Search tips
Search criteria 


Logo of aacPermissionsJournals.ASM.orgJournalAAC ArticleJournal InfoAuthorsReviewers
Antimicrob Agents Chemother. 2010 March; 54(3): 1082–1093.
Published online 2010 January 19. doi:  10.1128/AAC.01218-09
PMCID: PMC2825962

Delineation of a Bacterial Starvation Stress Response Network Which Can Mediate Antibiotic Tolerance Development[down-pointing small open triangle]


This study aimed at elucidating the physiological basis of bacterial antibiotic tolerance. By use of a combined phenotypic and gene knockout approach, exogenous nutrient composition was identified as a crucial environmental factor which could mediate progressive development of tolerance with markedly varied drug specificity and sustainability. Deprivation of amino acids was a prerequisite for tolerance formation, conferring condition-specific phenotypes against inhibitors of cell wall synthesis and DNA replication (ampicillin and ofloxacin, respectively), according to the relative abundances of ammonium salts, phosphate, and nucleobases. Upon further depletion of glucose, this variable phase consistently evolved into a sustainable mode, along with enhanced capacity to withstand the effect of the protein synthesis inhibitor gentamicin. Nevertheless, all phenotypes produced during spontaneous nutrient depletion lacked the sustainable, multidrug-tolerant features exhibited by the stationary-phase population and were attributed to complex interaction between starvation-mediated metabolic and stress protection responses on the basis of the following reasons: (i) the nutrition-dependent tolerance characteristics observed suggested that adaptive biosynthetic mechanisms could suppress but not fully avert tolerance under transient starvation conditions; (ii) formation of specific phenotypes could be inhibited by suppressing protein synthesis prior to nutrient depletion; (iii) bacteriostatic drugs produced only weak tolerance in the absence of starvation signals; and (iv) the attenuation of the stringent and SOS responses, as well as the functionality of other putative tolerance determinants, including rpoS, hipA, glpD, and phoU, could alter the induction requirement and drug specificity of the resultant phenotypes. These data reveal the common physiological grounds characteristic of starvation responses and the onset of antibiotic tolerance in bacteria.

Bacteria constantly harbor a subpopulation refractory to killing by bactericidal drugs (2, 3, 16). The size of this tolerant subpopulation increases in adverse growth conditions such as depletion of nutrients (6, 11, 32). The physiological basis of this phenomenon, termed antibiotic tolerance, remains poorly understood. Using a systematic approach to screen knockout mutants of Escherichia coli, Hansen et al. recently identified a number of genetic loci involved in the upstream control of a vast tolerance response network (8), suggesting that antibiotic tolerance was not simply due to passive shutdown of drug target activities as a result of retarded bacterial growth in an unfavorable environment. In Hansen's study, most of the genes for which deletion had a detectable effect on antibiotic tolerance were global regulators of nutrient metabolism, stress sensing, and protection as well as essential cellular processes such as transcription. We envision, however, that genetic determinants responsible for mediating response to specific stress may not be readily identified by deletion studies if the test population is subjected to exposure to multiple induction factors, since bacteria may sense alternative stress signals and develop tolerance via redundant pathways. Hence, any attempt to screen for tolerance mutants under uncontrolled conditions, such as those involving stationary-phase culture in which a number of growth parameters become unfavorable, may allow identification of only the rather upstream regulators that control a number of independent tolerance mechanisms or some common downstream protection mechanisms activated by different induction pathways. Indeed, this idea is consistent with the observation that only a small number of highly redundant and upstream regulatory genes were identified through screening of stationary-phase cultures of gene knockout and other mutant libraries (8, 17).

To overcome the limitation of conventional tolerance assays using a stationary-phase population and test our concept on the intermingling network of stress sensing and protection, we strived to dissect and depict the relative roles of the major branches of such network in tolerance development through analysis of the differential effects of progressive nutrient exhaustion, considering that starvation represented a major stress that bacteria encountered during stationary phase. In this work, a tolerance induction assay was designed in such a way that the test conditions were tightly controlled to evaluate the induction effect of a panel of nutrient compositions. Through a series of conditioned experiments, we were able to identify the key nutritional factors and determine their relative roles in tolerance induction and demonstrated that starvation of each nutrient category rapidly elicited matching metabolic and physiological changes which might in turn be reflected by phenotypic tolerance to antibiotics.

On the basis of the bacterial tolerance induction criteria gathered in the phenotypic experiments, we further attempted to perform a preliminary functional mapping of several major bacterial stress defense mechanisms in mediating formation of the nutrient-sensitive antibiotic tolerance phenotypes. These include the stringent (19, 26, 27), SOS (10, 21), and RpoS (13) mediated responses, which may be elicited when bacteria encounter nutrient starvation, physiological stress, or cellular damages as well as several other key genetic loci, such as hipA (12, 22), phoU (17), and glpD (30), which have been implicated in the formation of bacterial persister subpopulations that exhibit markedly reduced death rates during antimicrobial treatment. Currently, the relative effects of the depletion of different nutrient classes on the induction of these stress responses remain unclear. For example, it has previously been shown that starvation of glucose, amino acids, and ammonium salts could each result in differential expression or stability of the RpoS factor (9, 25), which is known to regulate the expression of a range of protection mechanisms, such as thermotolerance (18) and oxidative stress defenses (20); however, it is not clear whether different types of nutritional stress can activate such RpoS-controlled mechanisms, as well as the stringent and SOS responses, on an equal basis. Apart from a lack of information on the induction criteria of different tolerance responses, the pattern of interaction between these systems remains poorly defined. For example, the stringent response is known to enhance the activity of the RpoS sigma factor (9, 19), yet information regarding whether stringent response can also elicit the SOS response is not available. On the other hand, a number of recent studies suggested that bactericidal antibiotics could trigger the production of deleterious hydroxyl free radicals by destabilizing membrane integrity and eliciting the onset of Fenton reactions (14, 15). In view of the wide range of stress responses which can regulate protein translation efficiency as well as production of antioxidant defense actions, we hypothesize that formation of antibiotic tolerance during nutrient depletion is at least partly due to the fact that reactive oxygen species produced during the process of drug-induced physiological changes are neutralized by the protection effects of stress responses elicited under nutrient starvation. Upon testing specific deletion mutants against a range of conditions which were known to produce antibiotic tolerance, we were able to assess the relative role of each stress response component and the redundant nature of the key systems involved in tolerance formation. Our data showed that deletion of a specific gene often resulted in condition-dependent variations in tolerance phenotypes. Such data confirm the presence of a complex, redundant, and hitherto undefined nutrient-sensing and tolerance response network which can regulate bacterial physiological changes and the concomitant formation of drug-specific tolerance phenotypes in accordance with the nutritional status in the environment.


Bacterial strains and culture media.

The Escherichia coli K-12 strains and derivatives thereof listed in Table Table11 were used in all experiments. These strains included a ΔrelA ΔspoT double knockout mutant and the wild-type strain from which this mutant was constructed (30) as well as deletion mutants of specific genes derived from the Keio collection of E. coli knockout strains (1). All cultures utilized rich defined medium (RDM) (Teknova, Holister, CA) unless stated otherwise. This medium, developed by Neidhardt et al. (23) and later modified by F. R. Blatter and coworkers (, comprised six major components: (i) MOPS (morpholinepropanesulfonic acid) base, (ii) glucose, (iii) ammonium salts, (iv) inorganic phosphate, (v) nucleobases, and (vi) an amino acid mixture with trace vitamins. The composition of this growth medium could be precisely manipulated according to the manufacturer's instructions ( On the basis of the RDM recipe, a panel of defined media consisting of different compositions as denoted in the present study was constructed for the tolerance induction assays. Standard LB agar (Difco, Leeuwarden, The Netherlands) was used for assessing the proportion of the bacterial population that survived in the assay.

E. coli strains used in this study

Drugs and chemicals.

All antibiotics were purchased from Sigma (St. Louis, MO). Sodium azide and potassium cyanide were obtained from Merck and BDH Chemicals, respectively.

Growth and assay media.

Fresh bacterial colonies were inoculated into RDM and grown overnight at 37°C under constant agitation (200 rpm). The overnight culture was diluted 10,000-fold in RDM and cultivated for 4 h until the exponential growth phase was reached (~107 CFU/ml). Aliquots of this exponential-phase culture were washed and resuspended in the assay medium, followed by the tolerance induction assays.

(i) For analysis of the effect of different nutrient compositions on induction and maintenance of tolerance to specific bactericidal antibiotics, various recipes, ranging from complete RDM to one which was deprived of all test nutrients (MOPS), were prepared (Fig. (Fig.1).1). Specific growth rates (μ) prior to antibiotic treatments were determined by measuring the changes in cell density in the untreated controls, using the formula μ = [(log10 N3 − log10 N0) 2.303]/(t3t0), where N0 and N3 represent the initial and final cell densities, respectively, within the first 3 h of incubation upon the switch to the test medium (t3t0).

FIG. 1.
Effects of differential nutrient compositions on the induction and sustainability of antibiotic tolerance. Exponentially growing E. coli BW25113 bacteria were washed, resuspended, and incubated at 37°C for 2 h in a panel of defined media composed ...

(ii) For comparative assessment of the tolerance induction effect of bacteriostatic conditions in a nutrient-rich environment, complete RDM was supplemented with individual bacteriostatic agents (tetracycline [2 μg/ml], rifampin [16 μg/ml], 5 mM sodium azide, and 5 mM potassium cyanide) and used to resuspend the test population prior to the assays.

(iii) To determine whether the multidrug tolerance phenotypes observable in stationary-phase bacterial culture were attributable to diminished nutrient availability, filter-sterilized cell-free supernatant from RDM-grown overnight cultures was prepared and tested for its tolerance induction effect in tolerance induction assays.

(iv) For genetic mapping of tolerance induction pathways, a panel of selected deletion mutants was tested against specific recipes of nutrient compositions to determine whether deletion of specific genes would affect the ability by which bacteria produced specific tolerance phenotypes under conditions which were known to induce tolerance formation. The gene knockout mutants and the corresponding tolerance induction conditions tested are shown in Fig. Fig.55 and and6.6. Potential development of suppressor mutants from the guanosine 3′,5′-bispyrophosphate (ppGpp)-null mutant was assessed by plating out all test populations of this strain on M9-salt agar (29) supplemented with 0.4% (wt/vol) glucose.

FIG. 5.
Assessment of the relative role of stringent response in nutrient depletion-mediated drug tolerance. Exponentially growing CF1943 (WT; ppGpp+) and CF1946 (ΔrelA ΔspoT; ppGpp0) populations were preincubated in RDM, RDM without amino ...
FIG. 6.
Evaluation of effects of stress response regulators and putative antibiotic tolerance genes on nutrient starvation-induced drug tolerance. BW25113 (WT) and its isogenic gene deletion mutants (ΔhipA, ΔphoU, ΔglpD, ΔrecA ...

Tolerance induction assay.

Bacterial populations were preincubated with the test medium at 37°C for 2 h and subjected to antibiotic challenge using ampicillin (100 μg/ml), ofloxacin (0.75 μg/ml), and gentamicin (6.25 μg/ml) at a working concentration of 25 times the respective MIC of the test strain (BW25113). Cells were incubated for 48 h at 37°C under constant shaking (200 rpm). Standard serial dilution and plating on LB agar was performed at 0, 3, and 48 h. The plates were incubated at 37°C, and colony counts were recorded for two consecutive days to account for possible discrepancies due to residual drug effects. Changes in the sizes of surviving cell populations over time were recorded and compared to the level observed prior to antibiotic challenge for assessment of short-term (3-h) and long-term (48-h) drug tolerance. At least three independent experiments were performed in each assay to assess the reproducibility of the induction effect. Surviving cells were routinely tested for antibiotic susceptibility by using the agar dilution method to determine the MICs of isolates randomly selected from the tolerant population.

Tolerance inhibition assay.

To test the effect of transcription and protein synthesis inhibitors on starvation-induced tolerance development, tetracycline (4 μg/ml) was added to an RDM-grown log-phase culture and incubated for 2 h at 37°C, and the treated population was subjected to centrifugation, washed once, resuspended in MOPS base which also contained tetracycline at 4 μg/ml, and incubated at 37°C for 2 h. The cell suspension was then subjected to antibiotic challenge as described for the tolerance induction assay.


This study involved the development of a phenotypic assay to probe the induction criteria and physiological mechanisms governing bacterial antibiotic tolerance formation. In the assay, a panel of specially formulated media with known nutrient compositions was used to analyze the effects of limitation of different combinations of five key nutrients (amino acids, glucose, ammonium salts, phosphate, and nucleobases) in eliciting tolerance development in an actively growing bacterial population which was fully sensitive to the bactericidal effects of the three test drugs (ampicillin, ofloxacin, and gentamicin) prior to the assay. Figure Figure11 shows the relative survival rates of the test population recorded under each assay condition before and after drug treatment for 3 and 48 h and thereby the relative capacity of a bacterial population to develop and maintain tolerance to a specific drug under each assay condition. The ability of the test population to maintain viability or propagate in the assay medium prior to or without drug treatment, illustrated as specific growth rate and alteration in population size before and after treatment, respectively, was also shown to depict the relationship between the growth capacity and the corresponding drug susceptibility profiles of the test organisms. In all assays, the bacterial population size was standardized to approximately 107 cells per ml prior to drug treatment.

Four major classes of drug tolerance responses.

The diverse spectrum of nutrient compositions tested was found to produce complex and condition-dependent antibiotic susceptibility phenotypes which could be divided into four major categories according to their drug specificity and sustainability (Fig. (Fig.1).1). First, a total of five conditions, including the full RDM recipe, failed to support survival against the bactericidal effects of the three test drugs (Fig. (Fig.1,1, category A). These conditions mainly involved depletion of a single nutrient except amino acids but also included one in which three nutrient classes (glucose, ammonium salts, and nucleobases) were simultaneously depleted from the growth medium. Second, two conditions could specifically produce transient tolerance to ofloxacin but not to the other two drugs, and in both cases, phosphate was depleted from the medium (Fig. (Fig.1,1, category B). Third, a total of 11 recipes produced mixed tolerance responses to ampicillin and ofloxacin (Fig. (Fig.1,1, category C). The recipes that produced category C responses ranged from a lack of amino acids alone (Fig. 1, C1iii) to simultaneous depletion of up to four nutrients (Fig. 1, C3). Apart from specificity and sustainability, the strength of drug tolerance, recorded as the size of the tolerant population produced and maintained during the assay, also differed extensively between different nutrient compositions. Finally, eight conditions, all involving media which lacked glucose and amino acids, were found to elicit tolerance to all the three test drugs. The phenotypes that belonged to this category typically included sustainable tolerance to both ampicillin and ofloxacin and transient tolerance to gentamicin (Fig. (Fig.1,1, category D).

Inductive and regulatory effects of amino acids and nucleobases.

Detailed analysis of the relationship between nutrient compositions and the corresponding drug susceptibility phenotypes revealed several important tolerance induction characteristics which could be made apparent only by comparing the induction effects of differential nutrient compositions. First, we noticed that depletion of amino acids from the growth medium was a prerequisite for the onset of the vast majority of tolerance phenotypes, with the exception of several phosphate-limiting recipes which could induce short-term tolerance to ofloxacin or to ofloxacin and ampicillin even in the presence of amino acids (Fig. (Fig.1,1, Bi and -ii and C1ii, -vi, and -vii). Hence, depletion of amino acids produced a much more diverse and far-reaching effect than the lack of any other compounds in eliciting antibiotic tolerance, although the sustainability and drug specificity of the phenotypes produced were highly dependent on the nature of the nutrients which were concomitantly depleted.

The second important feature observable during nutrient depletion-mediated tolerance induction is that several induction processes appeared to be highly sensitive to the relative abundances of nucleobases. In the case of ampicillin but not ofloxacin, simultaneous depletion of amino acids and nucleobases consistently produced a much weaker induction effect than a medium which was deprived of amino acids only (Fig. (Fig.1,1, compare C1i and C1iii). Although this finding appeared to indicate that the presence of nucleobases could enhance the development of ampicillin tolerance mediated by depletion of amino acids, a contradictory role for these molecules in tolerance regulation was observable in assays with other nutrient compositions. Notably, a prolonged ampicillin tolerance phenotype conferred by a lack of amino acids and ammonium salts could not be supported if nucleobases were present (Fig. (Fig.1,1, compare C1iv to C2ii), suggesting a suppressive effect of nucleobases on specific tolerance induction pathways. In contrast, prolonged ampicillin tolerance triggered by glucose limitation was not affected, as long as amino acids were absent (Fig. (Fig.1,1, Di and -iv). Apart from ampicillin tolerance, the specific inhibitory effects of nucleobases were also observable in ofloxacin tolerance induction; in this case, simultaneous depletion of amino acids and phosphate produced sustainable ofloxacin tolerance only in the absence of nucleobases (Fig. (Fig.1,1, compare C1v to C4i).

Induction of prolonged tolerance by depletion of multiple nutrients.

The third important feature of nutrient-sensitive tolerance regulation is that bacteria were able to exhibit prolonged tolerance to antibiotics under specific conditions in which key nutrients were further removed from an amino acid limitation background. This finding was mainly based on our observation that highly sustainable tolerance of ampicillin and/or ofloxacin could be produced only by simultaneous depletion of amino acids and glucose (Fig. (Fig.1,1, category D), or amino acids and ammonium salts, plus either phosphate or nucleobases (Fig. 1, C2, -3, and -4). Absence of amino acids and glucose was also a prerequisite for induction of short-term tolerance to gentamicin, a phenotype which consistently emerged along with sustainable tolerance to ampicillin and ofloxacin (Fig. (Fig.1,1, category D).

Our data indicated that prolonged antibiotic tolerance was mediated by complex adaptive and stress responses inducible when specific nutrients other than amino acids became limiting. Such responses, apart from being subjected to negative regulation by amino acids, appeared to produce induction effects synergistic with those of amino acid depletion. First, we noted that simultaneous limitation of glucose, ammonium salts, and phosphate did not support even short-term ampicillin tolerance when amino acids were present (Fig. (Fig.1,1, Bii and C1ii), yet the presence of these three nutrients could reverse most of the MOPS-induced phenotypes even in the absence of amino acids (Fig. (Fig.1,1, compare C1i with category D conditions). We therefore hypothesized that a lack of one or more of these compounds could induce a tolerance response which was negatively regulated by amino acids and distinguishable from the one triggered by amino acid depletion alone, which was nucleobase dependent (Fig. (Fig.1,1, compare C1i to C1iii). We noticed that, although each of these components could not fully reverse MOPS-induced ampicillin tolerance (Fig. 1, C3i and Dvi and -vii), further depletion of ammonium salts or glucose (Fig. 1, C2ii and Div), but not phosphate (Fig. 1, C4i), from a medium background which failed to induce the putative amino acid depletion-mediated ampicillin tolerance response (lack of amino acids and nucleobases simultaneously) (Fig. 1, C1i) could induce sustainable ampicillin tolerance. Corroborating the idea that prolonged tolerance was regulated by derepressible mechanisms was the observation that ofloxacin tolerance was inducible by two independent pathways. First, a lack of either amino acids or phosphate induced primary tolerance to ofloxacin after 3 h of treatment (Fig. (Fig.1,1, Bi and C1iii), the phosphate depletion-mediated response being insensitive to suppression by amino acids. Second, a stronger and more persistent phenotype resulted from a combined depletion of amino acids, phosphate and nucleobases, or amino acids and carbon sources (Fig. 1, C3i, C4i, and all category D conditions). It should also be noted that, unlike amino acids and nucleobases, glucose, phosphate, and ammonium could not individually confer negative regulation on antibiotic tolerance formation. In this study, we confirmed that long-term tolerance to ampicillin and ofloxacin was not due to selection and proliferation of resistant mutants, since all tolerant populations were found to contain isolates which remained susceptible to the test drugs in MIC determination.

Phenotypic difference between starvation- and stationary-phase-induced tolerances.

To assess the relative contribution of nutrient deprivation in producing multidrug tolerance in stationary-phase populations, we performed a spent-medium assay to determine if the cell-free supernatant obtained from stationary-phase culture, which had been grown overnight in RDM, could induce formation of a phenotype which matched those produced by the tested recipes. Our results showed that this spent medium supported tolerance development in a manner similar to that of MOPS base (Fig. (Fig.2).2). We showed that this induction effect was due to nutrient limitation but not accumulation of toxic metabolites produced during prolonged culture, as it could be abolished by replenishment of nutrients to the spent medium (results not shown). These findings confirmed that nutrients in an overnight culture have been depleted to a level that would support development of tolerance phenotypes comparable but not entirely similar to those observed in stationary-phase populations, which exhibited prolonged gentamicin tolerance.

FIG. 2.
Analysis of the tolerance induction effect of cell-free supernatant of multidrug-tolerant stationary-phase culture on a log-phase, drug-sensitive population. Exponentially growing cells were preincubated in filter-sterilized supernatant of the stationary-phase ...

Growth rate and antibiotic tolerance.

In addition to the relationship between nutrition recipes and tolerance formation, we further analyzed the correlation between growth rate and phenotypic tolerance under the test conditions. We found that a high specific growth rate (an average of 1 h−1 or above between 0 and 3 h) (Fig. (Fig.1,1, category A) was not compatible with tolerance development, whereas conditions that suppressed bacterial growth (with specific growth rates of 0 h−1 or below) generally conferred sustainable tolerance to both ampicillin and ofloxacin as well as short-term tolerance to gentamicin (Fig. (Fig.1,1, Di to -iv and -viii). However, the relationship between growth rate and drug susceptibility became much less apparent under conditions in which bacteria grew at a relatively slow pace, when induction of antibiotic tolerance in a growth rate-independent manner was often observed. For example, a medium which was depleted of amino acids and ammonium salts induced significant short-term tolerance to ampicillin (Fig. 1, C1iv), yet another medium, which lacked glucose, ammonium salts, and nucleobases (Fig. (Fig.1,1, Av), did not confer detectable ampicillin tolerance, despite the fact that the test populations exhibited similar growth rates in these media (exhibiting specific growth rates of 0.69 h−1 and 0.62 h−1, respectively). Likewise, a number of other conditions that supported identical growth rates (ranging from 0.32 to 0.36 h−1) were found to induce a spectrum of tolerance phenotypes with significant variation in drug specificity and degree of sustainability (Fig. 1, C1v, C4i, and Dv). In addition, we do not observe significant correlation between the capacity to exhibit prolonged growth or regrowth in the test medium (exhibited by increase in population size between 3 and 48 h) (Fig. (Fig.1),1), which is a hallmark of adaptive biosynthesis, and development of sustainable tolerance.

Tolerance induction potential of bacteriostatic agents.

To determine whether growth inhibition alone could contribute to tolerance formation, we examined the effects of bacteriostatic agents in eliciting tolerance in an actively growing population. Potassium cyanide and sodium azide, both inhibitors of aerobic respiration, were the only agents which continued to exert bacteriostatic effects throughout 48 h of treatment (Fig. (Fig.3).3). In general, the tolerance profiles elicited by bacteriostatic agents were highly dissimilar to those conferred by MOPS or limitation of major nutrients, which consistently supported prolonged tolerance to ampicillin and ofloxacin and short-term tolerance to gentamicin (Fig. (Fig.1,1, category D). Potassium cyanide was the only agent which produced prolonged tolerance to both ampicillin and ofloxacin and elicited short-term gentamicin tolerance in a situation where starvation signals and detectable bacterial growth were both absent (Fig. (Fig.3),3), although the tolerant population was smaller than that observed under nutrient depletion conditions. Interestingly, the bacteriostatic effects triggered by sodium azide were not associated with any detectable tolerance phenotypes (Fig. (Fig.33).

FIG. 3.
Induction of drug tolerance by bacteriostatic agents. Exponentially growing cells were washed and preincubated in full RDM containing different classes of bacteriostatic agents (Tet, tetracycline [2 μg/ml]; Rif, rifampin [16 μg/ml]; Azide, ...

Effect of translation inhibitor on MOPS-induced tolerance.

We further tested the possibility that active cellular mechanisms might be involved in tolerance formation during nutrient starvation by assessing the effects of tetracycline, a protein synthesis inhibitor, on the potential of MOPS base to induce and maintain tolerance upon corruption of the bacterial protein translation machinery. To avoid potential development of tolerance to tetracycline itself, which would otherwise prevent us from accurately assessing its effects on tolerance to other drugs, we added tetracycline to RDM-grown log-phase culture and continued incubation at 37°C for 2 h prior to switching the culture medium to a MOPS base which also contained the agent, followed by treatment with the three test drugs and assessment of population survival. As shown in Fig. Fig.4,4, the presence of tetracycline was found to confer a significant but incomplete inhibitory effect on MOPS-induced long-term tolerance to ampicillin as well as short-term tolerance to gentamicin. However, formation of ofloxacin tolerance was not affected by prior inhibition of protein synthesis. It should be noted that this inhibitory effect was not observed when tetracycline was added only after the switch to MOPS instead of 2 h prior to the switch (results not shown), indicating that bacteria might also develop tolerance to tetracycline during starvation. In view of this finding, we speculated that the inhibitory effect of this protein synthesis inhibitor might have been suppressed when the cells became tolerant; hence, the detrimental effect of corrupting protein synthesis on MOPS-induced tolerance development might have been underestimated.

FIG. 4.
Effects of protein synthesis inhibitors on MOPS-induced antibiotic tolerance. Tetracycline (4 μg ml−1) was added to an RDM-grown log-phase culture, incubated at 37°C for 2 h, and then reconstituted in MOPS basal medium containing ...

Identification of tolerance response determinants.

In a preliminary attempt to identify the genetic determinants responsible for induction of the specific tolerance phenotypes that we observed in the tolerance induction assays, we selected a panel of putative tolerance genes which had previously been implicated in tolerance formation and tested whether deletion of these genetic components would undermine the ability of bacteria to develop antibiotic tolerance under conditions known to induce formation of differential phenotypes. In particular, we included a ΔrelA ΔspoT double knockout mutant, which was defective in producing the stringent response, to determine if the tolerance induction potential of amino acid depletion and the regulatory effects of nucleobases on tolerance formation were mediated through this major starvation response pathway. As shown in Fig. Fig.55 and and6,6, all the six gene knockout mutants tested displayed tolerance phenotypes which differed from each other as well as from that of the wild-type strain in terms of induction criteria, drug specificity, and the size of the resultant tolerant population that emerged under specific test conditions. In particular, the ΔrecA mutant failed to develop ofloxacin tolerance under all conditions, including stationary phase, yet production of ampicillin and gentamicin tolerance was generally not affected, except under specific conditions, such as the use of overnight culture or a medium in which amino acids alone were depleted, where development of sustainable gentamicin tolerance and transient ampicillin tolerance, respectively, was affected (Fig. (Fig.66).

The ΔrelA ΔspoT double knockout mutant, which was defective in producing the stringent response, exhibited pleiotropic effects on tolerance formation. First, this mutant displayed a reduced capacity to produce sustainable ofloxacin tolerance under two conditions (involving MOPS and RDM minus amino acids and nucleobases) which were known to induce prolonged tolerance to this drug in its wild-type counterpart (CF1943) (Fig. (Fig.5).5). Second, the phenomenon of nucleobase-dependent induction of ampicillin tolerance was not observable in this mutant background, as the proportions of the test population that became tolerant to ampicillin were similar regardless of availability of nucleobases in the growth medium. Third, the level of prolonged ampicillin tolerance was reduced in a medium in which both amino acids and glucose were simultaneously depleted. However, such effect was not observable in MOPS base. It should also be noted that, in the absence of antibiotics, the survival rate of this mutant was significantly lower than that of the wild-type strains when nucleobases were supplemented in an amino acid-depleted medium (Fig. (Fig.5).5). However, this mutant exhibited prominent growth when nucleobases were absent in the amino acid depletion medium. On the other hand, the ΔrelA ΔspoT mutant was not capable of surviving for 48 h in a medium depleted of amino acids and ammonium salts; hence, we were not able to examine if the suppressive effect of nucleobases on the ampicillin tolerance phenotypes inducible by a lack of amino acids and ammonium salts was observable in the absence of stringent response. Nevertheless, the ΔrelA ΔspoT mutant was capable of expressing short-term ampicillin tolerance in this medium (results not shown). We confirmed that the test populations of this strain did not harbor RNA polymerase suppressor mutants under all treatment conditions, as no viable colonies were recoverable in minimal glucose agar; hence, all phenotypic characteristics of this strain could be attributed to the lack of ppGpp production.

Apart from observations in the ΔrecA and ΔrelA ΔspoT mutants, deletion of the rpoS gene was also found to consistently result in a slight defect in formation of ampicillin tolerance (Fig. (Fig.6).6). However, development of ofloxacin and gentamicin tolerance was not affected in the ΔrpoS mutant. Finally, deletion of several other genes which were suggested to be key regulators of antibiotic tolerance, including phoU, glpD, and hipA, was also found to produce detectable effects on antibiotic tolerance induction in a condition- and drug-specific manner (Fig. (Fig.66).


Overview of tolerance regulatory network.

This study established a functional link between bacterial starvation responses and development of antibiotic tolerance. Our data suggested the existence of complex cellular mechanisms that regulated bacterial physiology in accordance with the nutritional status in the environment and effectively exerted sensitive control over the strength and specificity of antibiotic tolerance induction. The major regulatory strategies are summarized as follows. First, deprivation of amino acids, alone or combined with limitation of other compounds, can trigger both primary and synergistic tolerance responses, producing an optimal ratio of actively growing and dormant populations which best enhances population survival under the prevailing environmental condition; second, nucleobases confer either positive or negative regulatory effects on the development of the basal phenotypes, depending on the triggering factors, thus fine-tuning the core responses; third, several nutrient classes, including carbon and nitrogen sources, either individually or concomitantly with each other, can selectively suppress tolerance triggered by depletion of other nutrients, thereby imposing sophisticated negative control on tolerance development. Amino acids are nevertheless the key regulator, as their presence inhibits formation of most phenotypes. The differential, interdependent, counterinteractive, and summative features of tolerance induction indicate that bacteria rely heavily on nutrient-sensing pathways to activate matching metabolic adaptation and stress protection mechanisms, which in turn produce differential antibiotic tolerance phenotypes.

Effect of adaptive metabolism.

Our phenotypic data reflect to some extent the result of interplay between the multiple physiological effects exerted by changing levels and composition of endogenous intermediate metabolites during the dynamic process of bacterial adaptation to transient starvation of different nutrients. Such interactive effects on growth rate, drug target metabolism, and production of signaling molecules that regulate stress responses render it impossible to determine the exact role of starvation-induced adaptive mechanisms in tolerance formation. Nevertheless, analysis of the differential phenotypes concerning the availability of nucleobases suggests that they may play a role in enhancing the mRNA synthetic ability and thereby the active production of stress protection proteins, which have been compromised as a result of repression and slow induction of purine and pyrimidine nucleotide biosynthesis under various test conditions (24). In fact, the intrinsic ability of the test organisms to derepress biosynthetic pathways that replenish the depleted ingredients often fails to compensate for the induction effects elicited by the lack of specific nutrients, especially under amino acid depletion conditions. It is envisaged that the regulatory effects of adaptive metabolism in tolerance development diminish upon prolonged starvation, giving way to active stress defense which then plays a key role in conferring sustainable tolerance.

Relative role of stringent response in tolerance development.

Being one of the key cellular systems that mediate stress protection in bacteria and one that is activated in an amino acid-limiting environment through induced synthesis of the alarmone guanosine 3′,5′-bispyrophosphate (ppGpp) (26), the stringent response may best account for the nutrient-dependent drug susceptibility phenotypes observable in this study. Given its inhibitory effects on the synthesis of macromolecules such as DNA, phospholipids, and cell wall peptidoglycan (19, 31), initiation of the stringent response in the absence of amino acids is therefore highly consistent with the onset of ampicillin and ofloxacin tolerance under such conditions. This idea may further explain the observation that nucleobases were required for elicitation of amino acids starvation-induced ampicillin tolerance (Fig. 1, C1i and -iii). The results of our gene knockout experiments, however, suggest an alternative story, because deletion of the genetic determinants responsible for producing the stringent response had mixed effects on the differential tolerance phenotypes observed in the wild-type strains (Fig. (Fig.5).5). First, ofloxacin tolerance was not sustainable in this mutant, suggesting that induction of the stringent response was nevertheless required for formation of tolerance to fluoroquinolones. The fact that tolerance to ampicillin and gentamicin could still be produced in the absence of stringent response, however, suggested that its effect was drug target specific. Second, the nucleobase dependency in formation of transient ampicillin tolerance when amino acids were depleted from the growth medium was relieved in the ppGpp-null background, suggesting either that stringent response did not play an important role in formation of ampicillin tolerance under amino acid-limiting conditions or that failure to produce stringent response resulted in derepression of an unidentified mechanism which could then mediate formation of primary or transient tolerance to ampicillin in the absence of both amino acids and nucleobases. These results complement those reported by Hansen et al. (8), which showed that deletion of dksA, a determinant of a ppGpp-dependent modulator of RNA polymerase, resulted in a significant defect in tolerance formation.

The ppGpp-null mutant exhibited enhanced growth fitness in an amino acid-limiting background if nucleobases were further depleted (Fig. (Fig.5).5). It has previously been suggested that a lack of stringent response would undermine the ability of a bacterial population to switch to a defense mode at the expense of continued synthesis of macromolecules during nutrient starvation, thereby enabling the ppGpp-null mutant to grow at a normal or even enhanced rate under transient nutrient-limiting conditions (7, 26). Evidence gathered in this study shows that nucleobases have a significant negative impact on bacterial growth fitness under specific nutritional stress; whether such effect is mediated through complex, stringent-response-independent growth capacity/stress response regulatory pathways entails further investigation. Nevertheless, the intriguing findings regarding the multiple effects of nucleobases on growth rate and tolerance development under different genetic and nutritional backgrounds complement our limited knowledge on the role of ppGpp in eliciting antibiotic tolerance in response to nutrient starvation (27, 28).

Molecular basis of prolonged antibiotic tolerance.

In this work, we obtained evidence that depletion of nutrients other than amino acids was involved in expression of secondary responses that triggered onset of prolonged tolerance. Importantly, the tolerance induction effects of glucose-limiting conditions appeared to be suppressed if amino acids were available; once this suppression was lifted, the respective stress responses triggered by depletion of these nutrients could be superimposed on each other, producing a synergistic effect and thereby prolonged tolerance. The stringent response was found to play a role in regulating such secondary mechanisms in a drug-specific manner, as only formation of prolonged tolerance to ofloxacin, but not ampicillin, was affected when the ability to launch stringent response was abolished (Fig. (Fig.5).5). The role of stringent response in development of sustainable antibiotic tolerance was also analyzed in the context of the functional relationship of stringent response with the stationary-phase sigma factor RpoS, which is known to be activated during limitation of carbon and nitrogen sources, as well as phosphate, and to play roles in regulating cellular metabolism during nutrient starvation (25) and in upregulating the transcription of genes involved in stress protection (33). First, carbon starvation was shown to enhance ppGpp production (26), which might in turn activate RpoS-dependent responses (19). Second, our data clearly demonstrated that distinct but overlapping physiological changes were elicited when amino acids and glucose were individually or concomitantly depleted, yet deletions of RpoS and the stringent-response determinant exerted differential effects on the development of prolonged tolerance. These findings indicate that each of these two major stress response systems plays a partial and nonidentical role in formation and maintenance of antibiotic tolerance.

Tolerance induction specificity of SOS response and other key determinants.

The finding that RecA function was essential for ofloxacin-specific tolerance was consistent with the role of RecA in mediating DNA repair and therefore alleviation of the damaging effects of fluoroquinolones (5) (Fig. (Fig.6).6). Further investigation is required to determine whether RecA-mediated SOS response plays a role in tolerance formation by promoting DNA repair only or activation of additional protection mechanisms. Currently, there is little evidence which illustrates that SOS response can be spontaneously turned on upon nutrient depletion. Our data suggest a need to examine the possible relationship between the stringent and SOS responses in tolerance formation, as both the ΔrelA ΔspoT and the ΔrecA mutants were defective in exhibiting prolonged tolerance to ofloxacin but not ampicillin (Fig. (Fig.55 and and66).

Apart from the major stress responses, the finding that deletion of several putative tolerance determinants which have been postulated to execute a diversity of physiological roles (hipA, glpD, and phoU) (Fig. (Fig.6)6) could in each case affect nutrient-sensitive induction of tolerance in a condition and drug specific manner supports our original hypothesis that the true nature of cellular responses that produced the sustainable phenotypes might involve multiple stress-defense mechanisms regulated by some hitherto un-identified and overlapping signaling pathways. The multiplicity of genetic and nutritional factors observable in bacterial antibiotic tolerance induction confirmed that the underlying regulatory processes were nutrient sensitive, redundant, and additive in nature. The phenotypic difference between starvation-induced tolerance and those observable in an overnight-grown population also suggests that nutrient depletion is not the only tolerance induction factor. Whether the genetic determinants which were found to alter the tolerance strength in a stationary-phase population (8) play a common role in both starvation-specific and nutrient-independent tolerance induction remains to be elucidated.

Concluding remarks.

Findings in this work support a tentative conclusion that short-term antibiotic tolerance may develop under amino acid-limiting conditions, which cause reduction in growth rate and activation of stress responses such as ppGpp-related inhibition of drug target function (28, 31), rendering bactericidal drugs ineffective. The inverse relationship between reduced drug target activities and tolerance formation was demonstrated by the effects of bacteriostatic agents, which invariably induced transient tolerance (Fig. (Fig.3).3). The findings that mere inhibition of bacterial growth under specific nutrient-limiting conditions or upon treatment with bacteriostatic agents did not necessarily support tolerance formation are consistent with the idea that cellular integrity could not be maintained for a prolonged period when cell wall synthesis or DNA replication/synthesis was inhibited, unless additional protection mechanisms, such as RecA-mediated target site repair, were also activated. On the other hand, the finding that gentamicin-mediated bactericidal action was not tolerated in the long term implied that the ability to produce starvation-induced tolerance might be compromised without the normal protein synthesis machinery. Such finding is also in agreement with the suppressive effects of bacteriostatic protein synthesis inhibitor, and lack of nucleobases or raw materials for mRNA biosynthesis, on tolerance formation, presumably as a result of reduced capability to launch active protection mechanisms such as antioxidant defense (4) under such conditions. Arguably, the combined effect of nutrient starvation and treatment with protein synthesis inhibitor may impose a dilemma in which bacteria have to balance between the benefits of activation of stress protection mechanisms, which inevitably require active protein synthesis, and stringent-response-related shutdown of the protein synthesis apparatus, which may indirectly alleviate the detrimental effects of mistranslated proteins caused by protein synthesis inhibitors (15). However, the fact that stationary-phase populations exhibit prolonged tolerance to gentamicin (Fig. (Fig.2)2) suggests that bacteria may also possess tolerance strategies that either prevent entry of this drug or circumvent the need to synthesize stress protection proteins. In summary, our data have provided an important framework for further characterization of the molecular basis of bacterial antibiotic tolerance and delineation of the complex signal sensing and protection mechanisms involved in these processes. Such information shall facilitate development of more-effective antimicrobial strategies through assessment of the nutritional impact of in vivo microenvironment on tolerance formation, as well as devising antibacterial stress defense agents that suppress tolerance induction in specific pathogens.


We thank Michael Cashel and Hirotada Mori for providing the stringent-response mutants and the Keio Collection strains, respectively.


[down-pointing small open triangle]Published ahead of print on 19 January 2010.


1. Baba, T., T. Ara, M. Hasegawa, Y. Takai, Y. Okumura, M. Baba, K. A. Datsenko, M. Tomita, B. L. Wanner, and H. Mori. 2006. Construction of Escherichia coli K-12 in-frame, single-gene knockout mutants: the Keio collection. Mol. Syst. Biol. 2:2006.0008. [PMC free article] [PubMed]
2. Balaban, N. Q., J. Merrin, R. Chait, L. Kowalik, and S. Leibler. 2004. Bacterial persistence as a phenotypic switch. Science 305:1622-1625. [PubMed]
3. Bigger, J. W. 1944. Treatment of staphylococcal infections with penicillin. Lancet ii:497-500.
4. Debbia, E. A., S. Roveta, A. M. Schito, L. Gualco, and A. Marchese. 2001. Antibiotic persistence: the role of spontaneous DNA repair response. Microb. Drug Resist. 7:335-342. [PubMed]
5. Drlica, K., M. Malik, R. J. Kerns, and X. Zhao. 2008. Quinolone-mediated bacterial death. Antimicrob. Agents Chemother. 52:385-392. [PMC free article] [PubMed]
6. Eng. R. H., F. T. Padberg, S. M. Smith, E. N. Tan, and C. E. Cherubin. 1991. Bactericidal effects of antibiotics on slowly growing and nongrowing bacteria. Antimicrob. Agents Chemother. 35:1824-1828. [PMC free article] [PubMed]
7. Gummesson, B., L. U. Magnusson, M. Lovmar, K. Kvint, O. Persson, M. Ballesteros, A. Farewell, and T. Nystrom. 2009. Increased RNA polymerase availability directs resources towards growth at the expense of maintenance. EMBO J. 28:2209-2219. [PubMed]
8. Hansen, S., K. Lewis, and M. Vulic. 2008. Role of global regulators and nucleotide metabolism in antibiotic tolerance in Escherichia coli. Antimicrob. Agents Chemother. 52:2718-2726. [PMC free article] [PubMed]
9. Hengge-Aronis, R. 2002. Signal transduction and regulatory mechanisms involved in control of the sigma(S) (RpoS) subunit of RNA polymerase. Microbiol. Mol. Biol. Rev. 66:373-395. [PMC free article] [PubMed]
10. Janion, C. 2008. Inducible SOS response system of DNA repair and mutagenesis in Escherichia coli. Int. J. Biol. Sci. 4:338-344. [PMC free article] [PubMed]
11. Keren, I., N. Kaldalu, A. Spoering, Y. Wang, and K. Lewis. 2004. Persister cells and tolerance to antimicrobials. FEMS Microbiol. Lett. 230:13-18. [PubMed]
12. Keren, I., D. Shah, A. Spoering, N. Kaldalu, and K. Lewis. 2004. Specialized persister cells and the mechanism of multidrug tolerance in Escherichia coli. J. Bacteriol. 186:8172-8180. [PMC free article] [PubMed]
13. Klauck, E., A. Typas, and R. Hengge. 2007. The sigmaS subunit of RNA polymerase as a signal integrator and network master regulator in the general stress response in Escherichia coli. Sci. Prog. 90:103-127. [PubMed]
14. Kohanski, M. A., D. J. Dwyer, B. Hayete, C. A. Lawrence, and J. J. Collins. 2007. A common mechanism of cellular death induced by bactericidal antibiotics. Cell 130:797-810. [PubMed]
15. Kohanski, M. A., D. J. Dwyer, J. Wierzbowski, G. Cottarel, and J. J. Collins. 2008. Mistranslation of membrane proteins and two-component system activation trigger antibiotic-mediated cell death. Cell 135:679-690. [PMC free article] [PubMed]
16. Lewis, K. 2007. Persister cells, dormancy and infectious disease. Nat. Rev. Microbiol. 5:48-56. [PubMed]
17. Li, Y., and Y. Zhang. 2007. PhoU is a persistence switch involved in persister formation and tolerance to multiple antibiotics and stresses in Escherichia coli. Antimicrob. Agents Chemother. 51:2092-2099. [PMC free article] [PubMed]
18. Loewen, P. C., and R. Hengge-Aronis. 1994. The role of the sigma factor sigma S (KatF) in bacterial global regulation. Annu. Rev. Microbiol. 48:53-80. [PubMed]
19. Magnusson, L. U., A. Farewell, and T. Nystrom. 2005. ppGpp: a global regulator in Escherichia coli. Trends Microbiol. 13:236-242. [PubMed]
20. Merrikh, H., A. E. Ferrazzoli, A. Bougdour, A. Olivier-Mason, and S. T. Lovett. 2009. A DNA damage response in Escherichia coli involving the alternative sigma factor, RpoS. Proc. Natl. Acad. Sci. U. S. A. 106:611-616. [PubMed]
21. Miller, C., L. E. Thomsen, C. Gaggero, R. Mosseri, H. Ingmer, and S. N. Cohen. 2004. SOS response induction by beta-lactams and bacterial defense against antibiotic lethality. Science 305:1629-1631. [PubMed]
22. Moyed, H. S., and K. P. Bertrand. 1983. hipA, a newly recognized gene of Escherichia coli K-12 that affects frequency of persistence after inhibition of murein synthesis. J. Bacteriol. 155:768-775. [PMC free article] [PubMed]
23. Neidhardt, F. C., P. L. Bloch, and D. F. Smith. 1974. Culture medium for enterobacteria. J. Bacteriol. 119:736-747. [PMC free article] [PubMed]
24. Neuhard, J., and P. Nygaard. 1987. Purines and pyrimidines, p. 445-473. In J. L. Ingraham, K. B. Low, B. Magasanik, F. C. Neidhardt, M. Schaechter, and H. E. Umbarger (ed.), Escherichia coli and Salmonella typhimurium: cellular and molecular biology. American Society for Microbiology, Washington, DC.
25. Peterson, C. N., M. J. Mandel, and T. J. Silhavy. 2005. Escherichia coli starvation diets: essential nutrients weigh in distinctly. J. Bacteriol. 187:7549-7553. [PMC free article] [PubMed]
26. Potrykus, K., and M. Cashel. 2008. (p)ppGpp: still magical? Annu. Rev. Microbiol. 62:35-51. [PubMed]
27. Rodionov, D. G., and E. E. Ishiguro. 1995. Direct correlation between overproduction of guanosine 3′,5′-bispyrophosphate (ppGpp) and penicillin tolerance in Escherichia coli. J. Bacteriol. 177:4224-4229. [PMC free article] [PubMed]
28. Rodionov, D. G., A. G. Pisabarro, M. A. de Pedro, W. Kusser, and E. E. Ishiguro. 1995. Beta-lactam-induced bacteriolysis of amino acid-deprived Escherichia coli is dependent on phospholipid synthesis. J. Bacteriol. 177:992-997. [PMC free article] [PubMed]
29. Sambrook, J., and D. W. Russell (ed.). 2001. Molecular cloning: a laboratory manual, 3rd ed. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY.
30. Spoering, A. L., M. Vulic, and K. Lewis. 2006. GlpD and PlsB participate in persister cell formation in Escherichia coli. J. Bacteriol. 188:5136-5144. [PMC free article] [PubMed]
31. Srivatsan, A., and J. D. Wang. 2008. Control of bacterial transcription, translation and replication by (p) ppGpp. Curr. Opin. Microbiol. 11:100-105. [PubMed]
32. Tuomanen, E., R. Cozens, W. Tosch, O. Zak, and A. Tomasz. 1986. The rate of killing of Escherichia coli by beta-lactam antibiotics is strictly proportional to the rate of bacterial growth. J. Gen. Microbiol. 132:1297-1304. [PubMed]
33. Weber, H., T. Polen, J. Heuveling, V. F. Wendisch, and R. Hengge. 2005. Genome-wide analysis of the general stress response network in Escherichia coli: sigmaS-dependent genes, promoters, and sigma factor selectivity. J. Bacteriol. 187:1591-1603. [PMC free article] [PubMed]
34. Xiao, H., M. Kalman, K. Ikehara, S. Zemel, G. Glaser, and M. Cashel. 1991. Residual guanosine 3′,5′-bispyrophosphate synthetic activity of relA null mutants can be eliminated by spoT null mutations. J. Biol. Chem. 266:5980-5990. [PubMed]

Articles from Antimicrobial Agents and Chemotherapy are provided here courtesy of American Society for Microbiology (ASM)