viability of VBNC cells when the efficacy of antimicrobials is tested against dormant cells (39). Given the breadth of the presence of VBNC cells under experimental conditions, it is likely that many microbiological experiments are stealthily obscured by the presence of these culture-negative cells. While laboratory medium is one of the most useful and affordable means of microbial detection, it is important to be cognizant of the potential presence of dormant, unculturable cells and how that may confound experimental results. Whenever possible (e.g., when antibiotic effectiveness is being tested) researchers should incorporate viability detection in their protocols (40–42). A simple yet preliminary alternative to testing for the presence of VBNC cells would be to allow potential VBNC cells to resuscitate in a nutrient- and antibiotic-free medium for at least 24 h immediately after antibiotic testing.

Does human serum induce persisters and VBNC cells?

To test whether persisters and VBNC cells also coexist in a clinically relevant stressful environment, we tested their existence and dynamics in human serum. Serum is a stressful condition that many pathogens (such as V. vulnificus) encounter during disease progression and has previously been shown to limit the growth of certain bacteria (43). Interestingly, a recent study by Putrinš et al. reported that population heterogeneity allows a subpopulation comprised of persisters to survive a combination of antibiotics and human serum (44). Thus, as with antibiotic treatment, we hypothesized that exposure of bacteria to human serum would also induce both the persister and VBNC states. To test whether human serum increases the proportion of persisters and VBNC cells, V. vulnificus was incubated in active or heat-inactivated human serum for 1 h at 37°C to allow for a physiological response to serum. Ampicillin was then added to the cultures in serum, which were allowed to incubate at 37°C for another 4 h (as done in classical persister experiments). After each step of the experiment (log phase, post-serum treatment, and post-ampicillin treatment in serum), aliquots of the samples were washed, and culturability was measured. To quantify persisters, cells were plated on solid medium immediately after ampicillin treatment in serum. To quantify VBNC cells, washed cultures were allowed to resuscitate at 20°C for 24 h before culturability was quantified. Figure 4A shows the percentage of culturable cells after ampicillin treatment (representing the persister population) in active and inactive serum and after 24 h of resuscitation (representing the persister population and resuscitated VBNC cells). The percentages were calculated relative to the amount of cells exposed to serum for 1 h (pre-ampicillin treatment). Results revealed a significantly higher percentage of persisters present in active serum than in inactive serum (P = 0.0035), indicating that some of the heat-labile factors in human serum, such as complement, may be inducing persistence. Furthermore, there were significantly more persisters in both active and inactive serum than persisters isolated from antibiotic-treated log-phase cultures without serum (P = 0.0006 and P = 0.0357, respectively), indicating that there may be heat-stable factors in human serum that could potentially induce persistence as well.

FIG 4.

Human serum induces persister and VBNC cells. (A) Cells were incubated in active and inactive human serum (HS) for 1 h, before supplementation with ampicillin for 4 h to isolate persisters. Serum and antibiotics were then removed, and cells were allowed to resuscitate for 24 h at 20°C. Error bars represent standard errors of three replicates. *, P < 0.05; **, P < 0.01; ***, P < 0.001 (individual unpaired t test). (B) Persister cells were isolated at 30°C and 37°C. Error bars represent standard errors of three replicates. *, P = 0.0385 (unpaired t test). Cont Pers, control persister; Resus, resuscitated.

Since human serum experiments were performed at 37°C while all other persister studies were performed at 30°C, we tested whether temperature has an effect on the level of persistence. Results showed that there were significantly (P = 0.0385) fewer persisters formed at 37°C than at 30°C, indicating that incubation at a higher temperature during human serum experiments was not a factor in the increased level of persistence (Fig. 4B).

To test whether human serum induced the VBNC state, cells that were exposed to antibiotics in human serum were allowed to resuscitate for 24 h prior to measuring culturability. Results show that both active and inactive serum harbored a substantial number of VBNC cells (Fig. 4A). However, there were significantly more resuscitated cells after active-serum treatment than after the inactive-serum treatment, indicating that heat-labile factors in human serum may be inducing the VBNC state. Furthermore, significantly more cells (P = 0.0051) resuscitated after ampicillin treatment in human serum (Active HS Resus.) than after ampicillin treatment alone (Cont. Pers. Resus.), further supporting that active human serum induces the VBNC state.

These results indicate that human serum, containing important components of the innate immune system, promotes the survival of bacteria in the face of antibiotics. This phenomenon is similar to the antagonism observed when cells are exposed to bactericidal and bacteriostatic antibiotics simultaneously. This study demonstrates that human serum induces a nongrowing state (VBNC state and persistence), which is tolerant to the bactericidal antibiotic ampicillin, an antibiotic that is most effective against rapidly growing cells.

Are TA systems induced in VBNC cells?

Toxin-antitoxin (TA) systems were first proposed to play a role in persistence over a decade ago (45). It has previously been shown that type II TAS are important mediators of persistence (7, 10). Type II TA systems are composed of a protein toxin that is neutralized by a protein antitoxin by direct interaction. When levels of antitoxin are reduced through the action of activated intracellular proteases such as Lon and Clp, type II toxins are liberated and go on to inhibit translation. RelE and HipA are type II toxins that inhibit translation by direct cleavage of mRNA at the ribosomal A site and by phosphorylation of glutamyl-tRNA synthetase, respectively (7).

The VBNC state is known to be a nonreplicating physiology with reduced protein synthesis, and it has previously been suggested that TA systems may play a role in controlling the VBNC state (46–50). Considering that HipA and RelE proteins are known to induce persistence through inhibition of translation, we examined gene expression of several hipA and relE homologs after V. vulnificus cells entered into the VBNC state.

Relative expression levels of the CMCP6 loci hipA2 and hipA3 were found to be 16.8- and 7.0-fold higher, respectively, in VBNC cells than in log-phase cells, while the expression levels of hipA1 were no different between the two cell states (Fig. 5). Relative expression levels of several relE homologs were also found to be markedly elevated in VBNC cells. Notably, a 111-fold increase was evident for one locus (relE5), while others were elevated 20-fold (relE1-relE2) and 23-fold (relE3) (Fig. 6A). The three hipA loci tested here have sequence homology to the hipA toxin gene implicated in the growth arrest of E. coli persister cells (46). Furthermore, increased transcript abundance for relE family genes is characteristic of transcriptional profiles for E. coli persister cells (51). The presence of multiple copies of hipBA and relBE operon homologs in V. vulnificus, along with evidence for increased transcript abundance in VBNC cells, prompts the consideration that these systems may play a role in the VBNC state. Overall, these findings suggest that both VBNC and persister states are nonreplicating physiologies linked by a characteristic increase in transcript level for hipBA- and relBE-related TAS.

FIG 5.

The VBNC state induces expression of hipA toxins. Expression of hipA homologs hipA2 and hipA3 are elevated in VBNC cells relative to levels of culturable cells in log phase. Fold expression values were calculated relative to log phase expression and normalized using rplT as the reference gene. Error bars indicate standard deviations of three biological replicates and three technical replicates. Asterisks represent statistically significant differences assessed using a Kruskal-Wallis nonparametric ANOVA and Dunn's post hoc test. **, P < 0.01; ****, P < 0.0001.

FIG 6.

The VBNC state induces expression of relE toxins. (A) Expression of relE homologs relE1-relE2, relE3, and relE5 are elevated in VBNC cells relative to levels in culturable cells. (B) Relative expression of relE5 (VV1_0110) is elevated in cells after 60 min of exposure to VBNC-inducing conditions. For both A and B, culturable cells are in log phase, and fold expression values were calculated relative to log-phase expression using rplT as the reference gene. Error bars indicate standard deviations of three biological replicates and three technical replicates. Asterisks represent statistically significant differences assessed using a Kruskal-Wallis nonparametric ANOVA and Dunn's post hoc test. *, P > 0.05; **, P < 0.01; ****, P < 0.0001.

Evaluation of relE5 expression in a sample of cells obtained after 60 min of exposure to cold temperatures showed rapid (4-fold) elevation though not to the extent manifested in the fully VBNC cells (Fig. 6B). It is possible that that the culturable cell population sampled after this 60-min period is composed of cells that have modulated their transcriptional profiles in response to the environmental perturbation. As such, they may represent a population of cells in the early phases of transition to a VBNC lifestyle. It is important to note that the VBNC state was induced under a stressful condition (4°C), and any early change in gene expression may be a simple result of the stress and not necessarily a hallmark of the VBNC state or the transition to such a state. However, this does not take away from the finding that a fully VBNC population had an increased level of TAS expression. Alternatively, it is likely that populations are phenotypically multistable and harbor some fraction of VBNC cells and persisters at all times. If the stressful conditions selectively eliminate those cells not already in the VBNC and persister states, the RNA pool from the population present in the 60-min sample would have an increase in the proportional representation of VBNC and persister transcripts in the collective population, leading to an observed difference in relative expression levels. The fact that both explanations are in agreement with this observation underscores the need for experimental tools that afford single-cell resolution for the study of populations entering the VBNC state. The genes identified in this study as having substantially elevated transcript abundances in VBNC cells could potentially be leveraged for the creation of reporter strains that would indicate individual cells in this state. Such an approach is not without precedent and has contributed greatly to the field of persister cell research (10, 46, 52).

Are toxin-antitoxin systems induced in human serum?

Since human serum was found to be an effective inducer of both VBNC and persister cells, we tested whether human serum induces these dormant states through an increase in the expression of relE and hipA in V. vulnificus. Results show that there were approximately 3-fold and 2-fold increases in the expression levels of relE and hipA, respectively, in active serum relative to levels in a log-phase culture, whereas there was lower expression of these toxin genes in inactivated serum than in active serum (Fig. 7). These results indicate that a heat-labile component of serum may induce the expression of these type II toxin genes. Since gene expression was evaluated on the entire population, including active cells, persisters, and VBNC cells, a modest fold increase in expression may in fact be very biologically significant since there are more RNA transcripts produced by active cells than by persisters and VBNC cells. It is also possible that active serum induces other TAS loci not investigated here which contribute to the higher level of dormancy that is attained by active serum. In order to further substantiate these findings, gene expression studies on purely dormant populations in human serum should be performed.

FIG 7.

Human serum induces expression of TAS. Expression of relE5 and hipA2 was evaluated in cells incubated in active or inactive human serum for 4 h. Fold expression values were calculated relative to log-phase expression and normalized using gyrB as the reference gene. Error bars indicate standard deviations of three technical replicates. Asterisks represent statistically significant differences assessed using one-way ANOVA and Bonferroni's post hoc test. **, P < 0.01; ***, P > 0.001; ****, P < 0.0001.

Concluding remarks.

We have previously proposed that VBNC cells and persisters are closely related and that they share molecular mechanisms that modulate their existence (33). A concise summary of the dormancy continuum hypothesis is depicted in Fig. 8. Importantly, we have discovered that VBNC cells, like persisters, may exist prior to exposure to stressful environments. We also demonstrate that VBNC cells coexist with persisters after antibiotic treatment and (if given the appropriate time) are able to resuscitate when the antibiotics are removed. These findings not only highlight the inadequacies of conventional antibiotics but also demonstrate that the culture-based methods often used to evaluate antibiotic efficacy are insufficient in detecting a substantial portion of cells (VBNC), which tolerate high-dose antibiotics and have the potential to resuscitate within a host. We also demonstrated that a simple, accessible, and widely available viability staining protocol could be used to more effectively evaluate antibiotic efficacy by allowing the detection of all viable cell types in a heterogeneous population.

FIG 8.

The microbial dormancy continuum hypothesis. Environmental stress induces cellular processes that lead to the degradation of antitoxins, causing the liberation of their cognate toxins (triangles). This affects cellular metabolism and, consequently, growth. In the initial stages of dormancy, this produces antibiotic-tolerant nongrowing cells (persister cells) that can quickly resuscitate upon removal of the stress. Resuscitated or active cells have relatively low free-toxin levels and proliferate on medium. However, if cells are exposed to prolonged stress, more free toxins accumulate, the extent of dormancy increases, metabolic activity further decreases, and cells require more time to resuscitate, rendering cells viable but nonculturable (VBNC cells). When VBNC cells are allowed to resuscitate after removal of the inducing stress, their free-toxin levels slowly decrease while metabolic activity increases. At a particular free-toxin threshold, cells are defined as persisters because they are nongrowing but can quickly resuscitate into active cells and become culturable on medium. (Reprinted from reference 33 with permission from Elsevier.)

Since TA systems are strongly implicated in mediating persistence, we further corroborate the relatedness between VBNC cells and persisters by showing that VBNC cells express higher levels of toxin genes than growing cells. To verify these findings in a medically relevant environment, we tested the ability of V. vulnificus to form persisters and VBNC cells in human serum and found that human serum induced the formation of persisters and VBNC cells and that cultures incubated in human serum expressed significantly more toxin genes. These findings further support the role of TAS in the modulation of both persistence and the VBNC state. Overall, this study provides a stronger link between persisters and VBNC cells, highlights the potential clinical significance of both states, and emphasizes the need to use proper viability assays rather than culturability when bacterial cell survival is being evaluated.