Subinhibitory Concentrations of Bacteriostatic Antibiotics Induce relA-Dependent and relA-Independent Tolerance to β-Lactams

ABSTRACT The nucleotide (p)ppGpp is a key regulator of bacterial metabolism, growth, stress tolerance, and virulence. During amino acid starvation, the Escherichia coli (p)ppGpp synthetase RelA is activated by deacylated tRNA in the ribosomal A-site. An increase in (p)ppGpp is believed to drive the formation of antibiotic-tolerant persister cells, prompting the development of strategies to inhibit (p)ppGpp synthesis. We show that in a biochemical system from purified E. coli components, the antibiotic thiostrepton efficiently inhibits RelA activation by the A-site tRNA. In bacterial cultures, the ribosomal inhibitors thiostrepton, chloramphenicol, and tetracycline all efficiently abolish accumulation of (p)ppGpp induced by the Ile-tRNA synthetase inhibitor mupirocin. This abolishment, however, does not reduce the persister level. In contrast, the combination of dihydrofolate reductase inhibitor trimethoprim with mupirocin, tetracycline, or chloramphenicol leads to ampicillin tolerance. The effect is independent of RelA functionality, specific to β-lactams, and not observed with the fluoroquinolone norfloxacin. These results refine our understanding of (p)ppGpp's role in antibiotic tolerance and persistence and demonstrate unexpected drug interactions that lead to tolerance to bactericidal antibiotics.


Preparation of thiostrepton stock solutions
Several crystals of thiostrepton (Tocris Bioscience, validated by mass spectrometry) dissolved in ≈300 μl of 100% DMSO (Sigma Aldrich) and the concentration (≈600 mM) is measured by absorbance at 280 nm with ε = 0.027 µM -1 cm -1 . DMSO stock is then used to prepare the working stock in 1x Hepes:polymix buffer (25 mM Hepes-KOH pH 7.5, 15 mM MgCl2, 0.5 mM CaCl, 95 mM KCl, 5 mM NH4Cl, 8 mM putrescine, 1 mM spermidine, 5 mM K3PO4 pH 7.3 and 1 mM DTT (1)) supplemented 0.1% Pluronic F-127 (Sigma Aldrich) as follows. Thiostrepton stock in DMSO was mixed with 20% (v/w) Pluronic F-127 in DMSO and 1x Hepes:polymix buffer was gradually added to final volume to achieve the final concentration of 12.5 µM. The final DMSO concentration should be kept below 3% and all the preparations should be done at room temperature to avoid precipitation. Working stock was briefly centrifuged (21,000 rcf, 5 min) at room temperature in order to remove any traces of precipitated thiostrepton, supernatant transferred into a new tube and the final concentration of thiostrepton re-measured using absorbance at 280 nm using 1x Hepes:polymix supplemented with 3% DMSO and 0.1% Pluronic F-127 as a blank reference.

Preparation of ppGpp
ppGpp was produced enzymatically using either RelSeq enzyme from Streptococcus equisimilis (2) or RelQ enzyme from Enterococcus faecalis (3) essentially as described in Mechold et al. (4). Reaction mixture containing 5 mM GDP, 10 mM ATP, 15 mM MgCl2, 30 mM Tris-HCl (pH 8.0), 100 mM NaCl in MilliQ H2O was preincubated for 5 min at 37 °C followed by the addition of 50 µM RelSeq or 5 µM RelQ. Since RelSeq is strongly inhibited by free Mg 2+ ions, therefore the total nucleotide concentration (GDP + ATP) should be equal to that of MgCl2; in the case of using RelQ free Mg 2+ is not a concern since due to RelQ's insensitivity. After 2 hours of incubation at 37 °C with constant shaking, nucleotides were extracted by the addition of acidic phenol (pH 4.5) (Amresco); directly after phenol addition to the reaction (1:1), the mixture was subjected to vigorous vortexing followed by centrifugation (16k rcf, 4 °C, 30 min) in order to separate the phases. To completely remove the traces of phenol this step was repeated. Upper phase was collected and applied onto MonoQ TM 5/50 GL (GE Healthcare) column on equilibrated in buffer A (2.5 mM Tris-HCl (pH 8.0), 0.5 mM EDTA and 0.5 mM LiCl in MilliQ H2O). Nucleotides were separated by gradient elution increasing concentration of LiCl up to 2 M. We have collected the ppGpp peak that eluts at ≈250 mM LiCl, and the nucelotide was precipitated by the addition of 1 M LiCl (f.c.) and 3 volumes of cold 96% EtOH followed incubation either at -20 °C (overnight) or at -80 °C (2 hours). The precipitate was vortexed, transferred to centrifugation tubes and pelleted at 16k rcf, 4 °C for 30 min. Supernatant was discarded, the pellet washed with -20 °C 70% EtOH in order to remove traces of LiCl and the pellet (ppGpp) dissolved in MilliQ H2O. The concentration of ppGpp was determined with extinction coefficient Ea = 13,600 μM -1 cm -1 at 252 nm.

Preparation of RelA and EF-G
C-terminally 6His-tagged RelA (5) was overexpressed in E. coli BL21(DE3) from pET28a plasmid; a colony from overnight grown on LB/Kanamycin agar plate was inoculated into 3 ml of LB media and grown shaking overday at 37 °C in the presence of 25 µg/ml Kn to OD600=0.5. After that the culture was transferred into 200 ml of fresh pre-warmed 2x YT media (with 25 µg/ml Kn) and grown shaking at 37 °C to OD600=0.5 followed by the induction of RelA overexpression by adding 1 mM IPTG (Sigma). Straightly after induction the temperature was decreased to 30 °C and bacteria was left to grow for 1.5 hours. Afterwards the bacterial cell mass was collected by centrifugation. For the lysis step bacteria were resuspended in a buffer, consisting of 25 mM Tris (pH 7.5), 2 mM MgCl2 and 1 mM 2-mercaptoethanol (βME) with the addition of 100 μM PMSF and 1 U/ml DNase I (Thermo Scientific). Cells were lysed with Stansted SPCH-10 homogenizer and the lysate centrifuged with following application for Ni-NTA affinity chromatography in GE Healthcare HisTrap HP column on ÄktaPrimePlus (GE Healthcare). The running buffer consisted of 25 mM Tris (pH 7.5-8), 1 M KCl, 350 mM NaCl, 2 mM MgCl2, 5 mM imidazole, 1 mM βME while in the elution buffer the concentration of imidazole was increased to 300 mM. RelAcorrespondent peak was collected and the protein concentrated using Amicon Ultra-15 (Millipore) centrifugal filter devices with a 50 kDa cut-off. The final preparations were aliquoted by 15 μl, shock-frozen with liquid nitrogen and stored at -80 °C in buffer containing 25 mM HEPES-KOH, pH 7.5, 0.7 M KCl, 2 mM DTT, 15 mM MgCl2 and 10% glycerol.
N-terminally 6His-tagged EF-G protein was overexpressed in E. coli BL21(DE3) from pQE30 plasmid provided by J. Remme (6). The overnight-grown culture was transferred into 200 ml of fresh 2x YT media (with 100 µg/ml Amp) and grown shaking at 37 °C to OD600=0.5 followed by the induction of EF-G overexpression by adding 1 mM IPTG (Sigma). After the induction the bacteria were left to grow shaking for 2 more hours at 37 °C with the cell mass being collected by centrifugation in the end. Cells were lysed with Stansted SPCH-10 homogenizer in buffer A (50 mM Tris-HCl pH 7.5, 10 mM MgCl2, 200 mM PMSF and DNase I) and clarified by centrifugation at 40k rpm in Ti 50 rotor for 40 min (Beckman & Coulter). Clarified lysate was loaded on HisTrap HP Ni-NTA affinity chromatography column (GE Healthcare) and the protein was eluted by a gradient of buffer B (A supplemented with 300 mM imidazole). EF-G was concentrated and stored -80 °C in buffer containing 25 mM HEPES-KOH, pH 7.5, 0.7 M KCl, 2 mM DTT, 15 mM MgCl2 and 10% glycerol.

Isolation of peptidoglycan and UPLC analysis
Peptidoglycan of E. coli was isolated as described previously (8), with minor modifications. E. coli strains were grown in MOPS media (9), supplemented with 0.4% glucose and 25 µg/ml of 20 amino acids. 600 ml of media was inoculated from 1 ml of overnight culture and grown at 37 °C with vigorous aeration until OD600 = 0.5, then the corresponding antibiotics were added for further incubation. Cells were harvested by 10 min at 5000 g at room temperature, washed with 10 ml PBS, harvested again and snap frozen in liquid nitrogen. Pellets were re-suspended in 3 ml of PBS, added to an equal volume of 10% SDS in a boiling water bath and vigorously stirred for 3 h, then stirred overnight at 37 °C. The insoluble fraction (peptidoglycan) was pelleted at 60000 rpm, 15 min, 20 °C using a TL100 Beckman rotor in an Optima MAX-TL ultracentrifuge (Beckman Coulter), and then washed 3 times with MQ-water. Samples were digested for 1 h at 60 °C with pronase E (100 µg/ml) in 10 mM Tris-HCl pH 7.5 and NaCl 0.06% (w/v), to remove Braun's lipoprotein. The reaction was heat-inactivated by adding SDS to a final concentration of 1% (w/v) and boiling for 10 min in a water bath. SDS was then removed by washing 3 times with MQ-water. Purified peptidoglycan was re-suspended in 100 µl of 50 mM NaPO4 buffer, pH 4.9 and digested overnight at 37 °C with 100 µg/ml muramidase. The digestion was stopped by 15 min incubation in a boiling water bath, and coagulated proteins were removed by 10 min centrifugation at 14,000 rpm. The supernatants, containing the digested muropeptides, were reduced by adding 15 µl 0.5 M sodium borate pH 9 and 5 µl 2M sodium borohydride and incubating at room temperature for 30 min. Finally, samples were adjusted to pH 3.5 with phosphoric acid.  (10) and confirmed genetically (11,12).
HPLC analysis of peptidoglycan composition revealed that LD crosslinks between DAP moieties in the third position are significantly enriched in both wild type and relaxed strains upon antibiotic pre-treatment, especially in the case of mupirocin and trimethoprim combination (Supplementary Figure 9AB). Remodeling the cell wall depends on both de novo synthesis of peptidoglycan and remodeling of the existing one. Therefore we performed a kinetic analysis of DAP-DAP accumulation in relaxed strain exposed to either to mupirocin alone or mupirocin together with trimethoprim. In the latter case the fraction of DAP-DAP crosslink increases more than six times from 1.8 % prior to antibiotic challenge to 11% after five and a half hours (Supplementary Figure 9C).
However, a significant increase in the DAP-DAP crosslink (two and a half times) is observed upon mupirocin challenge of relaxed strain without the concomitant increase in ampicillin tolerance (Supplementary Figure 9D), indicating that the relationship between peptidoglycan remodeling and ampicillin tolerance is unlikely to be direct. To directly disprove the causal connection between accumulation of LD crosslinks and antibiotic-induced ampicillin tolerance we took advantage of bactericidal β-lactam antibiotic imipenem which is a potent Ldt inhibitor (15): if accumulation of the LD crosslinks is, indeed, essential for E. coli survival upon PBP inhibition by ampicillin, then a combination of imipenem and ampicillin should result in bacterial lysis. However, trimethoprim and mupirocin showed the same relA-independent protective effect in killing assays employing simultaneous challenge by imipenem and ampicillin (Supplementary Figure 9E). This suggests that, first, the pretreatment leads to tolerance to βlactams in general rather than specifically to ampicillin, and, second, that accumulation of the LD crosslinks is not the cause of the effect. Finally, we used an E. coli strain lacking functional Ldt genes ycbB (ldtD) and ynhG (ldtE) and therefore unable to form DAP-DAP crosslinks (16). Pre-treatment with trimethoprim combined with either chloramphenicol or tetracycline protects this strain from ampicillin, directly confirming that LD crosslinks are not necessary for the effect (Supplementary Figure 9E).

Supplementary Figure 1 | Experimental strategy for bactericidal antibiotic killings.
At OD600 of 0.5 bacteriostatic antibiotics were added at concentrations that reduce the growth rate two times and incubated for additional 30 minutes. After that bactericidal antibiotic ampicillin was added to final concentration of 200 μg/ml and the time course of antibiotic killing was followed by determining the surviving fraction using LB plating and colony count. The slower killing phase or plateau is defined as persistence, and slower killing of the bacterial population as a whole is defined as antibiotic tolerance. If not stated otherwise, the experiments were performed at 37°C in MOPS media supplemented with 0.4% glucose and 25 μg/ml amino acids using BW25113 E. coli strains. The antibiotics were used at concentrations causing growth rate reduction of two times. The growth experiments were performed at 37°C in MOPS medium supplemented with 0.4% glucose and 25 μg/ml of each amino acid using BSB1 B. subtilis wild type strain. Cells were pre-grown to OD600 of 0.5 prior to addition of mupirocin and the consequent growth was followed. Antibiotic concentrations in nM are indicated on the figures, e.g. mup70 corresponds to 70 nM mupirocin. The growth experiments were performed at 37°C in MOPS medium supplemented with 0.4% glucose and 25 μg/ml of each amino acid using BW25113 E. coli wild-type strain (A and B) and an isogenic relA knock-out (ΔrelA, C and D). Cells were pre-grown to OD600=0.5, antibiotics (mup = mupirocin, cam = chloramphenicol, tet = tetracycline, trim = trimethoprim, + designates antibiotic combinations) were added, and the consequent growth was followed. Antibiotic concentrations in μM are indicated on the figures, e.g. mup70 + tet2 designates treatment with 70 μM mupirocin and 2 μM tetracycline. Experiments were performed at least three times and error bars indicate standard error of the mean.  The antibiotics were used at concentrations causing complete inhibition of growth and maximum reduction in the ppGpp levels (20 μM chloramphenicol, 2 μM tetracycline and 16 μM trimethoprim), except for mupirocin (70 μM), which was used at a concentration causing a two-fold reduction of the growth rate and serves as a comparison for the results on Figures 6A and 6B. Antibiotic concentrations in μM are indicated on the figures, e.g. mup70 designates pretreatment with 70 μM mupirocin. The antibiotic pre-treatment was performed for 30 minutes at 37°C in MOPS medium supplemented with 0.4% glucose and 25 μg/ml of each amino acid, followed by addition of ampicillin to final concentration of 200 μg/ ml. The surviving fraction was determined by LB plating and colony count. Experiments were performed at least three times and error bars indicate standard error of the mean. To probe the role of DAP-DAP crosslinked dipeptide in ampicillin resistance induced by mupirocin combined with trimethoprim, ampicillin (AMP, 200 μg/ml) killing was followed in conditions when DAP-DAP formation was abrogated either chemically by addition of saturating concentrations of bactericidal antibiotic Ldt inhibitor imipenem (IMP, 4 μg/ml) (D) or genetically by using ΔldtE ΔldtD E. coli strain (E). Antibiotics concentrations in μM are indicated on the figures, e.g. mup70 designates pretreatment with 70 μM mupirocin. Error bars indicate standard deviation of three independent experiments. As indicated by brackets, the P-values were calculated using two-tailed Welch's t-test in relationship to killing time courses of untreated culture.