Antibiotic-induced DNA damage results in a controlled loss of pH homeostasis and genome instability

Extracellular pH has been assumed to play little if any role in how bacteria respond to antibiotics and antibiotic resistance development. Here, we show that the intracellular pH of Escherichia coli equilibrates to the environmental pH following treatment with the DNA damaging antibiotic nalidixic acid. We demonstrate that this allows the environmental pH to influence the transcription of various DNA damage response genes and physiological processes such as filamentation. Using purified RecA and a known pH-sensitive mutant variant RecA K250R we show how pH can affect the biochemical activity of a protein central to control of the bacterial DNA damage response system. Finally, two different mutagenesis assays indicate that environmental pH affects antibiotic resistance development. Specifically, at environmental pH’s greater than six we find that mutagenesis plays a significant role in producing antibiotic resistant mutants. At pH’s less than or equal to 6 the genome appears more stable but extensive filamentation is observed, a phenomenon that has previously been linked to increased survival in the presence of macrophages.

Survival assay-nalidixic acid. Acute survival was measured by survival following the removal of the DNA damage induced by nalidixic acid. The strains were grown (37 °C, 200 rpm) in quadruplet in media (LB or pH adjusted LBK, 3 ml in a 15 ml tube) to OD 600 = 0.8. The first sample was taken before addition of nalidixic acid and designated as t = 0. Samples were centrifuged (21,500 g, 20 s) media aspirated and the pellet resuspended (PBS) to remove the nalidixic acid. The samples were then serially diluted (10 ×, 100 µl, PBS or pH adjusted PBS) in 96 well plates before samples (10 µl) were transferred to 24 well plates containing only LB-agar or pH adjusted LBK-agar (1 ml/well, 1.5% agar). The droplets were allowed to dry, then incubated (overnight, 37 °C) in the dark before enumeration the following day. For each sample the well containing 10-100 colonies was counted and used to calculate CFU/ml. Samples were taken at the time points indicated on the graphs.
Flow cytometry-Intracellualar pH determination. The strains were grown (37 °C, 200 rpm) in quadruplet in pH adjusted media (LBK, 3 ml in a 15 ml tube) containing 200 µM arabinose to OD 600 = 0.8 (Nanodrop, ND-1000). The arabinose maintains high levels of expression and keeps the majority of active GFP in the cytoplasm 21 . The first sample was taken before addition of nalidixic acid and designated as t = 0. The samples (2 µl) were diluted into several pH buffered PBS solutions (100 µl), one at the pH of the LBK and a calibration curve of five pH buffered PBS solutions containing the membrane permeable weak acid sodium benzoate (60 mM). Benzoate collapses the ΔpH across the inner membrane resulting in the internal pH equalization to that of the external solution. The extracellular pH's were chosen to ensure that at least one of the solutions was higher and one was lower in pH compared to the intracellular pH of the bacteria. The fluorescence signal of the TorA-GFPmut3* protein from 10,000 bacteria/sample was then measured on an AccuriC6 flow cytometer in FLA-1. Subsequent samples were taken at the time points indicated after exposure to nalidixic acid. The intracellular pH values were calculated using the two solutions on the calibration curve that straddled the fluorescence intensity value of the unknown LBK sample.
Complementation at pH 6 was achieved using the relevant RecA producing strain (Table S1) using IPTG due to the T5\lac promoter. IPTG was added during subculturing at the same time as arabinose at the start of the experiment.
Flow cytometry-SOS induction and recA, lexA & umuDC-gfp measurement. The strains were grown (37 °C, 200 rpm) in quadruplet in pH adjusted media (LBK, 3 ml in a 15 ml tube) to OD 600 = 0.8 (Nanodrop, ND-1000). The first sample was taken before addition of nalidixic acid and designated as t = 0. The samples (2 µl) were diluted into PBS (100 µl) buffered to the same pH as the incubating media and simultaneously at pH 8 supplemented with sodium benzoate (60 mM) at room temperature. The GFP signal is maximised and standardised by using PBS at pH 8 together with sodium benzoate (60 mM) and eliminates the GFP signal Stopped-flow assays and data analysis. The experiments were performed essential as described previously using an SFM-300 stopped-flow machine (Bio-Logic) fitted with a MOS-200 monochromator spectrometer (Bio-Logic) with excitation wavelength set at 545 nm 24 . Fluorescence measurements were collected with a 550 nm long pass emission filter. The machine temperature was maintained at 25 °C with a circulating water bath. For all experimental setups, a master mix containing all common reaction components for each of the two syringes was prepared. To these mixtures a variable RecA amounts were added. Since equal volumes were injected into the mixing chamber from each syringe, the two solutions became mutually diluted. Therefore, all reaction components common to each syringe were prepared at the final concentration, whereas reaction components present in only one syringe were added at twice the desired final concentration. All concentrations quoted represent final concentrations after mixing. Components of each syringe were pre-incubated for 10 min before the start of experiments to allow the contents to reach equilibrium.
All reactions were performed in Stopped Flow Buffer (50 mM phosphate/citrate buffer (pH 5-8), 10 mM MgCl 2 , 50 mM NaCl, 1 mM ATP). All reactions contained 20 nM 5′-Cy3 fluorescently labelled (dT79). RecA was added directly from concentrated stocks. For all experiments, control reactions were also performed for buffer alone with and without DNA to confirm fluorescence signal stability over the time course of the experiments (data not shown).
Fluorescence measurements for most experiments were collected according to the following protocol: (1) every 0.00005 s from 0 to 0.05 s; (2) every 0.0005 s from 0.05 to 0.56 s; (3) every 0.02 s from 0.56 to 120.54 s. For each condition analysed, traces were collected from between three and nine independent reactions (n = 3-9) and averages were generated. Importantly, all conclusions from this study are made on the magnitude and rate of changes in fluorescence over time and how these vary with RecA concentration, and therefore the absolute fluorescence values were converted to arbitrary units by a normalisation procedure to facilitate comparison. For all experiments the raw data were normalised to the same fluorescence value for the 0 s time point.
For analysis, for all experiments ten-point moving averages were calculated on each individual normalised trace, which were used to define initial (0 s), final (120.54 s) and maximum fluorescence (and corresponding time point), and ΔCy3 fluorescence values for each experiment. To simplify direct comparison of conditions for various pH conditions we calculated observed overall half-times, measured from time points where the fluorescence from moving averages was closest to the value calculated for ΔCy3 fluorescence midpoints.
LexA degradation assay. RecA (0.6 µM) was mixed with ssDNA (30-mer, 70 nM) in citrate-phosphate buffer with various pH containing 35 mM NaCl, 10 mM MgCl 2 and either 1 mM ATP or ATPgammaS. RecA nucleofilaments were formed for 10 min at 37 °C and then mixed with LexA protein. Reactions were stopped at 0, 60 or 120 min by addition of Laemmli buffer. Samples were resolved in 15% SDS-PAGE gels and stained with Coomassie stain. The intensity of bands for full length LexA protein were analysed by MultiGauge Software. The average signal is shown with standard error for 3 independent experiments. Lactose reversion assay. FC29 (Rif S , F′ ∆(lacI lacZ)) E. coli cannot utilise lactose as a carbon source or revert from Lacto Lac + . FC40 (Rif R , F′ lacIΩlacZ) eliminates the coding sequence of the last four residues of lacI, all of lacP and lacO, and the first 23 residues of lacZ 25 . Individual colonies on M9 minimum glycerol plates were used to seed M9 minimal glycerol (30 °C, 200 rpm, 1 ml in 15 ml tubes, three days). pH adjusted M9 minimum lactose agar plates were inoculated with 10 10 FC29 to scavenge each plate for sources of carbon other than lactose. The next day, the plates were overlaid with 1:10 (FC29:FC40) suspended in pH adjusted M9 minimal lactose top agar (0.5% agar) and incubated at 37 °C.
The frequency of Lac + revertants was determined by counting Lac + colonies every 24 h for 12 days and calculating the mean and standard deviation (n = 3). www.nature.com/scientificreports/ MIC assay. Wild type E. coli (MG1655) was grown overnight (37 °C, LB agar plates) before being resuspended in pH adjust LBK media and adjusted to an OD 600 = 0.03. The antibiotics were dissolved to a final concentration of 128 µg/ml in pH adjusted LBK and serially diluted (× 2) to 0.25 µg/ml (giving 10 concentrations of each antibiotic). A single µl of adjusted bacterial culture was added to each well (100 µl). The plates were then incubated (37 °C) and the OD 595 read the following day. The experiments were repeated in quadruplet to determine the mean and standard deviation.
Forward UV induced rifampicin resistance assay. pH adjusted PBS and LBK media were used throughout. Wild type E. coli (MG1655) were grown overnight (37 °C, 1.5% agar) before being diluted and resuspended (< 1000 bacteria / culture) in liquid media for overnight growth (37 °C, 200 rpm). The following day the bacteria were diluted (× 50) and grown to OD 600 = 0.25 before being centrifuged and resuspended in PBS and exposed to UV (20 J/m 2 ) or not. Subsequently all cultures were kept in the dark. The samples of the cultures were then plated on Rifampicin (100 µg/ml) plates and also on LBK to determine CFU. The cultures were then centrifuged and resuspended in LBK before incubation (37 °C, 200 rpm, ON). The following day the cultures were plated on Rifampicin (100 µg/ml) and LBK to determine CFU. The following day the concentration of rifampicin resistant bacteria was calculated. The mean and standard deviation were calculated from three independent experiments.
Bacterial viability assay with propidium iodide. Bacteria were grown overnight (37 °C, LB agar plates) before being resuspended in pH adjust LBK media and adjusted to an OD 600 = 0.03. The strain(s) were grown (37 °C, 200 rpm) in pH adjusted media (LBK, 3 ml in a 15 ml tube) to OD 600 = 0.6 (Nanodrop, ND-1000). The first sample was taken before addition of nalidixic acid and designated as t = 0. The samples (2 µl) were diluted into PBS (100 µl) buffered to the sample pH as the incubating media and containing PI (2 µl). The fluorescence signal of PI from 10,000 bacteria/sample were then measured on an AccuriC6 flow cytometer in channel FLA-3. Simultaneous readings of FSC-A were recorded and used to calculate the density of PI staining in the bacteria. Subsequent samples were taken at the time points indicated after exposure to nalidixic acid. All experiments were run on at least three independent days.
To examine the effect of pH alone on PI staining the above experiment was carried out but without the addition of nalidixic acid. To account for growth phase variation the experiments were carried out both by allowing growth to continue unrestricted into the stationary phase and using serial dilution to keep the cultures at lower densities.

Results
DNA damage leads to the loss of pH homeostasis. To induce DNA damage we used the antibiotic nalidixic acid a quinolone that binds DNA gyrase and topoisomerase IV inhibiting their ligation activities leading to DNA double strand breaks and chromosome fragmentation 26 . To explore the hypothesis that antibioticinduced DNA damage affects the proton gradient of the bacterium we first confirmed that nalidixic acid led to the recA-dependent loss of electrical potential (Supplementary Figure S1 and "Results" section) 27 . Thereafter, we tested the potential of a pH sensitive reporter, GFPmut3* (Supplementary Table S1), to follow intracellular pH changes (Supplementary Figure S2) 28 . E. coli expressing GFPmut3* were grown to exponential phase before treating with Nalidixic acid (100 µg/ml). The fluorescence of GFPmut3* was recorded by flow cytometry both before and after treatment and simultaneously the GFPmut3* signal of each sample was calibrated over the possible pH range. We observed that the calibration curve changed over time following the introduction of the antibiotic and determined that continuous calibration of GFPmut3* was necessary (Supplementary Figure S3). The intracellular pH of E. coli was monitored for four hours following antibiotic-induced DNA damage where the external pH was buffered to pH 7 (Fig. 1A). Following a delay of 30 min, the intracellular pH fell reaching the pH of the extracellular environment about 120 min after the addition of nalidixic acid.
E. coli is the predominant aerobic organism in the caecum and colon of the human gastrointestinal tract where the pH varies from 5.7 to 6.8 and 6.1-7.2 respectively 29,30 . Additionally, pathogenesis can lead certain E. coli strains to be exposed to the urinary tract (pH 4.5-8 31 ) and the eukaryotic intracellular environment which includes macrophages and phagosomes (pH 5-7.5 32 ). Due to these various ecological niches occupied by E. coli, we were curious as to how the external pH may affect the kinetics of loss of proton homeostasis in more acidic environments. Buffering the media to either pH 6 or 5.5 led to the decrease of intracellular pH to almost that of the external media over 180-240 min (Fig. 1A). As RecA coordinates the DNA damage induced-stress response in E. coli, by inducing the expression of LexA sensitive genes in a dose-dependent manner 33 , we used a recA mutant to demonstrate a significant reduction in proton depolarisation after treatment with nalidixic acid as compared to wild type (Fig. 1B). To confirm we complemented the recA strain with the RecA producing plasmid from the ASKA collection 13 . This plasmid has previously been shown to successfully complement recA mutants 14 . By inducing with 0.1 mM IPTG we were able to clearly demonstrate complementation using a culture buffered to pH 6 ( Fig. 1C). Additionally, we noted in the wild type that electrical depolarisation never fully equilibrated at pH 7 whilst proton equilibration was achieved after two hours. Identical experimental setups allowed for a direct comparison illustrating that the two processes appear to be kinetically independent (Fig. 1D). Together these results show both ΔΨ and ΔpH equilibrate with their external environment in a recA-dependent manner. However, the kinetics of equilibration indicate the two processes appear be independent of each other.
External pH affects filamentation and recA transcription. The recA dependency of proton equilibration following DNA damage suggested that various SOS processes may also be dependent on the external pH. Filamentation following induction of the SOS response is a classical morphological response to DNA damage 34 www.nature.com/scientificreports/ Therefore, we investigated if the external pH would influence the dynamics and extent of filamentation following DNA damage using flow cytometry by following forward scatter (FSC-A), as a proxy for cell volume 37 . Surprisingly, when the extracellular pH was buffered to pH 7 after exposure to nalidixic acid cell filamentation showed an initial increase followed by a gradual decline ( Fig. 2A). Next we examined how the external pH buffered to 6 or 5.5 would affect the dynamics of filamentation. Both pH conditions showed a greater extent of filamentation ( Fig. 2A). These data demonstrate that following DNA damage the external pH can affect internal cellular processes, such as filamentation and loss of mean cell volume via changes in the intracellular pH. As the expression of RecA is immediate following the introduction of a substantial amount of DNA damage 38 we next examined the RecA-mediated transcriptional SOS response over the replicative permissive pH range of E. coli. To monitor the induction and kinetics of the SOS response we used a strain expressing an optimised genomic recA-gfpmut2 fusion protein which responds quickly to SOS induction and minimizes steric interference of the tag (recA-gfp, Supplementary Table S1) 39 . We calculated GFP fluorescence density by dividing fluorescence The cytoplasmic pH of E. coli was measured using calibrated GFPmut3* fluorescent intensity measurements using flow cytometry following Nalidixic acid (100 µg/ml) induced DNA damage at 0 min. (A) pH depolarisation was followed over 4 h, the extracellular media was buffered to 5.5, 6 and 7, as indicated on the graph. (B) pH equilibration is not complete in a recA mutant following exposure to Nalidixic acid at pH 7. (C) Complementation of the recA mutant to produce a recA + strain restores the pH depolarisation phenotype following exposure to Nalidixic acid. (D) Intracellular pH equilibrates with the extracellular pH before electrical polarity as measured by DiBAC4 (3) following DNA damage in wild type E. coli. All data points shown are mean ± s.e.m., n = 4 (independent replicates). † p > 0.05, *p ≤ 0.05, **p ≤ 0.01 (t-test, two tail, unpaired equal variance).  Using an identical experimental setup as utilised to determine the loss of pH homeostasis, the recA-gfp strain was grown to the exponential phase before exposure to nalidixic acid in four different medias buffered to pHs 5.2, 6, 7 and 8 (Fig. 2B). pH 6 promoted the most substantial expression of RecA-gfp while pH 8 showed the least expression. The FSC-A data also confirmed our previous findings and demonstrated the largest changes in cell size appear at the lowest pH after exposure to nalidixic acid (Supplementary Figure S4B). These data illustrate that following DNA damage pH homeostasis is impaired and, consequently, affects RecA expression and downstream processes. These data also reinforce the fact that the external pH, via homeostasis loss, leads to altered morphological plasticity following DNA damage.
Transcription of later induced SOS genes are repressed by an external pH of 5.2. Autoproteolytic degradation of the LexA transcription factor is a key step in the regulation and induction of the SOS response genes. To observe the induction of lexA an ectopic lexA promoter gfpmut2 fusion was used (plexA-gfp, Supplementary Table S1) 40 . The plexA-gfp system was investigated as outlined above for recA-gfp. The general pattern is similar to recA-gfp with a maximum expression around pH 6 with the following key differences: at pH 5 the lexA promoter activity is poorly induced and declines from180 to 300 min Induction of the lexA promoter at pH 6-8 shows slightly sigmodial dynamics whereby induction accelerates to a maximum between 60 and 120 min whereas recA starts at maximum induction only slowing from 120 min (Fig. 2C). Thus we show that lexA promoter expression is strongly influenced by the external pH with an optimal expression at pH 7.
The LexA protein is known to have asymmetric affinity to the heterogeneous SOS operator sequences and as such different genes are induced at varying LexA concentrations 41 . As pH equilibration occurs gradually following DNA damage we hypothesised that the various SOS genes and their products could be differentially affected 42 . PolV mediated mutagenesis is detrimental to the fitness of most bacteria and as a result the two SOS active components umuD and umuC of PolV are induced late in the SOS response 43 . To monitor the induction of umuDC an ectopic umuDC promoter gfpmut2 fusion was selected (pumuDC-gfp, Supplementary Table S1) 40 . The pumuDC-gfp system was tested as outlined above for recA-gfp. The fluorescent signals of UmuDC-gfp start to increase after about 90 min of incubation, confirming umuDC is induced later than recA. Further, following nalidixic acid treatment pumuDC showed the highest and lowest induction in media at pH 6 and 5.2, respectively (Supplementary Figure S5A). Analysis of FSC-A in the umuDC promoter Gfpmut2 expressing cells at the indicated pH's revealed the same filamentation patterns as recA-gfp expressing cells (Supplementary Figure S5B) confirming the previous findings ( Fig. 2A). After adjustment of the Gfpmut2 fluorescent intensity signals to cell size, pumuDC-gfp showed the strongest induction at pH 8 and practically no induction at 5.2 (Fig. 2D), which is in contrast to recA-gfp (Fig. 2B). To directly compare the pH dependency of early and late induced SOS genes, we plotted the 300 min after nalidixic acid exposure data points across the permissible pH range (Fig. 2E,F). These data indicate early induced genes (recA and lexA) show maximum expression at pH 6 and 6.5 while late induced genes (umuD and umuC) showed maximum expression over a broader pH-range (6-8) (Fig. 2D). Of the genes tested, only recA was found to be significantly induced at pH 5.2.
Finally, the identical experimental setup allowed temporal comparisons of pH equilibration against recAgfp, plexA-gfp and pumuDC-gfp induction. Comparative fluorescence density measurements (Fig. 3A) clearly confirmed the earlier induction of recA-gfp whose expression begins < 30 min after exposure and up to 90 min before umuDC-gfp expression. The relative expression of umuDC is much faster and plateaus at 180 min, similar to recA-gfp. plexA-gfp shows an intermediate response, with a delayed start compared to recA-gfp but fast relative expression that plateaus after 180 min of exposure to the antibiotic. Comparison of intracellular pH and fluorescence density measurements also demonstrate induction of early-induced SOS genes such as recA and lexA occurs whilst the pH is at or still near physiological pH (Fig. 3B,C). However, early expression for recA indicates that it is exposed to a large gradient of pH's while later induced genes are expressed in an environment where the pH is already approaching that of the external environment (Fig. 3D). Together, these results show that following DNA damage late induced SOS genes are more susceptible to a reduction or loss of induction at low external pHs whilst at pH 8 the opposite is observed.

The biochemical activity of RecA is strongly dependent on pH. The strong correlation between
RecA expression, the strength of SOS induction and their dependence on the external pH prompted us to investigate how pH affects known biochemical activities of RecA. First, we used an electrophoresis mobility shift assay to monitor the formation of RecA: DNA complexes. The binding of RecA to fluorescently labelled ssDNA is strongly pH dependent with almost tenfold higher affinity observed at pH 5 compared to pH 8 (Fig. 4A). This was accompanied also by a change in the mobility of the RecA-ssDNA complexes, indicating a possible conformational change of the filament (Supplementary Figure S6A). Second, a kinetic study using stopped-flow allowed us to further investigate the DNA binding properties of RecA at different pH's by monitoring increasing fluorescence of a 5´Cy3-labelled dT79 oligonucleotide after binding with ATP pre-incubated RecA protein (Fig. 4B,C and Supplementary Figure S6B) 24 . Quantification of the binding parameters confirmed that decreasing pH led to higher affinity (Fig. 4B) as well as a massive change in binding half-times (Fig. 4C). Strikingly, the shift of pH from 5 to 8 resulted in a more than 3000 times increase in half-times. In other words, while at pH 5 it takes RecA less than one second to bind DNA, at pH 8 it needs several minutes. Finally, we used a D-loop formation assay to directly correlate the pH dependency of RecA-DNA binding with its ability to search for the homology within the donor dsDNA plasmid. Indeed, reactions at lower pH resulted in maximal activity, indicating that less RecA is required for the recombination reaction ( Fig. 4D and Supplementary Figure S6C). Based on altered SOS response we also elucidate previously reported RecA coprotease activity on LexA repressor 44 . Our experiments show that RecA filaments stimulate LexA cleavage only at pH 6 in the presence of ATP (Fig. 5A). www.nature.com/scientificreports/ pH dependent DNA three-strand exchange as well as a slow general recombination DNA repair phenotype 45 . We hypothesised that this residue is responsible, at least partially, for the pH sensitivity we had observed. Initially we used RecA K250R to show a significantly impaired LexA autocatalytic activity (Fig. 5B). As this mutant binds to ssDNA in a pH-dependent manner (Supplementary Figure S7), similar to the wt protein, it is plausible that impaired ATPase activity leads to a loss of coprotease activity. Taken together, these data show that pH strongly affects the biochemical activities of RecA and are remarkably similar to the in vivo SOS induction data.

Scientific Reports
RecA K250 is important for SOS gene transcription at low pH. Subsequently, we wished to understand if recA K250R altered the observed pH depolarisation of the wild type strain. To do this the recA K250R strain (Supplementary Table S1) was transformed with the pH sensitive reporter to create recA K250R GFPmut3* and intracellular pH changes following antibiotic exposure were examined by flow cytometry (Fig. 6A). We observed accelerated depolarisation kinetics compared to wild type recA where equilibration with the external media was achieved 1 h earlier. This suggests that recA K250R was also having an effect on the SOS genes, and www.nature.com/scientificreports/ cytometry (Fig. 6B). The transcriptional activity of pumuDC showed an increase in pH sensitivity as compared to wt recA, with the greatest difference between the two strains observed at pH 6-6.5 (Fig. 6C), indicating that the K250 residue is responsible for increasing the transcriptional activity of pumuDC at lower pH.

Mutagenesis is strongly influenced by pH. DNA damage-induced loss of pH homeostasis and cor-
responding changes in RecA activities prompted us to hypothesise that genome stability may be affected over the range of studied pH. The rate of mutagenesis was initially monitored using a lactose reversion assay where the media was buffered across the permissible replicative range in the E. coli FC40 strain (Supplementary  Table S1) 46 . Since the lactose reversion assay detects frameshift mutations, base substitutions, and gene amplifications/ genome rearrangements, which in turn are dependent on a variety of SOS genes including recA¸ recBC, dinB, polB, and umuDC 47 , this allowed us to also monitor the cumulative effect of the individual pathways in mutagenesis. The data illustrates the strong influence external pH has over the rate of mutagenesis (Fig. 7A). The frequency of lactose reversions is close to zero at the extremes, pH 5 and 8, but reaches a maximum between pH 6.5 and 7 (Fig. 7B). Colonies formed at pH 7.5 arise due to lac-amplification rather than mutations as they occurred only gradually. In contrast, most of the colonies at pH 6.5 and 7 form relatively quickly over the first few days, suggesting point mutations are the predominant mechanism 48 . These data demonstrate how the external environment could affect the rate of mutagenesis in niches of different pH. Subsequently, we examined the pH dependency of a forward mutation assay as opposed to the lactose reverse mutation assay. We used UV to directly induce DNA damage and increase mutagenesis (Fig. 7C), since all antibiotic minimum inhibitory concentration assays tested showed pH dependency (Figures S8A-H). In this case mutagenesis showed pH dependency with an observed minimum at pH 6 and a maximum at pH 8. We concluded that mutagenesis is significantly influenced by pH but that the specific stimuli the bacteria are exposed to alter the response. In our examples, the lactose reversion assay, which is dependent on both entering into the stationary phase and the DNA damage response results in the highest mutational frequency from pH 6.5 to 7 whilst when only the SOS response is required the lowest frequencies were found at pH 6.
To explain the differences in mutagenesis over the viable pH range for E. coli we hypothesised that bacterial viability was pH dependent. We used propidium iodide (PI) staining to monitor bacterial viability whereby PI enters the bacteria only upon loss of membrane integrity. The results showed that the bacteria lose membrane integrity at acidic pH after nalidixic acid exposure, whereas at pH 7-8 the membrane of the bacteria remains impermeable to PI (Fig. 8A). An identical experiment using a recA mutant showed the loss of membrane integrity at low pH was significantly dependent on recA (Fig. 8B). Strikingly, at pH 8, where we recorded the highest number of rifampicin resistant mutants there was both the lowest PI staining and highest transcription of pumuDC-gfp. To differentiate between the effect of pH on PI staining and the effect of pH in combination with nalidixic acid we monitored PI staining of cultures grown for 5 h from the exponential phase (Fig. 8C) and cultures maintained at lower densities for 5 h by subculturing (Fig. 8D) in the absence of Nalidixic acid. The absence of increased PI staining at low pH with either approach suggests that the increase in PI staining is Naladixic acid dependent and not solely a pH effect. Additionally we investigated viability using a classical colony forming survival assay (Supplementary Figure S9), however, we observed large variations at and after two hours that prevented any interpretation of the significance of external pH.
Together our results clearly demonstrate that the environmental pH where bacteria and antibiotics meet plays a significant role in antibiotic resistance development. This appears to occur due to proton equilibration across the plasma membrane so that the intracellular pH becomes that of the environment. This pH change subsequently affects many downstream processes of the SOS DNA damage response system ultimately leading to pH-dependent mutagenesis.

Discussion
In this work we aimed to understand the physiological consequences of environmental pH on the SOS response. We initially discovered, in response to the DNA-damaging antibiotic nalidixic acid, a failure of pH homeostasis in the bacterial population. This failure occurred in a controlled fashion over several hours as the cytoplasmic pH equilibrated with the extracellular pH environment. Subsequently, we showed the pH of the external media affects the strength of induction of the SOS genes recA, lexA and umuDC, and the rate of mutagenesis resulting in antibiotic resistant mutants. We also examined how pH affects the biochemical function of the RecA protein in vitro and show that at pH 8 binding of ssDNA was severely compromised, a finding which correlates well with the in vivo weak induction of recA. Additionally, we found that RecA K250, a pH sensitive residue in RecA, was partially responsible for RecA inducing pumuDC-gfp expression and essential up to a pH of 6.5. Taken together our results indicate that the pH of the surrounding environment can influence and enhance mutagenesis and thus accelerate antibiotic resistance development (Fig. 9). We also suggest that pH depolarisation across the inner membrane could be a mechanism whereby bacteria directly sense the environmental pH following a lethal challenge to their DNA integrity. By this mechanism the environmental pH would automatically be sensed by all intracellular components including all translated proteins.
Environmental pH influences antibiotic resistance development. Specifically, our findings show the following pH-dependent responses to nalidixic acid treatment: At an external pH of 5.2 genetic instability is not detectable. This could be explained by the lack of pumuDC transcription and significant loss of membrane integrity. The primary survival response at this pH appears to be filamentation. As the external pH increases, membrane integrity improves so that it is roughly the same as before the antibiotic was applied at pH's 7 and 8, at the same time the transcriptional activity of the genes studied increases, increasing the probability that translesion polymerases will be active. This could explain the increased genome instability seen at the higher pH's. www.nature.com/scientificreports/ These increases in genome instability at pH's above 6.5 also imply that the development of resistance to antibiotics will be increased under these circumstances.
The activities of RecA are sensitive to pH. Early transcriptional activity leads RecA to be exposed to all possible pH's from homeostatic to environmental pH. Previous in vitro studies have shown pH 6.2 is optimal for stable RecA filamentation on ssDNA 49 . Experiments comparing the binding of SSB protein and RecA on ssDNA at pH's 7.5 and 6.2 showed the lower pH greatly favoured RecA over SSB 50 . DNA independent hydrolysis of ATP was significant at pH 6 whilst undetectable at pH 7 51 . More recent work has demonstrated both RecA nucleation and subsequent filament extension on SSB-ssDNA complexes are strongly pH dependent, where pH 6.5, the lowest examined pH, gave the highest nucleation and filamentation rates 52 . We also found pH dependencies www.nature.com/scientificreports/ regarding RecA's interaction with ssDNA. Affinity was strongest and rate of binding fastest at pH 5 and became respectively weaker and slower as pH increased. Ultimately we discovered almost no D-loop formation at pH 8 with the implication that recombination is unlikely to occur in vivo once the intracellular pH reaches 8. Optimal D-loop formation was seen at pH's 6-7 indicating that recombination could potentially be an important resistance development mechanism at these pH's. The origins of the pH sensitivity have been postulated to involve an electrostatic interaction between the negatively charged core domain of one RecA with the positively charged N-terminal domain of the next 53 . Accordingly, we examined the effect of the K250R RecA residue predicted to be www.nature.com/scientificreports/ involved in inter-protein interactions. We found this mutant to be highly influenced by pH, this was most clearly evident at pH 6 when compared to the wild-type, at pH 6 there was a significant reduction in LexA degradation and no expression of umuDC. Since the comparative stability of RecA vs RecA K250R is unknown we cannot exclude that altered biophysical stability could at least be partly the basis of our observations. We conclude, therefore, that K250 and monomer-monomer interactions are important at low pH and influential up to pH 7.5 whilst assuming the stability of RecA and RecA K250R are similar.
A controlled loss of pH homeostasis. The kinetic profile of proton depolarisation after DNA damage indicates how the mechanism may proceed. The observed process contrasts with a physical breach of the membrane which results in a near instantaneous pH equilibration. Indeed, even a rapid shift in external pH without exposure to DNA damaging agents, for example, a shift from 7.5 to 5.5 can result in an intracellular pH shift immediately to 6, with a biphasic recovery beginning after 10 s and most of the physiological pH recouped after 30 s, the remaining being recovered over the next 4 min 28 . Based on our data we concluded, therefore, that the observed mechanism did not involve any rapid disruption of the inner membrane, temporary or permanent, but a gradual and controlled loss of pH homeostasis. www.nature.com/scientificreports/ The bacterium expends considerable resources in maintaining pH homeostasis and optimal internal environment for the metabolism of the cell 54 . However, our data point to situations where the loss of homeostasis and thereby direct environmental sensing of pH may be an advantage. As such the pHs used for the study of bacterial proteins outside of the homeostatic range of 7.5-7.7 may be biologically relevant in such cases. In an optimal environment the maintenance of ion gradients, to generate a pmf, is important for energy production. However, under DNA damaging conditions, it could be an advantage for the cell to enter a state of persistence, a dormant condition where energetic processes are reduced to a minimum. Recently, dependency between intracellular ATP concentrations and persister cell formation has been established 55 . Following DNA damage we show that, in addition to the loss of ΔΨ, the cell also suffers a loss of ΔpH thus greatly reducing the capacity of the bacterium to generate ATP.
Dynamic bacterial morphology following filamentation. We found the reduction in bacterial filamentation after 1 h of exposure to Nalidixic acid surprising as we assumed that filamentation would continue until either the DNA damage stress is removed or the bacterium succumbs to the stress. Yet, recent work suggests cell division could be a response to severe DNA damage-induced stress. In order for cell division to occur replication must be completed. When irreparable damage occurs on one DNA strand this dooms the bacterium that receives this permanently damaged chromosome but allows the other bacterium with an undamaged chromosome to continue proliferation 56,57 . Double strand breaks as produced by inhibition of topoisomerase ligation activities as in the case for nalidixic acid could lead to a serious impact on bacterial viability for both chromosomes. Yet it was interesting to note that reducing cell size following filamentation correlated with the start of the slow phase of bacterial killing attributed to the presence of persister cells in response to an overwhelming challenge to DNA integrity. We see this time point as a change of survival tactics to better suit a chronic challenge to DNA integrity.
A low external pH environment leads to enhanced filamentation with implications for phagocytosis. Varying intracellular pH should lead to altered biochemical activities of the cellular proteins. Filamentation, or morphological plasticity, is a physiological change seen in bacteria in response to DNA damage. The mechanism involves many components but initiation is caused by the SOS protein SulA. Specifically SulA binds FtsZ and thereby inhibits polymerisation of FtsZ which would otherwise form a ring at mid-cell and thus inhibits septation 58 . Other constituents involved in filamentation are proteins participating in cell membrane and cell wall synthesis. From our results, it is clear the sum of the pH effects on these components together leads to longer filaments at lower pH, specifically 5.2 and 6 whilst at pH 7 and 8 the filaments are shorter. It is interesting to note in uropathogenic E. coli SulA is essential for pathogenesis in immunocompetent hosts 59,60 . It has been postulated an influx of neutrophils following bacterial emergence from epithelial cells results in the phagocytosis of bacillary forms of uropathogenic E. coli but not filamented forms. Filamented E. coli appear to be resistant to phagocytosis and exposure to an acidic environment, such as in macrophages leads to a stronger filamentation phenotype so if the bacterium survives it is unlikely to be engulfed again.
In conclusion we find that the pH of the environment will significantly affect the mechanisms of antibiotic resistance and the rate at which they occur in response to antibiotic treatment.