Peptoids successfully inhibit the growth of gram negative E. coli causing substantial membrane damage

Peptoids are an alternative approach to antimicrobial peptides that offer higher stability towards enzymatic degradation. It is essential when developing new types of peptoids, that mimic the function of antimicrobial peptides, to understand their mechanism of action. Few studies on the specific mechanism of action of antimicrobial peptoids have been described in the literature, despite the plethora of studies on the mode of action of antimicrobial peptides. Here, we investigate the mechanism of action of two short cationic peptoids, rich in lysine and tryptophan side chain functionalities. We demonstrate that both peptoids are able to cause loss of viability in E. coli susceptible cells at their MIC (16–32 μg/ml) concentrations. Dye leakage assays demonstrate slow and low membrane permeabilization for peptoid 1, that is still higher for lipid compositions mimicking bacterial membranes than lipid compositions containing Cholesterol. At concentrations of 4 × MIC (64–128 μg/ml), pore formation, leakage of cytoplasmic content and filamentation were the most commonly observed morphological changes seen by SEM in E. coli treated with both peptoids. Flow cytometry data supports the increase of cell size as observed in the quantification analysis from the SEM images and suggests overall decrease of DNA per cell mass over time.


Results and Discussion
In Vitro Antibacterial Activity. We have previously reported the broad spectrum activity of a library of cationic tryptophan rich peptoids against six bacterial strains and compared those activities with the ones of similar antimicrobial peptides 12,16 . In addition, the hemolytic activity and cytotoxicity against HeLa cells were analysed, altogether demonstrating good selectivity profiles for these peptoids. Two peptoids showed good selectivity index values with respect to their activity against E. coli and their haemolytic values and were therefore chosen for further studies (Table 1). Both peptoids have relatively short sequences (9 residues, Fig. 1) and exhibit MIC ( ≤ 22.2 μ M) values that are comparable to that of the antimicrobial peptide indolicidin (Table 1) and are within the range of other antimicrobial peptoids reported in the literature 8,17 . E. coli killing kinetics. In order to differentiate between bacteriostatic and bactericidal mode of action of peptoids 1 and 2, E. coli cultures were challenged with different concentrations in a period of 6 hours and growth The selectivity index is the quotient of 10% hemolysis and the lowest and highest MIC of the bacterial strains. d Cytotoxicity is reported as IC 50 , designated as concentration found to inhibit the metabolic activity of HeLa WT cells using the colorimetric tetrazolium salt based MTS assay. e Extended spectrum beta-lactamase. f multidrug resistant P. aeruginosa. inhibition patterns were compared. Data supports the inhibition capacity of both peptoids which show clear concentration dependent inhibition of E. coli. Both Peptoids 1 and 2 at 1 × MIC concentrations exhibit bacteriostatic mode of action on E. coli growth ( Fig. 2A,B). The slow killing for Peptoids 1 and 2 may indicate that at MIC concentration these peptoids target intracellular metabolic processes that cause bacterial growth inhibition. With respect to this observation, delayed killing kinetics have been observed for AMPs that target intracellular compartments 18,19 . These could be inhibition of DNA synthesis or cell filamentation as observed for bacteriostatic antimicrobial compounds 20 . At 2 × MIC concentrations different killing kinetics are observed, where Peptoid 2 acts via bactericidal mechanism ( Fig. 2A). The observed difference may be partly due to the different hydrophobicity profiles of these peptoids measured by retention times on HPLC. Here, Peptoid 2 is globally more hydrophobic and this property could enhance its insertion into bacterial membranes. Similar killing kinetic profiles as Peptoid 2 have been previously observed for other antimicrobial peptidomimetics 13,[21][22][23] . At 4 × MIC concentrations, both peptoids show significant bactericidal behaviour which is marginally higher than that shown for ciprofloxacin, an antibiotic that prevents DNA replication, recombination and repair ( Fig. 2B) 24 . The fast killing kinetics is still not higher than that reported in the literature for polymyxin B, a fast permeabilizing antimicrobial peptide, where upon reaching a threshold concentration, the accumulated peptide molecules of polymyxin B can rapidly cause bacterial death 25 .
Quantification of membrane permeabilization. The mechanism of action of antimicrobial peptides and related mimetics that carry net positive charge and moderate hydrophobicity on bacteria is generally associated with fast killing kinetics as a result of membrane permeabilization 18,23,[26][27][28] . To investigate whether the bactericidal effect observed from the killing kinetics experiments is related to the ability of the peptoids to disrupt bacterial membranes, we calculated the percentile of viable and non-viable cells using Live/Dead quantification assay. In this assay two nucleic acid dyes, syto 9 green and propidium iodide (PI) red, that have different membrane permeabilities were used as indicators of live bacteria with intact membranes versus dead bacteria with compromised membranes, respectively. Peptoid 1 caused gradual decrease of the number of PI stained cells at 1 × MIC concentration. In response to Peptoid 2 treatment, 90 percent of E. coli showed high membrane permeabilization within one hour of treatment, that supports the observation for the higher potency exhibited by this peptoid from the killing kinetics experiments (Fig. 3). Importantly, the activity of both peptoids remains stable after 5 hours of incubation. This is in contrast to the activity observed for the peptide from which these peptoids are derived, where re-established growth is observed within 2 hours of peptide administration 16 .
Release of fluorescent dyes from different liposome compositions. Two fluorescent dye leakage assays were used to assess the disruptive ability of peptoids. The two assays use similar fluorescent dyes, carboxyfluorescein (CF) and calcein to measure the leakage from the liposomes. These two assays offer information on the electrostatic interaction with membranes mimicking bacterial surfaces and interaction with membranes with different fluidities. Bacterial membranes have a highly negative transmembrane potential (approximately − 120 mV) in contrast to the weak membrane potential maintained by mammalian cells 29 . Thus they attract positively charged compounds such as cationic peptoids. Therefore, to test the existence of a specific interaction between Peptoid 1 and the glycerol polar headgroup of palmitoyl-2-oleoyl-sn-glycero− 3-phospho-(1′ -rac-glycerol) (POPG) or, if the interactions are purely electrostatic, carboxyfluorescein (CF) leakage assays from liposomes of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) containing 20 mol % of 1-hexadecanoyl-2-(9Z-octadecenoyl)-sn-glycero-3-phosphate (POPA) or POPG were performed initially. Both POPA and POPG are negatively charged phospholipids but with different polar headgroup structures; while PA phospholipid only has a phosphate group linked to C3 of the glycerol moiety, PG has a phosphoglycerol group at this same position. Liposomes containing 20 mol % of negatively charged POPA showed to be more susceptible to Peptoid 1 when compared to liposomes with 20 mol % of POPG (Fig. 4). This indicates that there is no specificity in peptoid binding to PG. Again, the greater extent of dye release from liposomes with 40 mol % of POPG compared to 20 mol % of POPG shows the importance of electrostatic interactions between the peptoid and liposomes and that its activity increases with the increase of overall net surface charge (Fig. 4). Furthermore, besides electrostatic interactions, hydrophobic side chain functionalities contribute greatly to the peptoid-membrane interactions. Tryptophan-like residues have strong membrane disruptive activities due to their ability to interact with the membrane interface. This activity is assigned partly due to the flat and rigid structure of the indole ring that may favour alignment of the side-chain in the membrane interface region 4 . To evaluate the permeability of the peptoids to membranes, we used the calcein leakage assay where calcein was encapsulated in two different combinations of selected lipids POPC:POPG (7:3) and POPC, POPG and Cholesterol (5:2:3). The membranes of mammalian cells have sterol molecules such as Cholesterol in their membrane compositions and this property can prevent peptoid insertion in the lipid bilayers. We used membrane composition containing POPC, POPG and Cholesterol (5:2:3), as reported previously, to mimic membrane environment with decreased fluidity while retaining negative charge to ensure peptoid binding 16,30 . In addition, the liposomes composed of POPC, POPG and Cholesterol also mimic membrane compositions of some of the membranes of tumour cells  and neurons, even though POPG is not found as a part of the composition of mammalian membranes 31 . The total amount of released calcein was monitored and the degree of membrane permeabilization of the peptoids was calculated using Equation 2. Distinct tendency to lyse liposomes mimicking bacterial membranes (POPC:POPG) over liposomes mimicking less fluidic membranes, was observed for Peptoid 1, in contrast to the indiscriminative membrane permeabilization activity of the control peptide melittin that acts via the toroidal pore formation mechanism (Fig. 5A,B) 32 . Membrane permeabilzation is expected for the peptoids in this study because of their structural compositions. Short tryptophan rich cationic peptides have been reported to exhibit high membrane activity on E. coli where the peptide with the highest tryptophan content (2 W) and charge (+ 9) showed highest percentage of membrane permeabilization (approximately 60%) 33 . Similarly, the peptoids in this study are rich in charged (+ 4) and tryptophan-like residues (4 W) which are expected to contribute significantly to the observed membrane permeabilizing properties. The permeability of peptoid 1 was greater for liposome compositions that mimic bacterial membranes (Fig. 5). Maximum calcein leakage (< 30%) was observed from sub MIC concentrations of Peptoid 1 and an increase in the peptoid concentration from 6.25 to 25 μ g/ml did not contribute to an increase in the calcein release from these liposomes (Fig. 5B), thus indicating that Peptoid 1 only has a weak membrane activity. The membrane activity is also much lower than the one observed for structurally similar GN-4 antimicrobial peptide we have published previously 16 . The unique physiochemical properties of peptoids such as loss of hydrogen donation and backbone chirality may be expected to impact the membrane interactive properties when compared with peptides. However, not all antimicrobial peptides insert and disrupt the membranes or form pores. Studies on mechanism of action of cationic tryptophan rich peptides have shown that they either primarily lyse the membrane of Gram-negative bacteria by pore formation or they translocate and act on internal targets 34 . The overall total thickness of the membrane of Gram-negative bacteria varies from 30-50 nm. In order for a peptide or peptidomimetic to form a pore in the membrane, it has to be long enough to span the membrane e.g. barrel-stave model. Some peptides are too short to span the membranes 35,36 . They usually induce pore formation via a toroidal model by stabilizing the lipid pore and forming a peptide-lipid complex 37 . Indolicidin, a short (13 residue), tryptophan rich, extended structure cationic peptide, is able to cause leakage of calcein and carboxyfluorescein from negatively charged liposomes via non-membrane pore formation mechanism, but rather via translocation of dye molecules across the membrane in the form of dye-peptide complexes 35 .
Concentration dependent calcein release was observed from liposomes containing Cholesterol, but with a significantly slower release rate and a lower threshold, which is in agreement with the low toxicity reported for Peptoid 1 at concentration as high as 4 × MIC. Calcein leakage in a concentration dependent manner is also observed for antimicrobial peptides 38 . This concentration dependent activity of antimicrobial peptides has been explained by a two-state model 37 . This model illustrates initial binding of peptide monomers parallel to the plane of the membrane. The peptides are then dispersed all across the membrane without causing pore formation. As the concentration of the peptide increases, thinning and formation of transmembrane pores takes place 37 . Similar membrane thinning has been reported for indolicidin 35,39 . Our findings suggest that the release of calcein upon addition of Peptoid 1 is concentration dependent and significantly lower than the total calcein release observed for melittin (Fig. 5A,B) 40 . The total dye leakage together with the results on the bacterial viability quantification at concentrations around 1× and 2 × MIC of Peptoid 1, are in agreement with the killing kinetics experiments indicating that the growth inhibition is partially exerted by some degree of membrane damage that does not include toroidal pore formation.
Changes in Membrane Morphology of bacteria upon peptoid treatment. E. coli were inspected using scanning electron microscopy to visualize any morphological changes (membrane disruption, cell elongation, etc.) after peptoid treatment. Untreated bacterial cells appear with smooth surfaces whereas considerable roughening and leakage of cytoplasmic content is observed for peptoid treated bacteria (Figs 6 and 7). The SEM micrographs show membrane damage upon treatment with both peptoids at 1× and 4 × MIC, with a more pronounced effect at 4 hours (Figs 6 and 7). These cellular changes are related to the growth inhibitory activities observed at 1× and 4 × MIC concentrations. Membrane disintegration is expected for these peptoids due to the presence of four tryptophan like side chains which are strongly associated with high affinity for the interfacial region of bilayers 41 . Similar changes in bacterial surfaces, such as membrane damage and blebs have been observed both for antimicrobial peptides and for metal oxide nanoparticles with negative surface potential as a result of their strong interaction with P. aeruginosa and E. coli membranes, respectively 5,42 . Bleb morphology in E. coli has also been reported for the human antimicrobial peptide defensin 5, where the cause of bleb formation was localization of this peptide at the site of cell division and site poles indicating that the antibacterial activity is exerted in the cytoplasm 43 . Additionally, E. coli treated with the cationic antimicrobial peptide, gramicidin, illustrated a high degree of blister and dimple formation 44 . Filamentation, a continuous cell elongation which does not result in cell division, is often observed in bacteria as a result of various stress responses such as those to beta lactams and fluoroquinolones antibiotics 45,46 . Bacterial filamentation has been reported as a consequence of induction of SOS response mechanism which is responsible for regulation of DNA damage repair 47,48 . Elongated  (Figs 6E and 7B). In addition, from the quantitative data, the majority of the E. coli with cell length of 3 μ m were those treated with 1 × MIC concentration of peptoid 2 (Fig. 8). Antimicrobial compounds that act via bacteriostatic mechanisms (e.g. inhibition of protein and RNA synthesis) have been shown not to induce filamentation or increase cell size 49 . Therefore the increase in cell size in bacteria treated with peptoids could be due to possible inhibition of DNA processes inside the cell. These results indicate that peptoids can destabilize the membrane of E. coli causing pronounced damage that does not always lead to complete lysis and cell death, but to cell elongation.
Changes in size and DNA content. We used flow cytometry to measure the changes in cell size and nucleic acid content reflected by increased light scatter and fluorescence, respectively. Exponentially growing cells were challenged with both peptoids and ciprofloxacin and stained with propidium iodide (PI) cell impermeant dye. PI stains nucleic acids in cells with damaged membranes. Untreated cells retain constant DNA per cell size (Fig. 9). Bacteria treated with ciprofloxacin appear significantly bigger with overall decrease of DNA per cell size. Certain quinolone antibiotics such as ciprofloxacin or norfloxacin interact with DNA gyrase and topoisomerase IV causing DNA replication fork arrest and induction of SOS response in E. coli. This leads to inhibition of cell division and filamentation [50][51][52] . Ciprofloxacin is also shown not to affect the membrane integrity of E. coli even at high concentrations 46,53 . The bactericidal effect of ciprofloxacin demonstrated in the killing kinetics experiments is in agreement with the decrease of DNA per cell mass obtained from the flow cytometry measurements. This effect is thought to be related to the release of free DNA ends from DNA gyrase-ciprofloxacin complex that ultimately leads to chromosomal DNA fragmentation 20,54 . The fluorescence signal from PI binding to nucleic acids in the cells may therefore indicate a different pathway of membrane disturbance to be involved in making the bacteria permeable to the dye 55 . Investigation into the effect of the peptoids on the cell size and DNA content demonstrated that cell size increased considerably with an increase in concentration for both peptoids. The overall effect of the peptoids was a substantial decrease in DNA per cell mass, which was concentration dependent and it occurred at faster rate for peptoid 2 when compared to that of ciprofloxacin. A similar dose dependent interaction between peptoid and bacterial plasmid DNA was confirmed using a gel retardation assay (Fig. 2, see Supplementary Material). At 1 × MIC the inhibition pattern was somewhat less pronounced for the peptoids in regards to ciprofloxacin demonstrating a possible distinctive antibacterial effect.  Conclusions Many antimicrobial peptides have been referred to as "dirty drugs" due to their ability to act via multiple mechanisms which involve membrane permeabilization and inhibition of intracellular processes 36,56 . While dissecting the mode of action of lysine and tryptophan rich peptoids that mimic the structure of antimicrobial peptides, we have demonstrated that the growth inhibition ability of peptoids that show different killing kinetics is most likely based on the different charge density distribution and overall hydrophobicities. We show that the killing mechanism for peptoid 1 is not fully attributed to its ability to cause leakage from the membranes as observed in the dye leakage assays. However, the killing mechanism is definitely supported by some degree of membrane damage exerted by this peptoid. Higher degree of bacterial lysis through membrane damage is observed for peptoid 2. The peptoids exhibited potent killing kinetics which are not explained solely by their membrane permeabilization ability, measured by calcein leakage from lipid vesicles mimicking bacterial membranes. However, the high degree of membrane disturbance measured by Live/Dead assay is not directly translated into a decrease of viable cells. The membrane disruptive ability is also supported by scanning electron micrographs of E. coli treated with both peptoids. Although the mechanism behind the observed increase in size is not clear, it can be reasonably assumed that besides the peptoids effect on the outer membrane they enter the cytoplasm and bind to internal targets thus inhibiting DNA, RNA or protein synthesis. Taken together we propose that these peptoids act via combined mechanism of action involving membrane disruption with probable intracellular targets. Further investigations are needed to support their effects via intracellular targets. The peptoids in this study have low toxicity, high stability and low cost of synthesis, properties which contribute to their high potential as new peptidomimetics with therapeutic applications.

Experimental Section
Materials and Bacterial strains. The strains used in the present study were from our laboratory strain collection. Escherichia coli (ATCC 25922) was obtained from the American Type Culture Collection (ATCC, Rockville, Md.). Other strains include Pseudomonas aeruginosa PAO1 H103 strain and Liverpool epidemic clinical strain H1027, and Escherichia coli clinical isolate expressing extended spectrum β -lactamases (ESBL) 57 . All tested bacterial strains are categorized as biohazard level 2 pathogens. Peptoid synthesis. Peptoids were synthesized using standard submonomer solid-phase synthesis, as described previously 12 . In vitro susceptibility studies. Minimum inhibitory concentrations (MIC) for the peptoids have been measured as described previously 12 . Killing kinetics. The kinetics of antimicrobial activity against E. coli (ATCC 25922) were assessed at a peptoid concentration corresponding to 1× , 2× and 4 × MIC. Briefly, an overnight culture of bacteria was diluted 1:50 in fresh Mueller Hinton broth and regrown to an OD 600 of 0.4 before diluting to a turbidity of 0.1. The bacterial suspension was added to a 96-well polystyrene flat bottomed plate containing the peptoid of interest in addition to known antibiotics. The plate was incubated without shaking at 37 °C for 180 minutes. Samples (100 μ l) were taken at time 20, 40, 80, 120 and 180 minutes and diluted in ice cold 0.9% NaCl from which 100 μ l was plated on LB agar plates. The plates were incubated for 18-24 hours at 30 °C and colony forming units (CFU) were counted. Sterility control was performed by plating 100 μ l of MH broth and 0.9% NaCl. Plates where no detectable bacterial growth was observed were left for additional 18 hrs incubation.
Live/Dead Staining. Overnight culture of E. coli ATCC 25922 was diluted 100-fold in fresh MH broth and allowed to grow until it reached OD 600 of 0.1. Duplicate samples of 1 ml each were transferred to eppendorf tubes and pelleted at 10.000 g for 8 minutes. One sample was resuspended in 1 ml 0.9% NaCl and the other in 50 μ l 0.9% NaCl followed by addition of 950 μ l 70% isopropyl alcohol to kill the bacterial. The tubes were left for 1 hour on ice with manual shaking every 15 minutes. The tube containing dead bacteria was pelleted at 12.000 rpm for 8 min and resuspended with 1 ml 0.9% NaCl to remove the isopropyl alcohol. Five different ratios corresponding to 0, 10, 50, 90 and 100% of live bacteria were prepared to obtain the standard curve for the analysis of bacterial Figure 9. Flow cytometry data of exponentially growing E. coli exposed to Peptoids 1 and 2 and ciprofloxacin at concentrations corresponding to 1× and 4 × MIC. Fluorescence corresponds to cellular DNA content and lightscatter is equivalent to the bacterial cell size, thus fluorescence/lightscatter is a measure of the DNA content per cell mass. Results are from 3 independent experiments with SEM error bars. viability when challenged by antimicrobials (Fig. 1, see Supplementary Material). Briefly, 100 μ l of the staining solution (3 μ l of 3.34 mM SYTO 9 and 3 μ l of 20 mM propidium iodide in water) was added to 100 μ l of the different bacteria ratios containing wells and incubated in dark for 15 minutes before reading on multi-detection macroplate reader Synergy HT. The green fluorescence (SYTO 9) corresponds to the amount of live bacteria and was excited at 485 nm and the emission detected at 528 nm. The red fluorescence (propidium iodide) was excited at 530 nm and detected at 645 nm which correspond to the amount of dead bacteria in the sample. Bacteria suspension challenged with antimicrobial for viability analysis were prepared by incubating 90 μ l of bacteria (OD 600 of 0.1) and 10 μ l antimicrobials for 5 hours during which period samples were extracted in PCR tubes at 0, 10, 20, 40, 80 and 120 minutes and immediately put on ice. All PCR tubes were pelleted at 10.000 g for 8 minutes before being resuspended in cold 0.9% NaCl and placed on ice again. The content of each PCR tube was then added to individual wells of flat bottomed polystyrene 96 microtitter plate and mixed with 100 μ l staining solution. After 15 minutes incubation in dark the fluorescence was measured using multi-detection microplate reader Synegry HT. The percentage of living cells was calculated using Equation 1 below. Liposome quantification. Quantification of lipid content was performed using standard protocols described previously 58 . Initially, standard curve was obtained using 0, 20, 40, 60, 80, and 100 nmols of standard potassium phosphate solution (1 mM). Briefly, the tubes containing 0, 20, 40, 60, 80 and 100 μ l of potassium phosphate buffer were dried in a heating block at 120 °C and 400 μ l of perchloric acid (HClO 4 ) was added. For determination of the lipid content of the prepared liposomes, 50 nmols were dried and 400 μ l of perchloric acid was added. Digestion of the lipids was achieved by heating the suspension at 180 °C for up to 2 hours. After all samples have been cooled to room temperature, 1 ml of MilliQ water is added, followed by addition of 400 μ l of ammonium molybdate (1.252 g/100 ml). All the tubes are vortexed and 400 μ l of freshly prepared ascorbic acid (3% w/v) is added. The tubes are put to boiling water for 10 minutes and after they cool down, 1 ml is transferred to a 1 cm cuvettes and absorption is read using standard spectrophotometer at 797 nm. Using the standard curve equation, the concentration of liposome suspension was calculated before use.
Calcein release assay. Calcein release was done in a 96-well plate with shielded wells (MicroWell 96 optical bottom plate, NUNC, Roskilde, DK) as previously described 16 . Briefly, peptoids were diluted in 10 mM HEPES buffer and 100 μ l were added to wells followed by addition of 80 μ l of 45 μ M liposomes suspension for a final liposome concentration of 20 μ M. Immediately after addition the fluorescence was read using multi-detection microplate reader Synergy HT at an excitation wavelength of 485 nm and an emission wavelength of 520 nm over a course of 1 hour at 37 °C. Maximum calcein release was acquired using 10% Triton X-100 and release following peptoid exposure was calculated using equation 2, where F and F o represent the initial and the final levels of fluorescence before and after peptoid addition, respectively, and F max is the fluorescence level after complete disruption of liposome by addition of detergent, 10% Triton X-100. After the acetone step, the sample was further dried in a Leica EM CPD300 onto a square piece of silicon wafer. The silicon substrate with the dried sample was attached onto an aluminium stub with a double sided C tape, and the sample was coated with 2 nm Pt in a Cressington 208HR High Resolution Sputter Coater. The sample was then imaged in an FEI Helios dual beam scanning electron microscope by monitoring the secondary electron signal at 2 keV and 43 pA with the through the lens detector. The size analysis was performed using the NIH public domain Image J.
Flow cytometry. Overnight culture of Escherichia coli ATCC 2592 was diluted 1:50 in fresh MH broth and allowed to grow until an OD 600 of 0.1. This suspension is further 10-fold diluted and regrown till OD 600 of 0.1 to ensure a uniform bacterial population. A volume of 90 μ l of bacterial suspension was loaded to a flat bottomed 96-well Greiner plate. After extraction at zero minutes, 10 μ l of the peptoid and antibiotics corresponding to 1× and 4 × MIC concentrations was loaded to the bacteria containing wells for a total volume of 100 μ l. Samples were collected at 20, 40, 80,120 and 180 minutes and put on ice. The plate was sealed and incubated at 37 °C between sample extractions. Samples from wells containing bacteria without peptoids or antibiotics were taken at each time point. The samples were, whenever appropriate, centrifuged at 10.000 × g for 5 min at 4 °C. The supernatant was carefully removed and the bacteria were resuspended in 100 μ l 10 mM Tris HCl pH 7.4. 1000 μ l of ice cold 77% Ethanol was added before storing the samples at 4 °C until analysis. Rifampicin/Cephalexin samples were prepared by collecting 200 μ l from the wells containing bacteria and the appropriate antimicrobial in an E-tube containing 45 μ l of a mixture of Rifampicin (300 μ g/ml) and Cephalexin (36 μ g/ml). The tube was placed in a 37 °C shaking water bath for 2½ hours after which the samples were treated in the same way as the other samples before storing them at 4 °C. To monitor the bacterial growth the plate was read before each sample was extracted in a multi-detection plate reader Synergy HT. The following day, the samples were centrifuged at 10.000 × g for 5 min and the supernatant was carefully removed. The samples were stained for flow cytometry analysis with 140 μ l of staining solution (90 μ g/ml Mitramycin and 20 μ g/ml Ethidium Bromide in 10 mM Tris pH 7.4, 10 mM MgCl 2 ). For flow cytometry analysis A10 Bryte Flow Cytometer was used and around 20.000 events were included.