Suicidal chemotaxis in bacteria

Surface-attached bacteria move towards antibiotics via twitching motility

We used microfluidic devices and automated cell tracking to quantify the movement of P. aeruginosa cells as they are exposed to well-defined spatial gradients of antibiotics in developing biofilms (Fig. 1). We began with the antibiotic ciprofloxacin, which is widely used to treat P. aeruginosa infections^27,28. To set a baseline, we first determined the minimum inhibitory concentration (hereafter MIC) of ciprofloxacin for P. aeruginosa (strain PAO1) in shaking cultures, which agrees with the published MIC of this strain (Fig. S1²⁹). We then exposed surface-attached cells to an antibiotic gradient in a microfluidic device where the antibiotic concentration ranged from zero to 10 times the MIC (Fig. 1A, B, Methods). After approximately 5 h of unbiased movement, we were surprised to see that twitching cells began to bias their movement towards increasing concentrations of ciprofloxacin (Fig. 1B, D, Movie 1). The movement bias, β, defined as the number of cells moving up the gradient divided by the number of cells moving down the gradient, peaks after approximately 10 h and then decays as the surface becomes crowded with cells (Movie 1) and tracking becomes difficult (Methods). The flow through the device also has a small influence on the direction of cell movement because it tends to pull cells in the downstream direction (Fig. 1B, C, E). However, this fluid flow is orthogonal to the direction of the antibiotic gradient, and so does not explain the movement towards antibiotics.

We explored gradients that varied in steepness, always starting at a ciprofloxacin concentration of zero and up to a maximum of 0.1, 1, 10, 100 and 1000 times the MIC, whilst keeping the length scales of the gradient constant. Whenever the maximum concentration was higher than the MIC, cells biased their movement towards the antibiotics (peaking at β ≈ 2.5 to 4, Fig. 1D and S2). These experiments revealed that cells are capable of entering regions of extremely high concentration (up to 1000 times MIC) and remain motile for the remainder of the experiments (≈ 5 h; Fig. 1D, S3, Methods). In contrast, the direction of cell movement in a control without antibiotic was approximately random (β ≈ 1, Fig. 1C). Importantly, we also see a similar response with other strains of P. aeruginosa (PAK and PA14; Fig. S4 and Movie 2).

This motility response is not limited to ciprofloxacin: we found that cells also bias their movement up gradients of antibiotics belonging to different antibiotic classes that have different chemistries and mechanisms of action and do so on a similar timescale to the response seen with ciprofloxacin (Fig. 1E, S5, S6). Ciprofloxacin causes DNA damage by inhibiting an enzyme (gyrase) involved in DNA coiling, but we found that cells also move towards: rifampicin, which targets an RNA polymerase; streptomycin, which targets the ribosome; and mitomycin C, which is again a DNA damaging agent but one that acts by cross-linking DNA. We also tested carbenicillin and other β-lactam antibiotics, but these caused extreme cell elongation that inhibited cell movement in our assay and so we did not study them further (Fig. S7). The diversity in both the chemistry and mechanism of action of the antibiotics that elicit biased movement suggests a general response to the cell damage caused by antibiotics, rather than to the specific antibiotics themselves. Indeed, we also find that cells move towards the antimicrobial compound hydrogen peroxide (Fig. S8). Further consistent with the importance of cell damage, we found that sub-MIC concentrations of ciprofloxacin do not bias cell motility (Fig. 1D and S2).

The movement of cells in antibiotic gradients appeared to conform to the definition of chemotaxis: the ability of cells to actively bias their motion in response to chemical gradients¹⁶. How, though, are cells biasing their motility towards antibiotics? From previous work, we know that cell movement in our assays is driven by twitching motility and that these twitching cells can actively bias their movement towards beneficial nutrients by reversing their movement direction more frequently when moving away from a nutrient source²³. Is biased motility in antibiotic gradients driven by the same behaviour? To test this hypothesis, we began by quantifying cellular reversal rates and found that cells moving away from ciprofloxacin actively reversed direction more frequently than cells moving towards it (Fig. S9). Our previous study showed that cells lacking the key response regulator of twitching chemotaxis, PilG, remain motile but have significantly reduced reversal rates compared to wild-type (WT) cells and are unable to chemotax towards beneficial nutrients and other chemoattractants^23,24,25. Consistent with this previous report, we find here that cells with an in-frame deletion of pilG (ΔpilG) also have a significantly reduced movement bias towards increasing ciprofloxacin concentrations, despite remaining motile in this assay and having an identical MIC as WT cells (Fig. S1, S10 and Movie 3). As a result, the pilG mutant, unlike the WT, is unable to accumulate in regions containing high ciprofloxacin concentrations (Fig. S11). It is worth noting that the ΔpilG cells do show reduced motility in the assay relative to the WT cells but, importantly, our measurements of movement bias only use cells that demonstrate appreciable movement, thus preventing non-motile cells of either genotype from influencing our measurements (Methods). When this ΔpilG strain is complemented at the pilG locus, its ability to move towards ciprofloxacin is restored (Fig. S10 and Movie 3). Taken together, these results suggest a common behavioural basis for twitching chemotaxis towards antibiotics and nutrients.

Nutrient gradients do not explain chemotaxis towards antibiotics

Cells only begin to bias their motility up antibiotic gradients after an initial ≈5 h period of nearly random motility (Fig. 1D and S2). This delayed response introduces an important complication, as secondary gradients in nutrients and other compounds potentially released by cells are expected to build up in the device over time, which might indirectly drive the movement towards antibiotics. In particular, cells situated in regions of the device with lower antibiotic concentrations can rapidly proliferate, while cells initially in regions with higher antibiotic concentrations either tend to detach or die (Movie 1). The different numbers of cells on either side of the device, both within the test section (Fig. 1A) and in the regions upstream, means that emergent chemical gradients can form because the cells consume nutrients and release compounds at different rates on either side of the device. Thus, cell movement towards higher antibiotic concentrations could be, in principle, driven by movement towards higher nutrient concentrations and towards lower concentrations of the diverse set of compounds released by cells. We, therefore, sought to confirm whether such secondary gradients could be responsible for the biased movement we observe, rather than it being a direct response to the antibiotic gradients themselves. We first tried switching the direction of the antibiotic gradient, after the cells had established themselves in the regions with lower antibiotic concentrations. However, the sudden change in antibiotic concentration caused cells to either stop moving or detach altogether. We, therefore, sought alternative approaches to control for the formation of secondary de novo gradients.

We focused first on whether putative emergent nutrient gradients could be responsible for the observed biased movement towards antibiotics. Previous work has shown that surface-attached P. aeruginosa cells undergo chemotaxis up gradients of the metabolisable carbon source succinate²³. Consistent with this, we found that twitching cells will also bias their motility up gradients of tryptone, the growth medium used in this study (Fig. 2A and Fig. S12), a response that has also been observed in swimming P. aeruginosa cells³⁰. To test whether the biased movement we see in our experiments with antibiotics could be driven solely by de novo gradients in tryptone, we created opposing gradients of tryptone and ciprofloxacin in our microfluidic device. Specifically, we injected full strength tryptone through one inlet of the device and ciprofloxacin mixed with tryptone at 10% of the regular concentration through the other inlet. We use 10% media rather than 0% because, with the latter, we find that cells quickly stop moving in regions without nutrients. If cells in our previous experiments are simply moving in response to nutrient gradients, we expect that cells would now move away from the antibiotic source. While cells initially move in the direction of increased tryptone, after ≈5 h this response rapidly drops off and after ≈7.5 h, cells again exhibit biased movement in the opposite direction towards increasing antibiotic concentrations (Fig. 2A and Fig. S12). The robust movement of cells towards low nutrient and high antibiotic concentrations so early in the experiment suggests that chemotaxis towards antibiotics (Fig. 1) is not driven by nutrient gradients.

Repulsion by cell-free supernatant does not explain chemotaxis towards antibiotics

We next explored the possibility that biased movement in antibiotic gradients could be explained by products released by cells that might trigger repulsion away from regions of the device with high cell density. Consistent with the potential for repulsion from areas of high cell density, we find that twitching cells are repelled by cell-free supernatant extracted from cells grown to high density under static conditions in the multi-well plate assays that are widely used to study biofilm physiology (Fig. 2B, S12, Methods^13,31). Specifically, we mixed cell-free supernatant collected from static well-plates with fresh media (20% supernatant, 80% tryptone broth) and injected it through one inlet of our device, whilst media mixed with the same proportion of water (20% water, 80% tryptone broth) was injected through the other inlet. These experiments showed that cells move away from the channel containing supernatant (Fig. 2B and Fig. S12, purple lines). This result is interesting as it suggests that in P. aeruginosa, twitching motility will guide cells away from regions of high cell density in developing biofilms, potentially facilitating the colonisation of new territory. However, it also raises the possibility that this repulsion could explain the migration towards antibiotics. We, therefore, performed a number of additional experiments to explore this possibility.

We reasoned that if a factor released by cells is responsible for movement towards antibiotics, then adding cell-free supernatant in the background of an antibiotic gradient should limit or stop the movement of cells toward higher antibiotic concentrations. We tested this hypothesis with two additional experiments. The first experiment exposed cells to a gradient of ciprofloxacin in a uniform background of cell-free supernatant. We again used cell-free supernatant collected from high-density static cultures at 20% concentration, which we expected to saturate any gradients in cell products that might occur de novo in the microfluidic device where cell densities are initially low (Fig. 2B and Fig. S12, green lines). The second experiment exposed cells to a ciprofloxacin gradient and supernatant gradient simultaneously, with the larger concentration of both on the same side of the device, so that repulsion of cells away from supernatant would oppose the putative movement towards ciprofloxacin (Fig. 2B and Fig. S12, yellow lines). We still observed movement towards ciprofloxacin in both of these experiments. In fact, we found that movement towards ciprofloxacin actually commences more rapidly in the presence of the cell-free supernatant than without (Fig. 2B and Fig. S12, red lines). This more rapid response could occur because the cell-free supernatant helps the cells to tolerate antibiotics or because a factor in the supernatant promotes the chemotaxis response itself. Either way, it shows that the delay before bacteria begin the chemotaxis response is labile and can be greatly reduced under some conditions (see also Fig. S4). Importantly, this rapid migration towards antibiotics occurred even when the cell-free supernatant gradient was oriented such that one would expect it to repel cells away from the antibiotics. In sum, repulsion due to factors present in cell-free supernatant can promote cell movement from regions of high cell density, but this effect does not explain the observed chemotaxis towards ciprofloxacin.

There remains the possibility that chemotaxis towards antibiotics is driven by a repulsive factor, or set of factors, that are not present in the cell-free supernatant used in the above experiments, which was collected from static cultures grown in multi-well plates. For example, it may be that cells growing under flow in the microfluidic device have different physiology, such that any secondary chemical gradients that develop within the device are not captured by adding cell-free supernatant collected from static cultures on the bench. We, therefore, repeated the cell-free supernatant experiments using media collected directly from cells growing within microfluidic devices (Fig. S13, Movie 4, Methods). To maximise the likelihood of observing a response, we used 100% cell-free supernatant for these experiments and flowed this through one inlet, with growth media (tryptone broth) through the other. This revealed that, as for the standing culture experiments, cells move away from the channel with cell-free supernatant obtained from microfluidic devices. With 100% cell-free supernatant, this response is likely to be a combination of both moving away from cell-free supernatant and towards an increasing nutrient concentration. When we combined the cell-free supernatant with 10X MIC ciprofloxacin, cell movement was initially biased away from the cell-free supernatant/ ciprofloxacin. However, importantly, after ≈7 h, cells began to move in the opposite direction, towards the cell-free supernatant and ciprofloxacin. The addition of cell-free supernatant on the antibiotic side of the device, therefore, does not prevent chemotaxis towards the antibiotic, as would be expected if cell products released in the device were driving the response. These experiments again suggest that repulsion by cell-free supernatant is not sufficient to explain migration towards antibiotics.

A factor released in response to antibiotics does not explain chemotaxis

An additional possibility is that chemotaxis is driven by a repulsive factor that is not present in the cell-free supernatants we studied. These supernatants were collected from experiments without antibiotic present, as any residual antibiotic remaining in the collected supernatant would complicate interpretation. However, this leaves open the possibility that surface-attached cells within our microfluidic experiments might release a compound in response to the antibiotics (i.e. antibiotic-induced cell products) that could drive directional cell movement, something that has been observed for swarming bacteria³². To examine this possibility, we used a classical result from the chemotaxis literature: movement bias is predicted to increase with the strength of the chemical gradient. This correlation has not only been demonstrated in chemotaxis occurring in twitching bacteria²³, but also in swimming bacteria¹⁵ and eukaryotic cells³³. Since the distribution of ciprofloxacin in our microfluidic devices is predicted to differ starkly from that of putative antibiotic-induced cell products (Figs. S14, S15), measuring how movement bias changes at different positions within the device gives us the ability to distinguish which of these alternatives is most likely to explain the observed patterns of cell movement.

In order to do this, we first wanted to estimate how the concentration of ciprofloxacin varies in space in the device. Here, a model of diffusion was used to quantify how the concentration of ciprofloxacin, and its spatial gradient, varies within our device (Methods). Using a diffusion coefficient of D = 200 µm² s⁻¹, the isocontour corresponding to the MIC of our strain (1X MIC) was found to closely match the position where biofilm growth becomes strongly suppressed by antibiotics (dashed line, Fig. 2C, E). If the mean flow speed within the device is increased by a factor of three (from 42.3 µm sec⁻¹ used as our baseline, to 127 µm sec⁻¹, see Methods), the 1X MIC isocontour was predicted to be pushed further downstream, which again closely matches the distribution of biofilm experimentally observed under this higher flow condition (Fig. 2E), suggesting our model is working as expected. While the diffusion coefficient of ciprofloxacin is not as well-known as some other compounds, our fitted value of D (where the 1X MIC isocontour follows the line where biofilm growth is suppressed) is within a factor of two of previous estimates for ciprofloxacin^34,35,36. We also tested how different values of D would affect the distribution of antibiotics in our microfluidic device (Fig. S16). These analyses indicate that the lower boundary of the thick biofilm in our device might occur at a slightly higher or lower concentration compared with the MIC measured in shaking liquid culture⁶. However, the general features of the distribution of antibiotics in our device are not sensitive to the precise value of D. In particular, no matter what value of D is used, the concentration is always half the maximum along the centreline and the steepest gradients always occur along the centreline. In the subsequent analyses, we assumed D = 200 µm² s⁻¹. These experiments also reveal that within the region close to the centreline (where the gradient is steepest), cell movement is highly biased towards antibiotics (Fig. 2C–G). This process can be observed dynamically in the inset of Movie 1, where cells along the centreline of the device move between a region of high cell density on the antibiotic-free side of the channel and a second region of high cell density that forms as cells begin to accumulate in high antibiotic concentrations.

Next, we calculated the movement bias, β, as a function of both time and the local ciprofloxacin gradient (∂C/∂y). These analyses used a high flow speed (mean speed = 127 µm sec⁻¹) so that chemotaxis could be observed further downstream in the device (Fig. 2E, F). This allowed us to simultaneously image cells in four different fields of view so that we had a sufficient number of cell trajectories in each bin (Fig. 2E, F, Methods). Our analyses reveal that the strength of the ciprofloxacin gradient (∂C/∂y) is an excellent predictor of β – after an initial period of random motility, cells experiencing a larger ∂C/∂y were observed to have a larger β (Fig. 2G and Fig. S17). These results indicate that the stimulus the cells are responding to closely matches the predicted gradient of ciprofloxacin within our device. While the gradient of ciprofloxacin decreases as one moves downstream, the gradient of a putative antibiotic-induced cell product would likely increase in the downstream direction. This is because as fluid passes through the device, it passes by more and more cells exposed to sub-MIC levels of antibiotics, which would steadily increase the concentration of the product on one side of the device (Figs. S14, S15). While the exact distribution of a putative cell product is hard to predict, the fact that it would differ strongly from that of ciprofloxacin suggests that the excellent correlation between ∂C/∂y and β (Fig. 2G and Fig. S17) is unlikely to be driven by antibiotic-induced cell products.

Taken together, our data suggest that neither nutrient depletion, nor products released either in the presence or absence of antibiotics, can explain the movement of cells towards ciprofloxacin. These results suggest that we are observing a direct response to the antibiotics. There is a growing body of work showing that flagella-based swimming can transport bacteria into regions of high antibiotic concentration^37,38,39,40. However, to the best of our knowledge, there is no evidence of chemotaxis in response to the antibiotic gradients themselves. Here, we observe that cells move towards antibiotics, actively reverse direction to facilitate this directed movement and, finally, that movement bias increases with the strength of the antibiotic gradient, which are all signatures of chemotaxis^15,16. As discussed above, this chemotaxis occurs towards a range of antibiotics (and hydrogen peroxide) that have different mechanisms of action, which suggests that the biased migration is driven by a general response to the effects of antibiotics on cells.

Cells migrate towards antibiotics and die

Our data show that P. aeruginosa cells will bias their motility to move into regions with extremely high antibiotic concentrations. Moreover, many cells remain motile at concentrations many times greater than that needed to prevent their growth in a standard liquid culture assay (Fig. S3). This ability to persist in the presence of antibiotics is likely to be enabled by stress responses, such as the SOS response to DNA damage in the case of ciprofloxacin. This response stalls cell division while cells attempt to repair DNA damage⁴¹. We wanted to establish the ultimate fate of the migrated cells: do they remain viable and gain tolerance to high concentrations of antibiotics, or are they terminal and doomed to die? Or indeed, is it possible that some cells evolve mutations that confer resistance in the short time-scale (≈20 h) of our experiments^37,39? The challenge with answering this question is that the cells in question are fully encased within PDMS-based microfluidic devices, which means it is not possible to isolate migrating cells and study them further. One can watch them for the duration of the experiment, but after ≈20 h, the devices become clogged with growing cells. We are, therefore, unable to ascertain whether cells that migrate into regions of high antibiotic concentration remain active and retain long-term viability.

In order to overcome this challenge, we moved to a novel assay that uses open “fluid-walled” microfluidics⁴², where twitching cells can be studied in a chemical gradient and then isolated after they have migrated, (Methods; for a detailed description of our approach, see⁴³). Briefly, microfluidic devices (Fig. S18) were inoculated with a 1:1 co-culture of YFP-labelled WT cells and unlabelled, chemotaxis-null ΔpilG cells, which allowed us to follow cells that can and cannot perform chemotaxis. We then followed cell movement under the microscope until a substantial number of WT cells had begun to migrate towards ciprofloxacin (C_MAX = 100X MIC, t = 24 h; Fig. 3A–C). To reconfigure the fluid-walled channels, we mounted a PTFE (polytetrafluoroethylene) stylus⁴⁴ to the microscope’s condenser and moved the device relative to the stylus using the microscope’s motorised stage. This technique allowed us to reconfigure the central channel into segments in real-time and with a high level of accuracy (tens of micrometres). Thus, we could physically isolate the cells that had performed chemotaxis within a fluid-walled chamber (Fig. 3D, E). Importantly, these cells were strongly biased towards WT rather than the chemotaxis-impaired ΔpilG cells (mean = 84% WT, Fig. 3E), confirming our ability to isolate cells that could undergo chemotaxis. To analyse the viability of these isolated cells, we extracted the contents of the chambers and plated them onto agar plates without antibiotics to monitor whether any of the cells were able to form viable colonies. Additionally, we replaced the chamber contents with fresh growth media without antibiotics to monitor the health of any cells that remained attached to the surface of the microfluidic chambers.

We reasoned that cell viability might depend on how far a cell has migrated up the ciprofloxacin gradient and, thus the ciprofloxacin concentration it has experienced. Therefore, we carried out three separate experiments in which the microfluidic channels were reconfigured at different positions, allowing us to isolate three different sub-populations of cells that had each experienced a different minimum concentration of ciprofloxacin. In the first position, we reconfigured the channel to isolate only those cells that had already undergone chemotaxis and thus begun to accumulate in regions of high ciprofloxacin concentration (“Position C”, Fig. 3A). The cells extracted from this sub-population were found to be incapable of forming colonies on antibiotic-free agar (Fig. 3F), and none of the cells that remained within the isolated fluid chamber underwent any further cell division in over 36 h (Fig. 3G). To confirm that the fluid chamber provided a suitable environment for cell growth, we inoculated ≈25 healthy cells into the chamber at the end of the experiment. These freshly-inoculated cells grew rapidly, completely covering the bottom surface of the chamber within 24 h (dotted line, Fig. 3G), demonstrating that the chamber did not contain residual antibiotics that could stifle growth.

We next reconfigured the channel to isolate both cells that had already undergone chemotaxis as well as the majority of cells that were still migrating (“Position B”, Fig. 3A). Once again, none of these cells showed evidence of long-term viability (Fig. 3F, G). Finally, we reconfigured the channel to isolate all cells that were undergoing or had already undergone chemotaxis, including those that had only very recently begun to start moving up the ciprofloxacin gradient (“Position A”, Fig. 3A). Once again, the extracted cells were found to be incapable of forming colonies (Fig. 3F). Whilst there was initially no detectable cell growth within the fluid chamber, after ~20 h, the early stages of a biofilm containing both WT and ΔpilG cells was observed forming on the surfaces of the fluid chamber (Fig. 3G). After a further 10 h, the growth of this biofilm appeared to have stalled and so the entire chamber contents were extracted and plated onto antibiotic-free nutrient-rich agar plates, on which a relatively small number of colonies were recovered (Fig. 3H). Interestingly, while the majority of cells seen to migrate towards antibiotics were WT, the majority of colonies that grew were chemotaxis-impaired ΔpilG cells, which were less likely to enter the regions of high antibiotic concentration. Taken together, these three separate channel reconfigurations suggest that whilst some cells retain long-term viability at the earliest stages of chemotaxis, they will continue to move into higher antibiotic concentrations where they rapidly lose this viability and will ultimately die.

Twitching bacteria move towards supernatant from competitors

Why does P. aeruginosa perform an ultimately fatal migration towards antibiotics? We hypothesized that the observed migration towards clinical antibiotics may have its evolutionary origins in the natural ecology of P. aeruginosa. Clinical antibiotics are widely reported to induce biofilm formation and this can be recapitulated in P. aeruginosa using the cell-free supernatant of other P. aeruginosa strains¹³. However, this effect is only observed when the supernatant is toxic to the focal strain, which is consistent with the hypothesis of competition sensing: an evolved ability to detect and respond to harmful competing strains⁷. Under this hypothesis, the presence of clinical antibiotics causes bacteria to act as though the attack is coming from ecological competitors¹³.

We, therefore, asked whether cells might bias their movement up gradients of toxic cell-free supernatant isolated from different P. aeruginosa strains, in a similar way to their response to clinical antibiotics. Based on previous work, we selected one strain (‘Strain 7’) that strongly inhibits the growth of our focal strain PAO1 in coculture and a second strain (PA14) that does not (Fig. 4A¹³). Consistent with our hypothesis, we observed biased movement towards the toxic supernatant that stifles cell growth, but unbiased movement in a gradient of supernatant from the non-toxic strain (both using a supernatant concentration of 10%, Fig. 4A and Fig. S19). These data again suggest that the response we observe with clinical antibiotics is a general response to a gradient of toxic components. But why would cells move towards toxic compounds? Moving towards toxic stimuli may be a maladaptive response driven by their adverse effects on cells. Alternatively, it may constitute part of a general strategy to attack and invade the territory of neighbouring genotypes that present a threat. Consistent with the latter, the evolution of mass suicide to release toxins was recently documented in non-motile Escherichia coli⁹. Such aggressive strategies also have a precedent in the animal world, particularly in the social insects that are well known for territorial behaviours, where workers will actively seek out and attack neighbouring colonies and other harmful species, and die in the process⁴⁵.

In this model, P. aeruginosa has evolved to move towards competitors in order to engage in a counter-attack. If this model is correct, we reasoned one should see the release of antimicrobials by P. aeruginosa as it migrates up the gradient of antibiotics or toxic cell-free supernatant. In order to test this hypothesis, we built a fluorescent transcriptional reporter for the Pyocin R2 of P. aeruginosa, which is induced by DNA damage and released via programmed cell suicide that causes cell lysis⁴⁶. We observed this strain in gradients of ciprofloxacin and either the toxic, or non-toxic, cell-free supernatant of another P. aeruginosa strain. These experiments confirmed that many cells expressed pyocins and often performed explosive lysis, which is known to release pyocins (Fig. 4B–F, see also⁴⁷). These lysis events can be seen clearly in the presence of propidium iodine which stains the free DNA released upon lysis, observed as a transient puff of colour (Fig. 4B, Movie 5). When combined with chemotaxis, the result is the biased movement of pyocin-releasing cells towards the source of toxins (Fig. 4C). The patterns of pyocin release are particularly striking in the case of ciprofloxacin (Fig. 4D, Movie 5). This observation fits with the known regulation of pyocins, which is DNA damage dependent^29,46, and suggests that DNA-damaging agents (like ciprofloxacin) will lead to a particularly aggressive response. Consistent with this, the release of colicin bacteriocins in E. coli is also known to be upregulated by DNA-damaging toxins, albeit in stationary cells⁹. Altogether, our experiments suggest that P. aeruginosa has evolved to bias its twitching motility towards harmful compounds while releasing bacteriocins, which is consistent with a counter-attack behaviour.

Source: Ecology - nature.com