HipA-Triggered Growth Arrest and β-Lactam Tolerance in Escherichia coli Are Mediated by RelA-Dependent ppGpp Synthesis

Abstract

Persistence is a phenomenon whereby a subpopulation of bacterial cells enters a transient growth-arrested state that confers antibiotic tolerance. While entrance into persistence has been linked to the activities of toxin proteins, the molecular mechanisms by which toxins induce growth arrest and the persistent state remain unclear. Here, we show that overexpression of the protein kinase HipA in Escherichia coli triggers growth arrest by activating synthesis of the alarmone guanosine tetraphosphate (ppGpp) by the enzyme RelA, a signal typically associated with amino acid starvation. We further demonstrate that chemically suppressing ppGpp synthesis with chloramphenicol relieves inhibition of DNA replication initiation and RNA synthesis in HipA-arrested cells and restores vulnerability to β-lactam antibiotics. HipA-arrested cells maintain glucose uptake and oxygen consumption and accumulate amino acids as a consequence of translational inhibition. We harness the active metabolism of HipA-arrested cells to provide a bacteriophage-resistant platform for the production of biotechnologically relevant compounds, which may represent an innovative solution to the costly problem of phage contamination in industrial fermentations.

INTRODUCTION

Persister cells are bacteria in a transient, growth-arrested state that provides tolerance of multiple classes of antibiotics (1, 2). The ability to resist antibiotic treatment enables persister cells to sustain bacterial infections. The persistent state, which likely describes a number of different growth-arrested physiologies (3), has been at least partially attributed to the activity of toxin/antitoxin (TA) modules (4). TA modules are composed of two genes; one of the two genes encodes a toxin protein whose expression slows or stops cell growth, and the other encodes its corresponding antitoxin protein or RNA, which either inactivates its toxin directly or prevents translation of the toxin protein. Toxins have diverse enzymatic activities and cellular targets (5), and it is thought that active toxins trigger the growth slowdown or arrest that is characteristic of the persistent state by inhibiting a central cellular process. A link between TA modules and persistence has recently been demonstrated in a laboratory setting using Escherichia coli MG1655. In that study, deletion of five or more chromosomally encoded TA modules reduced the number of persister cells generated (6).

The first toxin protein to be linked to persistence in E. coli was the serine/threonine kinase HipA (7), which forms a TA module with its cognate antitoxin, HipB. HipA overexpression within growing E. coli bacteria causes multiple responses that are the hallmarks of persistence, including growth arrest and tolerance of certain classes of antibiotics, as well as attenuation of DNA replication, transcription, and translation (8). Some insight into the mechanism of HipA-induced growth arrest was gained when it was discovered that HipA phosphorylates the protein EF-Tu (9), an essential translation factor that catalyzes the binding of aminoacyl-tRNA to the ribosome. Phosphorylation by HipA is expected to deactivate EF-Tu (10), which may account for the inhibition of translation within HipA-arrested cells. How HipA expression also inhibits DNA and RNA synthesis and provides tolerance of certain antibiotics when translation is inhibited is unclear. Translational inhibition is insufficient to explain the general arrest of macromolecular synthesis, as translation inhibition by other means (e.g., by addition of the ribosome inhibitor chloramphenicol) does not inhibit DNA replication or RNA transcription (11) or provide antibiotic tolerance (12). While it is possible that HipA inhibits DNA and rRNA synthesis directly, such as by phosphorylating protein targets other than EF-Tu, we and others have observed that the growth arrest triggered by HipA is similar to the effects of the alarmone guanosine tetraphosphate (ppGpp) (8). ppGpp allosterically inhibits enzymes that are central to an incredible variety of cellular processes in E. coli, including priming of DNA replication (13), rRNA transcription (14), translation (15), phospholipid synthesis (16), and certain metabolic enzymes (17). A link between the stringent response and persistence has been postulated on the basis of the finding that an E. coli strain lacking the enzymes required to synthesize ppGpp (RelA and SpoT) generates fewer persister cells (18). One possible interpretation of this finding might be that ppGpp is necessary for inducing a cellular response, such as expression of toxins or degradation of antitoxins, that triggers persistence upon application of stress. An alternative explanation is that ppGpp directly confers resistance by inhibiting cellular processes itself (18), an aspect of regulation by ppGpp that is often unappreciated (19).

Understanding the mechanisms by which toxin activity leads to growth arrest and antibiotic tolerance is critical for informing efforts to eradicate persisters (3). We sought to determine if HipA expression triggers ppGpp synthesis and if ppGpp is directly responsible for the general attenuation of macromolecular synthesis and development of antibiotic tolerance. We also wished to examine some aspects of the metabolic response to HipA-mediated growth arrest, as recent reports indicate that persister cells maintain an active metabolism (20). We find that HipA expression activates ppGpp synthesis by the ribosome-associated ppGpp synthetase RelA, leading to inhibition of macromolecular synthesis, triggering growth arrest, and enabling resistance to β-lactam antibiotics. This establishes the mechanistic link between HipA expression and several major traits of persistence. We also found that HipA-arrested cells maintain metabolic activity and attenuate anabolism by feedback inhibition.

MATERIALS AND METHODS

Strains, media, and plasmids.

Oligonucleotides used in this study are listed in Table 1. Plasmids and strains used in this study are listed in Table 2. Escherichia coli DH10B, grown in LB medium, was used for cloning and plasmid amplification. Unless otherwise noted, for all HipA arrest experiments we used morpholinepropanesulfonic acid (MOPS)-M9 medium with 0.4% glucose and NH₄Cl supplemented to 2 g/liter, with 50 μg/ml kanamycin added when appropriate. E. coli MG1655 (ATCC) was used for all persistence experiments. Experiments to determine the involvement of RelA in growth arrest were performed using E. coli strains CF1648 (MG1655, here referred to as the WT [wild type] strain) and CF12510 (MG1655, here referred to as the ΔrelA strain), which were generously provided by Michael Cashel. hipA was cloned from MG1655 genomic DNA using the primers GB081210-hipA-BB.F and GB081210-hipA-BB.R. The PCR fragment was digested using BglII and XhoI enzymes and ligated into BglII-XhoI-digested BglBrick plasmid pBbS2k to make pHipA. For measurement of HipA expression and ppGpp timing, an operon containing both hipA and gfpmut2 genes was constructed. Green fluorescent protein (GFP) was amplified from pUA66 using primers GB042111-GFPmut2.f and GB042111-GFPmut2.r, and the PCR product was digested with BglII and XhoI and ligated into pHipA digested with BamHI and XhoI, creating pHipA-GFP. Plasmid pBbA5c-MTSA, carrying genes atoB (codon optimized for E. coli [21]), Staphylococcus aureus mvaA, and S. aureus mvaS (22) encoding a heterologous mevalonate pathway, was assembled on the pBbA5c backbone using standard BglBrick cloning procedures (23). atoB was amplified from the pBbA5c-MevT-MBIS (fixed) plasmid using primers atoB-fw and atoB-rw. mvaA was amplified from plasmid pTSAmvaASP using primers mvaA_fw and mvaA_rv. mvaS was amplified in two PCRs using primer pair mvaS_fw and EcoRIdel_Rv and primer pair EcoRIdel_Fw and mvaS_rv in order to remove an internal EcoRI restriction site, and the results were linked by splice-by-overlap extension (SOE) PCR (24) using primers mvaS_fw and mvaS_rv.

Table 1.

Oligonucleotides used in this study

Oligonucleotide	Sequence (5′→3′)
GB081210-hipA-BB.F	TTC AAA AGA TCT CAA CAG AAC AGC AAA ATC TGG AGT GG
GB081210-hipA-BB.R	TCC TTA CTC GAG TTA GGA TCC TCA CTT ACT ACC GTA TTC TCG GCT TAA C
GB042111-GFPmut2.f	TAA TCC AGA TCT AGG AGG AAA AAA AAA TGA GTA AAG GAG AAG AAC TTT TCA CTG G
GB042111-GFPmut2.r	ATA CCA CTC GAG TTT GGA TCC TTA TTT GTA CAA TTC ATC CAT ACC ATG GGT AAT C
AtoB-fw	GAA TTC AAA AGA TCT TAG GAG GAA TAT A
AtoB-rv	CTC GAG TTT GGA TCC TCA ATT CAA CCG TTC AAT
mvaA_fw	GAA TTC AAA AGA TCT ATT CAG GAA ACA GAC CAT GTC CAT GCA A
mvaA_rv	CTC GAG TTT GGA TCC TTA TTG TTG TCT AAT
mvaS_fw	GAA TTC AAA AGA TCT AGG AGA AAC CTT ATG
mvaS_rv	CTC GAG TTT GGA TCC TTA CTC TGG TCT GTG AT
EcoRIdel_Fw	TGG ATT GAA CTC AGG CGG CGA GC
EcoRIdel_Rv	GCT CGC CGC CTG AGT TCA ATC CAT AAC GTG CT

Table 2.

Plasmids and strains used in this study

Strain or plasmid	Description	Reference or source
Strains
Escherichia coli DH10B	E coli strain for cloning and plasmid amplification	Invitrogen
Escherichia coli MG1655	E coli strain for growth arrest and antibiotic resistance studies	ATCC
Escherichia coli CF1648 (MG1655)	E coli strain MG1655 for direct comparison with ΔrelA strain	56; Michael Cashel, National Institutes of Health
Escherichia coli CF12510 (MG1655 ΔrelA)	E coli strain MG1655 relA256 (unmarked)	56, Michael Cashel, National Institutes of Health
Plasmids
pHipA	HipA under the control of Ptet promoter	This work
pHipA-GFP	HipA GFPmut2 transcriptional fusion	This work
pUA66	Source of gfpmut2 gene	57
pBbS2k	Plasmid backbone for pHipA	23
pBbA5c-MTSA	Encodes mevalonate pathway	This work
pBbA5c-Mevt-MBIS(fixed)	Source of codon-optimized atoB	21
pTSAmvaASP	Source of S. aureus mvaS and mvaA genes	S. W. Kim, Kyunsang National University, South Korea

ppGpp measurements performed with high-performance liquid chromatography (HPLC).

ppGpp and nucleotides were extracted from cells grown on agar-supported nylon filters, following a protocol inspired by the rapid-quenching and metabolite extraction approach described in reference 25. We adopted this technique to avoid perturbing ppGpp levels with lengthy cell harvesting and extraction procedures, as the enzymes that generate and degrade ppGpp are known to respond rapidly to external conditions (26). Briefly, cells grown in liquid medium to an optical density at 600 nm (OD₆₀₀) of ∼0.1 were transferred to nylon filters and adhered to solid agar plates containing our modified MOPS-M9 medium. Persistence was induced when filter-supported cells reached an OD₆₀₀ of 0.5 by transferring the filters to agar plates containing the same medium supplemented with 50 ng/ml anhydrous tetracycline (aTc). To induce rapid growth arrest by isoleucine starvation, filters were transferred to solid medium containing 40 μg/ml valine. Metabolism was rapidly quenched by transferring the filters into 5 ml of 2 M ice-cold formic acid for 30 min, after which the solutions were collected and neutralized by the use of 1 ml of 15% ammonium bicarbonate, flash-frozen in liquid nitrogen, and lyophilized. Lyophilized cells were resuspended in 0.2 ml cold 50% methanol, centrifuged again to precipitate insoluble debris, and spin filtered using a Amicon UFC500396 centrifugal filter unit (molecular weight cutoff [MWCO] of 3) in a refrigerated centrifuge set to −2°C to remove soluble proteins. Concentrations of ppGpp and GTP were measured using anion-exchange chromatography and a Partisil 5 Sax Rac II column (catalog no. 422-227; Hichrom Limited), with a silica Nova-Pak guard column (catalog no. WAT046845; Whatman) per a protocol described previously (26) and modified for brevity. Buffers A and B consisted of 0.05 M and 1 M ammonium phosphate (pH 3.4), respectively. Filtered extracts (25 μl) were injected, and the mobile phase was ramped to a buffer B level of 100% over 4 min and a flow rate of 1 ml/min. At 4.5 min, the rate was increased to 1.5 ml/min and the process was run until 10 min, at which point the mobile phase was ramped back down to 0% buffer B and was run through the column for 20 min. The column temperature was maintained at 50°C, and absorbance was monitored at 254 nm. The ppGpp standard was purchased from Trilink Biotechnologies (N-6001).

For measurements of hipA expression using the hipA-gfpmut2 transcriptional fusion, GFP measurements were made in parallel with ppGpp measurements, which were performed as described above using MG1655/pHipA-GFP cells. To measure GFP, translation was quenched by transferring filters to 5 ml of H₂O with 30 μg/ml chloramphenicol. GFP fluorescence measurements were performed 30 min after translational quenching to enable translated GFP to become fluorescent. GFP fluorescence (excitation, 488 nm; emission, 510 nm) was measured using a Spectromax M2 reader (Molecular Devices).

Measurement of DNA and RNA concentrations.

Cultures of MG1655/pHipA cells were inoculated into 40 ml of MOPS-M9 medium in 250-ml unbaffled Erlenmeyer flasks and incubated with shaking at 37°C. When the OD₆₀₀ reached ∼0.3, aTc was added to reach 50 ng/ml. After 1 h of HipA induction, a sample was removed for DNA or RNA extraction, the cultures were split into two 15-ml cultures, and 30 μg/ml chloramphenicol was added to one of the cultures. For comparison of DNA and RNA synthesis results in WT and ΔrelA strains expressing HipA, biological triplicates of the WT/pHipA and ΔrelA/pHipA strains were grown in MOPS-M9 medium at 37°C to an OD₆₀₀ of 0.2 to 0.4, at which point HipA expression was induced by addition of aTc to 50 ng/ml. Samples were processed per a previously published protocol (27) using Picogreen (Invitrogen) for DNA quantification. RNA was quantified by measurement of fluorescence using the Quant-It assay (catalog no. Q33140; Invitrogen) following the manufacturer's instructions. Fluorescence was measured using a Spectromax M2 reader.

β-Lactam susceptibility of HipA-arrested cells.

Biological triplicates of MG1655/pHipA cells were inoculated into 30 ml of MOPS-M9 medium and grown at 37°C until an OD₆₀₀ of 0.5 was reached. aTc was added to reach 50 ng/ml, and the cultures were split into two 12-ml subcultures. At 2.5 h later, carbenicillin (100 μg/ml) was added to one subculture and optical densities were measured. For comparison of the levels of β-lactam tolerance of WT and ΔrelA cultures expressing HipA, biological triplicates of the WT/pHipA and ΔrelA/pHipA strains were inoculated into 5 ml of MOPS-M9 medium and grown at 37°C. HipA expression was induced with 50 ng/ml aTc when the OD₆₀₀ reached ∼0.3. Ampicillin was added to reach a final concentration of 100 μg/ml 1.5 h later, and optical densities were monitored.

Measurement of glucose consumption and respiration.

Cultures were incubated in a Respiration Activity MOnitoring System (RAMOS; HiTec Zang) at 37°C in Erlenmeyer flasks with shaking in a Lab Therm LT-X incubator (Kuhner) (28). Overnight cultures of the MG1655/pHipA strain were inoculated at 1/100 into 20 ml of pre-equilibrated MOPS-M9 glucose medium. Samples were periodically removed for glucose concentration and growth measurements. Glucose concentrations were quantified via HPLC, using an Aminex HPX-87H column maintained at 50°C while running 4 mM H₂SO₄ at a constant rate (0.6 ml/min) on an Agilent 1200 system equipped with a refractive index detector.

Metabolite concentration measurements.

Intracellular metabolites were measured using an isotopic ratio-based approach, largely as described previously (25), with slight modifications. Briefly, cells were grown on agar plates with our modified MOPS-M9 medium with 0.4% glucose fully labeled with ¹³C (CLM-1396; Cambridge Isotope Laboratories). Cells were grown in liquid medium and transferred to nylon filters, and the nylon filters were adhered to solid medium as described above. Persistence was induced by transferring filters to agar plates containing 50 ng/ml aTc, and metabolism was rapidly quenched by transferring the filter to a petri dish containing 40:40:20 acetonitrile:methanol:water (with 0.1 M formic acid) and isotopically unlabeled metabolite standards. After 30 min of quenching, the extract was used to wash cellular debris from the filters and collected into a Falcon tube. An additional 2.5 ml of quenching solution (without metabolite standards) was used to wash the filter and the petri dish once more and collected into the same Falcon tube. Finally, 0.5 ml of 15% ammonium bicarbonate was added to the extract to neutralize the solution, and the solution was flash-frozen in liquid nitrogen and lyophilized. Lyophilized samples were resuspended in 200 μl of 50% methanol and filtered as with the ppGpp extracts (described above), and the filtered extract was analyzed using tandem liquid chromatography/mass spectroscopy (LC/MS). LC/MS protocols for measuring nucleotide, organic acids, and sugar phosphate concentrations were used as previously described (29).

For measurement of intracellular amino acids, liquid chromatographic separation was conducted at 30°C with a Kinetex hydrophilic interaction LC (HILIC) column (Phenomenex, Torrance, CA) (100-mm length, 4.6-mm internal diameter, 2.6-μm particle size) using a 1200 Series HPLC system (Agilent Technologies, Santa Clara, CA). The injection volume for each measurement was 2 μl. The mobile phase was composed of 20 mM ammonium acetate–water (solvent A) and 10 mM ammonium acetate–90% acetonitrile–10% water (solvent B) (HPLC grade; Honeywell Burdick & Jackson). Ammonium acetate was prepared from a stock solution of 100 mM ammonium acetate and 0.7% formic acid (Sigma-Aldrich, St. Louis, MO) (98% to 100% chemical purity) in water. Amino acids were separated with the following gradient: 90% to 70% solvent B for 4 min followed by a hold at 70% solvent B for 1.5 min, 70% to 40% solvent B for 0.5 min followed by a hold at 40% solvent B for 5.5 min, and 40% to 90% solvent B for 0.5 min followed by a hold at 90% solvent B for 2 min. The flow rate was adjusted as follows: 0.6 ml/min for 12.5 min and 0.6 ml/min to 1 ml/min for 0.5 min followed by a hold at 1 ml/min for 2 min.

The HPLC system was coupled to an Agilent Technologies 6210 time-of-flight mass spectrometer (TOF MS) with a 1:6 postcolumn split. Nitrogen gas was used as both the nebulizing gas and the drying gas to facilitate the production of gas-phase ions. The drying and nebulizing gases were set to 12 liters/min and 25 lb/in², respectively, and a drying gas temperature of 320°C was used throughout. Fragmentor, skimmer, and optical coherence tomography 1 (OCT1) radio frequency (RF) voltages were set to 100 V, 50 V, and 170 V, respectively. Electrospray ionization (ESI) was conducted in the positive-ion mode with a capillary voltage of 3.5 kV. MS experiments were carried out in the full-scan mode (m/z 70 to 340) at 0.86 spectra per second for the detection of [M + H]⁺ ions. The instrument was tuned for a range of m/z 50 to 1,700. Prior to LC-ESI-TOF MS analysis, the TOF MS was calibrated with the Agilent ESI-Low TOF tuning mix.

The separation of nucleotides was conducted on a ZIC-pHILIC column (Merck SeQuant [distributed via The Nest Group, Inc., Southborough, MA) (150-mm length, 2.1-mm internal diameter, and 5-μm particle size) using an Agilent Technologies 1200 Series HPLC system. Injection volumes for the measurement of chemical standards and metabolites from biological extracts were 2 and 4 μl, respectively. The column compartment was set to 30°C. All other chromatographic conditions were as described above. The mobile phase was composed of (A) 100 mM ammonium carbonate in water and (B) acetonitrile. A flow rate of 0.23 ml/min was used unless otherwise stated. Nucleotide separation was conducted as follows: 78% B (0 min), 48% B (17 min), 78% B (20 min), 78% B (20.5 min), and 78% B (25 min). The flow rate was increased from 0.23 to 0.3 ml/min from 20 to 20.5 min and held for a further 4.5 min. The HPLC system was coupled to TOF MS with a 1:3 postcolumn split. ESI was conducted in the negative ion mode, and a capillary voltage of 3.5 kV was utilized. Fragmentor, skimmer, and OCT1 RF voltages were set to 200 V, 65 V, and 300 V, respectively. MS experiments were carried out in the full-scan mode (m/z 335 to 810). All other MS conditions were as described above. The LC/MS protocol for measuring pyruvate concentrations was taken from reference 30.

Kinetic flux measurements.

Flux through amino acid pools was measured using kinetic flux profiling as described in reference 31. Briefly, filter-grown cells induced into the persistent state were transferred to solid medium containing medium with aTc and 0.4% ¹³C-labeled glucose. At set times, filters were removed and quenched as described above for the metabolite concentration measurements. Ratios of unlabeled metabolites to metabolites labeled at any position were determined for each time point as described above.

For determining flux for each metabolite, an upstream precursor metabolite was found and its replacement rate was determined. This rate was used as a fixed parameter in a subsequent fitted equation for the amino acid as described previously (31). Suitable precursor metabolites could not be found for all amino acids (e.g., shikimate for the aromatic amino acids) because of instability of the precursor under our extraction conditions. Amino acids and the corresponding precursors used are listed in Table 3.

Table 3.

Amino acids and their precursors used for kinetic flux profiling

Amino acid	Precursor metabolite
Alanine	Pyruvate
Threonine	Aspartic acid
Methionine	Aspartic acid
Proline	Glutamate
Arginine	Glutamate
Asparagine	Aspartic acid
Lysine	Pyruvate
Isoleucine	Pyruvate
Glutamine	Glutamate

Antibiotic resistance.

Persister cells were prepared in triplicate by ofloxacin treatment of stationary-phase cultures based on a previously described protocol (32). Overnight cultures of MG1655 cells grown in LB medium for 20 h were treated with ofloxacin for 3 h at 37°C, after which cells were pelleted, washed with fresh LB medium, pelleted again, and resuspended in antibiotic-free LB medium. After resuspension, chloramphenicol (30 μg/ml) and ampicillin (100 μg/ml) were added singly or in combination, and cells were incubated at 37°C with shaking for 4 h. Samples were removed, pelleted once, washed with LB medium, resuspended, and serially diluted. A 10-μl volume of the serial dilutions was spot plated on an LB agar plate, which was incubated at 37°C overnight. Colonies were counted the following day.

Mevalonate production.

MG1655/pHipA cells were transformed with pBbA5c-MTSA. Overnight cultures were inoculated at 1/50 into 50 ml modified MOPS-M9 medium (Teknova) with 1% glucose, 50 μg/ml kanamycin, 30 μg/ml chloramphenicol, and 1 μg/ml nystatin and grown at 30°C. At an OD of 0.18, isopropyl-β-d-thiogalactopyranoside (IPTG) was added to the cultures to induce expression of the mevalonate pathway. At 5 h later (OD of ∼0.8), 100 ng/ml aTc was added to induce persistence, and 3 h later carbenicillin was added to reach 50 μg/ml to prevent the growth of cells that resisted HipA-induced arrest. Additional medium was added after 35 and 82 h to sustain mevalonate production with fresh glucose. At 99 h, nystatin was added to the cultures (final concentration, 100 μg/ml) to prevent mold growth. For measurement of extracellular mevalonate, culture samples were removed and the cells were pelleted by centrifugation. The supernatant was transferred to an equal volume of ice-cold methanol, and the mixture was spin filtered as described above. Mevalonate was measured using an LC-MS method for organic acids (29).

Phage resistance.

Overnight cultures of MG1655/pHipA/pBbA5c-MTSA cells were inoculated at 1/50 into 50 ml medium (as described for mevalonate production) plus 10 mM MgSO₄ and 5 mM CaCl₂ to enable phage binding and lysis and were grown with shaking at 30°C. At an OD₆₀₀ of ∼0.2, IPTG was added to reach a final concentration of 250 μM. At 5 h later (OD of ∼0.8), the cultures were split into two 25-ml subcultures, and aTc was added to 100 ng/ml to one of the subcultures. At 4 h later, carbenicillin (100 μg/ml) was added to the aTc-induced cultures to prevent growth of mutants that resisted growth arrest, and the cultures were further split into 5-ml tubes. At this time, 1 μl of a chloroform-sterilized P1vir lysate was added to the appropriate tubes. At 17 h later, all cultures were subcultured at 1/20 into 5 ml of fresh medium with additives identical to those used with the parent tubes. Samples were removed, cells and cell debris were pelleted, the supernatant was removed and spin filtered, and mevalonate in the filtrate was measured as before.

RESULTS

HipA expression causes growth arrest and ppGpp biosynthesis by RelA.

We introduced plasmid pHipA, a low-copy-number plasmid carrying hipA under the control of a tetracycline-inducible promoter (P_Tet), into an E. coli MG1655 derivative (CF1648, here referred to as the WT strain) and a relA-knockout MG1655 strain (CF12510, here referred to as the ΔrelA strain). HipA expression was induced during exponential growth with anhydrous tetracycline (aTc), causing nearly complete growth arrest in the WT strain. In contrast, HipA expression in the ΔrelA strain sharply inhibited growth but did not arrest growth completely (Fig. 1A). Increasing the concentration of the chemical inducer aTc did not suppress the residual growth observed in the ΔrelA strain, as the specific growth rates after HipA induction did not significantly change at concentrations above 2 ng/ml aTc (Fig. 1B; average growth rates, 0.025 ± 0.010 h⁻¹ for the WT strain and 0.066 ± 0.012 h⁻¹ for the ΔrelA strain).

Fig 1.

HipA expression activates ppGpp synthesis by the enzyme RelA, which is required for complete growth arrest and the persistent state. (A) Growth curves of WT and ΔrelA E. coli MG1655 (solid black and red, respectively) bearing the plasmid pHipA. At the indicated time, 50 ng/ml anhydrous tetracycline (aTc) was added to the cultures, inducing growth arrest after ∼45 min (dotted lines). (B) Specific growth rates of WT and ΔrelA strains measured 1 h after addition of various amounts of aTc. Error bars represent averages of at least 32 data point pairs. (C) Representative chromatogram of nucleotide extract of WT and ΔrelA plate-grown cultures after 90 min of HipA induction. mAu, milliabsorbance units. (D) ppGpp concentration measured in cells during exponential growth, cells starved for isoleucine by the addition of valine (+Val), and cells experiencing HipA-mediated growth arrest (+aTc). Error bars represent the standard deviations of the results of at least three measurements. (E) Timing of ppGpp synthesis in response to HipA expression and growth arrest. A hipA-gfpmut2 operon enabled HipA expression to be indirectly monitored with GFP fluorescence.

The residual growth observed in the ΔrelA strain during HipA expression suggests that growth arrest by HipA requires ppGpp synthesis by RelA. To determine if ppGpp is produced by RelA in HipA-expressing cells, we measured ppGpp concentrations 1.5 h after HipA induction in both WT and ΔrelA strains. HipA expression triggered ppGpp biosynthesis in WT cells but not in ΔrelA cells (Fig. 1C). ppGpp concentrations in HipA-arrested WT cells approached the levels measured in cells starved for an amino acid (Fig. 1D). In order to establish the timing of ppGpp synthesis relative to HipA expression, we constructed a transcriptional fusion of hipA and gfpmut2 by inserting a gfpmut2 gene immediately after hipA in a single operon. We found that ppGpp was synthesized quickly after hipA induction and reached its maximum concentration once growth arrest had completed (Fig. 1E).

HipA activity does not induce ppGpp biosynthesis by causing amino acid starvation.

Because ppGpp synthesis by RelA can be triggered by amino acid starvation, we asked whether HipA expression somehow increases ppGpp by causing a reduction in amino acid concentrations. We measured the intracellular concentrations of amino acids 1 h after induction of hipA expression (approximately 15 min after growth arrest). We found that most amino acid concentrations were elevated severalfold over those measured in actively growing cells (Fig. 2, light gray and dark gray bars) and did not significantly change after 1 h of growth arrest, indicating that HipA expression does not trigger ppGpp levels by inducing starvation. We attribute the increase in amino acid levels to the decreased translational activity of HipA-arrested cells.

Fig 2.

Growth arrest by HipA induction causes intracellular accumulation of amino acids, indicating that HipA does not trigger ppGpp synthesis by causing amino acid depletion. Light gray and dark gray bars represent log₂ amino acid concentration ratios of HipA-arrested cells to growing cells, as measured in cells 1 and 2 h after HipA induction by addition of aTc. Black bars indicate ratios (HipA-arrested cells/growing cells) of flux through indicated amino acid pools determined using kinetic flux profiling. A value of 0 indicates no change. For concentration measurements, error bars represent the standard deviations of the results of three biological replicates. Error bars in flux bars represent uncertainty in flux determination, as described in Materials and Methods. Asterisks indicate amino acid fluxes for which values could not be determined.

Translation inhibition likely attenuates amino acid biogenesis.

To understand how HipA-arrested cells might regulate anabolism in the absence of growth, we used kinetic flux profiling (31) to determine if the increase in amino acid concentrations would suppress their biosynthesis. Briefly, filter-supported cells were quickly transferred from solid medium containing unlabeled glucose to medium containing isotopically labeled glucose, and the washout rate of unlabeled metabolites was determined in both growing cells and HipA-arrested cells. We confirmed that flux through several of the amino acid pools was dramatically reduced (Fig. 2, black bars), which we suspect was a result of feedback inhibition exerted directly by accumulated amino acids upon their respective biosynthetic pathways.

ppGpp inhibits DNA and RNA biosynthesis in HipA-arrested cells.

Having established that HipA expression triggers the production of ppGpp via RelA activation, we next sought to confirm that ppGpp inhibits DNA and RNA synthesis in HipA-arrested cells. Chloramphenicol, a bacteriostatic drug that inhibits translation, is also known to suppress ppGpp levels in cells starved for amino acids (33). We found that chloramphenicol treatment caused the ppGpp concentration within HipA-arrested cells to decrease by 60% (±20%) within 15 min. Chloramphenicol treatment relieved inhibition of both DNA synthesis and RNA synthesis (Fig. 3A and B), indicating that ppGpp blocks DNA and RNA synthesis within HipA-arrested cells. Furthermore, HipA expression was not able to inhibit RNA synthesis in the ΔrelA strain, as RNA concentrations continued to increase for over 2 h after growth arrest (Table 4).

Fig 3.

RelA-dependent synthesis of ppGpp is responsible for the inhibition of macromolecular synthesis observed in HipA-arrested cells. (A and B) Chloramphenicol relieves the inhibition of DNA replication initiation (A) and RNA synthesis (B) imposed by ppGpp. DNA/OD and RNA/OD values were normalized to the value at 1 h. Data represent the results of biological duplicates. (C) Deactivation of ppGpp synthesis by chloramphenicol resensitizes HipA-arrested cells to β-lactam antibiotics. Carbenicillin (100 μg/ml) was added to all cultures at the time indicated. Error bars represent standard deviations of the results of biological triplicates. (D) HipA expression does not enable ΔrelA cells to resist ampicillin (added to 100 μg/ml). Shown are the results of biological triplicates.

Table 4.

DNA/OD and RNA/OD ratios observed after HipA arrest (1 h after addition of aTc) in WT and ΔrelA strains, normalized to ratio determined at 1 h after addition of aTc

Ratio for indicated strain	Avg ratio to value at 1 h after HipA induction	SD
DNA/OD (6 h after arrest)
WT	1.05	0.02
ΔrelA	1.09	0.02
RNA/OD (2 h after arrest)
WT	0.99	0.02
ΔrelA	1.54	0.15

We observed only slight increases in intracellular DNA concentrations in both WT and ΔrelA cultures over 6 h after HipA arrest (Table 4). Similar results have been observed in wild-type (WT) and ΔrelA strains subjected to translation inhibition by treatment with an amino acid analog (34), indicating that non-antibiotic-induced translational inhibition also contributes to regulation of DNA replication, as discussed below.

ppGpp provides resistance to β-lactam antibiotics in HipA-arrested cells.

The induction of ppGpp synthesis by toxins may also provide the mechanistic basis for the tolerance of β-lactam antibiotics by some persister cells, as high ppGpp concentrations protect cells from lysis by β-lactams during amino acid starvation (35, 36) via inhibition of phospholipid biosynthesis (12). If ppGpp also protects HipA-arrested cells from β-lactams, deactivating ppGpp synthesis with chloramphenicol should sensitize the cells to β-lactam antibiotics, and HipA expression should not provide resistance against β-lactams in a ΔrelA strain. We found that adding the β-lactam carbenicillin alone did not induce lysis in HipA-arrested WT cells, confirming the protective nature of the HipA-arrested state (37). However, the addition of carbenicillin in combination with chloramphenicol triggered lysis (Fig. 3C), and expression of HipA did not confer ampicillin resistance to the ΔrelA strain (Fig. 3D).

Chloramphenicol treatment of naturally occurring persister cells does not increase killing by β-lactam antibiotics.

Our finding that combining chloramphenicol with β-lactam treatment results in killing of HipA-arrested cells suggests that therapeutic interventions that deactivate ppGpp synthesis (or accelerate ppGpp hydrolysis) could render persistent infections readily treatable by antibiotics. We tested this possibility by first isolating naturally occurring persister cells by treating overnight cultures with the fluoroquinolone antibiotic ofloxacin for 4 h and subsequently subculturing the cells in fresh medium containing chloramphenicol, ampicillin, or both compounds. We found that chloramphenicol-ampicillin combination treatment did not significantly increase the killing rate of isolated persisters compared to ampicillin alone (28% ± 14% reduction in CFU by combination treatment compared to 41% ± 13% with ampicillin alone [P = 0.3, unpaired t test]).

HipA-arrested cells consume glucose and oxygen and maintain high ATP levels.

We considered the alternative possibility that HipA enables tolerance by depleting ATP levels, which might also be expected to inhibit macromolecular synthesis. We found that the ATP charge was not significantly affected by HipA arrest (Table 5), suggesting to us that the cells maintain active respiration, as has been observed in naturally isolated persister cells (20). We measured glucose and oxygen consumption by cultures expressing hipA during exponential growth in glucose medium. As might be expected from our observations of high ATP concentrations, the HipA-arrested cultures continued to consume glucose and oxygen during growth arrest (Fig. 4), indicating that HipA-arrested cells maintain an active metabolism despite their lack of macromolecular synthesis.

Table 5.

ATP charge of cells during growth and after HipA induction

Cell status	Avg ATP charge valuea	SD
Preinduction	0.93	0.02
HipA uninduced, 1 h	0.90	0.02
HipA induced, 1 h	0.91	0.01

^aData were calculated as follows: ([ATP] + 0.5 · [ADP])/([ATP] + [ADP] + [AMP]).

Fig 4.

HipA-arrested cells maintain consume glucose and oxygen. Data represent glucose and oxygen consumption by a growing culture (top) and a HipA-arrested culture (bottom). At the indicated time (vertical dashed line), HipA expression was induced by the addition of aTc. Examples shown are representative of the results of three biological replicates.

HipA-arrested cells can maintain production of chemicals via synthetic pathways and are refractory to phage infection.

Having found that HipA-arrested cells are fully metabolically active, we next sought to exploit the active carbon flux within HipA-arrested cells to produce useful compounds via synthetic pathways. One advantage of using growth-arrested bacteria as a platform for the production of chemicals might be their higher resistance to phage-induced lysis compared with that of growing cells (38). Industrial-scale fermentations are extremely vulnerable to contamination by phage, which often causes considerable production losses. Prophylactic measures to prevent phage infection are costly and rarely completely effective, and decontamination often requires drastic measures, including shutdown of the affected plants (39–41). Technology to address the threat from phage could potentially relieve some of the need for these measures and decrease fermentation costs (42).

As a proof of concept, we sought to determine if HipA-arrested cells could produce mevalonate, an important precursor to the isoprenoid family of compounds, which includes pharmaceuticals, specialty chemicals, and potential biofuels (43, 44). We transformed a plasmid encoding a mevalonate synthesis pathway into E. coli MG1655 bearing pHipA and activated expression of the pathway before inducing HipA. Mevalonate production in HipA-arrested cells was surprisingly robust, continuing for 7 days (Fig. 5A). The mevalonate production rate per cell in the growth-arrested culture closely matched the rate observed in the nonarrested culture [0.33 ± 0.04 μg/h/(OD₆₀₀ · ml)] before declining slowly after the third day of production [0.23 ± 0.04 μg/h/(OD₆₀₀ · ml) at 3.5 days]. Over 8 days, the HipA-arrested cells produced a final mevalonate titer approximately one-third that produced by growing cells (0.38 ± 0.01 mg versus 1.15 ± 0.07 mg produced by growing cells), which might have been due to translational inhibition in the HipA-arrested cells preventing the regeneration of pathway enzymes lost to proteolytic degradation or deactivation. In order to demonstrate that mevalonate production can be protected by HipA arrest from lysis by phage, we added phage P1vir to mevalonate-producing HipA-arrested cells and uninduced stationary-phase cells. Lysis was triggered by diluting the infected cultures into fresh medium, causing the stationary-phase cells to resume growth. After 2 days, phage infection decreased mevalonate production by nearly 50% in the nonarrested culture, while production by the HipA-arrested culture was unaffected (Fig. 5B). Absolute titers in the HipA-arrested cells were lower (232 ± 44 μM for uninfected and 151 ± 13 μM for phage-infected growing cells versus 22 ± 2 μM for uninfected and 22 ± 3 μM for phage-infected HipA-arrested cells) due to the low cell density of HipA-arrested cells.

Fig 5.

HipA-arrested cells can maintain a heterologous pathway that enables production of mevalonate and are unaffected by infection by phage. (A) Production of mevalonate by HipA-arrested cells. A plasmid-encoded mevalonate pathway was induced with IPTG (at 0.17 days), and cells were subjected to growth arrest by HipA induction 5 h later (0.36 days). At the times indicated with arrows, fresh medium was added to the culture to replenish the supply of glucose for further production. (B) Production of mevalonate in HipA-arrested cells is unaffected by phage infection. Data represent mevalonate titers of phage-infected cultures expressed as ratios to titers obtained from uninfected cultures.

DISCUSSION

The mechanism by which a single protein kinase, HipA, affects so many cellular processes and causes growth arrest has been unknown since the discovery of the hipA gene 30 years ago (45). The finding that HipA phosphorylates EF-Tu in vitro has been thought to explain how HipA expression attenuates translation, while the inhibition of other processes has led to the proposal that HipA phosphorylates targets in addition to EF-Tu (9, 46). Our results do not preclude the possibility that there exist other protein targets for phosphorylation by HipA or that ppGpp induces further protective responses. However, we find that ppGpp synthesis by RelA is necessary and sufficient for maintaining inhibition of DNA replication initiation, rRNA transcription, and β-lactam tolerance in HipA-arrested cells.

Inhibition of translation by HipA may also play some role in controlling DNA replication, as HipA expression within ΔrelA cells did not result in a large increase in intracellular DNA, seemingly contradicting the results obtained by chloramphenicol treatment. However, translational inhibition by an amino acid analog (serine hydroxamate), which induces ppGpp synthesis by RelA in wild-type cells, also does not cause a dramatic accumulation of DNA within ΔrelA cells (34). We suspect that while ppGpp inhibits DNA replication at defined points (namely, replication initiation [13, 47] and chromosome segregation), other factors also contribute to modulating the elongation stage of DNA synthesis. However, because serine hydroxamate is also unable to trigger runaway DNA synthesis in ΔrelA cells, our results do not imply a role for HipA in directly regulating DNA replication through any mechanism outside either ppGpp synthesis or translation inhibition. We also note that ppGpp may also contribute to the translational inhibition observed in HipA-arrested cells, as ppGpp itself directly inhibits protein synthesis (15), and the residual growth observed in HipA-expressing ΔrelA cultures may indicate that EF-Tu phosphorylation by HipA alone is not sufficient to completely inhibit translation.

While we have not determined the precise mechanism by which HipA expression leads to ppGpp biosynthesis by RelA, we speculate that inhibiting formation of the EF-Tu·GTP·aminoacyl-tRNA ternary complex blocks the primary route by which aminoacyl-tRNA is delivered to the A site of the ribosome. This allows deacylated tRNA to bind to unoccupied A sites, a process that is typically prevented during growth by EF-Tu-catalyzed delivery of aminoacyl-tRNA, which is far more rapid (48). Binding of deacyl-tRNA to the ribosomal A site triggers ppGpp biosynthesis by the enzyme RelA (49).

The dependence of HipA toxicity on ppGpp synthesis explains the susceptibility of HipA-arrested cells to aminoglycoside antibiotics (50), as aminoglycosides also suppress ppGpp synthesis (51). However, not all toxins found in E. coli trigger ppGpp synthesis, which implies that not all persister cells rely upon ppGpp to induce and sustain growth arrest. For instance, overexpression of the RNase toxin RelE in E. coli leads to a decrease in the ppGpp concentration (52). The molecular mechanisms by which the other toxins provide protection against antibiotic treatments remain unclear (3).

Our demonstration that chloramphenicol sensitizes HipA-arrested cells to β-lactam antibiotics implied to us that the protective effect of persistence could be directly eliminated by chemically inhibiting ppGpp biosynthesis. This would be consistent with a recent report that generation of persister cells within biofilms relies on the stringent response (ppGpp synthesis) (53). Unfortunately, we found that treating persister cells (as distinct from cells arrested by artificial HipA expression) with a β-lactam antibiotic in combination with chloramphenicol did not result in more efficient killing. This may have been due to the following possibilities.

1. ppGpp levels in naturally occurring persister cells are maintained in a way that is not affected by chloramphenicol.

2. While HipA is a bona fide toxin, it may be that it does not, in fact, contribute to the development of persister cells, at least under the conditions tested.

3. TA modules are highly redundant (6). Removal of more than five toxin/antitoxin modules was required before a reduction in the number of persisters generated could be detected. If HipA generated a small fraction of the total number of persister cells under the conditions we tested, their elimination would be difficult to detect.

4. HipA activity triggers ppGpp, which may then induce the development of persistence through the activation of other toxins which do not rely upon the maintenance of high ppGpp for antibiotic tolerance. It was recently shown that overexpression of HipA triggers the transcriptional activation of other TA modules (54).

While progress has been made in establishing the mechanisms by which toxins trigger growth arrest, how naturally activated toxins lead to persistence remains unclear. It may be that true persister cells arise only via the overlapping activation of multiple toxins with different activities rather than with a single toxin. Further studies will be required to determine which, if any, of these possibilities is correct.

ppGpp has been previously linked with persistence by Gerdes and Maisonneuve (4). In their model, ppGpp activates toxin expression indirectly by inhibiting the enzyme Ppx, which catalyzes the hydrolysis of intracellular polyphosphate. Inhibition of Ppx is expected to cause the buildup of polyphosphate, a signal that allosterically activates the protease Lon. Because antitoxins are rapidly eliminated by Lon activity, toxin inhibition should be relieved. Our results demonstrate another link between persistence and ppGpp, indicating that toxin activity can use ppGpp to drive a global cellular response. Linking our results with the model proposed above, it is possible that HipA may activate other toxins in a cascade by initiating ppGpp synthesis. The activity of downstream toxins may be the basis of some of the stress-resistant phenotypes that are activated by ppGpp but that can be maintained without ppGpp, as alluded to above.

We found that artificial growth arrest induced by HipA leads to a suspended state of metabolism where carbon is still consumed but is not significantly directed to biomass accumulation, illustrating how cells continue to regulate their metabolism during a state of attenuated translation by relying on posttranslational mechanisms. Active metabolism of toxin-arrested cells has been observed previously, and has been used to direct the specific synthesis of proteins, using the RNase MazF (55). Here, we have shown that toxin-arrested cells can also render a culture resistant to phage. We acknowledge that toxin arrest would not completely eliminate the vulnerability of a production culture to phage attack, as the culture would still need to be grown to a high density, during which time it would be susceptible. However, once a high density of cells is established, toxin-arrested cultures could enable both phage- and antibiotic-resistant fermentation.