Bacteria living as biofilm are frequently reported to exhibit inherent tolerance to antimicrobial compounds, and might therefore contribute to the persistence of infections. Antimicrobial peptides are attracting increasing interest as new potential antimicrobial therapeutics; however, little is known about potential mechanisms, which might contribute to resistance or tolerance development towards these compounds in biofilms. Here we provide evidence that a spatially distinct subpopulation of metabolically active cells in Pseudomonas aeruginosa biofilms is able to develop tolerance to the antimicrobial peptide colistin. On the contrary, biofilm cells exhibiting low metabolic activity were killed by colistin. We demonstrate that the subpopulation of metabolically active cells is able to adapt to colistin by inducing a specific adaptation mechanism mediated by the pmr operon, as well as an unspecific adaptation mechanism mediated by the mexAB-oprM genes. Mutants defective in either pmr-mediated lipopolysaccharide modification or in mexAB-oprM-mediated antimicrobial efflux were not able to develop a tolerant subpopulation in biofilms. In contrast to the observed pattern of colistin-mediated killing in biofilms, conventional antimicrobial compounds such as ciprofloxacin and tetracycline were found to specifically kill the subpopulation of metabolically active biofilm cells, whereas the subpopulation exhibiting low metabolic activity survived the treatment. Consequently, targeting the two physiologically distinct subpopulations by combined antimicrobial treatment with either ciprofloxacin and colistin or tetracycline and colistin almost completely eradicated all biofilm cells.
In nature, bacteria are constantly exposed to a variety of small bioactive natural products, also referred to as secondary metabolites. Secondary metabolites are produced by virtually all living organisms on earth. The ecological role of these chemical compounds is still a matter of debate, and might range from waste product removal to signalling or defence. Some of these compounds can exhibit bacteriocidal or bacteriostatic activities and are now frequently used in medical settings as antibiotics to treat bacterial infections. However, bacteria have evolved different strategies to sense, respond and adapt to these small chemical compounds. In the present study, we investigate the spatiotemporal-dependent effects and mechanisms taking place in an established microbial community of Pseudomonas aeruginosa upon exposure to the antimicrobial peptide (AMP) colistin.
Pseudomonas aeruginosa is a metabolically versatile Gram-negative bacterium, which can be found as natural inhabitant of terrestrial and aquatic environments as well as associated with animals and plants (Ramos, 2004). In humans, it can frequently cause life-threatening infections under conditions where the host is injured, the immune system is compromised, or in individuals who are afflicted with cystic fibrosis (CF) (Bodey et al., 1983; Lyczak et al., 2002).
Haagensen et al. (2007) recently presented evidence that the pmr operon is involved in the development of a colistin-tolerant subpopulation of cells in P. aeruginosa PAO1 flow-chamber-grown biofilms. In addition to colistin tolerance, the cells of this particular subpopulation were found to exhibit surface-associated motility. However, whether there was a link between tolerance development to colistin and motility, or whether the apparent correlation was indirect was not clear. Here we provide evidence that development of tolerance to colistin in mature P. aeruginosa biofilms is dependent on cellular metabolic activity but independent of cellular motility. The metabolically active cells are able to adapt to colistin treatment by inducing a specific adaptation mechanism mediated by the pmr operon, as well as an unspecific adaptation mechanism mediated by mexAB-oprM. Moreover, we show that by specifically targeting physiologically distinct subpopulations present in a biofilm, it is possible to eradicate most of the cells in the biofilm.
When P. aeruginosa PAO1 is grown under continuous culture conditions in flow-chambers for 4 days with glucose as carbon source, it can form mature multicellular structures of mushroom-like shape (e.g. Fig. 1A, and Klausen et al., 2003a). If this biofilm is exposed to 25 μg ml−1 colistin for 24 h, only part of the biofilm is eradicated, as can be visualized by staining of dead cells with the fluorescent indicator propidium iodide: cells in the interior part close to the substratum are killed by colistin, whereas cells in the upper layer survive the treatment (Fig. 1B, and Haagensen et al., 2007). Experiments with colistin concentrations of 15 and 50 μg ml−1 gave similar killing patterns in the biofilms (data not shown).
Laboratory and clinical strains of P. aeruginosa exhibit a similar colistin tolerance phenotype in biofilms
Bacterial strains, which have been repeatedly passaged in the laboratory (as the strain used here), can under some conditions exhibit different phenotypes compared with environmental or clinical isolates (Velicer et al., 1998; Branda et al., 2001; Lee et al., 2005). In order to investigate whether the PAO1 strain differed from clinical isolates with respect to colistin tolerance in biofilms, we investigated a number of P. aeruginosa strains, which had been isolated from sputum of patients afflicted with CF, with respect to their sensitivity to colistin when grown as biofilm. The clinical isolates exhibited the same phenotype as the laboratory strain, namely a colistin-sensitive subpopulation close to the substratum and a colistin-tolerant subpopulation on top (two examples are shown in Fig. 1C–F). Because genetic variants evidently can arise at high frequency in some biofilms (Rainey and Travisano, 1998; Boles et al., 2004), it might be speculated that the cells, which survive the colistin treatment under our experimental conditions, are genetic variants inherently resistant to colistin. However, when tested on agar plates, cells harvested from colistin-treated biofilms formed by the laboratory strain or the clinical isolates did not exhibit any change in sensitivity to colistin compared with the cells used to initiate a biofilm (data not shown). This provides evidence that neither laboratory nor clinical strains develop inherent resistance to colistin during biofilm growth in the flow-chamber system. Instead, some cells exhibit phenotypic tolerance to colistin in biofilms and apparently only under circumstances when they are situated in the upper part of the biofilm.
Cellular migration per se is not involved in tolerance development towards colistin in mature biofilms
Studies employed to unravel the developmental stages and genetic determinants involved in formation of the mushroom-like structured P. aeruginosa biofilm have shown that this biofilm is composed of at least two distinct subpopulations: a stalk-forming subpopulation situated at the substratum, and a cap-forming subpopulation on top (Klausen et al., 2003b; Haagensen et al., 2007; Pamp and Tolker-Nielsen, 2007). The studies demonstrated that the stalk-forming subpopulation is composed of non-motile cells, whereas the cap-forming subpopulation is formed by motile cells. In addition, the previous study on colistin tolerance in P. aeruginosa biofilms suggested that the colistin-sensitive subpopulation spatially corresponds to the subpopulation of non-motile cells, whereas the colistin-tolerant subpopulation spatially corresponds to the subpopulation of motile cells (Haagensen et al., 2007). This suggests that cellular migration might be involved in tolerance development to colistin. To test this hypothesis, we examined a number of motility-defective isogenic P. aeruginosa mutant strains with respect to tolerance development towards colistin in 4-day-grown biofilms. However, all motility-defective mutants exhibited the same phenotype as the wild type: a sensitive subpopulation close to the substratum, and a tolerantsubpopulation on top (Fig. S1, and data not shown). This suggests that, although bacterial migration is involved in structural development of the mushroom-like structured biofilm, it is not per se required for tolerance development to colistin in mature biofilms.
Metabolic/physiological activity is highest in cells situated in the upper layer of the biofilm
The observations described above suggested that the physiology of cells situated in the upper layer of the flow-chamber-grown biofilm (cap-forming subpopulation) significantly differs from the physiology of the cells situated in the area closer to the substratum (stalk-forming subpopulation). To learn more about the physiology of P. aeruginosa cells grown in flow-chambers, we examined their metabolic/physiological activity by the use of a fluorescent reporter. Similar to as previously described (Andersen et al., 1998; Sternberg et al., 1999), we introduced the growth rate-dependent fluorescent reporter fusion rrnBP1–gfp[AGA] into our P. aeruginosa strain. In this case, the gfp[AGA] gene, encoding for a Gfp protein with a short half-life, is placed under the transcriptional control of the ribosomal promoter rrnBP1. From cells of P. aeruginosa rrnBP1–gfp[AGA], which exhibit a high metabolic/physiological activity, one can therefore expect a high fluorescent signal, whereas cells exhibiting a low metabolic/physiological activity would give no or a low fluorescent signal. As control, we used P. aeruginosa rrnBP1–gfp, a strain expressing a stable version of Gfp, which should give an equal high fluorescent signal from all cells independent of the actual physiological state. When P. aeruginosa rrnBP1–gfp[AGA] was grown as a biofilm and CLSM images were acquired at intervals over a course of time of 4 days, a high fluorescent signal from cells situated in the upper layer could be detected at any time point (Fig. 2A). This suggests that the cells in the upper layer of the biofilm exhibited high metabolic/physiological activity. In contrast, cells located in the interior part of the multicellular structure exhibited a lower fluorescent signal (Fig. 2A), indicating lower cellular metabolic/physiological activity. In some cases, a small fraction of cells exhibiting a high fluorescent signal could be observed in the stalk part close to the substratum. The exact reason for this high signal is unknown at present. One reason could be that a high fluorescent signal is detected from this area because particularly here the cells are situated in very close proximity, resulting in a higher average number of cells compared with the rest of the structure, as can be seen using a higher magnification (data not shown; Fig. 1 in Haagensen et al., 2007). As expected, an almost homogenous fluorescent signal from the entire biofilms of the control strain P. aeruginosa rrnBP1–gfp was detected (Fig. 2B).
To get supporting evidence for the result involving the growth activity reporter suggesting that specifically the biofilm cells in the upper layer of the biofilm are metabolically active, we chose to treat the biofilm with a compound that acts specifically on metabolically active cells. Ciprofloxacin was chosen, as it was shown to be significantly more effective on growing than non-growing planktonic cells of P. aeruginosa, and because penetration of this compound was shown not to be restricted in a colony biofilm of P. aeruginosa (Davey et al., 1988; Walters et al., 2003). We exposed a 4-day-grown biofilm of P. aeruginosa rrnBP1–gfp[AGA] to 60 μg ml−1 ciprofloxacin in the presence of the dead cell indicator propidium idodide, and followed the effect over 13 h. As shown in Fig. 3A, ciprofloxacin killed only the cells in the upper layer of the biofilm, which exhibited high fluorescent signal. Ciprofloxacin killed neither the cells exhibiting a low fluorescent signal, nor the small fraction of cells close to the substratum, which exhibited a higher fluorescent signal. Also, prolonged exposure to ciprofloxacin did not lead to a killing effect in the stalk-forming subpopulation of cells (data not shown). This provides evidence that only the cells in the upper layer of the flow-chamber-grown biofilm exhibit high metabolic/physiological activity.
To investigate if the killing effect of colistin is confined to a distinct metabolic/physiological subpopulation of biofilm cells, a biofilm of the strain harbouring the growth activity reporter, P. aeruginosa rrnBP1–gfp[AGA], was grown for 4 days and subsequently exposed to 25 μg ml−1 colistin in the presence of the dead cell indicator propidium iodide. CLSM image acquisition showed that colistin specifically targeted the cells exhibiting low fluorescent signals, indicating that it preferably killed cells with low metabolic activity (Fig. 3B). In contrast, cells exhibiting high fluorescent signals survived the colistin treatment, indicating that biofilm cells with high metabolic activity exhibited tolerance towards colistin (Fig. 3B).
Biofilm cells depleted of metabolic/physiological energy are sensitive towards colistin
The results described above suggested that an energy-driven active process might be involved in tolerance development towards colistin of the distinct upper subpopulation of cells in biofilms. To obtain additional evidence for this proposition, we chose to deplete the biofilm cells of metabolic energy using the uncoupler of oxidative phosphorylation carbonyl cyanide 3-chlorophenylhydrazone (CCCP). Biofilm cells treated with CCCP should not be able to develop tolerance towards colistin in the upper layer. When a 4-day-grown biofilm of P. aeruginosa was simultaneously exposed to 30 μg ml−1 CCCP and 25 μg ml−1 colistin (in the presence of propidium iodide), all cells in the biofilm were killed within 13 h (Fig. 4A). However, no cells were killed in a 4-day-grown biofilm of P. aeruginosa when solely exposed to 30 μg ml−1 CCCP (and propidium iodide) (Fig. 4B). This finding substantiates the suggestion that an active cellular process is required to develop tolerance towards colistin in P. aeruginosa biofilms.
Biofilm cells exhibiting high metabolic activity induce pmrHFIJKLME expression upon colistin treatment and hence survive
It is known that pmrHFIJKLME-mediated LPS modification is involved in tolerance towards colistin, and using the fluorescent promoter fusion PpmrH–gfp, Haagensen and colleagues have recently shown that this operon was induced in a 4-day-grown biofilm of P. aeruginosa, which had been exposed to colistin for 24 h (Haagensen et al., 2007). In the light of the results described in the previous sections, we investigated if biofilm cells depleted of metabolic energy are able to induce the pmr operon. In agreement with the previous study (Haagensen et al., 2007), we found that the pmr operon was not expressed in the absence of colistin in a 4-day-grown biofilm of P. aeruginosa PpmrH–gfp (Fig. 5A, t0). However, after 4 h exposure to 25 μg ml−1 colistin, the pmr operon was induced in the biofilm, but only in the subpopulation of cells comprising the upper layer of the multicellular structures, which had just been identified to exhibit high metabolic activity (Fig. 5A). Cells in the interior part of the biofilm did not induce the pmr operon during the first hours upon colistin exposure, and were killed by the colistin subsequently (Fig. 5A). When a 4-day-grown biofilm of P. aeruginosa PpmrH–gfp was exposed simultaneously to 30 μg ml−1 CCCP and 25 μg ml−1 colistin in the presence of propidium iodide, no induction of the pmr operon was observed (Fig. 5B). Consequently, the cells in the upper layer of the biofilm, as well as in the interior part were killed (Fig. 5B). In agreement with the previous report by Haagensen et al. (2007), we found that a pmrF mutant was unable to develop tolerance towards colistin in biofilms, and therefore almost all cells were killed within 13 h (Fig. 5C). Colistin killing in the CCCP-treated P. aeruginosa PpmrH–gfp biofilm (Fig. 5B), as well as in the P. aeruginosa pmrF biofilm (Fig. 5C), started in the upper layer and proceeded to the inner parts, most likely reflecting the kinetics of colistin penetration into the biofilms and killing of cells which are unable to induce the pmr genes.
mexAB-oprM is required for tolerance towards colistin
The uncoupler of oxidative phosphorylation CCCP was used in the experiments described above to interfere with energy-driven cellular metabolic/physiological processes by disrupting the proton motive force (PMF). However, a disruption of PMF in bacterial cells does not only inhibit the regeneration of redox equivalents, it also interferes with the energy-driven transport of small molecules across the membrane such as the transport mediated by some efflux pumps. CCCP is therefore also frequently used as efflux pump inhibitor to study the role of efflux pumps in resistance to antimicrobial compounds (e.g. Takiff et al., 1996; Bogdanovich et al., 2006). P. aeruginosa is known for its intrinsic resistance to antimicrobials, to a large degree mediated by efflux pumps (Schweizer, 2003; Poole, 2005). However, although so far no efflux pump has been described to confer tolerance towards colistin in P. aeruginosa, we wanted to assess this. Consequently, we constructed a number of RND-efflux-pump knock-out mutants, namely mexPQ-oprE (PA3521–3523), mexAB-oprM (PA0425–0427), mexGHI-oprD (PA4205–4208) and yegMNO-opmB (PA2525–2528), and tested their ability to develop tolerance to colistin when grown as biofilms. We found that specifically the mexAB-oprM mutant exhibited a significant decrease in tolerance to colistin in biofilms. When a 4-day-grown biofilm of the P. aeruginosa mexAB-oprM mutant was exposed to 25 μg ml−1 colistin, it was unable to develop tolerance towards colistin (Fig. 6). Already after 4 h of exposure to colistin, some cells within the upper layer of the biofilm were killed, and within 13 h of exposure almost all cells of the biofilm were eradicated (Fig. 6). To verify this finding, we tested the independently created mexAB-oprM mutant P. aeruginosa PAO200, and found that it also exhibited a significant decrease in tolerance towards colistin in biofilms compared with its isogenic wild-type strain (data not shown). In addition, we found that P. aeruginosa mexAB-oprM mutants complemented with plasmid-borne mexAB-oprM genes displayed a wild-type phenotype in colistin-treated biofilms (data not shown). Together, these experiments indicate that the MexAB-OprM efflux pump is involved in tolerance development to colistin in P. aeruginosa biofilms.
Expression of mexAB-oprM is increased in the upper layer of a biofilm upon colistin exposure
To assess whether mexAB-oprM is constitutively expressed in our biofilm or induced upon colistin exposure, we introduced the fluorescent reporter fusion pmexA–gfp into the P. aeruginosa wild-type strain and followed expression of the reporter in biofilms formed by this strain. A 4-day-grown biofilm of P. aeruginosa pmexA–gfp exhibited a homogenous low fluorescent signal, indicating a weak basal expression of mexAB-oprM in the biofilm (Fig. 7). However, upon exposure to 25 μg ml−1 colistin, the fluorescent signal of the cells situated in the upper layer of the biofilm increased over time (Fig. 7). Cells in the interior part of the biofilm did not exhibit an increased expression of mexAB-oprM and were killed within 13 h upon exposure to colistin (Fig. 7). Biofilms of the wild-type strain, harbouring the same vector but expressing gfp from the mexC promoter (regulating expression of the efflux pump MexCD-OprJ) instead, exhibited a lower basal expression compared with P. aeruginosa pmexA–gfp, and no induction of mexC upon colistin exposure (data not shown). Altogether, this indicates that colistin is a signal for upregulation of mexAB-oprM expression, and that MexAB-OprM contributes to tolerance development towards colistin in P. aeruginosa biofilms.
Specifically targeting distinct physiological subpopulations in biofilms enables eradication of most biofilm cells using combined antimicrobial treatment
The observations described so far indicate that the biofilm is composed of at least two distinct physiological subpopulations. The subpopulation of cells situated close to the substratum exhibited low metabolic activity and sensitivity to the AMP colistin. The subpopulation of cells comprising the upper layer within the biofilm exhibited high metabolic activity and sensitivity to ciprofloxacin, which preferably targets replicating cells. It was therefore of interest to investigate if a combined treatment using the two compounds ciprofloxacin and colistin would eradicate all cells in the biofilm. To examine this, 4-day-grown biofilms of P. aeruginosa were established and exposed to colistin or ciprofloxacin alone, or simultaneously to ciprofloxacin and colistin for 24 h (in the presence of propidium iodide respectively). As shown in Fig. 8, a combined treatment with ciprofloxacin and colistin was able to kill almost all cells in the biofilm: less than 10 cells ml−1 survived the combined antimicrobial treatment, compared with 3.80 × 105 and 2.25 × 107 cells ml−1 on average in separately colistin- or ciprofloxacin-treated biofilms (Fig. 8) (details regarding the determination of the number of surviving biofilm cells are described in the Experimental procedures section). To get additional evidence that it is possible to specifically target the two different subpopulations dependent on their metabolic activity, we chose to treat biofilms with another antimicrobial agent, which is supposed to act specifically on active cells. We exposed 4-day-grown biofilms to tetracycline, a compound interfering with bacterial translation, for 24 h (in the presence of propidium iodide). As expected, tetracycline alone only killed the subpopulation of cells in the upper layer of the biofilm, leaving 1.90 × 108 cells ml−1 on average surviving (Fig. 8). However, a combined treatment with tetracycline and colistin was able to eradicate most cells in the biofilm, with a surviving fraction of 9.60 × 102 cells ml−1 (Fig. 8). This indicates that by combined antimicrobial treatment using compounds that target separate physiological subpopulations within biofilms, it is possible to kill the majority of the cells in a biofilm.
Haagensen and colleagues recently reported that colistin kills preferably a distinct subpopulation of cells situated close to the substratum in P. aeruginosa PAO1 biofilms, whereas a distinct subpopulation situated on top exhibits phenotypic tolerance to colistin (Haagensen et al., 2007). In addition, our results here suggest that this might be a general feature of P. aeruginosa flow-chamber-grown biofilms, as in addition to the laboratory strain PAO1, clinical isolates also exhibited this phenotype under the same conditions.
Haagensen et al. (2007) previously presented results on 2-day-grown P. aeruginosa PAO1 biofilms, which strongly suggested that cellular migration is involved in tolerance development to colistin in P. aeruginosa PAO1 biofilms. Surprisingly, we found in the present study that mature biofilms of various mutant strains, impaired in cellular migration, exhibited a similar tolerance phenotype as the wild type, independent of the actual three-dimensional structure of the biofilm, i.e. a colistin-sensitive subpopulation close to the substratum and a colistin-tolerant subpopulation on top. In addition, it should be noted that the clinical isolates used here are also to different degrees impaired in cellular migration (data not shown). Altogether, this indicates that cellular migration per se is not involved in tolerance development in mature P. aeruginosa biofilms. It might be speculated that cellular migration could be a major factor required for developmental processes during the first 2 days of biofilm formation of P. aeruginosa PAO1, and might be involved in colistin tolerance development (Klausen et al., 2003b; Haagensen et al., 2007). However, cellular migration might slow down during the following stages and structural maturing processes of the cap-forming subpopulation might then be mainly due to cell proliferation. In agreement, our unpublished experiments showed that a PpilA–gfp(AGA) fusion (i.e. a fusion between the type IV pilin gene promoter and a gene uncoding an unstable Gfp) was expressed strongly in the outer layer of the microcolonies in 2-day-grown P. aeruginosa PpilA–gfp(AGA) biofilms, whereas only weak expression could be detected in the mushroom-shaped structures in 4-day-grown P. aeruginosa PpilA–gfp(AGA) biofilms. Because of technical limitations of CLSM, it has not been possible to assess whether the cells of the cap-forming subpopulation in mature biofilms are still migrating (data not shown). More detailed investigations will be required to understand the impact of cellular migration on colistin tolerance development in young P. aeruginosa PAO1 biofilms.
Our results here provide evidence that metabolic/physiological activity in the biofilm is highest in a distinct subpopulation of cells in the upper layer of the multicellular structures. This conclusion is based on experiments involving in situ gene expression studies using a fluorescent reporter fusion, expressing an unstable version of Gfp under the control of a ribosomal promoter. Moreover, this subpopulation exhibited increased sensitivity to antimicrobial agents interfering with bacterial replication processes (such as ciprofloxacin) and translation processes (such as tetracycline). In addition, the antimicrobial agent tobramycin, which interferes with bacterial translation, was recently found to target specifically cells in the upper layer of a P. aeruginosa biofilm (Hentzer et al., 2003). Altogether, these data provide evidence that metabolic activity is highest in the upper layer of P. aeruginosa flow-chamber-grown biofilms, whereas metabolic activity is low in the deeper layers. This seems plausible as cells in the upper layer of the multicellular structures can obtain oxygen and nutrients from the bulk liquid, in contrast to the cells in the deeper layers where concentrations of dissolved oxygen and nutrients are likely to be low. Because similar observations have been obtained for P. aeruginosa biofilms grown as colony or established in capillary glass tubes (Werner et al., 2004), the observed spatial distribution of active and non-active cells might be a general characteristic of P. aeruginosa biofilms.
We observed that the distinct subpopulation of cells, which exhibits high metabolic activity, is able to survive the colistin treatment, in contrast to the subpopulation exhibiting low metabolic activity. This suggested that an energy-driven adaptation response might be required which renders the cells in the upper part tolerant to colistin. Our hypothesis was supported by the finding that biofilm cells depleted of metabolic energy (using the protonophore CCCP) were unable to adapt to colistin exposure and therefore did not survive the treatment. Haagensen et al. (2007) recently reported that cells in the upper layer of P. aeruginosa biofilms exhibited an induced expression of the pmr operon after exposure to colistin for 24 h. Here we found, using in situ gene expression analysis, that the metabolically active cells in the upper layer of the biofilm were able to induce expression of the pmr operon, which allowed them to adapt to colistin and hence survive. In contrast, the subpopulation of cells situated closer to the substratum was not able to induce expression of the pmr operon, and hence did not survive. In accordance, when biofilm cells were exposed to CCCP, induction of the pmr operon did not occur in the presence of colistin, and therefore the cells in the confined upper part did not survive the treatment. However, disruption of the PMF by CCCP does not only inhibit metabolic/physiological processes within cells, it also inhibits H+- and ATP-driven efflux pumps. Although efflux pumps have so far not been found to confer tolerance to colistin in P. aeruginosa, we investigated if our CCCP treatment might have interfered with the function of efflux pumps rendering cells sensitive to colistin. We found that specifically mexAB-oprM mutants exhibited increased sensitivity to colistin compared with the wild type. The fact that two different P. aeruginosa mexAB-oprM mutants showed the same phenotype, and the finding that the wild-type phenotype was restored in mexAB-oprM mutants harbouring plasmid-borne mexAB-oprM genes, ruled out the possibility that the observed phenotype could be caused by secondary mutations. Moreover, using the fluorescent reporter fusion PmexA–gfp, we found that expression of mexAB-oprM occurred at a basal level in 4-day-grown P. aeruginosa PmexA–gfp biofilms, whereas exposure to colistin led to increased PmexA–gfp expression in the cells which were situated in the upper layer of the biofilm. In agreement with our results, DeKievit et al. (2001) have previously shown that the PmexA–gfp fusion is expressed at a basal level, decreasing over time, in P. aeruginosa PmexA–gfp biofilms. However, contrary to our observations, DeKievit et al. (2001) found that the PmexA–gfp fusion is expressed primarily in P. aeruginosa PmexA–gfp biofilm cells close to the substratum. At present, it is not clear whether the effect of CCCP on colistin-treated biofilms is due to depletion of metabolic energy or inhibition of efflux pumps, or both.
Various compounds (such as chloramphenicol, β-lactams, macrolides and SDS) have been identified as substrate for the MexAB-OprM efflux pump in P. aeruginosa, but to our knowledge no AMPs so far (Poole, 2001; Schweizer, 2003). Until now, efflux-pumps have been found to be involved in natural resistance to AMPs in the Gram-negative Neisseria gonorrhoeae and Neisseria meningitides (Shafer et al., 1998; Tzeng et al., 2005) and an efflux pump/potassium antiporter system has been found to be involved in resistance to polymyxin B in Yersinia enterocolitica (Bengoechea et al. 2000). We found that mexAB-oprM is not involved in inherent resistance to colistin in P. aeruginosa, as the minimal inhibitory concentration to colistin (as examined by macro-dilution method and E-test) does not significantly differ between the wild-type and mexAB-oprM mutant under our experimental conditions (data not shown). Instead, mexAB-oprM is involved in development of phenotypic tolerance to colistin in P. aeruginosa when grown as biofilm. The observation that an efflux pump might be involved in the development of colistin tolerance could indicate that colistin also has an intracellular target in addition to interfering with the membrane of P. aeruginosa. In agreement, results obtained from studying the interaction of polymyxins with membranes of P. aeruginosa have led to the conclusion that these compounds might also have cytoplasmatic targets (Zhang et al., 2000).
In some cases after prolonged colistin exposure of biofilms formed by either the mexAB-oprM mutants or different pmr mutants, we observed small randomly distributed aggregates of living cells in the biofilm (data not shown). This indicates that a small fraction of single cells had survived the colistin treatment and were then able to initiate proliferation also in the presence of colistin. These cells did not exhibit any inherent resistance to colistin (data not shown). Experiments with mutants harbouring reporter genes showed that the small fraction of surviving pmr mutant cells expressed mexAB-oprM and that the small fraction of surviving mexAB-oprM mutant cells expressed the pmr operon (data not shown). In addition, we found no living cells after prolonged colistin exposure of biofilms formed by mexAB-oprM pmr double mutants (data not shown). Together, this might indicate that in the majority of the cells, neither pmr nor mexAB-oprM alone is able to completely confer tolerance to colistin in biofilms, instead both systems seem to be required simultaneously under these conditions. The fact that mexAB-oprM is expressed to some degree also in unexposed biofilms might indicate that the MexAB-OprM-mediated efflux contributes to an intrinsic tolerance to colistin in biofilms. Increased expression of mexAB-oprM during prolonged exposure to colistin might indicate that MexAB-OprM in addition is involved in long-term survival during colistin exposure. In contrast, induction of LPS modification, mediated by the pmr operon, might be a specific adaptation response to colistin. The pmr and mexAB-oprM tolerance mechanisms might operate independently as neither PmrA- nor PhoP-binding sites are present upstream of the mexAB-oprM genes (McPhee et al., 2006); however, at present it can not be ruled out that other determinants regulate expression of both, the pmr operon and mexAB-oprM.
Whereas conventional antimicrobial compounds (such as ciprofloxacin and tetracycline) specifically target metabolically active biofilm cells in the upper layer, our data indicated that colistin apparently specifically targeted the biofilm cells exhibiting low metabolic activity situated in the deeper areas. To investigate if cells, which have a low growth activity, exhibit increased sensitivity to colistin, we compared exponentially growing planktonic cells with stationary phase planktonic cells with respect to sensitivity to colistin. However, the survival rate upon colistin treatment was similar under both conditions (data not shown). In a number of other batch culture experiments, we addressed the potential role of oxygen or energy depletion or pH on cellular sensitivity to colistin; however, also under these conditions, no significant difference was observed (data not shown). We did not examine conditions under which the expression of the pmr operon is induced (e.g. Mg2+ limitation), as we did not observe any induction of pmr genes in our biofilms in the absence of colistin. Moreover, we could not observe any significant differences in sensitivity to colistin in our batch culture experiments under non-inducing conditions, when we compared wild-type and a pmrF mutant (data not shown), which is in agreement with previous reports (McPhee et al., 2003). Altogether, this supports the notion that sensitivity to polymyxins is greatly dependent on the prevailing environmental conditions. With respect to our investigations on biofilms, this might indicate on the one hand, that a complex microenvironment of unknown composition in the deeper layers of the biofilm exist, which impacts on cellular metabolic activity and sensitivity to colistin, but might be difficult to simulate in batch cultures. On the other hand, the particular subpopulation of cells exhibiting low metabolic activity in the biofilms might at the same time exhibit another specific characteristic, independent of metabolic activity, which renders it sensitive to colistin. Interestingly, the chelator EDTA targets the same subpopulation of cells in P. aeruginosa biofilms as colistin (Banin et al., 2006; data not shown), and investigations on batch culture-grown cells have indicated that EDTA preferably targets stationary phase cells compared with exponential phase cells (Imamura et al., 2005). Investigations are currently ongoing in our laboratory to determine the factors which render cells in the deeper layers of biofilms sensitive to colistin.
Our observation that conventional antimicrobial agents (e.g. ciprofloxacin and tetracycline) specifically target the distinct subpopulation of biofilm cells in the upper layer, whereas colistin preferably targets the distinct subpopulation of biofilm cells in the deeper layers, prompted us to study the killing effect of a combined treatment with these antimicrobial compounds. The combined treatment with either colistin + ciprofloxacin or colistin + tetracycline reduced the number of biofilm cells significantly, compared with the single antimicrobial treatments. In particular, the combined treatment with colistin + ciprofloxacin was very effective, eradicating nearly all cells of the biofilm. Intriguingly, the administration of aerosolized colistin in combination with oral ciprofloxacin has been found to significantly reduce the onset of chronic P. aeruginosa infection in CF patients (Valerius et al., 1991). Furthermore, this treatment strategy is part of a recommended early intervention and prevention therapy of lung disease in CF according to a European consensus report (Döring et al., 2004; Høiby et al., 2005).
Altogether, our data indicate that in general, antimicrobial-tolerant cells in biofilms are not randomly distributed. Instead, antimicrobial tolerance seems to be confined to physiologically distinct subpopulations of cells within the multicellular structures, independent of the antimicrobial compound used. Tolerance development to colistin is confined to a distinct subpopulation of metabolically active cells, which is able to adapt to colistin exposure by: (i) reducing interaction with the antimicrobial compound (LPS modification mediated by the pmr operon) and (ii) export of the antimicrobial compound (efflux mediated by mexAB-oprM). Moreover, we found that a systematic combined antimicrobial treatment, specifically targeting distinct physiological subpopulations, enables eradication of almost all cells in a biofilm.
Bacterial strains and growth conditions
The bacterial strains used in this study are listed in Table 1. For routine strain manipulations, P. aeruginosa and E. coli strains were grown in Luria–Bertani (LB) medium at 37°C. For batch culture experiments with P. aeruginosa, AB minimal medium (Pamp and Tolker-Nielsen, 2007) supplemented with 1 μM FeCl3 and 10 mM glucose was used. The clinical non-mucoid isolates P. aeruginosa CFSJ208 and CFSJ234 were isolated from sputum obtained from 23- and 25-year-old patients respectively, who are afflicted with CF. Isolation was carried out using Pseudomonas isolation agar and the strains verified using phenotypic analysis and polymerase chain reaction (PCR). Where appropriate, antibiotics were used for bacterial cultures at the following concentrations: for P. aeruginosa strains, gentamycin (Biochrome AG, Germany) at 30 μg ml−1, streptomycin (Sigma, Germany) at 300 μg ml−1, potassium tellurite (Sigma, Germany) at 150 μg ml−1, carbenicillin (Sigma, Germany) at 200 μg ml−1 and tetracycline (Sigma, Germany) at 15 μg ml−1; for E. coli strains, ampicillin (Vepidan ApS, Denmark) at 100 μg ml−1, chloramphenicol (Sigma, Germany) at 25 μg ml−1, kanamycin (Sigma, Germany) at 50 μg ml−1, tetracycline (Sigma, Germany) at 10 μg ml−1 and gentamycin at 15 μg ml−1.
Table 1. Strains, plasmids and primers used in this study.
Strain, plasmid or primer
Relevant characteristics or sequence
Source or reference
The underlined sequence corresponds to the gene-specific region.
Plasmids and primers used in this study are listed in Table 1. DNA restriction enzyme digestions and modifications were performed according to the manufacturer's instructions (Fermentas, Invitrogen). Plasmids were transformed using electroporation if not otherwise stated: for P. aeruginosa, 25 mF, 200 Ω, < 5 ms, 2.5 kV; for E. coli, 25 mF, 400 Ω, < 5 ms, 2.5 kV. P. aeruginosa strains were fluorescently tagged at an intergenic neutral chromosomal locus (25 bp downstream of the glmS gene) in mini-Tn7 constructs, as described previously (Lambertsen et al., 2004). The insertion was verified by PCR using the primers Tn7-GlmS and Tn7-R109, as described previously (Lambertsen et al., 2004). The strain P. aeruginosa PAO1 ΔmexAB-oprM::Gmr was constructed as follows: a knockout fragment of ΔmexAB-oprM containing a gentamycin (Gm)-resistance cassette was generated by PCR overlap extension essentially as described by Choi and Schweizer (2005); shortly, primers MexA-UpF-GW, MexA-UpR-Gm, OprM-DnF-Gm, OprM-DnR-GW were used to amplify chromosomal regions of mexA and oprM respectively, and primer Gm-F and Gm-R were used to amplify the Gm cassette using plasmid pPS856 (Hoang et al., 1998) as template. The PCR fragments were fused together and amplified with primers GW-attB1 and GW-attB2 incorporating the attB1 and attB2 recombination sites at either end of the knockout cassette. Using the Gateway cloning system (Invitrogen), the resulting knockout fragment was first transferred via BP reaction into pDONR221 generating entry plasmid pDONR211mexA-oprM and subsequently transferred via LR reaction into pEX18ApGW generating the knockout plasmid pEX18ApmexA-oprM. The knockout fragment ΔmexAB-oprM::Gmr was transferred into P. aeruginosa PAO1 by triparental mating using helper strain E. coli HB101 pRK600. The resulting double cross-over in strain P. aeruginosaΔmexAB-oprM::Gmr was confirmed by PCR and the phenotype regarding sensitivity to conventional antimicrobial agents compared with the known phenotype of strain P. aeruginosa PAO200. The strains P. aeruginosa PAO1 ΔmexPQ-oprE::Gmr, P. aeruginosa PAO1 ΔmexGHI-oprD::Gmr and P. aeruginosa PAO1 ΔyegMNO-opmB::Gmr were constructed in exactly the same way, with the exception that the following gene-specific primers were used: MexP-UpF-GW + MexP-UpR-Gm and OprE-DnF-Gm + OprE-DnR-GW for the construction of the mexPQ-oprE mutant, MexG-UpF-GW + MexG-UpR-Gm and OprD-DnF-Gm + OprD-DnR-GW for the construction of the mexGHI-oprD mutant, and YegM-UpF-GW + YegM-UpR-Gm and OpmB-DnF-Gm + OpmB-DnR-GW for the construction of the yegMNO-opmB mutant.
Cultivation of biofilms
Biofilms were cultivated in flow cells with individual channel dimensions of 1 × 4 × 40 mm, covered with a glass coverslip (Knittel Gläser, Germany) as substratum for biofilm formation. The biofilm flow cell system was assembled and prepared as described elsewhere (Sternberg and Tolker-Nielsen, 2005). AB minimal medium (Pamp and Tolker-Nielsen, 2007) supplemented with 0.3 mM glucose as carbon source was used as growth medium. Individual flow cells were inoculated with 300 μl aliquots of overnight growth cultures of P. aeruginosa, which were adjusted to an optical density at 500 nm of 0.005. Overnight cultures of P. aeruginosa were grown in AB minimal medium supplemented with 30 mM glucose at 30°C under vigorous shaking. To allow attachment of the bacterial cells to the substratum, flow cells were left without flow for 1 h after inoculation at 30°C. Afterwards, a laminar flow with a mean flow velocity of 0.2 mm s−1 was achieved using a Watson Marlow 205S peristaltic pump. The P. aeruginosa strain containing plasmid pmexA–gfp was cultivated as biofilm without supplementation of carbenicillin to avoid any possible secondary effects during treatment with other antimicrobial compounds. Loss of plasmid during 4 days of biofilm growth was observed in a minor fraction of non-treated biofilms on average in less than 4% of cells, as examined by plating of dilutions of cells derived from biofilms on LB agar medium with and without carbenicillin. P. aeruginosa mexAB-oprM mutant strains containing either pRK415 or pRSP17 were cultivated in flow cells perfused with minimal medium supplemented with 15 μg ml−1 tetracycline.
Exposure of biofilms to antimicrobial and other compounds
Mature biofilms were exposed to the following compounds where indicated: 25 μg ml−1 colistin (Colimycin, colistin methanesulfonate; Lundbeck A/S, Denmark); 60 μg ml−1 ciprofloxacin (Bayer, Germany); 200 μg ml−1 tetracycline (Sigma, Germany); 30 μM CCCP (Sigma, Germany). This was achieved by supplementing biofilm media with the required compounds at appropriate final concentrations and addition of the fluorescent dead cell indicator propidium iodide (Sigma, Germany) at a final concentration of 0.3 μM. In non-treated control experiments, 0.3 μM propidium iodide was added alone to the biofilm media. The colistin used here is colistin methanesulfonate, which is a compound undergoing hydrolysis in aqueous solutions to form the active compound colistin sulphate (CS) in a time-dependent manner (Bergen et al., 2006). In control experiments, CS (Sigma, Germany) was added in concentrations raging from 0.5 to 8.0 μg ml−1 instead of colistin (i.e. colistin methanesulfonate). In these experiments, the same spatial distribution of dead and live cells was observed after 24 h of exposure, compared with biofilms treated with 25 μg ml−1 colistin (colistin methanesulfonate). We used colistin (colistin methanesulfonate) in this study, as this is the actual compound, which is mainly used in medical settings for treatment of infections, in particular for treatment of P. aeruginosa lung infections in CF patients (Littlewood et al., 2000; Høiby et al., 2005; Li et al., 2006). In some experiments, 2.5 μM Syto 9 was added for counterstaining of living cells. To determine the survival rate upon antimicrobial treatment, cells were harvested from treated and non-treated biofilms and the number of surviving cells determined as follows: the bulk liquid was carefully removed from the flow-chambers and the biofilm cells recovered by pumping 1 ml of 0.9% NaCl solution containing 50 μg of glass beads (size ≤ 106 μm) (Sigma, Germany) rapidly back and forth through the flow-chamber until all cells were removed from the substratum as examined by microscopic inspection. The cell suspension was mixed by vortexing and the number of living cells determined by plating dilutions of cells on LB agar medium and determination of the number of colony forming units upon incubation.
Microscopy and image processing
Image acquisition was performed with a Zeiss LSM 510 confocal laser scanning microscope (Carl Zeiss, Germany) equipped with an argon and a NeHe laser and detectors and filter sets for simultaneous monitoring of Gfp (excitation, 488 nm; emission, 517 nm) and propidium iodide (excitation, 543 nm; emission, 565 nm). Images were obtained using a 40×/1.3 Plan-Neofluar oil objective. Simulated multichannel cross sections were generated using Imaris software package (Bitplane AG, Switzerland).
We are very grateful to H. P. Schweizer for providing us with P. aeruginosa PAO200 and the isogenic wild-type strain. We thank B. Iglewski for providing us with plasmids pmexA–gfp and pmexC–gfp. K. Poole is gratefully acknowledged for the provision of plasmids pRK415 and pRSP17. Critical reading of the manuscript by J. A. J. Haagensen is greatly acknowledged. This work was supported by a grant from the Danish Research Council and by a grant from the Lundbeck Foundation.