The E. coli strains used in this study, a kind gift from Dr. William Margolin (Geissler et al., 2007), are listed in Table 1. A sulAp–GFP reporter strain regulated by DNA damage that was introduced into the wild-type WM1074 (WM1074+sulAp–gfp) chromosome served as a reporter of the SOS response (Geissler et al., 2007) [original source; (McCool et al., 2004)]. WM2026, WM1074, and WM2739 were grown on Luria-Bertani (LB) broth at 32°C and WM2760 was grown under the same conditions with the addition of 100 μg/mL ampicillin. WM2026 and WM2760 were induced with 40 and 10 μg/mL of IPTG, respectively, 2 h prior to harvest. An overnight culture was used as a stock for inoculating into all the test samples. A formula control, consisting of all formulation ingredients except 2,4-D, and sample controls containing deionized water in place of formulation, were tested in parallel.

The MIC of 2,4-D and the changes to cell length during 2,4-D exposure for E. coli WM1074 were determined as previously described (Bhat et al., 2015b). Growth curves for 1 mM and 4 mM 2,4-D were constructed after 30 h incubation in 96 well plates, with the OD₆₀₀ (optical density at 600 nm) measured every 30 min.

The extent of 2,4-D induced DNA fragmentation was tested using the agar diffusion method (Fernandez et al., 2008). E. coli WM1074 cells grown for 3–4 h were exposed to 0–4 mM 2,4-D (5, 10, 30, and 60 s, 3 h and overnight), H₂O₂ (6 h and overnight) and 50°C (2 h). The culture was diluted to approximately 0.1 OD₆₀₀ and 25 μL was mixed with 60 μL of 0.1% molten agarose at 37°C and vortexed thoroughly. A 20 μL aliquot of this mixture was then spotted onto slides pre-coated with agarose (pre-coating involves coating a clean, grease-free glass slide with 0.1% agarose and drying in an oven at 70°C for 2 h) and a coverslip was carefully placed on the sample to prevent air bubbles in the gel. The slide was incubated at 4°C for 10 min allowing the gel to solidify. The coverslip was removed carefully and the slides were incubated in lysis buffer (2% sodium dodecyl sulfate, 0.05 M EDTA, and 0.1 M dithiothreitol, pH 11.5.) at 37°C for 5 min. All samples were submerged in lysis solution in the same tray to avoid bias in treatment. The lysis buffer was removed and the slides carefully washed, without tilting, in deionized water (5×) and dried in ethanol (70, 95, and 100%, 3 min each at -20°C). The slides were then dried in an oven under vacuum overnight and stained with SYBR gold (Life Technologies) for 5 min, washed and mounted in TBE buffer with a 18 × 18 mm Zeiss precision coverslip and imaged by epifluorescence microscopy (Ex: 497 nm, Em: 537 nm, Zeiss Axio Observer Z1). Exposure to elevated temperature and H₂O₂ served as positive controls. The degree of DNA damage was determined by measuring the distance from the cell periphery to the edge of the halo around each cell, indicating DNA spreading. Images were processed to maximize contrast and remove background noise (Zeiss Zen software) and the diameter of DNA halos surrounding 100 cells for each sample were measured using ImageJ. Differences in samples were assessed using an unpaired t-test in GraphPad Prism 5.

To determine the effects of 2,4-D on membrane potential, a membrane depolarization assay was used as described in Te Winkel et al. (2016). Briefly, exponentially growing E. coli WM2026 cultures (O.D 0.2–0.3) were exposed to 0, 1 mM and 4 mM 2,4-D in a 96 well plate in LB media. The dye 3,3^′-diethylthiadicarbocyanine iodide [DisC2(3)] dissolved in DMSO was added 5 min before the addition of test compounds to a final concentration of 2 μM, with the final concentration of DMSO kept at 1%. Valinomycin at 30 μM was used as a positive control. To examine the possibility of DisC2(3) dye reacting with 2,4-D, wells with water, 4 mM 2,4-D and dye were tested simultaneously. Changes to fluorescence intensities were measured on a microplate reader (BioTek-Synergy) equipped with 560 nm excitation and 580 nm emission filters. The measurements were taken for 60 min after the addition of test compounds.

Immediately following 2,4-D exposure, 10 μL of the sample was placed on a microscope slide, covered with a glass coverslip and examined by LSCM (560/590 nm).

Samples from the DNA damage assay (above) and the E. coli WM2026 strain exposed to 4 mM 2,4-D after 3 h and overnight exposure were imaged (Ex: 488 nm, Em: 509 nm) on a Zeiss Axio Observer Z1 inverted wide-field fluorescence microscope to determine DNA damage, and the Z-ring and nucleoid positioning, respectively. The cells were stained with DAPI (100 μg/mL, Ex: 358 nm, Em: 461 nm) and mounted in 0.01 M PBS and imaged.

Similarly, E. coli WM2739 was imaged to quantify the oxidative stress-induced SOS response. Changes to SulAp-GFP intensity were measured using the ZEN software (Blue edition, 2.1 lite) intensity measurement function. Intensity values collected for each of 100 cells were corrected for background, and statistical analysis conducted with an unpaired t-test using GraphPad 5.

Polystyrene petri dishes with an 18 mm circular hole cut into the bottom were sealed with a Zeiss high precision coverslip placed at the bottom using epoxy resin. The coverslip was pre-cleaned (Bhat et al., 2015b) and then coated with Cell-Tak (Corning) using a slightly modified method (Louise Meyer et al., 2010). A 30 μL aliquot of freshly prepared Cell-Tak solution (145 μL of pH 8 NaHCO₃ buffer, 5 μL of 1 mM NaOH and 5 μL of Cell-Tak) was spread (1 sq. cm) onto coverslips, incubated (RT, 30 min), rinsed gently with ultrapure water and air dried. The coated coverslips were stored for up to 2 weeks at room temperature.

Approximately 500 μL of the culture was added to the coverslip and incubated (32°C, minimum of 30 min) in the dark. The sample was rinsed with LB diluted 1:1 with PBS (0.01 M, pH 7) and mounted with 500 μL of the same solution for imaging using the 63× oil immersion objective on the LSCM. The PBS and media were filtered (0.2 μm) and maintained at 32°C prior to sample preparation. The petri dish with sample was placed in a heated holder maintained at 32°C. Time lapse images were collected before the addition of 2,4-D, and 2,4-D (0.01–4 mM) was added in real-time during imaging, with images collected for up to 6 h post treatment. Formula solution, not containing 2,4-D, was added for imaging control samples.

To determine the effect of oxidative stress on the localization of FtsZ and FtsA, known ROS inducers – paraquat and H₂O₂ were added in real-time during imaging at various concentrations (0.01–20 mM). Similarly, the ionophores vali-nomycin (10–30 μM) and nigericin (5–10 μM) dissolved in DMSO were tested for their effects on the localization of the cell division proteins based on their ability to alter membrane potential. Controls contained the same volume of DMSO only.

Our AFM-LSCM setup consists of the Nano Wizard AFM (JPK, Germany) placed on an inverted LSCM 780 equipped with 34 channels of highly sensitive GaAsP detectors, steady state excitation lasers (458, 488, 514, 543, and 594 nm) and a Ti:Sapphire tunable femtosecond pulsed IR laser. The LSCM stage was replaced with the custom designed AFM stage purchased from JPK and the sample mounted in a manner similar to that described above. A camera was mounted onto the front port of the confocal microscope base and connected to the AFM-ECU to produce a low resolution (20×) DIC image of the sample to allow alignment of the laser on the AFM tip. The sample was subsequently viewed using a 63× oil immersion objective, with the focal plane close to the center of the cell. A suitable actively dividing (FtsZ-GFP ring at the mid cell) single cell, which appeared immobile by DIC and confocal, was chosen for imaging. The JPK AFM software optical calibration tool was used to precisely position the AFM tip in relation to the cell for generating optical overlay. A force constant calibration of the AFM cantilevers (Nanoscience, model no: HYDRA4V-100N), having a low nominal spring constant (k = 0.08 N/m and calibrated k = 0.05 ± 0.03 N/m in imaging media), was followed by a low resolution force map used to evaluate the integrity of the chosen cell. The height image from the force map was used to choose an appropriate area for collecting the QI^TM (quantitative imaging) image. Parameters such as set point, Z-length, and approach/retract speed were adjusted for live QI^TM to reduce noise. Since AFM-QI and LSCM images are collected over different time scales, LSCM images were collected before and immediately after the completion of QI^TM images to demonstrate simultaneous change. Approximately six images collected on different samples and days, at different time points, each with 16834 force curves, were used for processing images before 2,4-D treatment and three images collected on three separate samples and days per time point were used for processing time lapse images during 2,4-D treatment. The AFM and LSCM images taken approximately at the same time were overlaid using Photoshop 11. LSCM images were processed for contrast and digitally enlarged to fit the AFM height image.

QI^TM force curves obtained at each pixel on a 128 × 128 image were corrected (JPK software) for the cantilever force constant and baseline tip-sample separation. The adhesion was determined using the distance between the lowest point and baseline of the retract curve. Young’s modulus was determined using a Hertzian fit, which is an estimate of cell envelope elasticity (JPK software). Surface roughness was measured at the mid-point of the cell using the QI^TM height images as previously described (Bhat et al., 2015b). All the force curves in the image were batch processed using a 200 nm × 200 nm square in the middle of the cell. Histogram data was exported from the JPK software for statistical analysis and plotting. All data were statistically analyzed using unpaired student’s t-test and one-way ANOVA (GraphPad Prism 5).

The extent of DNA damage was determined by monitoring fluorescently labeled cells for DNA spreading as a function of DNA damage in E. coli WM1074. These changes were examined during short and long time exposures by measuring the DNA halo around single cells, a direct indication of fragmented DNA. After a 5 s exposure to 1 mM 2,4-D, there was a statistically significant (p < 0.03) increase in DNA damage, which further increased (p < 0.0001) after 10, 30, and 60 s exposures compared to controls after 60 s exposure (Figure 1 and Table 2). The DNA damage assay after 3 h and overnight exposure to 4 mM 2,4-D showed significantly increased DNA fragmentation (p < 0.0001, n = 100) compared to formula and sample controls (Figure 1 and Table 2). Positive controls, in which oxidative stress had been induced for 3 h and overnight using 4 mM H₂O₂ (p < 0.0001, n = 100) and elevated temperature at 37°C overnight (p < 0.03, n = 100) and 50°C for 2 h (p < 0.0001, n = 100), all showed increased DNA fragmentation (Figure 1 and Supplementary Figure S2).

Escherichia coli WM1074 (MG1655 lacU169) is a robust and fast growing strain (Archer et al., 2011), forming short rods visible during correlative imaging. Previous studies showed that 2,4-D at concentrations as low as 0.02 mM induces a filamentous phenotype and reactive oxygen species (Bhat et al., 2015b), indicating that 2,4-D impacts cell division in E. coli. In this study we exposed E. coli to 2,4-D while monitoring the localization of GFP-labeled cell division proteins FtsZ and FtsA in real-time using LSCM. Known ROS inducers, paraquat and hydrogen peroxide, were used as positive controls.

Time-lapse images taken in the absence of 2,4-D (sample control and formula control) show the presence of a distinct dynamic Z-ring at the mid-cell, partial rings and bright spots at sub-polar and polar regions (Figures 2A–C). Immediately following the addition of 1 mM 2,4-D, the Z-ring was perturbed within seconds, forming delocalized bright punctate fluorescence away from the center and toward cell periphery (Figures 2D–F). In the absence of 2,4-D, with <1 mM 2,4-D or formula treatment, GFP-FtsA localized in a manner similar to FtsZ-GFP, with a distinct ring structure at the mid-cell (Figures 2G–I). Immediately (<5 s) after the addition of 1 mM 2,4-D, GFP-FtsA delocalized and formed bright fluorescent masses (Figures 2J–L) more concentrated near the cell periphery. Relocalization of FtsZ-GFP and GFP-FtsA was observed for all cells exposed to 2,4-D (>100).

To determine whether the instant delocalization of FtsZ and FtsA could be induced by reactive oxygen species, we exposed E. coli FtsZ-GFP and GFP-FtsA to increasing concentrations of paraquat (0.01–20 mM) and hydrogen peroxide (0.01–10 mM), during live cell imaging in real time. Delocalization of FtsZ was not observed at any exposure level for either reagent (Supplementary Figure S5).

To determine the effects of membrane potential on the localization of FtsZ, and FtsA, E. coli was exposed to ionophores valinomycin and nigericin during real-time live cell imaging. At 30 μM valinomycin and at 10 μM nigericin there was complete delocalization of FtsZ and FtsA within 1 min (Figures 2M–T), resulting in diffuse fluorescence throughout the cell. Nigericin at 5 μM resulted in loss of the Z-ring structure in 50% of the cells, and at 10 μM the Z-ring was completely perturbed in the entire population within 1 min. DMSO had no effect on the localization of FtsZ and FtsA (Figure 2).

Following several hours of live imaging in the presence of 2,4-D (Supplementary Figure S3B) and overnight (15 h) exposure to 2,4-D, E. coli showed a significantly elongated phenotype (p < 0.001, n ≥ 60), (Figure 3) compared to the formula and sample controls (p < 0.03, n ≥ 60) (Table 2 and Supplementary Figure S3).

After 3 h exposure to 2,4-D, WM2026 (FtsZ-GFP) were stained with DAPI and imaged by epifluorescence to determine the location of the Z-ring in relation to the nucleoid, showing 70% of cells lacking a typical Z-ring (n ≥ 100) with a diffuse green fluorescence distributed throughout the cells but absent in nucleoid regions (Figures 3G–I). Elongated cells also showed large uneven nucleoid masses and multiple partially formed Z-rings along the axis (Figures 3D–F). Following overnight exposure (15 h) only 7% of the cells (n ≥ 100) had a typical Z-ring and a large number of cells showed irregular bright fluorescent spots mostly near the membrane. In the absence of 2,4-D, both the control and formula exposed cells had greater than 90% of cells with a centrally positioned Z-ring after 3 and 24 h.

After exposure to 1, 4 mM 2,4-D and 30 μM valinomycin, there was a significant increase in DisC2(3) fluorescence intensity over the course of 1 h (p < 0.0001), whereas unexposed E. coli, and wells containing water and dye or water, dye and 4 mM 2,4-D showed a slight decrease in fluorescence intensity (Figure 4). LSCM also showed cells having significantly increased DisC2(3) fluorescence signal in samples exposed to 2,4-D and valinomycin (Figures 4A–D).

There was no change to SulAp-GFP intensity immediately following exposure to 1 mM 2,4-D during live imaging. After an overnight exposure to 2,4-D, the SulAp-GFP intensity was significantly increased (p < 0.0001, n ≥ 100). Cells exposed to 2 and 4 mM 2,4-D showed a significantly (p < 0.0001) increased intensity of 303.2 ± 120.8 IU and 521.7 ± 159.0 IU, respectively, compared to those of the formula (137.1 ± 40.9 IU) and sample controls (154.3 ± 43.9 IU). The positive controls hydrogen peroxide (1 mM; 105.58 ± 29.3) and paraquat (1 mM; 102.9 ± 33.6 IU) also showed an increased SulAp-GFP intensity after overnight exposure compared to the control (38.3 ± 8.9 IU). In general elongated cells had a higher signal intensity compared to shorter cells (Supplementary Figure S4).

In the absence of 2,4-D, E. coli shows typical Z-ring localization at the mid-cell (Figures 5A–D), however after the addition of 4 mM 2,4-D, cell division arrested abruptly, regardless of the stage of cell division, and remained stagnant (Figures 5E–H). We removed 2,4-D and added fresh media after 4 h, but cell division did not resume even after washing (Figure 4H).

QI^TM height images (Figures 6A,C,E,G) showed a significant change (p < 0.0001, n = 30) in surface roughness after a 20 min exposure to 4 mM 2,4-D, from 5.24 ± 3.23 nm (control) and 7.67 ± 2.23 (formula control) to 22.07 ± 12.2 nm, which did not change significantly after 20 min (p > 0.05). The high standard deviation of surface roughness following 2,4-D exposure indicates high variability. Young’s modulus, an estimate of cell envelope elasticity, was determined from QI^TM force curves at the center of the cell. Values were in the range of 1–4 MPa (Figures 6B,D) for cells in formula and prior to 2,4-D treatment, and did not change significantly during the course of cell division. Elasticity was significantly altered (p < 0.0001, n > 2000) with the addition of 4 mM 2,4-D, such that the average elasticity reduced to over a 100-fold after 20 min (Figures 6F,H,I) and was highly variable. In general, elasticity decreased as a function of exposure time, but to different degrees in each cell. Conversely, surface adhesion increased after 20 min 2,4-D exposure (p < 0.0001, n > 2000), but with no statistically significant increase thereafter (Figure 6J).

Escherichia coli WM1074 (Geissler et al., 2007) is a robust and fast growing environmental strain that serves as a good model to study the impact of xenobiotics. We previously demonstrated that 2,4-D produces ROS and induces a filamentous phenotype in E. coli BL21 DE3 and several genotypically diverse environmental strains at very low concentrations, implicating an impact on cell division (Bhat et al., 2015b). This herbicide also induced elongation in the strain under study (Figure 3) which was more pronounced after an overnight exposure. The rate of elongation was not consistent for all the cells, as some cells appeared more elongated than others, and there was a small population of cells that appeared to be unaffected, rod shaped with a distinct Z-ring (Figures 3H, 5A–D). This discrepancy appears to be common, since individual cells react differently to external stress, including stress induced by mutation (Mileykovskaya et al., 1998; McCool et al., 2004; Justice et al., 2008; French et al., 2017). It is also interesting that compounds that directly interact and inhibit FtsZ produce a greater number of filamentous cells, nonetheless they produce cells with variable lengths (Araujo-Bazan et al., 2016). It is to be expected that not all cells in a given population are in the same metabolic state. If we consider for example stress associated with DNA damage, the cell has a number of different ways to respond (Uphoff et al., 2013) which will directly impact the level of stress experienced by the cell. Furthermore, every molecular process in the cell is in a constant state of flux, with only a snap shot observed during the imaging process. Consistent with this idea, 6 and 15% of the cells are filamentous after 3 and 15 h, respectively, greater than three times that of the control. The filamentous cells, as expected, showed increased DNA fragmentation and SOS response, indicating that the elongated cells were under greater stress compared to average sized cells (Supplementary Figure S4).

There was no immediate increase in SulAp-GFP during 2,4-D exposure, likely since any subtle immediate effects would be below the detection limit of the microscope and could be convoluted with photo-bleaching. The increased SulAp-GFP intensity after 3 h and overnight exposure indicates increased SOS response due to oxidative stress. Oxidative stress that induces DNA damage in bacteria activates RecA, a coprotease that helps autocleave the LexA repressor. In the absence of LexA, the operator sequences allow expression of more than 40 SOS genes, including SulA, which halt cell division and repair damaged DNA (Janion, 2008). E. coli WM2739 is a strain with SulAp-GFP, a reporter of the SOS response with the SulA promoter under the control of LexA regulation. SulAp-GFP showed a uniformly dim signal in the absence of 2,4-D, indicating low constitutive basal SulA expression, and an increase in expression indicative of oxidative stress and DNA damage (McCool et al., 2004; Nazir and Harinarayanan, 2015). SulA is a major repair protein in the SOS operon which inhibits cell division, prolonging the cell cycle during which the damaged DNA can be repaired (Imlay, 2013). An increase in the SulA:FtsZ ratio is known to cause cell filamentation, since SulA binds to FtsZ when present at higher concentrations, inhibiting cell division (Dajkovic et al., 2008). SulA works by interacting with the catalytic site on FtsZ, helping to disassemble the existing Z-ring and sequestering FtsZ monomers to prevent its further assembly (Huang et al., 1996; Chen et al., 2012), all consistent with the filamentous phenotype observed at longer exposure times (Figure 3). Elongated cells also produced more SulAp-GFP compared to average sized cells, lending further support for this idea (Supplementary Figure S4).

Cell division in E. coli is initiated by septum formation, facilitated by the accumulation of FtsZ and its colocalizing partners, which form a membrane associated complex called the divisome (Natale et al., 2013). FtsZ undergoes GTP-dependent self-polymerization to form a highly dynamic scaffold called the Z-ring, for which division proteins are in constant flux between the cytosol and the divisome (Anderson et al., 2004). FtsZ-GFP was observed as bright fluorescent spots that moved along the cell’s longitudinal axis and formed mid-cell bands that eventually constrict, all the while forming a second ring at the middle of daughter cells preparing for the next division (Adams and Errington, 2009). Cell constriction was clearly visible in the AFM images, for which constriction began following complete Z-ring formation observed by LSCM (Figure 4).

We monitored the effects of 2,4-D exposure by fluorescence assay using the carbocyanine dye, 3,3^′-diethylthiacarbocyanine iodide [DisC2(3)], known to be a good indicator of membrane potential (Epand et al., 2010; Te Winkel et al., 2016). E. coli membrane potential began to dissipate almost immediately following exposure (5 min) to 1 mM, 4 mM 2,4-D and 30 μM valinomycin, as indicated by a significantly increased fluorescence signal (Figure 4). There was reduced fluorescence in control cells, consistent with fluorescence quenching in polarized cells. A reduction in florescence with just water and 2,4-D indicates a possible reaction between the two, but regardless, treated samples had significantly increased fluorescence (p < 0.0001), indicating a rapid loss in membrane potential. It is well known that 2,4-D is lipophilic, altering the fluidity of the cell membrane (Fabra de Peretti et al., 1992; Viegas et al., 2005), and affecting oxidative phosphorylation (Bhat et al., 2015b). This is the first study to show 2,4-D directly causing membrane depolarization, resulting in an immediate loss of the Z-ring and cell division.

FtsZ-GFP and DisC2(3) were imaged simultaneously by LSCM to confirm the loss of the Z-ring, along with the dissipation of the membrane potential. Indeed, cells exposed to 1 mM, 4 mM 2,4-D and valinomycin had no Z-rings and increased fluorescence, the latter indicating depolarized membranes (Figures 4A–H). There was also variation in the DisC2 intensity in different cells, showing heterogeneity of the stress response.

The immediate change in FtsZ localization (Figure 2) when exposed to 1 mM 2,4-D, valinomycin and nigericin can be largely explained by changes to membrane potential. The positive control valinomycin is a K⁺ carrier ionophore that specifically dissipates the membrane potential. Nigericin facilitates electroneutral exchange of H⁺ and K⁺ ions, thereby depleting ΔpH between the cytosol and external environment (Ahmed and Booth, 1983). Rapid perturbation of the Z-ring during 2,4-D exposure in a manner similar to those of the positive controls indicates that 2,4-D disassembles the divisome complex through alteration to the membrane potential. Absence of the Z-ring during longer 2,4-D exposures indicates the cell’s inability to reverse divisome disassembly with increased DNA damage. Sub-lethal 2,4-D (1 mM) is insufficient to kill all cells in the population, some of which likely overcome the stress and continue to divide, consistent with some E. coli showing Z-rings following 2,4-D exposure after the 3 h mark (Figure 3H).

Our previous metabolomics study showed that 2,4-D inhibits oxidative phosphorylation in E. coli (Bhat et al., 2015b) known to be a direct consequence of membrane potential dissipation (Lewis et al., 1994). Since FtsZ is GTP dependent, loss of FtsZ assembly may be a direct consequence of changes in cellular respiration and membrane potential. Membrane potential is crucial for the stability of the divisome complex (Strahl and Hamoen, 2010), which could directly result in Z-ring collapse on the time scales observed in this study. Indole, a structural analog of 2,4-D, blocks E. coli cell division by disrupting the MinCD oscillation as a result of reduced membrane potential crucial for cell division (Chimerel et al., 2012). Consistent with this idea, we observed FtsA changing its localization in a manner similar to FtsZ following 2,4-D exposure (Figures 2G–L). It has been demonstrated that FtsA and ZipA compete for the C-terminal end of FtsZ to form membrane tethers and colocalize within the Z-ring structure. It has been shown that upon SulA induction and disassembly of the Z-ring, FtsZ, FtsA and ZipA appear as punctate, short polymeric structures, appearing as patches on the membrane (Rowlett and Margolin, 2014). Altered membrane organization induced by 2,4-D would have a direct impact on membrane-associated FtsA, which could lead to disassembly of the Z-ring. Changes to FtsA localization and its patchy distribution indicate that 2,4-D may alter the assembly of the entire divisome, possibly impacting other cell division proteins.

Previous studies show 2,4-D to bind directly and affect polymerization of purified neuronal tubulin in vitro (Rosso et al., 2000), so it could have a similar impact on its structural homolog FtsZ in E. coli. Changes to FtsZ structures have been previously documented in E. coli, showing punctate polymeric structures throughout the cell upon its disassembly, with the induction of SulA (Rowlett and Margolin, 2014; Vedyaykin et al., 2014). FtsZ in Bacillus subtilis exposed to benzamide appears as randomly distributed dynamic foci, having no specific localization (Adams et al., 2011). Resveratrol inhibits E. coli division and causes elongation by inhibiting Z-ring formation and FtsZ expression, causing DNA damage and an upregulated SOS response (Hwang and Lim, 2015). Together with this study, it can be concluded that 2,4-D disrupts the divisome by immediately dissipating the membrane potential, causes DNA damage which induces the SOS response, ultimately leading to cell elongation during longer exposures.

Although no prior study has characterized the effects of aromatic compounds on the bacterial surface in live cells, there are extensive studies on fixed cells imaged in air following exposure to antimicrobial peptides (Fantner et al., 2010), essential oils (La Storia et al., 2011) and antibiotics (Yang et al., 2006), all showing changes to surface ultrastructure as a primary response to external stress. Our results are consistent with the impact of other antimicrobial compounds, with alterations to surface roughness and envelope elasticity, but we have resolved the temporal changes. E. coli exposed to sub-lethal levels of 2,4-D had a 100-fold lower surface elasticity within 30 min, further reduced over a period of a few hours, indicating changes to envelope compliance over time. Removal of 2,4-D from the imaging medium did not reverse this effect, indicating an irreversible alteration to cell envelope physical properties and implying that 2,4-D likely perturbs a select group of cell surface molecules.

The outer membrane of E. coli is highly complex and asymmetric, made up of lipids, long chain lipopolysaccharides (LPS) and membrane proteins (Kleanthous and Armitage, 2015). Known to be lipophilic (Benndorf et al., 2006), 2,4-D likely reacts with LPS which could lead to the observed changes in surface elasticity and adhesion. The E. coli cell surface contains ∼90% LPS, making its surface negatively charged, consistent with lower adhesion between the negatively charged silicon tip and bacteria compared to the Cell-Tak^TM covered glass surface. We speculate that an increase in adhesion immediately after exposure to 2,4-D (Figure 6J) likely indicates a rearrangement of LPS, in accordance with more compliant and rougher cells. We did not observe increased adhesion after 20 min, indicating that 2,4-D irreversibly interacts in a short period of time, likely with the entire surface exposed LPS.

This is the first report showing that 2,4-D is capable of arresting cell division in E. coli within seconds by altering membrane potential, FtsZ and FtsA localization, accompanied by DNA damage and the SOS response. Longer exposures resulted in an even greater SOS response, irregular Z-ring formation and improperly divided cells, giving rise to cell filamentation. Simultaneous real-time AFM imaging of live E. coli showed time-dependent changes to surface roughness, adhesion and elasticity, indicating a direct interaction between 2,4-D and the envelope surface.