Research ArticleMicrobiology

Slow growth determines nonheritable antibiotic resistance in Salmonella enterica

See allHide authors and affiliations

Science Signaling  30 Jul 2019:
Vol. 12, Issue 592, eaax3938
DOI: 10.1126/scisignal.aax3938

Slow growth for bacterial persistence

Even bacteria that do not carry mutations or genes that confer resistance to specific antibiotics can survive antibiotic treatment, a phenomenon known as persistence (see the Focus by Kaldalu and Tenson). Several models have been proposed to account for bacterial persistence, including the activation of toxins in toxin-antitoxin modules, the production of the alarmone guanosine (penta) tetraphosphate [(p)ppGpp], and a reduction in intracellular adenosine triphosphate (ATP) abundance. Pontes and Groisman demonstrated that Salmonella exhibited persistence even in the absence of toxin-antitoxin modules or (p)ppGpp production and under conditions that increased intracellular ATP. These and additional findings show that slow growth alone is sufficient for persistence and may contribute to the difficulty in treating some bacterial infections.

Abstract

Bacteria can withstand killing by bactericidal antibiotics through phenotypic changes mediated by their preexisting genetic repertoire. These changes can be exhibited transiently by a large fraction of the bacterial population, giving rise to tolerance, or displayed by a small subpopulation, giving rise to persistence. Apart from undermining the use of antibiotics, tolerant and persistent bacteria foster the emergence of antibiotic-resistant mutants. Persister formation has been attributed to alterations in the abundance of particular proteins, metabolites, and signaling molecules, including toxin-antitoxin modules, adenosine triphosphate, and guanosine (penta) tetraphosphate, respectively. Here, we report that persistent bacteria form as a result of slow growth alone, despite opposite changes in the abundance of such proteins, metabolites, and signaling molecules. Our findings argue that transitory disturbances to core activities, which are often linked to cell growth, promote a persister state regardless of the underlying physiological process responsible for the change in growth.

INTRODUCTION

Bacteria can survive prolonged exposure to high concentrations of antibiotics despite lacking specific genetic mechanisms to counteract the bactericidal effects of these compounds. This behavior can be manifested population-wide by tolerant cells or by a small subpopulation of cells known as persisters (1, 2). Genetically identical to the original susceptible population, persistent bacteria prevent the eradication of bacterial infections and favor the emergence of antibiotic-resistant mutants (37). The occurrence of persisters also hinders pathogen elimination by the host immune system. For example, upon internalization by macrophages, the facultative intracellular pathogen Salmonella enterica serovar Typhimurium forms a slow-replicating subpopulation that is both resistant to killing by ampicillin and protected from clearance by macrophages (8). Because antibiotic persistence has been reported in phylogenetically diverse bacterial species (9, 10), the underlying basis for this phenomenon is likely to be conserved across bacteria.

In Escherichia coli, persistence was initially attributed to a signaling pathway that relied on the RelA and SpoT proteins (11, 12), the synthase and a combined synthase and hydrolase of the alarmone guanosine (penta) tetraphosphate [(p)ppGpp], respectively (13). Stochastically synthesized, (p)ppGpp indirectly activates the protease Lon, which degrades antitoxins from toxin-antitoxin (TA) modules. This allows the released toxins to inhibit core cellular processes, such as protein synthesis, resulting in tolerance to multiple antibiotics (11). Similarly, Salmonella has been reported to form persisters within 30 min of phagocytosis by murine macrophages through a pathway that depends on acidification of the Salmonella-containing vacuole, the synthesis of (p)ppGpp, and the resulting expression of genes encoding toxins from TA modules (14). In support of this model, mutants unable to synthesize (p)ppGpp or lacking the genes for particular TA modules are defective in persister formation, and this is also the case for wild-type Salmonella if phagosome acidification is inhibited (14). In addition, during growth in Luria-Bertani (LB) medium, the serine analog serine hydroxamate (Shx), which induces (p)ppGpp production, or a mildly acidic pH, a condition that partially mimics the environment Salmonella encounters inside an acidic macrophage phagosome (15), promotes the formation of persisters and the expression of genes encoding TA modules (14), an event often associated with the release of TA toxins (16).

However, TA modules were eventually shown to be dispensable for persister formation in both E. coli (17, 18) and Staphylococcus aureus (19). Moreover, the original report ascribing the formation of E. coli persisters to (p)ppGpp-mediated stochastic induction of TA modules was retracted (11, 20). Thus, an alternative model suggesting a causal connection between a reduction in adenosine triphosphate (ATP) amounts and persistence emerged. The latter model was supported by the low ATP amounts present in E. coli and S. aureus persisters (18, 19). That persister formation requires (p)ppGpp, TA modules, and acidified conditions in Salmonella (14), but not in the related Gram-negative species E. coli or the distantly related Gram-positive S. aureus (1719), raised the possibility of persister formation actually differing across bacterial species and prompted us to reexamine persister formation in Salmonella.

Here, we report that growth rate is the ultimate indicator of persister formation in Salmonella and most likely in other bacterial species. We established that (p)ppGpp, TA modules, and exposure to an acidified medium were all dispensable for persister formation in different media, demonstrating that Salmonella behaves like E. coli and S. aureus with regard to persister formation. We identified conditions in which antibiotic tolerance was achieved despite high ATP amounts, which rules out a low ATP concentration as being essential for this property. Our model accounts for the inextricable association between growth and the activity of core cellular processes, which are both the sources of biosynthetic activity and the targets of antibiotics. Moreover, the model suggests potential strategies to eliminate persister organisms.

RESULTS

Low cytoplasmic Mg2+ induces Salmonella tolerance to antibiotics independently of (p)ppGpp and TA modules

Three properties—slow growth, increased (p)ppGpp amounts, and low ATP abundance—have been implicated in persister formation and antibiotic tolerance (11, 18, 19, 2123). Given that low cytoplasmic Mg2+ triggers slow growth, increased (p)ppGpp amounts, and low ATP abundance (24, 25), we wondered whether it also promotes persister formation or antibiotic tolerance. Thus, we examined bacterial survival after exposure to bactericidal antibiotics before and after a drop in cytoplasmic Mg2+ concentration that reduces growth, increases (p)ppGpp, and decreases ATP. When placed in defined media containing low Mg2+ (10 μM), Salmonella grows rapidly, consuming the extracellular Mg2+ available in the medium. At about 4 hours of growth, when the culture reaches a density of about 4 × 107 cells/ml, the concentration of Mg2+ in the medium becomes limiting. This leads to a drop in the cytoplasmic Mg2+ concentration. Thus, at 2 hours in defined media containing 10 μM Mg2+, Salmonella replicates quickly because the cytoplasmic Mg2+ concentration is not limiting. In contrast, by 5 hours, Salmonella replicates very slowly due to a decrease in the cytoplasmic Mg2+ concentration, an event that is accompanied by a reduction in growth rate and ATP abundance and an increase in (p)ppGpp (25). Wild-type Salmonella grown in defined media with low Mg2+ for 5 hours were ~10,000-fold less effectively killed by the cell wall inhibitor cefotaxime or the DNA gyrase inhibitor ciprofloxacin than were cells grown for 2 hours in low Mg2+ (Fig. 1, A and B). Because the media were at neutral pH, these results indicate that a mildly acidic pH was not required for tolerance to cefotaxime or ciprofloxacin in Salmonella.

Fig. 1 Effects of low cytosolic Mg2+ on Salmonella antibiotic tolerance.

(A) Quantification of surviving wild-type (14028s), ∆12TA (MP1422), and relA::Tn10 spoT (MP342) Salmonella after a 2.5-hour exposure to cefotaxime or ciprofloxacin after 2 (2h) or 5 (5h) hours of growth in low (10 μM) Mg2+ medium. Error bars represent SDs, derived from at least eight biological samples and three independent experiments. Note the log scale of y axis. CFU, colony-forming units. (B) Representative dilution series of wild-type (14028s), ∆12TA (MP1422), and relA::Tn10 spoT (MP342) Salmonella after a 2.5-hour exposure to cefotaxime or ciprofloxacin after 2 or 5 hours of growth in low Mg2+ medium. Each spot represents a 5-μl aliquot.

A Salmonella relA spoT double mutant, which is unable to produce (p)ppGpp (13), exhibited wild-type tolerance to cefotaxime and ciprofloxacin at both 2 and 5 hours after growth in low Mg2+, and this was also the case for an engineered strain lacking all 12 bona fide TA modules (26) designated Δ12TA (Fig. 1, A and B). Complete genome sequencing of the Δ12TA strain and its wild-type parental strain revealed the presence of a single genetic change in addition to the engineered mutations (see Materials and Methods). This mutation, consisting of a 125–base pair (bp) deletion between the intergenic regions of the rrlB and rrfB genes, is not predicted to affect antibiotic tolerance. The wild-type parental strain, the relA spoT double mutant, and the Δ12TA mutant grew at similar rates (fig. S1A). These results indicate that (p)ppGpp and the 12 tested TA modules were dispensable for antibiotic tolerance triggered by low cytoplasmic Mg2+ in Salmonella.

Acidic pH, (p)ppGpp, and TA modules are dispensable for persister formation in LB medium

The results presented in the previous section imply that either growth in defined media of low Mg2+ induces tolerance by a different mechanism from that taking place in acidified LB broth (14) or tolerance promoted in LB broth does not actually require an acidic pH, (p)ppGpp, and TA modules as reported (14). Therefore, we reexamined persister formation after growth in LB medium using the same experimental conditions and parental strain as those of the previous report (14).

The relA spoT double mutant and the Δ12TA mutant formed persisters in five- to eightfold lower amounts than did wild-type Salmonella after a 24-hour exposure to cefotaxime or ciprofloxacin (Fig. 2A). Moreover, a mildly acidic pH, which was previously reported to increase the number of persisters by three orders of magnitude (14), did not affect the frequency of persisters (Fig. 2B). Furthermore, mutants lacking individual TA modules implicated in macrophage-induced persister formation (14) exhibited wild-type persistence when grown in LB broth (Fig. 2C). Note that the number of surviving bacteria reflects the number of persisters, as opposed to cells that acquired resistant mutations, because similar numbers of surviving cells were retrieved after a second antibiotic exposure (fig. S1B). The presence of mutations conferring resistance would be predicted to increase the number of cells surviving a second antibiotic exposure. These results demonstrate that (p)ppGpp and TA modules were not essential for antibiotic persistence in Salmonella and that an acidic pH did not increase the frequency of persisters formed in LB broth. Thus, they are in contrast to both the previous report on Salmonella (14) and a model in which E. coli persisters emerge through stochastic (p)ppGpp synthesis (11, 17, 19).

Fig. 2 Effects TA modules, (p)ppGpp biosynthetic and degradation genes, or acidic pH on Salmonella persistence.

(A) Quantification of surviving wild-type (14028s), ∆12TA (MP1422), and relA::Tn10 spoT (MP342) Salmonella after a 24-hour exposure to cefotaxime or ciprofloxacin in LB medium. (B) Quantification of surviving wild-type (14028s), ∆12TA (MP1422), and relA::Tn10 spoT (MP342) Salmonella after acid shock before a 24-hour exposure to cefotaxime or ciprofloxacin in LB medium. (C) Quantification of surviving wild-type (14028s), ∆TA3 [MP1454; corresponding to ∆TA8 in (14)], ∆TA5 [MP1455; corresponding to higBA2 in (14)], ∆TA7 [MP1456; corresponding to shpAB in (14)], ∆TA9 [MP1457; corresponding to phD/doc in (14)], and ∆TA10 [MP1458; corresponding to relBE1 in (14)] Salmonella after a 24-hour exposure to cefotaxime or ciprofloxacin in LB medium. Error bars represent SDs. n = 8 independent biological replicates. Note the log scale of y axis. For (A) and (B), two-tailed t test between populations at the edge of brackets. For (C), two-tailed t tests paired with wild-type populations: **P < 0.01 and ***P < 0.001.

Triggering (p)ppGpp synthesis with Shx promotes transient antibiotic tolerance but does not affect persister frequency

The serine analog Shx was reported to promote persister formation in stationary-phase cultures of Salmonella (14). This observation is puzzling for two reasons: First, Shx induces amino acid starvation by hindering the charging of seryl–transfer RNA, leading to inhibition of growth (27) and the stimulation of (p)ppGpp synthesis (28). However, we found that the (p)ppGpp-lacking relA spoT double mutant formed persisters at a similar frequency as did the wild-type strain (Fig. 2A). Second, logarithmically growing bacteria have active ribosomes that stall in response to Shx treatment; RelA is recruited to stalled ribosomes and synthesizes (p)ppGpp (28). Therefore, Shx should promote tolerance in logarithmically growing bacteria but not in stationary-phase cultures because the latter, already nutrient-starved, have high amounts of (p)ppGpp and reduced translation rates (27, 2931). As hypothesized, the addition of Shx to bacteria in the stationary phase altered neither growth (Fig. 3A) nor tolerance (Fig. 3B and fig. S2). This is in contrast to Shx treatment of logarithmically growing Salmonella, which elicited a temporary growth arrest (Fig. 3A) during which the bacteria were immune to killing by cefotaxime (Fig. 3B). We note that Shx did not increase persister frequency when added to either stationary-phase or logarithmically growing bacteria (Fig. 3B and fig. S2).

Fig. 3 Effect of Shx on bacterial growth and antibiotic tolerance.

(A) Growth curves of wild-type Salmonella (14028s) in LB medium in the absence (black) or presence (green) of Shx. Shx was added for the first 30 min of outgrowth (t = 0, before logarithmic growth) or at 120 min of growth (t = 120, during logarithmic growth). (B) Quantification of viable bacteria (log CFU/ml) in cultures of wild-type Salmonella (14028s) grown as in (A). Where indicated, bacteria were exposed to cefotaxime or ciprofloxacin for either 24 hours (top) or 22 hours (bottom). Error bars represent SDs. n = 6 independent biological replicates. Rightmost graph shows an enlargement of the area shaded in pink.

In contrast to the complete inhibition of killing by cefotaxime (Fig. 3B), Shx only slowed down killing by ciprofloxacin when compared to the buffer control (Fig. 3B). This likely reflects the summation of the inhibitory effects of (p)ppGpp on the initiation of chromosome replication (32) and preexisting replication forks, which are sensitive to the gyrase-inhibiting effects of ciprofloxacin. That is, (p)ppGpp provides partial protection from ciprofloxacin because it inhibits only new chromosomal replication events. Together, the results presented above demonstrate that the transient growth inhibition caused by Shx in logarithmically growing organisms increased antibiotic tolerance but did not affect persister frequency.

Growth inhibition promotes antibiotic tolerance despite increased ATP abundance

Wild-type Salmonella grown in 10 μM Mg2+ for 5 hours were 10,000-fold more antibiotic tolerant (Fig. 1, A and B) and contain 15 times less ATP (25) than when grown in the same media for 2 hours. These data support the inverse correlation between antibiotic tolerance and ATP concentration reported for E. coli (18, 22) and S. aureus (19). To independently test this correlation, we examined the behavior of wild-type Salmonella in which ATP abundance was decreased by expressing the soluble subunit of the F1Fo adenosine triphosphatase (ATPase) using the plasmid pATPase (33).

Expression of pATPase increased survival to both cefotaxime and ciprofloxacin (Fig. 4A and fig. S3A), whereas the vector control had no effect (Fig. 4A and fig. S2A). These results further support the inverse correlation between ATP abundance and survival to bactericidal antibiotics. However, they are also compatible with greater antibiotic survival resulting from a reduction in bacterial growth rate (34, 35), as opposed to ATP abundance per se. This is because bacteria are growing logarithmically at 2 hours but linearly at 5 hours (25) and also because a reduction in ATP abundance decreases growth rate (fig. S1A) (36).

Fig. 4 Effects of ATP depletion and inhibition of protein synthesis on antibiotic tolerance.

(A) Fraction of surviving wild-type Salmonella (14028s) carrying pATPase or the pVector control after a 24-hour exposure to cefotaxime or ciprofloxacin. (B) Bacterial growth (left y axis, solid lines) and ATP abundance (right y axis, dashed lines). ATP measurements are expressed as relative light units (RLUs). Fraction of surviving (C) wild-type (14028s) or (D) relA::Tn10 spoT (MP342) Salmonella after cefotaxime or ciprofloxacin exposure in the presence or absence of chloramphenicol. Error bars represent SDs. n = 8 independent biological replicates. For (A), (C), and (D), note the log scale of y axis. Wilcoxon rank-sum test between pVector and pATPase (A) or untreated and untreated populations (C and D): ***P < 0.001 and ****P < 0.0001.

To determine whether the distinct antibiotic tolerance displayed by wild-type Salmonella harvested at 2 and 5 hours in 10 μM Mg2+ reflected differences in growth rate versus ATP abundance, we investigated bacteria exposed to the bacteriostatic protein synthesis inhibitor chloramphenicol. Because protein synthesis is the activity that consumes most ATP in a cell (37), chloramphenicol treatment promoted a marked increase in ATP amounts (Fig. 4B) (38) that was accompanied by growth stasis (Fig. 4B) and a decrease in (p)ppGpp abundance (39).

Chloramphenicol rendered wild-type Salmonella highly tolerant to cefotaxime and ciprofloxacin (Fig. 4C and fig. S3B). This tolerance was still observed in the relA spoT mutant (Fig. 4D and fig. S3C), indicating that it was not mediated by (p)ppGpp. Thus, the pharmacological inhibition of protein synthesis leads to tolerance in Salmonella as observed in E. coli exposed to the protein synthesis inhibitor tetracycline (22). These data unequivocally demonstrate that bacteria can exhibit high antibiotic tolerance despite having high amounts of ATP. We conclude that a reduction in bacterial growth rate promotes antibiotic tolerance.

To further explore the notion that conditions leading to growth slowdown or stoppage promote antibiotic tolerance, we examined the behavior of nutritional auxotrophs deprived of their required nutrients. First, we determined that a tryptophan auxotroph (trpEDCBA) was virtually immune to cefotaxime and ciprofloxacin in medium lacking tryptophan (Fig. 5A). This mutant retained high cefotaxime tolerance in a relA spoT background (relA spoT trpA) (Fig. 5B and fig. S4B), indicating independence from (p)ppGpp. Second, the tolerance phenotype of the trpEDCBA mutant is due to growth stasis (as opposed to the tryptophan biosynthetic enzymes affecting tolerance by a mechanism unrelated to tryptophan synthesis), because the trpEDCBA mutant displayed antibiotic sensitivity when grown in the presence of tryptophan (Fig. 5A). Last, a mutant unable to catabolize the sugar mannose (manA) was rendered antibiotic-tolerant when placed in medium with mannose as the sole carbon source (Fig. 5A). This mutant lost its antibiotic tolerance when grown in glucose or in complex medium (Fig. 5A and fig. S4A). These results show that various conditions promote antibiotic tolerance by slowing down bacterial growth.

Fig. 5 Relationship between growth and antibiotic tolerance in nutritional auxotrophs and relA spoT strains.

(A) Fraction of surviving wild-type (14028s), manA (MP50), and trpEDCBA::Tn10 (AS301) Salmonella after a 24-hour treatment with either cefotaxime or ciprofloxacin under growth-permissive and growth-restrictive conditions (see Materials and Methods). Fraction of surviving (B) wild-type (14028s) and relA::Tn10 spoT trpA::kan (MP1494) or (C) wild-type (14028s) and relA::Tn10 spoT (MP342) Salmonella after a 24-hour treatment with cefotaxime under growth-permissive and growth-restrictive conditions. Error bars represent SDs. n = 8 independent biological replicates. Note the log scale of y axis. Wilcoxon rank-sum test between populations at the edge of brackets: ****P < 0.0001.

(p)ppGpp inhibits antibiotic tolerance during restricted growth

(p)ppGpp plays a critical role during nutritional downshifts by promoting reallocation of cellular resources, which ensures balanced growth (27, 29, 30). Specifically, (p)ppGpp inhibits cellular processes favoring bacterial growth until the necessary precursors become available. In downshift experiments, precursor availability is typically provided by the biosynthetic capability of the cell. Thus, although transiently inhibitory, (p)ppGpp does promote growth in the absence of certain nutrients, as evidenced by the fact that a relA spoT mutant displays multiple auxotrophies (13).

Given that growth inhibition promotes antibiotic tolerance in wild-type Salmonella, we postulated that a relA spoT mutant should display increased antibiotic tolerance under conditions in which growth is restricted. As hypothesized, the relA spoT mutant survived cefotaxime treatment better than wild-type Salmonella during growth in defined medium without amino acid supplementation (Fig. 5C and fig. S4C). These results identify a condition in which (p)ppGpp is actually detrimental to antibiotic tolerance.

DISCUSSION

We herein report that a slowdown or stoppage in bacterial growth is a direct driver of antibiotic tolerance. Using an array of growth conditions and genetic contexts, we determined that antibiotic tolerance was inversely correlated with bacterial growth in Salmonella (Figs. 1 and 3 to 5) (34, 35). In addition, we established that (p)ppGpp, TA modules, and acidic pH were all dispensable for persister formation during axenic growth in LB medium (Figs. 2 and 3). These results are in full agreement with those reported by others for E. coli (17, 18) and S. aureus (19).

The association between growth rate and antibiotic tolerance is supported by several independent findings. On the one hand, mutations in essential genes were historically isolated by enriching mutants that resisted killing by antibiotics because such mutants failed to replicate under particular conditions, such as in the absence of an essential nutrient or a cellular precursor or at nonpermissive temperatures (4043). On the other hand, multicopy plasmids expressing genes specifying proteins with unrelated biochemical functions render E. coli antibiotic tolerant by reducing bacterial growth (34, 35). These genes specify the chaperone DnaJ, the lipopolysaccharide-modifying PmrC, the transhydrogenase PntA, the sulfoquinovose isomerase YihS, the formate transport channel FocA, the transcriptional factor Zur, and a putative protein of unknown function termed YqjE. We posit that expression of a toxin from a TA module renders a bacterium tolerant to antibiotics by virtue of reducing bacterial growth (23, 44, 45), as happens when the DnaJ, PmrC, PntA, YihS, FocA, Zur, or YqjE proteins are produced in high amounts or when an organism runs out of an essential nutrient.

In the context of infection, naturally auxotrophic pathogens (46) may become tolerant when their required nutrients are scarce in host tissues. For instance, Mycobacterium tuberculosis is unable to synthesize several nutrients, relying on dedicated transporters to obtain amino acids such as asparagine and aspartate from the host (47, 48). Because it occupies a variety of nutritional microenvironments during infection (49), M. tuberculosis replication may be hindered when experiencing limitation for asparagine, aspartate, or another essential nutrient, rendering subpopulations of this pathogen antibiotic tolerant. Hence, although macrophage internalization leads to the formation of Salmonella persisters that are both antibiotic-tolerant and resistant to killing by host immune cells (8, 14, 50, 51), we ascribe the antibiotic tolerance to the emergence of a Salmonella subpopulation exhibiting a much lower growth rate (8).

Because it slows down growth (fig. S3A) (36), a reduction in ATP amounts can lead to tolerance (Fig. 4A and fig. S3A) (18, 19, 22). However, a reduction in ATP amounts is not a causa sui. That is to say, reduced ATP may contribute to tolerance or result from other processes that promote tolerance. ATP is critical in many cellular activities, and alterations in ATP abundance will inevitably result from changes in other physiological processes, such as metabolic switches and dissipation of membrane potential (23, 52). The activity of ribosomal RNA (rRNA) P1 promoters—a phenotypic marker used to identify E. coli cells harboring low cytosolic ATP abundance (18)—is also regulated by growth rate (31, 53, 54). Critically, the effects of ATP and growth rate on the rRNA P1 promoters can be uncoupled (55, 56). Therefore, low rRNA P1 transcription may reflect low ATP or slow growth, or both.

It is currently unclear why others found a requirement for (p)ppGpp, TA modules, and an acidic pH for persister formation in Salmonella (14) despite using the same experimental conditions and parental strains described in this paper. We established that (p)ppGpp actually reduces antibiotic tolerance under certain conditions (Fig. 5C). It is also puzzling that others reported a 100-fold increase in survival to antibiotics upon Shx treatment of bacteria in the stationary phase (14), when (p)ppGpp is abundant (27, 2931). In light of the findings reported in this paper, it should be noted that (p)ppGpp does not affect persister formation in Mycobacterium smegmatis, a close relative of M. tuberculosis (57). Because of the disparate results obtained with Salmonella grown in laboratory media and also because others found TA modules to be dispensable for persister formation in mice (51, 58), we chose not to reexamine the proposed requirement for (p)ppGpp and TA modules in persister formation triggered when Salmonella is phagocytized by macrophages (14).

How, then, does a halt or decrease in bacterial growth rate promote the high degree of antibiotic tolerance reported in a broad range of bacterial species (9, 10)? Antibiotics target core cellular processes (2), and these processes are essential for growth. These cellular processes are normally coordinated through feedback mechanisms that ensure congruent macromolecular synthesis, balanced growth, and survival (5962). Therefore, a variety of conditions that inhibit bacterial growth may give rise to a persistence state by transiently disrupting housekeeping processes. Given the nature of various core processes and the feedback mechanisms to which they are subjected during growth inhibition, it can be anticipated that variable frequencies of persisters may be recovered from exposure of bacteria to antibiotics that target distinct cellular processes (Figs. 1 to 5). For instance, although growth inhibition triggered by (p)ppGpp renders cells immediately immune to cell wall inhibitors, it will confer partial protection to DNA gyrase inhibitors, likely because active replication forks will not be inhibited (32) (Fig. 3B).

We propose that a condition will give rise to persistence if and only if it does not compromise a preexisting feedback mechanism and displays bistability to allow for the emergence of subpopulations (1, 2), that is, persistence normally arises from nonproductive interactions among core cellular components, which are often enhanced when a bacterium experiences stress. In other words, there is no evidence of a dedicated “genetic program” driving formation of antibiotic persistent cells, which is in contrast to the role that specific genes play in genetic resistance to antibiotics (63).

The proposed model provides a parsimonious explanation for persistence, a phenomenon found not only among bacteria but also in eukaryotes such as the phenotypically distinct cancer cells that are able to survive chemotherapy by entering into a quiescent state (6470). In the context of bacterial persisters, our model accommodates the conclusions of a large number of studies, each ascribing a distinct underlying cause to persistence (18, 19, 21, 44, 45). Furthermore, it explains why certain antibiotic combinations increase bacterial survival (Fig. 4, B to D) (22, 71), whereas others (for example, compounds that inhibit transcription and induce unspecific proteolysis) kill persister cells (72), that is, the latter drug combination likely disrupts a feedback control by degrading a combination of essential core proteins that cannot be resynthesized because of inhibition of transcription.

MATERIALS AND METHODS

Microbial strains, plasmids, and growth conditions

Microbial strains and plasmids used in this study are listed in table S1. All S. enterica serovar Typhimurium strains are derived from strain 14028s (73) and were constructed by lambda Red–mediated recombination (74), followed by phage P22–mediated transduction as described (75). Bacterial strains used in transformation or recombination were grown in LB medium at 30° or 37°C (74). For selection of recombinant mutants, LB medium was supplemented with ampicillin (100 μg/ml), chloramphenicol (20 μg/ml), tetracycline (20 μg/ml), kanamycin (50 μg/ml), apramycin (80 μg/ml), gentamycin (18 μg/ml), and/or spectinomycin (100 μg/ml). For antibiotic tolerance experiments, growth medium was supplemented with ampicillin (20 μg/ml), chloramphenicol (50 μg/ml), cefotaxime (200 μg/ml), ciprofloxacin (2 μg/ml), and Shx (100 μg/ml). Antibiotic tolerance experiments were conducted in complex LB (Miller) liquid medium or the following defined liquid medium formulations. For Mg2+ starvation experiments, cells were grown in MOPS minimal medium (76) lacking calcium and containing 10 and 2 mM K2HPO4 or 10 μM MgCl2 and 500 μM K2HPO4, 30 mM glucose, and 0.1% casamino acids. For experiments involving tryptophan auxotrophs, cells were grown in MOPS minimal medium lacking calcium and containing 5 mM MgCl2, 2 mM K2HPO4, 30 mM glucose, amino acid mixture 1 (AA1MIX: 800 μM arginine; 500 μM methionine, histidine, valine, threonine, and leucine; and 250 μM tyrosine and glycine), and the presence (permissive to growth) or absence (restrictive to growth) of 800 μM tryptophan. For experiments involving the mannose catabolic mutant, cells were grown in MOPS minimal medium lacking calcium and containing 5 mM MgCl2, 2 mM K2HPO4, AA1MIX, 800 μM tryptophan, and 30 mM of either glucose (permissive to growth) or mannose (restrictive to growth). For experiments with strains containing relA spoT mutations, cells were grown in LB medium (permissive to growth) or N-minimal medium (restrictive to growth) (24) with amino acid mixture 2 (100 μM methionine, histidine, valine, threonine, leucine, and isoleucine) and 37 mM glycerol as the sole carbon source. In experiments involving strains harboring plasmid pATPase (33), cells were grown in MOPS minimal medium containing 5 mM MgCl2, 2 mM K2HPO4, 30 mM glucose, and 0.1% casamino acids. Expression of the ATPase was induced by addition of 1 mM isopropyl β-d-1-thiogalactopyranoside (IPTG). For all physiological experiments, bacteria were grown in 50-ml conical tubes containing 6 ml of culture medium. Cultures were incubated at 37°C in an orbital shaker (330 rpm).

Determination of colony-forming units

After growth or exposure to cefotaxime or ciprofloxacin, cells were transferred to 5-ml centrifuge tubes and collected by centrifugation (13,500 rpm for 20 min at 4°C). Cells were washed with 1.2 ml of ice-cold phosphate-buffered saline, transferred to 1.5-ml microcentrifuge tubes, and collected by centrifugation (14,000 rpm for 20 min at 4°C). Last, cells were resuspended in 500 μl of LB medium, diluted, and plated onto LB agar (1.5%, w/v) plates.

Determination of minimal inhibitory concentration

Two antibiotics with distinct modes of action were used in tolerance and persistence experiments: the inhibitor of cell wall biosynthesis cefotaxime and the gyrase inhibitor ciprofloxacin. The minimal inhibitory concentrations (MICs) for ciprofloxacin and cefotaxime were determined in LB broth and MOPS minimal media (containing 10 mM MgCl2 and 2 mM K2HPO4) by serial dilution of antibiotic in 13-mm test tubes. Cultures were initiated in 1 ml of medium at a concentration of ≈1 × 107 CFU/ml and were allowed to grow at 37°C in an orbital shaker (330 rpm) for 24 hours. The MIC was determined as the lowest concentration of antibiotic that prevented bacterial growth. In LB medium, the MICs for cefotaxime and ciprofloxacin were 25 and 0.374 μg/ml, respectively. In MOPS medium, MICs for cefotaxime and ciprofloxacin were 12.5 and 0.187 μg/ml, respectively.

Mg2+ starvation and antibiotic exposure

After overnight growth in MOPS medium containing 10 mM Mg2+ and 2 mM K2HPO4, cells were collected by centrifugation (14,000 rpm for 1 min at room temperature), washed twice in MOPS medium containing no Mg2+ or phosphate, and inoculated (1:100) in MOPS medium containing 10 μM MgCl2 and 500 μM K2HPO4. Cultures were grown for 2 or 5 hours, aliquoted into separate tubes, and grown for an additional 2.5 hours in the presence or absence of antibiotics.

Acid-shock treatment and antibiotic exposure

Stationary-phase bacteria from overnight LB cultures were collected by centrifugation (14,000 rpm for 1 min at room temperature), resuspended in fresh acidified (pH 4.5) or neutral (pH 7.2; control) LB broth, and incubated for 30 min at 37°C. Cells were collected by centrifugation (14,000 rpm for 1 min at room temperature), resuspended in fresh LB broth (pH 7.2), and diluted 1:200 into fresh LB broth (pH 7.2) containing antibiotics. Cultures were allowed to grow for 24 hours before collection and estimation of CFU per milliliter.

Shx treatment and antibiotic exposure

Stationary-phase bacteria from overnight cultures were collected by centrifugation (14,000 rpm for 1 min at room temperature), resuspended in fresh LB broth containing Shx (100 μg/ml) or LB broth alone (control), and incubated for 30 min at 37°C. Shx (100 μg/ml) is sufficient to induce a large increase in (p)ppGpp synthesis (25), inhibit growth (Fig. 4A), and elicit the global changes in transcription known as the stringent response (27). After incubation, cells were collected by centrifugation (14,000 rpm for 1 min at room temperature), resuspended in fresh LB broth, and diluted 1:200 in fresh LB broth containing antibiotics. For logarithmically growing cells, Shx was added to cultures after 120 min of growth. After Shx addition, cultures were allowed to grow for an additional 15 min. Cultures were then subjected to a 24-hour exposure to cefotaxime or ciprofloxacin before estimation of CFU per milliliter.

Growth restriction of mannose catabolic mutant and antibiotic exposure

After overnight growth in minimal medium containing tryptophan, cells were diluted 1:200 into fresh medium and allowed to grow for 2 hours. Three milliliters of culture was collected on a 0.22 μM pore-size filter (Millipore), placed in a glass filter holder assembly with a funnel, using a vacuum pump. Cells on the filter were vacuum-washed with 10 ml of deionized water (dH2O), and filters were placed on a 50-ml conical tube containing 6 ml of fresh medium containing either glucose or mannose as the sole carbon source. Cells were allowed to grow for an additional 35 min to a concentration of 2 × 107 to 20 × 107 CFU/ml before the addition of antibiotics. Cultures were subjected to a 24-hour exposure to cefotaxime or ciprofloxacin before estimation of CFU per milliliter.

Growth restriction of tryptophan auxotroph and antibiotic exposure

After overnight growth in minimal medium containing glucose as the sole carbon source, cells were diluted 1:200 into fresh medium and allowed to grow for 2 hours. Three milliliters of culture was collected on a 0.22 μM pore-size filter (Millipore), placed in a glass filter holder assembly with a funnel, using a vacuum pump. Cells on the filter were vacuum-washed with 10 ml of dH2O, and filters were placed on a 50-ml conical tube containing 6 ml of fresh medium either containing or lacking tryptophan. Cells were allowed to grow for an additional 35 min to a concentration of 2 × 107 to 20 × 107 CFU/ml before the addition of antibiotics. Cultures were subjected to a 24-hour exposure to cefotaxime or ciprofloxacin before estimation of CFU per milliliter.

Growth restriction of relA spoT strains and antibiotic exposure

After overnight growth in LB medium, cells were diluted 1:200 into fresh LB medium and allowed to grow for 2 hours. Three milliliters of culture was quickly collected on a 0.22 μM pore-size filter (Millipore), placed in a glass filter holder assembly with a funnel, using a vacuum pump. Cells on the filter were vacuum-washed with 15 ml of dH2O, and filters were placed on a 50-ml conical tube containing 6 ml of fresh LB medium or defined minimal medium. Cells were allowed to grow for an additional 35 min to a concentration of 2 × 107 to 20 × 107 CFU/ml before the addition of antibiotics. Cultures were subjected to a 24-hour exposure to cefotaxime or ciprofloxacin before estimation of CFU per milliliter.

Expression of pATPase and antibiotic exposure

After overnight growth in minimal medium, cells were diluted 1:200 into fresh medium and allowed to grow for 2 hours. After the addition of IPTG (1 mM), cultures were allowed to grow for an additional 35 min before the addition of antibiotics. Cultures were subjected to a 24-hour exposure to cefotaxime or ciprofloxacin before estimation of CFU per milliliter.

Secondary antibiotic exposure

Stationary-phase bacteria from overnight LB cultures were collected by centrifugation (14,000 rpm for 1 min at room temperature), resuspended in fresh LB broth, and incubated at 37°C for 30 min. Cells were then diluted 1:200 into fresh LB broth in the presence or absence of antibiotics. After 24 hours of antibiotic exposure, cells were collected by centrifugation (14,000 rpm for 20 min at 4°C). Cells were washed with 1.2 ml of ice-cold phosphate-buffered saline, transferred to 1.5-ml microcentrifuge tubes, and collected by centrifugation (14,000 rpm for 20 min at 4°C). Last, cells were resuspended in 500 μl of LB medium, diluted, and plated onto LB agar (1.5%, w/v) plates. Cells were diluted 1:100 into fresh LB medium and allowed to grow overnight. Stationary-phase bacteria from overnight cultures were subjected to a second round of antibiotic exposure as described above.

Construction of plasmid pKD4-TetR

Phusion High-Fidelity DNA Polymerase (New England BioLabs) was used in reactions with primers W2990 and W2991 (table S2) and genomic DNA derived from Salmonella strain MS7953 (phoP::Tn10) (73). The polymerase chain reaction (PCR) product was ligated into pKD4 (77) digested with Bgl II and Afe I using NEBuilder HiFi DNA Assembly Cloning Kit (New England BioLabs). The plasmid was transformed into EC100D cells (Epicentre). The integrity of constructs was verified by DNA sequencing and the ability to render cells resistant to tetracycline.

Construction of plasmid pKD4-SpmR

Phusion High-Fidelity DNA Polymerase (New England BioLabs) was used in reactions with primers W2992 and W3083 (table S2) plasmid pSIM19 (72) as template. The PCR product was ligated into pKD4 (77) digested with Bgl II and Afe I using NEBuilder HiFi DNA Assembly Cloning Kit (New England BioLabs). The plasmid was transformed into EC100D cells (Epicentre). The integrity of constructs was verified by DNA sequencing and the ability to render cells resistant to spectinomycin.

Construction of plasmid pKD4-GenR

Phusion High-Fidelity DNA Polymerase (New England BioLabs) was used in reactions with primers W2994 and W3084 (table S2) and plasmid pSAM_Gent_Pbad (a gift from B. Kazmierczak) as template. The PCR product was ligated into pKD4 (77) digested with Bgl II and Afe I using NEBuilder HiFi DNA Assembly Cloning Kit (New England BioLabs). The plasmid was transformed into EC100D cells (Epicentre). The integrity of constructs was verified by DNA sequencing and the ability to render cells resistant to gentamycin.

Construction of plasmid pKD4-AprR

Phusion High-Fidelity DNA Polymerase (New England BioLabs) was used in reactions with primers W2996 and W3082 (table S2) and plasmid pMCS-7 (78) as template. The PCR product was ligated into pKD4 (75) digested with Bgl II and Afe I using NEBuilder HiFi DNA Assembly Cloning Kit (New England BioLabs). The plasmid was transformed into EC100D cells (Epicentre). The integrity of constructs was verified by DNA sequencing and the ability to render cells resistant to apramycin.

Recombineering protocol

Oligonucleotide sequences used in this study are presented in table S2. S. enterica strain 14028s (71) harboring plasmid pSIM6 (74) was grown overnight in LB medium supplemented with ampicillin (100 μg/ml) at 30°C and 250 rpm. Cells were diluted (1:100) in 30 ml of the same medium and grown for about 2.5 hours [optical density at 600 nm (OD600), ~0.35 to 0.4]. The culture flask was then grown in a water bath at 42°C and 250 rpm for 20 min (final OD600, ~0.6 to 0.75). Cells were immediately transferred to a 50-ml conical tube, collected by centrifugation (7000 rpm for 2.5 min at 4°C), and resuspended in 40 ml of ice-cold dH2O. Cells were collected again by centrifugation, and this washing procedure was repeated a second time. Last, cells were resuspended in 150 μl of ice-cold dH2O. Homologous recombination was obtained by electroporating 70 μl of cell suspension with 10 μl of purified linear DNA fragments.

Construction of strain relA spoT trpA strain

The ΔtrpA::kan mutation from a Salmonella NR-42835 E11 strain from the BEI NR mutant library (79) was transferred to Salmonella strain MP342 (relA::Tn10 spoT) (25) using P22-mediated transduction (75).

Construction of manA mutant

A ΔmanA::cat mutation was generated by recombineering using a purified linear DNA fragment generated from a PCR reaction using primers 12460 and 12461 and plasmid pDK3 as template (77). The location of each insertion was verified by PCR using primers 12462 and 12190, which flank the recombination points, and by the ability to use mannose as the sole carbon source. The ΔmanA::cat mutation was subsequently moved into S. enterica strain 14028s by P22-mediated transduction (75) to produce strain MP50.

Construction of strains containing deletions of single and multiple TA modules

Single ΔTA strains were generated by recombineering using purified linear DNA fragments generated from PCR reactions using primers listed in table S2 and plasmids pDK3, pKD4 (77), pKD4-TetR, pKD4-SpmR, pKD4-GenR, or pKD4-AprR as template. The location of each insertion was verified by PCR using primers listed in table S2. Because the expression of recombineering functions can be mutagenic (80), these mutations were individually transferred into S. enterica strain 14028s by P22-mediated transduction (75) to produce single-deletion strains. To generate strains containing multiple deletions, six deletions, each containing a distinct antibiotic resistant marker, were sequentially moved into S. enterica strain 14028s by P22-mediated transduction. Antibiotic markers were removed using plasmid pCP20, which expresses the flippase (FLP) recombinase (77) that acts on FLP recombinase target (FTR) sites flanking the genes specifying the proteins conferring antibiotic resistance. This procedure was repeated a second time to delete the 12 TA modules.

Verification of strain lacking 12 TA modules

To verify the genetic integrity of strain MP1422, harboring deletions of multiple TA modules, genomic DNA was extracted using the DNeasy kit (QIAGEN). Purified DNA was submitted for library construction and high-throughput sequencing at the Yale Center for Genome Analysis. Sequence fragments were assembled using Geneious software (Biomatters) and the S. enterica strain 14028s genome [National Center for Biotechnology Information (NCBI) reference sequence: NC_016856.1] as reference. We identified a total of 10 polymorphisms relative to the 14028s strain. Sequencing of PCR products flanking the polymorphic regions from the S. enterica 14028s parental strain established that nine of these nucleotide differences were actually errors generated in the original sequence of 14028s (81). Thus, strain sequence of MP1422 [NCBI Sequence Read Archive (SRA) accession ID no. PRJNA539960] contained a single mutation, consisting of a 125-bp deletion between the intergenic regions of rrlB and rrfB, corresponding to a deletion between base pair positions 4,369,814 and 4,369,939.

Estimation of intracellular ATP

Intracellular ATP was estimated as described (33). Briefly, 150 μl of cell cultures were aliquoted into PCR tubes, immediately heated at 80°C for 10 min, and subsequently placed on ice. At the same time, 1 ml of each of these cultures was transferred to a 1.5-ml tube and placed on ice. Using a multichannel pipette, 30 to 50 μl of heat-inactivated (80°C for 10 min) cells were transferred to 96-well black plates (Corning) containing 150 μl of BacTiter-Glo Microbial Cell Viability Assay Solution (Promega). Samples were mixed by pipetting, and their luminescence was measured in a Synergy H1 Reader (BioTek). Cells from the 1 ml of culture aliquots were collected by centrifugation (4°C at 14,000 rpm for 5 min), resuspended in 250 μl of 0.5% SDS, and lysed by boiling at 100°C for 10 min. The concentrations of proteins in the cell lysates were estimated using a Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Protein concentrations in these samples were used to calculate the protein amount present in the luminescence reactions. ATP abundances were calculated by normalizing the luminescence values of each sample by their protein content.

SUPPLEMENTARY MATERIALS

stke.sciencemag.org/cgi/content/full/12/592/eaax3938/DC1

Fig. S1. Determination of growth rates and of the emergence of antibiotic-resistant mutants in bacterial cultures.

Fig. S2. Effect of Shx treatment on the frequency of persisters.

Fig. S3. Effects of ATP depletion and inhibition of protein synthesis on antibiotic tolerance.

Fig. S4. Effect of growth inhibition on antibiotic tolerance in Salmonella.

Table S1. Microbial strains and plasmids used in this study.

Table S2. Oligonucleotides sequences used in this study.

References (82, 83)

REFERENCES AND NOTES

Acknowledgments: We thank J. Aronson for useful discussions and comments, B. Kazmierczak (Yale University) for providing plasmid pSAM_Gent_Pbad, C. Jacobs-Wagner (Yale University) for providing plasmid pMCS-7, X. Zhan (Penn State University) for discussions about statistical analyses, and S. Helaine (Imperial College London) for providing the sequence information of the TA modules described in her paper (14). Funding: This research was supported by NIH grant AI49561 to E.A.G. Author contributions: M.H.P. conducted all experiments. M.H.P. and E.A.G. analyzed data and wrote the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: The sequence of strain MP1422 was deposited into NCBI SRA database (accession ID number PRJNA539960). All other data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.
View Abstract

Stay Connected to Science Signaling

Navigate This Article