The antimicrobial polymer PHMB enters cells and selectively condenses bacterial chromosomes

To combat infection and antimicrobial resistance, it is helpful to elucidate drug mechanism(s) of action. Here we examined how the widely used antimicrobial polyhexamethylene biguanide (PHMB) kills bacteria selectively over host cells. Contrary to the accepted model of microbial membrane disruption by PHMB, we observed cell entry into a range of bacterial species, and treated bacteria displayed cell division arrest and chromosome condensation, suggesting DNA binding as an alternative antimicrobial mechanism. A DNA-level mechanism was confirmed by observations that PHMB formed nanoparticles when mixed with isolated bacterial chromosomal DNA and its effects on growth were suppressed by pairwise combination with the DNA binding ligand Hoechst 33258. PHMB also entered mammalian cells, but was trapped within endosomes and excluded from nuclei. Therefore, PHMB displays differential access to bacterial and mammalian cellular DNA and selectively binds and condenses bacterial chromosomes. Because acquired resistance to PHMB has not been reported, selective chromosome condensation provides an unanticipated paradigm for antimicrobial action that may not succumb to resistance.

Scientific RepoRts | 6:23121 | DOI: 10.1038/srep23121 capacity for both electrostatic and H-bonding interactions 14 , which could occur at many possible targets in cells. For example, PHMB binding to nucleic acid has been demonstrated in vitro 15 , raising at least one possible alternative mechanism of action.
To re-examine its mechanism(s) of action, we subjected PHMB to cellular, molecular and biophysical analysis, using both bacterial and mammalian cell systems. We examined its membrane activities and cellular effects using a PHMB-fluorophore conjugate together with cell growth, microscopy and physiological assays. Also, we examined PHMB/nucleic acid interactions using biophysical methods. Surprisingly, the results reveal that PHMB enters both bacterial and mammalian cells, condenses bacterial chromosomes and is excluded from mammalian nuclei. The outcome suggests a new model to explain its selective antimicrobial activities.

Results
Bacterial cell membrane activities of PHMB. If the antibacterial activity of PHMB (Fig. 1a) is due to membrane disruption, as widely reported [6][7][8][9][10] , it would be expected to permeabilise bacterial cell barriers at growth inhibitory and sub growth inhibitory concentrations. To test this model, we first established the minimal inhibitory concentrations (MIC) and time kill properties for PHMB against Escherichia coli (strains K-12 and MG1655) and Salmonella enterica serovar Typhimurium (strain LT2). As reported previously 2,4 , PHMB displayed potent growth inhibitory and cidal properties (Supplementary Table 2 and Supplementary Fig. 1). Also, following treatment, we examined cells using light microscopy. Unexpectedly, growth inhibitory concentrations of PHMB did not lyse cells, as monitored by bright-field microscopy. To assess cell barrier damage that could be invisible to microscopy, E. coli K-12 cultures were grown to mid-log phase, treated with PHMB in the presence of the fluorescent membrane integrity probe SYTOX ® Green, and then monitored using fluorimetry. This probe is useful as an indicator of membrane damage; because it is normally excluded from intact bacteria and its fluorescence quantum yield increases upon DNA binding. Therefore, intact bacteria are expected to display low fluorescence, and fluorescence is expected to increase following cell barrier damage 16 . As anticipated, freshly grown E. coli cultures displayed large increases in fluorescence following treatment with the known cell wall disruptor polymyxin B or heat treatment (Fig. 1b). Unexpectedly, PHMB treatment resulted in comparatively lower levels of fluorescence. Most strikingly, higher concentrations of PHMB resulted in fluorescence at background levels. These observations are not compatible with membrane disruption as the main antibacterial mechanism, and therefore raised further doubt about the established model.

PHMB enters bacteria.
If PHMB's primary target is not bacterial cell barriers, or not exclusively cell barriers, then it likely acts internally, and this would require cell entry. To test for bacterial entry, we synthesised a PHMB-FITC conjugate ( Supplementary Fig. 2a,b) and assessed uptake into Gram-positive (Staphylococcus aureus), Gram-negative (Escherichia coli and Salmonella enterica serovar Typhimurium) and acid-fast (Mycobacterium smegmatis) bacteria using microscopy and flow cytometry (Fig. 1c, Supplementary Fig. 3). Strong cell-associated green fluorescence was observed in all species tested (Fig. 1c). To examine cell localisation more thoroughly, the large-sized bacterium Bacillus megaterium was treated with PHMB-FITC, counter-stained with membrane localizing wheat germ agglutinin (WGA-red), and examined by fluorescence microscopy (Fig. 1d). Cell entry was observed in both live and fixed cells, and a fluorescence intensity profile analysis shows that PHMB-FITC localized within the cytoplasm, without accumulation at the cell barrier (Fig. 1e).
The observation that PHMB enters cells at low microgram/mL concentrations suggests that it may enter live cells. To investigate whether PHMB uptake into bacteria requires energy metabolism, mid-log phase E. coli cultures were incubated at 37 °C, or at 4 °C for 2 hours to reduce cellular ATP levels. Subsequently, cells were treated with PHMB-FITC (0-6 μg/ml) and further incubated on ice for 2 hours. Cell-associated PHMB-FITC fluorescence was quantified by fluorimetry ( Supplementary Fig. 3b). Cells held at 4 °C displayed reduced PHMB-FITC uptake relative to cells incubated at 37 °C, consistent with an energy dependent cell uptake process. Also, green fluorescent and motile bacteria were observed at several time points during PHMB treatment ( Supplementary Fig. 3). Because bacterial motility is energy dependent 17 , the evidence indicates that PHMB-FITC enters metabolically active cells. Therefore, PHMB enters diverse bacteria and entry was observed in motile cells.

PHMB arrests cell division and condenses bacterial chromosomes.
When examining E. coli by microscopy, we noted that PHMB treated cells often displayed an elongated morphology, which can be characteristic of cell division inhibition (Fig. 2a). To measure the effects of PHMB on cell elongation, we titrated PHMB into growing cultures of E. coli strain SS996 (vide infra) and measured cell lengths. At growth inhibitory concentrations, more than 80% of cells elongated ( Fig. 2b; Supplementary Table 2). Also, we observed that E. coli treated with growth inhibitory concentrations of PHMB or PHMB-FITC followed by DAPI-staining displayed blue fluorescent foci near the cell centre (Fig. 2c). These structures resembled nucleoids 18 . To ease visualisation of the DNA foci, we generated filamentous/multinucleated populations of E. coli by inhibiting cell division through RNA silencing of the essential cell division gene ftsZ 19 . RNA silencing was selected for this experiment because it enables specific and controllable repression of translation of essential gene transcripts 20 . Genes that are growth essential cannot be knocked-out by genome disruption methods, as this would result in non-viable strains. In the absence of PHMB, filamentous cells showed uniform DAPI staining, whereas PHMB treated cells displayed blue "strings of beads" (Fig. 2d). Similarly, in the large-sized Gram-positive bacteria B. megaterium we observed DAPI-stained foci following PHMB treatment (Fig. 2e). These results, in both Gram-negative and Gram-positive species, show that PHMB exposure leads to condensed chromosomes within bacteria.
PHMB-mediated antibacterial effects are independent of stress response pathways. Cell elongation and chromosome condensation are characteristic morphologies that are often associated with the bacterial Scientific RepoRts | 6:23121 | DOI: 10.1038/srep23121 SOS response 21,22 . Therefore, we considered that these effects could involve this response. However, in the case of PHMB-mediated effects, an SOS response seemed unlikely. First, the SOS response is typically associated with DNA damage and there is no evidence for PHMB-mediated genotoxic or epigenetic effects 23 . Second, the condensation observed following ftsZ silencing and PHMB treatment occurred in a recA − strain (TOP10), which is an SOS response mutant. Nevertheless, antimicrobial mechanisms are notoriously difficult to decipher and may involve multiple mechanisms. Therefore, we decided to assess the possible involvement of an SOS response and other stress response pathways using an SOS reporter strain and a panel of E. coli stress response pathway mutants.
To test whether PHMB-mediated effects on cell elongation and chromosome condensation are altered by mutations to the SOS response pathway, we evaluated morphological responses in three mutant E. coli strains. Strain SS996 is able to initiate an SOS response, but due to a sulB mutation the response does not lead to cell division inhibition. This is because SS996 has a mutant allele of ftsZ (sulB103), the product of which is insensitive to the action of the SOS-induced cell division inhibitor SulA 24 . Strain JW2669 does not produce functional RecA, and so is SOS deficient. Strain AB2474 has a mutation in the LexA repressor that renders it non-cleavable by RecA, and so is not able to induce an SOS response (additional strain details are given in Fig. 3 and Supplementary Information Table 3). Cultures in mid-log phase were treated with PHMB, DAPI stained and observed under a fluorescence microscope. As observed with E. coli K-12, the mutant strains displayed elongated morphologies and condensed chromosomes following PHMB treatment (Fig. 3a). Therefore, PHMB-mediated cell division and chromosome structure effects occur independently of an SOS programmed response.
To more directly measure whether PHMB induces an SOS response, we used the E. coli strain SS996, which is a reporter strain that contains a sulAp-gfp chromosomal SOS response/reporter system 24,25 . If an SOS response is caused by PHMB, then PHMB exposure should induce GFP expression in this strain. Cultures of SS996 were treated with PHMB for 18 hours and then green fluorescence was measured. Mitomycin C, which damages DNA, was included as a positive control, and triclosan, which inhibits fatty acid biosynthesis, was included as a negative control. As expected, mitomycin C induced a large increase in GFP expression and triclosan did not induce GFP expression. In contrast to mitomycin C, PHMB did not induce GFP expression, indicating that PHMB does not induce an SOS response (Fig. 3b).
We next tested whether strains with defective or deregulated SOS responses differ in susceptibility to PHMB. To test recA-mediated effects on susceptibility, we used a strain of E. coli that lacks recA (JW2669) and a strain that overexpresses recA upon addition of the inducer IPTG (ASKA JW2669), and determined MIC values. Neither deletion nor induced overexpression of recA altered susceptibility to PHMB (Supplementary Information Table 3, rows shaded in dark grey). In contrast, the recA − strain was 2-fold more susceptible to the SOS response inducing Cultures of E. coli, strain K-12 were treated with PHMB-FITC, counter stained with DAPI and examined using fluorescence microscopy. Chromosomes appear as condensed DAPI-stained foci; more apparent in the enlarged image. (d) Pattern of chromosome distribution in filamentous/multinucleated E. coli following PHMB exposure. RNA silencing of ftsZ expression was used to arrest cell division, and cells were then untreated or treated with PHMB, stained with DAPI and examined by fluorescence microscopy. (e) Pattern of chromosome distribution in B. megaterium cells that were untreated or treated with PHMB, stained with DAPI and WGA-red and examined using fluorescence microscopy. drug nalidixic acid, and recA overexpression reduced susceptibility to nalidixic acid 8-fold. To test lexA-mediated effects on PHMB susceptibility, we used the lexA1(Ind − ) strain AB2474, which is unable to induce an SOS response. Relative to the parent, AB2474 was 1-fold more susceptible to PHMB and 1-fold less susceptible to nalidixic acid (Supplementary Table 3, rows shaded in light grey). Therefore, none of the SOS response mutants tested displayed changes in susceptibility to PHMB that indicate involvement of an SOS response. and AB2474 (lexA1, mutation that prevents SOS response induction) following treatment with PHMB for 2 hours. Cells were DAPI stained to reveal DNA. (b) SOS response reporter expression, quantified by fluorimetry. The SOS reporter E. coli strain SS996 carrying a chromosomal sulAp-gfp fusion was untreated or treated with PHMB, mitomycin C, a known SOS inducer, or triclosan, which does not induce an SOS response. The MIC values against SS996 were PHMB, 0.75 μg/mL; triclosan, 2 μg/mL; mitomycin C, 0.06 μg/mL, and these values were used to calculate %MIC. Finally, we considered whether other (non-SOS) stress response pathways affect susceptibility to PHMB. We tested a series of known E. coli stress response mutants in parallel with their parent for susceptibility to PHMB. None of the mutants displayed changes in MIC values that suggest a functional involvement of any of the stress response pathways (Supplementary Table 3). Therefore, the antibacterial effects of PHMB occur independently of the panel of stress response mechanisms tested.

PHMB condenses bacterial chromosomes in vitro.
If PHMB condenses bacterial chromosomes inside cells, this could occur via direct or indirect effects on DNA. We suspected direct effects, because PHMB has been shown to bind to DNA fragments in vitro 15 . We decided to examine the DNA binding properties of PHMB using isolated E. coli chromosomal DNA. PHMB-DNA interactions were first examined using an electrophoretic mobility shift assay (EMSA) and a dye exclusion assay. PHMB was mixed with chromosomal DNA isolated from E. coli K-12, and the mixtures were fractionated in agarose/TBE gels, followed by DNA staining with ethidium bromide. PHMB:DNA mixtures having wt:wt ratios of ≥ 0.5 displayed clearly retarded electrophoretic mobility, as indicated by retention of DNA in the well (Fig. 4a). Similar results were obtained for PHMB-FITC. Retarded mobility and retention in wells is consistent with stable interactions between PHMB and DNA. Also, the EMSA assays indicated reduced ethidium bromide fluorescence in the presence of PHMB or PHMB-FITC, suggesting that ethidium bromide was prevented from binding to DNA due to formation of PHMB:DNA complexes. This observation was further investigated using the DNA binding dye SYTOX ® Green in a dye exclusion assay. In the absence of PHMB, SYTOX ® Green bound isolated E. coli DNA, as indicated by a large increase in fluorescence, relative to the addition of dye alone. However, prior addition of PHMB reduced fluorescence > 80% (Fig. 4b).
Therefore PHMB forms complexes with bacterial DNA in a way that retards electrophoretic mobility and masks DNA access to DNA ligands. The results of each of these experiments indicate that PHMB binds directly to DNA.
To learn more about how PHMB binding to DNA impacts the structure of chomosomal DNA, we used biophysical methods and microscopy. Combinations of PHMB and isolated E. coli chromosomal DNA were examined by circular dichroism (CD) spectroscopy. PHMB alone did not show a characteristic CD spectrum, whereas isolated chromosomal DNA showed a typical DNA spectrum with a positive maximum ellipticity around 260 nm, a negative cross over at 252 nm and a negative trough at around 245 nm. This allowed us to assess changes in the CD spectrum of DNA following addition of PHMB. Mixtures of PHMB and DNA displayed reduced ellipticity at 260 nm, indicative of structural changes to the DNA upon PHMB binding (Fig. 4c,d). Also, dynamic light scattering (DLS) revealed that PHMB binding to DNA results in the formation of nanoparticles of approximately 50-60 nm, with a low polydispersity index (Supplementary Fig. 4a). Finally, transmission electron microscopy (TEM) and fluorescence microscopy also indicated that PHMB binding to DNA results in the formation of nanoparticles ( Supplementary Fig. 4b,c). Therefore, these results confirm earlier reports that PHMB binds DNA 26 , and reveals that PHMB binds isolated bacterial chromosomal DNA and can condense chromosomes into a low polydispersity population of nanoparticles.
The antibacterial effects of PHMB are suppressed by a dsDNA ligand. Our results for PHMB effects on bacteria cannot be reconciled with the membrane disruption model for PHMB's primary antibacterial mechanism of action. Rather we propose a new model, where PHMB enters bacteria and then condenses chromosomes, as illustrated in Fig. 5a. If correct, the new model would also predict functional interactions between PHMB and other DNA ligands, and this idea provided us with a way to test the model. In brief, if this model were correct, small molecular weight DNA ligands would be expected to suppress PHMB's antibacterial potency by competing for DNA binding sites within chromosomes. To test this possibility, pairwise combinations of PHMB and Hoechst 33258 were used in growth susceptibility assays. Hoechst 33258 is a DNA ligand that binds preferentially to the minor groove of AT-rich sequences 27 and it is cell-permeable, making it a suitable choice for this competition experiment.
Drug interactions were calculated as fractional inhibitory concentration index values (FICI) using a panel of diverged bacterial species. The FICI values for PHMB:Hoechst were significantly higher than for PHMB combined with either of two non-DNA ligand antibacterials (Fig. 5b). Also, the FICI values for PHMB:Hoechst combinations show a positive correlation with chromosome AT-content (Fig. 5c). In other words, PHMB's antimicrobial effects depend on access to DNA inside cells. In B. megaterium, the effects of PHMB:Hoechst combinations were striking, where growth inhibition by PHMB was suppressed using subinhibitory Hoeschst 33258 concentrations (Fig. 5d). Therefore, the small molecule DNA ligand Hoechst 33258 rescued bacteria from inhibitory concentrations of PHMB.
These pairwise drug interactions reveal that the antibacterial effects of PHMB occur mainly via targeting DNA in bacteria. Consistent with the cell permeability change profile observed for PHMB (Fig. 1b); the results also indicate competition between PHMB and a DNA ligand for DNA binding sites inside cells. Therefore, the results from separate experiments involving two known DNA ligands are consistent with our new model for PHMB activity.
PHMB enters mammalian cells but is excluded from nuclei. The prevailing model for PHMB activity holds that PHMB kills bacteria through bacterial membrane damage and the polymer does not interact with mammalian cell membranes (see above). However, given the unexpected bacterial cell entry properties of  Table 4 for species list and inhibition values. PHMB, and our recent observations that PHMB enters cultured macrophage 28 and keratinocytes 29 , we decided to directly assess its ability to enter a panel of mammalian cells. PHMB-FITC was added to several mammalian cell lines and primary fibroblasts and uptake was assessed by fluorescence microscopy and flow cytometry. We observed clear uptake into all cell types tested (Fig. 6a,b). Also, these conditions did not lead to disruption of mammalian cell membrane integrity (Supplementary Fig. 5a). Close inspection of the microscopy images reveals that PHMB-FITC was contained within vesicles and excluded from nuclei (Fig. 6a). If it is true that endosomes entrap PHMB, then release into the cytoplasm should de-quench PHMB-FITC and lead to an increase in fluorescence. This is because FITC fluorescence is quenched at low pH and the late endosomal pH is < 6 30 , whereas the cytoplasmic pH is 7.4. We observed that the addition of chloroquine, an osmolytic/buffering agent 31 , increased the fluorescence of PHMB-FITC treated cells (Fig. 6c), consistent with polymer entrapment within endosomes. Therefore, PHMB efficiently enters mammalian cells, but is entrapped within endosomes, which appears to restrict entry into nuclei.

Discussion
The arsenal of registered antimicrobial agents (antibiotics + biocides) exploits only a small number of defined mechanisms 32 . Past studies on the mechanism of action of PHMB focused on its ability to interact with microbial membranes in preference to mammalian membranes. The present study, in contrast, suggests an alternative model. The results demonstrate that PHMB is able to enter bacterial cells, arrest cell division and condense chromosomes, resulting in intracellular foci of DNA. PHMB is revealed as the first example of a drug that binds and condenses bacterial chromosomes. Indeed, this is the first example of any drug that condenses chromosomes.
Chromosome condensation has not previously been considered as an antibacterial mechanism of action; however, the feasibility of using a cationic polymer to directly condense chromosomes in bacteria is suggested by several prior findings: (i) the regulation of bacterial DNA condensation and decondensation appears to involve polyamine and basic protein binding 33 , (ii) condensation aids chromosome partitioning during cell division 34 , (iii) overexpression of histone-like proteins leads to nucleoid condensation and bacterial cell death 35,36 , (vi) chromosomes in bacterial cell lysates are condensed by the addition of polylysine 37 , and (v) certain cationic polymers enter bacteria 38,39 . Therefore, while indirect mechanisms leading to chromosome condensation in bacteria remain formally possible, our data indicates a direct mechanism, where PHMB binds and condenses DNA following cell entry (Fig. 5a), and the prior evidence outlined above is consistent with this new model. A chromosome condensation model for the antibacterial action of PHMB (Fig. 5a) can explain how PHMB kills bacteria, but also raises a new and difficult question. How can chromosome condensation provide a selective antibacterial mechanism, given that all organisms have chromosomes? In other words, if PHMB enters cells and condenses bacterial chromosomes, why doesn't it also kill mammalian cells and display toxic effects when used in clinical applications? The data shown in Fig. 6 provide an answer by showing that PHMB's distribution within mammalian cells is partitioned. Specifically, it localizes within endosomes and is excluded from nuclei. Therefore, PHMB's antibacterial selectivity appears to involve differential target access through drug partitioning inside cells, rather than by the well-established principles of target recognition and structure conservation 40 .
Mammalian cell uptake and nuclear exclusion of PHMB are unexpected observations; however, they may reflect aspects of how mammalian cells evolved together with microbes. For example, cationic antimicrobial peptides are central to innate immunity, and our observations may indicate mechanisms that protect host cells against endogenous cationic antimicrobial peptides and other natural molecules that could bind cellular DNA. Also, these observations raise interesting questions about potential toxicity and how PHMB kills eukaryotic pathogens. Recently, we reported that PHMB is able to enter the parasite Leishmaniasis major, and following parasite cell entry, PHMB disrupts chromosomes within nuclei 28 . Therefore, we have observed similar mechanisms in both bacteria and parasites, suggesting that chromosome condensation or disruption appears to be a major mechanism of action of PHMB that is effective against diverse pathogens. Understanding antimicrobial mechanisms can aid further drug development 41 . We hope that the results of this study will inspire antimicrobial strategies designed to selectively condense microbial chromosomes -a mechanism that does not appear to succumb to acquired resistance 3 .

Determination of Minimal Inhibitory Concentrations (MIC), Minimal Bactericidal Concentrations (MBC) and Fractional Inhibitory Concentration Index (FICI). All E. coli strains,
S. enterica serovar Typhimurium LT2 and S. aureus NCTC 6571 were grown in Mueller Hinton broth (MHB, Fluka, Germany) at 37 °C overnight. Where required, MHB was supplemented with kanamycin or chloramphenicol/IPTG (Supplementary Table 3). MICs were determined by serial dilution of the antibacterial in 200 μl MHB containing 10 5 colony forming units (CFU)/mL using 96-well plates (Costar, UK). Plates were incubated for 18 hours at 37 °C in a BioTek Power-Wave X340I spectrophotometer with shaking for 5 seconds every 5 minutes followed by recording of the absorbance at 550 nm. The MIC was scored as the lowest concentration of compound at which no growth was observed. To determine the minimal bactericidal concentration (MBC; > 10 3 CFU/mL reduction), bacterial cultures at 10 5 CFU/mL were treated or not treated with PHMB and at specific time points samples were diluted and plated on LB agar plates. CFUs were counted after 18 hours of incubation at 37 °C. FICIs were determined using the checkerboard assay as described 42 , using Hoechst 33258 (Sigma-Aldrich, UK). Bacteria cell membrane permeability assay. E. coli K-12 from mid log phase (10 μl of culture, OD 600 adjusted to 0.1) were transferred to 96-well plates containing PHMB, polymyxin B or triclosan (0-8 μg/mL) in 100 μl phosphate buffered saline (PBS), and incubated at 37 °C for 60 minutes in a BioTek Power-Wave X340I spectrophotometer with shaking for 5 seconds every 5 minutes. To generate cells with maximum permeability to SYTOX ® Green, untreated cultures were incubated for 10 minutes in a heating block maintained at 70 °C. The dye SYTOX ® Green (Invitrogen, UK) was added to a final concentration of 1 μM, and changes in fluorescence emission were monitored at 535 nm upon excitation at 485 nm using a Wallac Victor 1420 Multi label counter (PerkinElmer, UK). SYTOX ® Green fluoresces strongly upon binding to DNA, and fluorescence was taken as an indication of membrane permeabilisation 43 .
Epi-fluorescence microscopy of bacteria and particles. Overnight bacterial cultures of E. coli K-12, SSPP6 and JW2669-1, S. enterica serovar Typhimurium LT2, and S. aureus NCTC 6571 were diluted 1:50 in MHB and incubated for 2 hours to reach mid log phase. M. smegmatis MC2155 was grown in Middlebrook 7H9 broth with 10% OADC enrichment (BD Biosciences, UK) for 18 hours. Aliquots from mid log phase cultures were diluted in MHB (E. coli and S. enterica) or 7H9 broth (M. smegmatis) to 0.1 OD 600 , and 100 μl aliquots were transferred to 1.5 mL tubes, PHMB-FITC (0-8 μg/mL) was added and the cultures were incubated for 90 minutes. Following incubation, bacteria were washed 3 times using 200 μl 1× PBS via centrifugation at 5000 rpm for 5 minute and removal of the supernatants. Bacteria pellets were resuspended in 100 μl of 1 μM DAPI (Invitrogen, UK) in 1× PBS and incubated at room temperature for 10 minutes, then mounted on a glass slide and observed under a fluorescence microscope (Leica DM4000B microscope with Zeiss "AxioVision" software).
For particle analyses, PHMB:DNA mixtures were prepared as described above for the dye exclusion assay and loaded on an agarose bed (1.5% in 1× PBS) prepared on a glass slide, overlaid with a glass coverslip and observed under an upright fluorescence microscope using a FITC filter set (400× magnification, Leica DM4000B microscope, "AxioVision" software). Dye exclusion assay. PHMB (0-4 μg/mL, final) and isolated chromosomal DNA from E. coli K-12 (1 μg/mL, final) were mixed in 100 μl 1× PBS in a 96-well plate and incubated at 37 °C for 30 minutes. Following incubation, SYTOX ® Green was added to 100 nM, and the plates were further incubated for 10 minutes at 37 °C and fluorescence was measured as described in the bacteria cell membrane permeability assay; see above.
Confocal microscopy of bacteria. To explore the spatial distribution of PHMB in live bacteria, one of the largest known Gram positive bacteria, B. megaterium, was exposed to PHMB-FITC and counterstained with a fluorescent membrane marker. B. megaterium was cultured in LB broth overnight before dilution to 10 8 CFU/mL in fresh broth containing a final concentration of 5 μg/mL of PHMB-FITC. Following incubation in a shaker for 120 minutes at 37 °C, the bacteria were washed and stained with DAPI as described above. Bacterial pellets were re-suspended in 1× PBS and 80 μl of the suspension was centrifuged in a Shandon Cytospin 2 cytocentrifuge at 800 rpm (75 × g) for 5 minutes before circumscribing the slide-deposited bacteria with a hydrophobic barrier pen (ImmEdge pen, Vector Laboratories). Cells were fixed in 4% formaldehyde (Thermo Scientific, Fisher Scientific) in 1× PBS at room temperature for 15 minutes. The formaldehyde was aspirated and the preparation were washed three times with 1× PBS at 5 minute intervals. To label the bacterial cell membrane, the fixed bacteria were immersed in 100 μl of Hank's balanced salt solution (HBSS) minus phenol red (Invitrogen) containing 5 μg/mL wheat germ agglutinin (WGA) conjugated to Alexa Fluor-555 (Invitrogen). Following 20 minute incubation at room temperature, the labeling solution was removed and the slides were washed twice at 5 minute intervals with fresh HBSS. The preparation was mounted with FluorSave reagent (Calbiochem) and a coverslip fixed in place with clear nail varnish. Slides were kept at − 20 °C. Laser scanning confocal microscopy was conducted using a Leica SP5 microscope (Leica Application Suite, Advanced Fluorescence 3.1.0 build 8587 Software). Sequential scan Z-stacks (115 slice 1024 × 1024) were compiled at a line average of 96. ImageJ 1.49r was used to produce, analyse and profile plot dual channel composite Z-stacks.
Generation of filamented bacteria. Overnight cultures of E. coli TOP 10 carrying a plasmid expressing lac promoter driven antisense RNA against bacterial cell division ftsZ mRNA (anti-ftsZ) were grown in MHB for 18 hours in the presence of chloramphenicol (30 μg/mL, Sigma-Aldrich, UK) to maintain the plasmid 19 . Overnight cultures were diluted 50-fold in MHB and IPTG (75 μm) was added to induce anti-ftsZ expression.
Scientific RepoRts | 6:23121 | DOI: 10.1038/srep23121 Cultures were further incubated for 90 minutes in an orbital shaker at 37 °C in the presence or absence of PHMB. Following incubation, bacteria were washed and stained with DAPI as described above and observed under a fluorescence microscope (630× magnification, Leica DM4000B microscope, Zeiss "AxioVision" software).

SOS response assay.
Overnight cultures of E. coli strain SS996 (10 5 CFU/mL) were sub-cultured in the absence or presence of PHMB, mitomycin C (Roche, UK) or triclosan (Ciba AG, Switzerland) in MHB for 18 hours in a 96-well plate in a Bio-Tek PowerWave X340I spectrophotometer as described above, and SOS response was assessed as described previously 25 . GFP expression was measured in a Wallac Victor 2 1420 Multi label counter using 485 nm (excitation) and 535 nm (emission).
Dynamic light scattering. PHMB:DNA mixtures (25 μg/mL:10 μg/mL) were diluted with 900 μl of 0.2 μm filtered water and mean particle size was measured by dynamic light scattering in a Zetasizer Nano ZS instrument (Malvern instruments, UK). Mean size represents the average of 20 readings. Experiments were repeated independently three times.
Flow cytometry. Cell associated fluorescence of PHMB-FITC treated or untreated cells was analysed by flow cytometry (FACSCalibur TM ; CellQuest ™ software, BD Bioscience) using an FL1 filter set. Adherent mammalian cells were trypsinised and diluted to 10 6 cells/mL prior to flow cytometry. Untreated cells were used to establish thresholds for gating fluorescence negative cells. Results were analysed using Flowjo7.6.5 software.
Endosome release assay. PHMB-FITC (3.5 μg/mL) was suspended in 1× PBS solutions in the pH range 4.06-7.4 (pH adjusted with 0.1 M HCl). The solutions were incubated at room temperature for 10 minutes and fluorescence was measured by recording emission at 535 nm upon excitation at 485 nm in a fluorimeter. HeLa cells were co-treated with 3.5 μg/mL PHMB-FITC and chloroquine (0-150 μM, Sigma-Aldrich, UK) for 2 hours at 37 °C and fluorescence was quantified by flow cytometry. The number of cells positive for uptake and the geometric mean (GM) was recorded to assess the cell-associated fluorescence from three independent experiments. Data analysis. Data are represented as mean ± standard deviation from at least three independent experiments. Data in Fig. 5 and Supplementary Fig. 5 were analysed by one-way ANOVA and data in Supplementary  Fig. 2 were analysed by two-way ANOVA, followed by Tukey's Post test (Prism 6, GraphPad, Inc., San Diego, CA). P < 0.05 was considered as statistically significant (*).