Monitoring the resuscitating cells with the protein-dilution method
One of the common persister-isolation techniques involves treating cultures with beta-lactams (such as ampicillin) and sedimentation of the non-lysed cell population [10]. Ampicillin inhibits transpeptidase enzymes involved in bacterial cell wall synthesis, thus targeting proliferating cells only [15, 16]. This was also verified by a fluorescent protein-dilution method here (Fig. 1), a strategy commonly used to monitor cell growth [3, 9]. In this method, we first induced mCherry expression during overnight growth of an E. coli strain that harbors a chromosomally integrated IPTG-inducible mCherry expression cassette. The mCherry-positive cells from the overnight pre-culture (Fig. 1, t = 0) were then inoculated into a fresh medium without the inducer. At time zero, all cells exhibited a high level of mCherry (red) fluorescence, which declined as the cells divided, except in a small subpopulation (~ 4% of the entire population at t = 150 min, OD600 = 0.25) whose fluorescence remained constant due to the lack of division (Fig. 1, subpopulations highlighted with red circles). As expected, the growing cells, exhibiting higher forward scatter (FSC) signals, became filamented and were lysed rapidly upon exposed to ampicillin; however, the non-growing cell population, which was shown to be enriched with persister and VBNC cells [8, 12], remained intact (Fig. 1).
Using the protein dilution and ampicillin-induced cell lysing techniques, we wanted to monitor persister resuscitation on flow-cytometry diagrams. Unlike the strategy described above, IPTG was kept in the media during the exponential growth phase and ampicillin treatment. Although persisters are largely assumed to be pre-existing non-growing cells, antibiotics are also known to induce cell dormancy and persistence in proliferating cells [17]. In fact, up to 20% of persister cells can arise from growing cell subpopulations [8]. Therefore, IPTG was used to maintain high fluorescent signals in these persister types. When we treated the mid-exponential-phase cells (OD600 = 0.25, Fig. 2a) with ampicillin, the growing cells eventually lost their membrane integrity and mCherry (Fig. 2a, t = 10 to 180 min), as expected. In contrast, live, intact cells, comprising persister and VBNC cells, retained high fluorescence (Fig. 2a, t = 180 min, the subpopulation highlighted with a red circle). Our flow cytometry images showed that a 3-h treatment is sufficient to lyse all antibiotic sensitive cells (Fig. 2a, t = 180 min). This treatment length was also found to be sufficient to obtain a bi-phasic kill curve of colony-forming unit (CFU) counts, which ensures the enrichment of persisters and the death of non-persister cells in the cultures (Fig. 2c). After the treatment, cells were washed to remove the antibiotic and IPTG, and then transferred to fresh Luria-Bertani (LB) broth to stimulate persister resuscitation. Persisters, unlike VBNCs, can exit from their non-proliferating phenotypic state and proliferate upon removal of antibiotics. The resuscitating cells were detected by monitoring single-cell mCherry levels using a flow cytometer. We observed that, although the surviving live cells initially exhibited high fluorescence, upon resuscitation in the absence of IPTG, flow cytometry revealed ongoing cell division as the dilution of mCherry protein (Fig. 2b, subpopulations highlighted with green circles). Forward scatter was also expected to increase due to the elongation characteristic of the growing cells. The fluorescence of the cells that did not resuscitate (i.e., VBNCs) remained constant due to lack of cell division (Fig. 2b, subpopulations highlighted with red circles).
Ampicillin persisters wakeup within one hour
Our flow-cytometry data indicates that persister cells started to resuscitate within 1 hour upon their transfer to fresh media (Fig. 2b). We observed a similar trend in the control cultures where stationary-phase cells from overnight pre-cultures were transferred to fresh media without receiving antibiotic treatments. These untreated cells also started to divide within 1 hour (Additional file 1: Fig. S1A). The doubling time of resuscitating persisters (estimated from the fluorescence decay equation; see Methods) was obtained as 23.30 ± 2.54 min (Fig. 2d), which was found to be consistent with the doubling time of normal, untreated cells (23.91 ± 1.7 min) that grew in LB (Additional file 1: Fig. S1B). We also estimated the initial number of persister cells (No) that woke up in fresh liquid cultures, using the classical exponential-growth equation, \( N={N}_o{2}^{\left(t-{t}_o\right)/{t}_d} \), where td is the doubling time (calculated from the fluorescence decay equation) and N is the number of resuscitated cells at time t. For N, we used the flow cytometry cell counts obtained at t = 240 min (Fig. 2b, highlighted with a green circle). Although it is hard to predict the exact initial resuscitation time (to) of persister cells, our flow-cytometry diagrams showed that these cell subpopulations became visible at t = 60 min (Fig. 2b, highlighted with a green circle), indicating that their wake-up time (to) should be less than 60 min. The same trend was also observed during the wake-up of untreated, stationary phase cells (Additional file 1: Fig. S1A). Here, we used two scenarios to calculate the No levels from the flow-cytometry diagrams: (i) assuming to = 0, a scenario that underestimates No levels since resuscitation is not expected to happen at this time point, or (ii) assuming to = 60 min, a scenario that overestimates No levels since persister cells have already resuscitated (Fig. 2e, patterned orange columns). The actual No should be between these two calculated values. Since the resuscitating cells at 60 min were measurable by the flow cytometer, we compared the calculated cell levels at to = 60 min with the experimentally determined values to further validate our computational analysis. As expected, we did not see a significant difference between the calculated and experimental data (Additional file 1: Fig. S2A). We also quantify No levels using the standard agar plating method; briefly, samples after ampicillin treatments were collected, washed to dilute antibiotics to sub-MIC levels, and plated on agar plates. Once a persister cell starts dividing, it forms a colony; thus, the CFU levels correlate with the number of resuscitating cells on agar plates. No estimates from the flow-cytometry analysis and the CFU measurements of agar plates were found to be consistent and within the expected range (Fig. 2e). Still, resuscitating cells, as measured by CFUs, were calculated as ~ 4% of the non-lysed cell subpopulation. The remaining, ~ 96% cells, classified as VBNCs, did not resume replication. Consistent with the previously published results [8, 9, 12], VBNC cells were found to be more abundant than persisters in our cultures as well. We note that the dilution of mCherry in resuscitating cells is not due to the protein leakage caused by compromised membranes. In order to verify that these cells are alive, we used a pQE-80 L expression system, enabling us to tightly regulate the expression of a green fluorescent protein (GFP) with IPTG. Unlike the mCherry-dilution method, GFP was not induced initially. Cells were first treated with ampicillin for 3 h in the absence of the inducer, and then, transferred to fresh media with IPTG. As expected, a similar fraction of cell subpopulation started to resuscitate and overexpress GFP (Additional file 1: Fig. S3ABC), verifying that resuscitating cells are live cells with active metabolic mechanisms.
Long-term ampicillin treatment did not affect the persister resuscitation and doubling time
To elucidate the effect of long-term antibiotic treatments on persister resuscitation, we treated the mid-exponential-phase cells with ampicillin for 16 h. Although, non-lysed cell levels quantified by flow cytometry in short- and long-term persister assay cultures (3-h treatment vs. 16-h treatment) were found to be similar (Fig. 3c), persister levels obtained from CFU measurements were slightly lower (P < 0.05) in 16-h treatment cultures (Figs. 2e and 3d; gray columns representing the 3-h or 16-h ampicillin treatments). Since the slower-killing phase in the biphasic kill curves is the hallmark of the persistence phenotype, it is possible that longer exposure to the antibiotic has killed more persister cells or converted some of them into VBNC cells. However, the observed increase in VBNC levels in the long-term persister assay cultures is not statistically significant (Additional file 1: Fig. S4), which may be expected considering that VBNCs are orders of magnitude more abundant than persisters.
Persister cells similarly started to resuscitate in fresh LB broth within 60 min after the removal of antibiotics (Fig. 3a, t = 60 min, the subpopulation highlighted with a green circle) with a doubling time (21.43 ± 0.24) (Fig. 3b) similar to that of untreated cells (Additional file 1: Fig. S1B). CFU measurements from agar plates were found to be within the range of No levels estimated from our flow-cytometry analysis (Fig. 3d and Additional file 1: Fig. S2B), although persister cells still constitute a small fraction of the non-lysed cell subpopulation. Overall, these results unveiled that long-term treatments do not pose any impact on non-lysed cell levels as well as persister-wake up time or doubling time.
Long-term pre-culturing enhances VBNC cell levels
Persister cell populations are heterogeneous and exhibit diverse, but poorly characterized metabolic and gene expression activities. Whereas some persisters can arise from growing cell subpopulations [8], many persister cells are found to be at the non-proliferating state before antibiotic treatments, and they are largely formed by passage through the stationary phase [18]. To analyze the effects of stationary-phase lengths on persister resuscitation, we cultured the overnight pre-cultures up to 9 days. On certain days (1, 2, 3, 5, 7, and 9 days), cells from overnight pre-cultures were transferred to fresh media, cultured until they reached the mid-exponential growth phase (OD600 = 0.25), and then treated with ampicillin for 3 h or 16 h. Although our results clearly showed that long-term pre-culturing did not significantly impact the persister resuscitation and doubling time (Fig. 4a,b,d,e and Additional file 1: Fig. S5ABCD and Fig. S6ABCD), it significantly increased the non-lysed cell (VBNCs) levels (Fig. 4f,g). Approximately, 10% percent of exponential phase cells obtained from 9-day overnight pre-cultures were not lysed by ampicillin (Fig. 4f,g). We note that mCherry levels of the non-lysed cells from long-term pre-cultures are lower than those from short-term pre-cultures (Figs. 2a vs. 4a); this is because of the increased mCherry protein degradation and/or leakage in long-term pre-cultures (Additional file 1: Fig. S7), known characteristics of the late-stationary phase cells [19, 20]. Interestingly, we observed a significant difference between persister levels obtained from flow-cytometry diagrams and CFU measurements (Fig. 4h,i and Additional file 1: Fig. S2CD). Our analysis indicates that at least 10-fold more cells (based on to = 0 estimates from flow-cytometry analysis, Fig. 4h,i) resuscitated in liquid cultures compared to solid media. This interesting phenomenon, which has been observed in both 3-h- and 16-h-ampicillin-treated cultures (Fig. 4h,i), indicates the importance of culture conditions to persister cell resuscitation.
Stationary-phase lengths in pre-cultures significantly affected the ability of persister cells to wake up in solid media. Persisters levels (as measured by CFU counts on agar plates) in long-term pre-cultures were found to be almost 50-fold less than those identified in short-term pre-cultures (Figs. 2c vs. 4c). However, this was not observed in liquid media. In fact, our flow cytometry analysis showed that a similar amount of persister cells started to wake up within 1 hour in liquid cultures, regardless of the pre-culturing lengths or ampicillin treatment times (persister data highlighted with patterned orange columns in Figs. 2e, 3d and 4hi). The cell growth profiles of the resuscitating cells in liquid cultures were also found to be very similar in all conditions tested (Fig. 5). Doubling times obtained from these growth profiles (Fig. 5) were found to be consistent with those obtained from mCherry-dilution method (Additional file 1: Table S1). Overall, these results indicate that although long-term culturing does not significantly affect the persister resuscitation and doubling time in liquid cultures, it impacts the non-lysed cell levels. Our study also clearly shows that the culture environment affects the ability of non-lysed cells to wake-up.
Ampicillin cannot lyse the pre-proliferating cells in the presence of arsenate
A correlation between ATP depletion and persistence has been consistently shown in previous studies using arsenate-treated cell cultures [21,22,23]. Arsenate, which competes with phosphate [24], is thought to increase persistence in bacteria by reducing ATP production [21, 22]. To elucidate the resuscitation characteristics of cell populations exhibiting increased persistence, we treated the exponential-phase cultures (t = 150 min, OD600 = 0.25) with 10 mM arsenate for 30 min followed by ampicillin treatment with arsenate (co-treatment), as described previously [21, 22]. Thirty-min pretreatment was found to be sufficient to inhibit cell proliferation (Fig. 6b, the cell subpopulations highlighted with dark-green circles) and to reduce ATP levels (Fig. 6d), consistent with the studies published by Kim Lewis’ group [21,22,23]. As expected, the persister levels of arsenate-treated cultures were significantly higher than those of the control groups (Fig. 6c). However, our flow cytometry analysis indicates that this increase in persistence may be due to the bacteriostatic effect of arsenate during antibiotic treatment, as ampicillin was not able to lyse the pre-proliferating cells in the presence of arsenate (Fig. 6b, the cell subpopulations highlighted with dark-green circles). Interestingly, all these non-lysed, pre-proliferating cells started to resuscitate within an hour in fresh media after the removal of ampicillin and arsenate (Fig. 6b, the cell subpopulations highlighted with light-green circles), and reached the stationary phase very quickly (Fig. 6b, t = 240 min and Additional file 1: Fig. S8). We note that the cultures at t = 240 min were further diluted for flow-cytometry analysis; this explains the reduction in the number of non-growing cells shown in the flow diagram (Fig. 6b, t = 240 min and Additional file 1: Fig. S8). Due to their transition to the stationary phase, FSC of resuscitating cells decreased at t = 240 min (Fig. 6b). This is expected as the reductive division resulting in small spherical cells is known to take place when cells enter the stationary phase [19]. We also note that the chromosomal mCherry expression cassette is slightly leaky during the stationary phase even in the absence of IPTG; therefore, mCherry levels of resuscitated cells at t = 240 min (Fig. 6b) are slightly higher than expected. This might be due to the up-regulation of CRP/cAMP complex during the stationary phase, as this complex tightly regulates the lac promoters. However, this phenomenon does not impact our current analysis; therefore, we are not planning to investigate this further. The control group has undergone similar procedures without receiving any arsenate treatment. All proliferating cells were lysed by ampicillin, and only a small fraction of intact cells from the non-growing cell subpopulation was able to resuscitate in the control group (Fig. 6a), consistent with our aforementioned results.
Antibiotic treatments have been performed in the presence of arsenate in previous studies [21, 22], which makes it ambiguous whether the arsenate-induced persistence is indeed due to the ATP depletion. To figure out this, we performed the persister assays mentioned above in the absence of arsenate. After the pretreatment with arsenate, we washed the cells with Phosphate Buffered Saline (PBS) solution to remove the chemical and transferred the washed cells to fresh media with ampicillin. This washing procedure did not affect the exponential-phase persister (Figs. 6c vs. 7c) and ATP levels (Figs. 6d vs. 7d) in control groups. Our flow cytometry analysis showed that growing subpopulations in both pretreatment and control groups were lysed within 3 h by ampicillin (Fig. 7a,b). Unexpectedly, both groups have similar persister levels (Fig. 7c), despite the significant reduction in ATP levels observed in arsenate-pretreated cultures (Fig. 7d). Also, persisters from non-growing cell subpopulations started to resuscitate within an hour in both pretreatment and control groups when they were exposed to fresh media (Fig. 7a,b). Overall, these results indicate that the ATP-depleted cultures do not necessarily exhibit increased persistence (Fig. 7c); in fact, the enhanced persistence shown in Fig. 6c is potentially due to the synergistic effect of arsenate in co-treated cultures.