Rhamnolipids mediate detachment of Pseudomonas aeruginosa from biofilms

Authors


Summary

The process of detachment, through which bacteria use active mechanisms to leave biofilms and return to the planktonic (free-living) state, is perhaps the least understood aspect of the biofilm life cycle. Like other stages of biofilm development, detachment is a dynamic, regulated process, controlled by specific genes, and induced by particular environmental cues. In previous work we discovered Pseudomonas aeruginosa variants that exhibit accelerated biofilm detachment. These hyper-detaching variants arise spontaneously from biofilms at a high frequency, and they exhibit robust detachment under different biofilm growth conditions. Here we show that these variants detach by a mechanism requiring the biosurfactant rhamnolipid and that this detachment mechanism rapidly restores antibiotic sensitivity to separating bacteria. We also show that rhamnolipids can bring about detachment in wild-type P. aeruginosa biofilms. These findings raise the possibility that this detachment mechanism may be useful as a treatment to disrupt established biofilms. Interestingly, the rhamnolipid-mediated detachment mechanism involves the formation of cavities within the centre of biofilm structures. Our data suggest a model to explain detachment that occurs via this pattern.

Introduction

Many bacterial species are capable of two general growth modes. In planktonic growth, organisms exist as independent, free-living cells, usually suspended in a liquid medium. Bacteria can also form biofilms in which the cells live clustered together in matrix-encased groups (Costerton et al., 1999; Hoiby et al., 2001). Biofilms cause some of the most recalcitrant human infections (Costerton et al., 1999; Parsek and Singh, 2003) and are the predominant growth mode for bacteria in the environment.

Biofilm growth provides important advantages to organisms. Biofilm bacteria manifest marked resistance to killing by antimicrobial agents (Costerton et al., 1999; Stewart, 2002), increased synthesis of protective matrix materials (Costerton et al., 1999; Friedman and Kolter, 2004) and in some cases, enhanced metabolic cooperation (Shapiro, 1998). Recent work also suggests that horizontal gene transfer and intercellular communication may be facilitated in biofilms (Hausner and Wuertz, 1999; Parsek and Greenberg, 2000), and that biofilm growth increases the genetic diversity of bacterial populations (Boles et al., 2004). All of these characteristics could enhance the survival of bacterial communities in harsh conditions.

While the benefits provided by biofilm growth are impressive, these advantages come at some cost. For example, matrix production imposes a synthetic burden on the bacteria and the nutrient gradients within biofilms can limit growth and the cells’ ability to sustain defences, such as the production of antioxidants, and metabolic enzymes (Huang et al., 1998; Hassett et al., 1999). Perhaps the most significant risk for biofilm bacteria occurs when local conditions deteriorate. This could occur because of nutrient depletion, the accumulation of wastes, the appearance of immune cells or antibiotics, or other threats. Biofilm bacteria have a reduced ability to evade stresses because they are physically confined by the matrix and their motility functions are repressed (Whiteley et al., 2001; Sauer et al., 2002).

The costs associated with biofilm growth make it vital that bacteria possess mechanisms to separate from biofilms and assume planktonic life. This process is referred to as detachment (some investigators use the term dispersion). Unlike sloughing in which shear forces physically dislodge bacteria (Picioreanu et al., 2001; Stoodley et al., 2002), in detachment, environmental conditions trigger active mechanisms that bring about bacterial separation (Hunt et al., 2004; Sauer et al., 2004; Thormann et al., 2005). The fact that detachment can be triggered by several different cues could allow organisms to regulate their movement between the biofilm and planktonic growth states as environmental conditions change.

In addition to its importance for biofilms in the environment, detachment also plays an important role in infection pathogenesis. For instance, organisms like Vibrio cholerae, Legionella pneumophila and Pseudomonas aeruginosa are thought to live in soil or aquatic biofilms (Parsek and Singh, 2003). The process of detachment creates mobile bacteria (single cells or aggregates) that are capable of causing infection. Detachment may also promote dissemination from an initial infection point to other sites in the body. A well-characterized example is the devastating embolic events of endocarditis caused by detachment of the complex biofilm growing on heart valves (Parsek and Singh, 2003). In addition, many cases of hospital-acquired pneumonia are caused by bacteria detached from biofilms that form in a patient's endotracheal tube or oropharnyx (Bergmans et al., 1998; Adair et al., 1999; Feldman et al., 1999).

The process of detachment is complex. Even within a single bacterial species, multiple detachment patterns have been observed, suggesting that different detachment mechanisms may exist. In the pathogen P. aeruginosa, detachment can involve the discharge of individual bacteria (Stoodley et al., 2001), the piecemeal separation of cell clusters (Stoodley et al., 2001) and the mass detachment of whole colonies seen early in biofilm development (P. Singh and B. Boles, unpubl. obs.). Perhaps the most interesting pattern, often observed in aged biofilms (∼10–12 days of growth in our system), involves the formation of internal cavities that eventually fracture to release motile bacteria. This pattern has been termed ‘central hollowing’ (Sauer et al., 2002; Hunt et al., 2004). Each of these detachment patterns may involve multiple steps including degradation of the matrix, activation of motility and physiological changes that prepare cells for conditions outside the biofilm.

In addition to this mechanistic complexity, the timing and extent of biofilm detachment is likely to be under tight and local regulation. Under favourable conditions, most wild-type biofilms release a relatively small number of cells continually; they also sporadically undergo major detachment events after prolonged periods of growth. These major detachment events can be very heterogeneous in terms of timing, the regions of the biofilm affected and perhaps, the mechanism of detachment employed. This heterogeneity poses challenges to the investigation of detachment mechanisms. One approach that has been used to overcome this challenge involves exposing biofilms to major, sudden changes in environmental conditions in order to induce detachment. This approach has led to greater understanding of the process by which detachment is triggered (Hunt et al., 2004; Sauer et al., 2004; Thormann et al., 2005).

Here we took a different approach. In a previous study, we found that variants with a hyper-detaching phenotype spontaneously arise at a high frequency after short-term growth of wild-type P. aeruginosa in biofilms (Boles et al., 2004). The variant we chose to study (hereinafter called ‘variant’) showed a normal pattern of initial biofilm formation; however, after 2 days of growth, the biofilm rapidly disperses (see Fig. 1 and Boles et al., 2004).

Figure 1.

Schematic of biofilm detachment by wild-type P. aeruginosa and the hyper-detaching variant (also see Boles et al., 2004). Some biofilm structures formed by the wild type detach via the central hollowing pattern after 10–12 days of growth (top). The variant exhibits a similar detachment pattern; however this strain detaches much more extensively and detachment occurs after only 2 days of biofilm development (bottom).

We decided to study the variant's detachment mechanism because its detachment phenotype is robust; most of the biofilm biomass simultaneously disperses after only 2 days of growth and it detaches under different biofilm growth conditions (Boles et al., 2004). Furthermore, detachment of variant biofilms is spontaneous. This eliminates the need for an exogenous disturbance to induce detachment that could produce pleiotropic physiological effects. It also allowed us to focus on an effector mechanism that brings about bacterial separation, rather than the triggering mechanisms that activate the detachment process. Here we show that the variant detaches via the central hollowing pattern, and this detachment mechanism rapidly restores aspects of planktonic physiology to detaching cells. Our investigation of the variant's detachment mechanism revealed that the bio-surfactant rhamnolipid is required for the hyper-detachment phenotype. Importantly, using three independent methods, we found that rhamnolipids also produce detachment (via central hollowing) in wild-type P. aeruginosa biofilms.

Methods

Strains, plasmids and growth conditions

The P. aeruginosa strains used in this work were derived from the wild-type strain PA01 (obtained from B. Iglewski). The isolation of the hyper-detaching variants, which spontaneously arise from wild-type PA01 biofilms, is described in the study by Boles et al. (2004). For visualization by confocal microscopy, pMRP9-1 (which expresses green fluorescent protein) was transformed into appropriate strains (Davies et al., 1998). For the variant–wild type coculture experiments a red fluorescent protein (rfp) expressing plasmid (pMTRFP) was constructing by ligation of polymerase chain reaction (PCR)-amplified rfp from PRSETB (Campbell et al., 2002) into pUCP18. In this vector rfp expression is controlled by the lac promoter and has a T7gene10 ribosome binding site. In some experiments mucoid and non-mucoid P. aeruginosa isolates from cystic fibrosis patients (from the University of Iowa Cystic Fibrosis Clinic) were used.

A rhlAB insertion/deletion mutation was generated by inserting a gentamicin cassette into the P. aeruginosa rhlAB genes digested with DraIII and MluI. The rhlAB mutation was moved on to the chromosome of PA01 and the variant by homologous recombination. For complementation studies, rhlAB including the rhlA promoter was PCR-amplified from PA01 and cloned into pEX1.8 (Pearson et al., 1997). A construct for the inducible expression of rhamnolipids was made by cloning rhlAB without its native promoter into pSW195 (obtained from D. Wozniak), which contains mini-CTX1 (Hoang et al., 2000; Boles et al., 2004) with an arabinose inducible promoter pBAD (Newman and Fuqua, 1999). This was then moved onto the chromosome at the phage attachment site attB. Expression was induced with 0.5%l-arabinose. Trypticase Soy Broth (Difco) was used as the growth medium unless otherwise specified. Swarming motility plates consisted of 8 g l−1 nutrient broth, 5 g l−1 dextrose and 0.5% Bacto Agar (Difco). For orcinol assays, bacteria were grown in a medium composed of 6 g l−1 Na2HPO4, 3 g l−1 KH2PO4, 0.5 g l−1 NaCl, 0.1 mM CaCl2, 2 mM MgSO4, 0.2% glucose and 0.05% glutamate. Antibiotic concentrations (per millilitre) were: 300 µg carbenicillin, 100 µg gentamycin, 60 µg tetracycline for P. aeruginosa.

Biofilm experiments

Flow cell experiments and confocal microscopy were performed as previously described (Davies et al., 1998). Biofilms were grown in continuous culture flow cells similar to those described in the study by Davies et al. (1998). The size of the flow channel was 5 × 35 × 1 mm. An overnight culture diluted to 107 cells per millilitre in fresh biofilm medium was used as the inoculum and flow was arrested for 45 min. Flow of biofilm medium was then initiated at a rate of 170 µl min−1. Images were obtained using a Bio-Rad (Hercules, CA) scanning confocal microscope. Biofilms were grown in 1% TSB unless otherwise specified. To measure the proportion of each strain in the wild type and variant coculture biofilms, either the wild type or the variant was marked with tetracycline resistance by insertion of mini-CTX1 at the strains chromosomal phage attachment site. Just before detachment, the biofilm was scraped from the growth surface, homogenized and plated on LB agar with and without tetracycline in order to enumerate each strain in the biofilm. In some flow cell studies, 3-day-old wild-type biofilms were exposed to 0.5 mg ml−1 of purified rhamnolipid (obtained from the Jeneil Biosurfactant, Saukville, WI) or 0.2% SDS (Sigma) by replacing the biofilm medium with medium containing either of these surfactants. Viability of planktonic P. aeruginosa in the presence of purified rhamnolipid was assayed by incubating 106 stationary-phase cells in a range of rhamnolipid concentrations in 1 ml of PBS for 6 h.

For antibiotic susceptibility testing of detached cells and intact and dispersed biofilms, biofilms were grown using the rotating disk biofilm reactor as previously described (Boles et al., 2004). To determine the antibiotic sensitivity of detached bacteria, the discs (with attached biofilm) were washed in PBS (to remove growth media), and incubated in 1 ml of PBS for 4 h at 25°C. The overlying fluid (containing detached cells) was then moved to a new container and incubated with tobramycin at indicated concentrations for 4 h. The cell suspension was then homogenized, and cell counts were determined by plate counting. To determine the antibiotic sensitivity of intact biofilms, discs (with attached biofilm) were washed with PBS and immediately treated with tobramycin for 4 h. The biofilm was then homogenized to produce a cell suspension, and cell counts were determined by plate counting. To determine the antibiotic sensitivity of mechanically disrupted biofilms, discs with attached biofilms were placed in 1 ml of PBS and homogenized for 30 s to disrupt the biofilm (prior to antibiotic treatment). The tobramycin sensitivity of these dispersed bacteria was then determined as described above.

The rotating disk biofilm reactor (Boles et al., 2004) was also used to generate biofilms for the static detachment assay. In this assay, discs (with attached biofilm) were removed from the reactor, washed in PBS and incubated in 1 ml of PBS for 4 h. In some experiments 0.5 mg ml−1 of purified rhamnolipid was added to PBS during the 4 h detachment period. The bacteria in the overlying fluid (detached cells) and bacteria that remained on the disk (biofilm cells) were counted as described above. The detachment fraction was calculated by dividing the number of detached cells by the total number of cells (detached + biofilm cells). Detachment of the P. aeruginosa strain inducibly expressing rhamnolipids was analysed by collecting 1 ml aliquots of the biofilm effluent on ice (in triplicate) from flow cell biofilm reactors at the indicated times and enumerating cell numbers in each aliquot by plate counting.

Rhamnolipid assays

Several methods were used to assay rhamnolipid production. For rhamnolipid detection on plates, methylene blue-containing plates were prepared as previously described in the study by Kohler et al. (2000). The orcinol assay (Koch et al., 1991) of ether-extracted culture supernatants was used to measure rhamnolipids in liquid cultures. Cultures were grown for 48 h at 37°C, and 0.5 ml of culture supernatant was extracted twice with 1 ml of diethyl ether. The ether fractions were pooled and evaporated to dryness, and reconstituted in 0.5 ml H2O. Samples were diluted 1/10 in a solution containing 0.19% orcinol in 53% H2SO4. The sample was then placed in boiling water for 30 min, cooled at room temperature for 15 min, and the absorbance (A421) was measured. Rhamnolipid concentrations were calculated by comparing the data with a standard curve generated from purified rhamnolipid. Rhamnolipid gene expression was measured using a rhlA::gfp fusion construct by Y. Lequette and E. P. Greenberg (Lequette and Greenberg, 2005). To measure rhlA::gfp expression in liquid cultures, fluorescence was measured with a microtitre plate fluormeter (Tecan). The excitation wavelength was 435 nm and the emission length was 535 nm. The relative level of rhlA::gfp expression in nascent biofilms was determined using confocal microscopy. Images of 12 different 2-day-old variant and wild-type biofilms of similar size were acquired using identical confocal microscope settings. A line profile of fluorescence pixel intensity across each biofilm colony was generated using Laser Sharp software (Bio-Rad, Hercules, CA). These values were normalized for colony dimensions and averaged.

Results

Detachment rapidly produces physiological changes in separating bacteria

To begin our investigation of the variant's detachment mechanism, we performed time-lapse microscopy during the process of detachment (see Movie S1 in Supplementary material). These images revealed that the variant reproducibly detached via a characteristic sequence of events. First, the biofilm began to slacken; individual bacteria were seen moving to and fro within the previously solid structure. Soon after this, cavities formed in the centre of biofilms and these became filled with rapidly swimming bacteria (Fig. 2A). The cavities increased in size over time, and they eventually ruptured to release motile bacteria. Finally, the remaining biofilm biomass detached from the growth surface leaving only a monolayer of cells behind. In variant biofilms, the entire detachment process was completed in a short period of time (2–3 h), and it occurred nearly simultaneously in most all of the biofilm structures present within the reactor.

Figure 2.

The central hollowing pattern of biofilm detachment. Confocal images of variant (A) and wild-type (B) biofilms during detachment. In both strains, central cavities form (demarked with white dashed lines) and these become filled with rapidly swimming bacteria. Images are X-Y views, scale 10 µm.
C. The variant's detachment mechanism restores antibiotic sensitivity to detached bacteria. Susceptibility to tobramycin killing of variant bacteria in intact biofilms (•) and variant bacteria immediately after detachment (▴). Biofilms that have been mechanically disrupted (▪) show intermediate sensitivity. Graph shows the mean of three experiments; error bars show SEM.

This pattern of detachment is very similar to the central hollowing detachment pattern observed in aged wild-type biofilms by several other groups and us (Sauer et al., 2002; Hunt et al., 2004) (Fig. 2B; see Movie S2 in Supplementary material). In the wild type, however, this type of detachment occurs after much longer periods of growth (∼10–12 days in our growth conditions versus 2 days for the variant), and it affects only some areas of the biofilm at any given time. Nevertheless, the close resemblance of these detachment patterns suggests that a common mechanism may be involved in both variant and wild-type biofilms.

The presence of motile bacteria within the central cavities indicated that the variant's detachment mechanism produces some aspects of planktonic physiology in the separating bacteria; previous work has shown that swimming motility is generally repressed during biofilm growth (Whiteley et al., 2001; Sauer et al., 2002). This led us to investigate whether detachment reversed another key biofilm phenotype, antibiotic resistance (Stewart, 2002; Mah et al., 2003). To test this, newly detached bacteria were collected in a saline buffer lacking any nutrient source in order to prevent bacterial growth. This was important because the daughter cells of detached bacteria would likely be antibiotic-sensitive even if the detachment process produced no change in physiology. Figure 2C shows that whereas 1 µg ml−1 of tobramycin killed most detached cells, 100 µg ml−1 of tobramycin was required to kill the intact variant biofilm. Of note, mechanical disruption of variant biofilms (by homogenization) produced cells that had an intermediate resistance to tobramycin as compared with detached bacteria and intact biofilms (Fig. 2C). These results suggest that the variant's detachment mechanism restores aspects of bacterial physiology including antibiotic sensitivity and motility. These findings are consistent with recent work showing that the protein expression profiles of detached P. aeruginosa more closely resembled planktonic bacteria than biofilm cells (Sauer et al., 2002).

The variant overproduces rhamnolipids, a candidate detachment factor

To explore the possibility that the variant produces a secreted detachment factor or signal we grew mixed biofilms containing the variant and the wild type. In coculture (1:1 ratio of variant to wild type), the variant's phenotype was dominant; the mixed biofilm detached at a similar time and extent as the variant in pure culture (Fig. 3A–C). To eliminate the possibility that the variant was somehow excluding wild-type bacteria from the biofilm (or killing them), we grew mixed biofilms in which either the wild type or the variant carried an antibiotic-resistant marker, and counted viable bacteria before detachment occurred. Both strains were present in approximately equal proportions. These experiments suggest that the variant produces a secreted factor or signal that is potent enough to cause the detachment of both strains, even though variants comprised only approximately half the biofilm population.

Figure 3.

The variant's detachment phenotype is dominant in coculture biofilms. Confocal images of wild-type (A) and variant (B) biofilms after 3 days of growth. The wild type has developed tower-like biofilm structures whereas the variant has detached, leaving only a monolayer of cells behind. A mixed biofilm formed by equal ratios of wild-type and variant bacteria (C) detaches to a similar degree as pure-culture variant biofilms. Images are X-Z views, scale 10 µm.

To determine the relative location of the two strains within the mixed biofilm (prior to detachment) we used wild-type and variant bacteria expressing red and green fluorescent proteins respectively. These experiments revealed two surprising results. First, we found that the two strains occupied distinctly different locations within the mixed biofilm prior to detachment; the wild type localized to the interior, and the variant to the exterior of the mixed biofilm (Fig. 4A and B). The peripheral location of the variants in the cocultures makes it unlikely that they physically dislodge wild-type bacteria during detachment. Second, in spite of the fact that the centre of the mixed biofilms consisted primarily of wild-type bacteria, detachment occurred via central cavity formation (Fig. 4C). This was unexpected because pure-culture wild-type biofilms do not undergo detachment via central hollowing until much later times (∼10–12 days) in these growth conditions. The formation of cavities within the central region (populated by wild-type bacteria) could be explained by at least two different mechanisms: the peripherally located variants could activate a detachment mechanism in the centrally located wild-type bacteria; alternatively, the central regions may be inherently more susceptible to the action of a detaching agent, produced in this case by peripherally located variants (see below).

Figure 4.

The variant and the wild type inhabit different regions in mixed biofilms.
A and B. Differential labelling of wild-type (red) and variant (green) bacteria show that the variant covers the wild type in mixed biofilms; wild-type bacteria are localized in the centre of biofilm structures whereas the variants are on the periphery.
C. Detachment of the mixed biofilm occurs via the central hallowing pattern after 3 days of biofilm growth. Even though wild-type bacteria (which would not detach at this time) occupy the biofilm centre, and the hyper-detaching variant occupy the exterior, detachment begins in the central regions. Images A and C are X-Y views and B is an X-Z view, scale 10 µm. Results are representative of three independent experiments. Similar results were obtained with the red and green fluorescent markers switched to the opposite strains.

Our observation that the variant physically covered wild-type bacteria in coculture biofilms led us to hypothesize that the variant may show greater motility than the wild type. Previous work showed that P. aeruginosa can utilize surface motility to reach the topmost locations in a biofilm (Klausen et al., 2003). To test this we compared the swimming, twitching and swarming motility of the wild-type and variant strains. Whereas the swimming and twitching motility of the two strains was similar, the variant exhibited markedly increased swarming motility (Fig. 5A). Swarming is a type of surface motility that is facilitated by the biosurfactant, rhamnolipid (Kohler et al., 2000). Rhamnolipids are amphipathic glycolipids with surface-active properties that decrease adhesive interactions (Neu, 1996; Desai and Banat, 1997); as such they are candidate detachment factors.

Figure 5.

The variant produces more rhamnolipids than wild-type P. aeruginosa.
A. Swarming motility of wild-type (top colonies) and variant bacteria (bottom). Three point inoculations of each strain were made on the surface of 1% agar plates (at arrow heads) and swarming motility observed after 1 day.
B. When grown on methylene blue-containing plates the variant (right panel) produces a larger halo than the wild type (left), indicating increased surfactant production.
C. Rhamnolipid production of the wild type (black bars) and variant (hatched bars) after 2 days of planktonic culture as determined by the orcinol assay.
D. Relative expression of rhlAB genes by the variant (▪) and wild-type (•) P. aeruginosa in planktonic culture as determined by rhlA::gfp expression reporter. Error bars showing SEM are obscured by data points.
E. Relative expression of the rhlA::gfp reporter by wild-type (top) and variant (bottom) biofilms before detachment. Images are X-Y views and are representative of three different experiments, scale 10 µm.
F. Average fluorescence intensity of 12 similarly sized wild-type and variant biofilm colonies expressing the rhlA::gfp reporter. Measurements were obtained just above the biofilm growth surface using identical confocal microscope settings (see Methods). Error bars show SEM.

This finding prompted us to investigate rhamnolipid production by the variant using several methods. The variant produced a larger halo than the wild type when grown on indicator plates containing methylene blue dye (Fig. 5B). The halo is caused by precipitation of this cationic dye by an anionic surfactant, and in P. aeruginosa this is likely to be rhamnolipids (Kohler et al., 2000; 2001). We also measured rhamnolipid production using the orcinol method on ether-extracted culture supernatants (Koch et al., 1991; Zhang and Miller, 1994; Ochsner and Reiser, 1995; Pearson et al., 1997; Van Delden et al., 1998; Rahim et al., 2001). This assay showed that rhamnolipid production by the variant was fivefold higher than the wild type after 2 days of planktonic growth (Fig. 5C).

To confirm these results by an independent method, and to determine whether increased rhamnolipid production resulted from higher gene expression levels or some post-translational mechanism, we utilized a chromosomal green fluorescent protein reporter linked to the promoter of the rhamnolipid biosynthetic genes (rhlAB). During planktonic exponential phase growth, expression of the rhlA::gfp fusion by the two strains was identical (Fig. 5D). In stationary phase, however, expression by the variant exceeded that by the wild type, eventually reaching approximately threefold higher levels. These data are consistent with the results of the orcinol assay (Fig. 5C).

Rhamnolipid gene expression is regulated by a number of factors, not all of which are known (Medina et al., 2003). Thus, the increased level of rhamnolipid expression exhibited by the variant in planktonic culture (Fig. 5C and D) may not reflect the situation during biofilm growth. To examine this, we compared rhlA::gfp expression levels in variant and wild-type biofilms, just prior to the variant's detachment. Consistent with the results seen in the planktonic state, 2-day-old biofilms formed by the variant expressed the rhlA::gfp fusion at a approximately twofold higher levels than the wild type (Fig. 5E and F). Even after 5 days of biofilm growth, expression by the wild type remained significantly lower than that seen in the variant at 2 days (not shown). Analysis of the variant's rhlAB gene sequence (including the promoter region) showed it to be identical to the wild type. Therefore, increased rhamnolipid gene expression by the variant is likely caused by some regulatory mechanism rather than a functional change in this gene.

Rhamnolipids are required for the variant's hyper-detachment phenotype

To determine whether rhamnolipids are required for the variant's accelerated biofilm detachment, we inactivated rhlAB in this strain, and performed two different detachment assays. We confirmed that rhlAB inactivation eliminated rhamnolipid production using the orcinol method. Consistent with previous work in wild-type P. aeruginosa (Davey et al., 2003), inactivation of rhamnolipid genes changed biofilm architecture. Biofilms formed by variant rhlAB strain were more mat-like than variant biofilms (compare top panels, Fig. 6A versus B).

Figure 6.

Effect of rhlAB genes on the detachment phenotype of variant biofilms. Biofilms were established from the variant strain, and the variant with rhlAB deleted. In flow cells, the variant biofilm (A) detached after 2 days of growth. Inactivation of rhlAB genes caused loss of the detachment phenotype (B). Complementation of rhlAB restored the detachment phenotype (C). Images are X-Z views, scale 10 µm. Variant rhlAB biofilms (•) also exhibited lower detachment than the variant (▪) in the static detachment assay (D). Graph shows the mean of three experiments; error bars show SEM.

Inactivating rhamnolipid genes eliminated the variant's accelerated detachment phenotype. Whereas the variant detached after 2 days of biofilm growth (Fig. 6A), variant rhlAB biofilms remained solid, and in fact increased in size over a 7 day period of observation (Fig. 6B). At later time points (∼8–12 days) some loss of biomass did occur but the central hollowing pattern of detachment was not observed. To confirm that the absence of early detachment (occurring at day 2) was caused by rhamnolipid gene inactivation, we complemented the variant rhlAB strain using plasmid-born rhlAB genes. Nascent biofilms formed by the complemented mutant detached to a similar extent as the variant (Fig. 6C), whereas variant rhlAB biofilms carrying a control vector exhibited no discernable detachment during 7 days of growth. The effect of rhlAB mutation was also seen in the static detachment assay [the detachment rate decreased eightfold (Fig. 6D)] and in mixed biofilm experiments involving the variant rhlAB and wild-type strains (Fig. S1 in Supplementary material).

Rhamnolipids produce detachment in wild-type P. aeruginosa biofilms

The data above show that rhamnolipids are required for the hyper-detachment phenotype of the variant. To begin investigating whether rhamnolipids also function in detachment of wild-type P. aeruginosa, we inactivated rhlAB in this strain. The rhlAB mutation decreased the detachment rate of wild-type biofilms in the static detachment assay (Fig. 7A). Interestingly, the wild-type rhlAB strain had an almost identical detachment rate as the variant rhlAB strain [compare, Fig. 6D (•) versus Fig. 7A (•)]. Furthermore, central hollowing was not seen in aged wild-type rhlAB biofilms grown for 14 days (aged wild-type biofilms typically show this after 10–12 days in these conditions). However, as reported by Davey et al. (2003), rhlAB inactivation had a marked effect on biofilm architecture; the characteristic wild-type tower and mushroom shape was lost. The change in biofilm architecture caused by rhlAB inactivation raises the possibility that reduced detachment is a consequence of biofilm structure and not rhamnolipid inactivation per se.

Figure 7.

Effect of rhannolipids on the detachment of wild-type biofilms.
A. Wild-type rhlAB biofilms (•) exhibited a lower rate of detachment than the wild type (▪) in the static detachment assay. Graph shows the mean of three experiments; error bars show SEM.
B. Rhamnolipid production of the wild type, variant and wild-type rhlAB expression strain (with and without arabinose induction). Rhamnolipids were assayed after 2 days of planktonic culture using the orcinol assay. Graph shows the mean of three experiments; error bars show SEM.
C. Central hollowing of biofilm structures produced by inducing rhamnolipid gene expression. Effects of arabinose induction in P. aeruginosa carrying a control construct (top panel) and inducible rhlAB genes (bottom). Images are X-Y views, scale 40 µm. Results are representative of three experiments.
D. Effect of rhamnolipid gene expression on the number of detached bacteria in the effluent medium from flow cell biofilms. Biofilms formed by wild-type P. aeruginosa carrying inducible rhlAB genes (▪) showed an increase in detachment after arabinose induction. Biofilms formed by bacteria carrying a control construct (•) were not affected by arabinose induction. Graph shows the mean of three effluent collections from one experiment; error bars show SEM. Results are representative of three separate experiments.

To circumvent the confounding effects of biofilm architecture, we used two additional approaches to determine whether rhamnolipids can produce detachment in wild-type biofilms. First, we expressed rhamnolipid genes in 3-day-old wild-type biofilms using an arabinose inducible system. Induction of rhlAB gene expression using 0.5% arabinose (in planktonic culture) produced rhamnolipid concentrations very similar to those made by the variant (Fig. 7B). Within ∼20 h of rhlAB induction in biofilms, central cavities began to form, and these enlarged with time (Fig. 7C). Similar to the detachment pattern seen in variant and aged wild-type biofilms, the central cavities contained large numbers of free-floating bacteria. However, unlike the spontaneous detachment pattern of the variant and aged wild-type biofilms, the detached bacteria within central cavities moved only slightly, as if by Brownian motion – a marked stimulation of swimming within the central cavities was not seen.

To confirm that detachment was induced by rhlAB expression, we counted detached bacteria in the effluent medium. Activating rhlAB gene expression produced a marked increase in detachment; 60 h after induction the concentration of bacteria in the effluent medium was increased by ∼10-fold (Fig. 7D). Arabinose induction had little effect on the strain carrying a control vector. It is notable that complete dispersion of the biofilm was not produced; whereas the central cavities enlarged somewhat over time, an outer shell of the biofilm structures remained even after 7 days of continuous arabinose induction.

We also exogenously added purified P. aeruginosa rhamnolipid to 3-day-old wild-type biofilms, and assayed for detachment. The optimal rhamnolipid concentration to test would be that present inside variant biofilms; however, such measurements are not technically possible. Thus, we used the concentration made by the variant in stationary-phase planktonic cultures (0.5 mg ml−1). This concentration had no effect on the viability of planktonic wild type or variant P. aeruginosa. In the static detachment assay, rhamnolipids increased detachment of wild-type biofilms by fourfold in 4 h (not shown).

Analysis in continuous culture flow cells (Fig. 8A) confirmed this finding. Similar to the pattern seen after induced expression of rhamnolipid genes (Fig. 7C), exogenous addition of rhamnolipids caused central cavities to form, which contained free-floating bacteria that were moving slightly (not swimming vigorously). Both mucoid and non-mucoid clinical isolates also readily detached upon addition of rhamnolipids (not shown). Of note, endogenous rhamnolipid production was not required for rhamnolipid-induced detachment as wild-type rhlAB biofilms were detached by rhamnolipid treatment and by the variant strain in biofilm coculture (Figs S2 and S3 in Supplementary material).

Figure 8.

Exogenous addition of surfactants produce central hollowing in wild-type biofilms.
A. Exogenous addition of rhamnolipids produces detachment of wild-type P. aeruginosa biofilms that begins in the centre of biofilm structures. Images are X-Z views, scale 50 µm. Results are representative of three experiments.
B. Exogenous addition of SDS to mature biofilms also produces central hollowing in biofilm structures. Images are X-Y views, scale 50 µm. Results are representative of four experiments.

To determine if central hollowing could be produced by another surfactant, we added SDS (0.2%) to wild-type biofilms. Like rhamnolipids, SDS caused central cavities to form within biofilm structures (see Fig. 8B and Movie S3 in Supplementary material). Taken together, these experiments show that rhamnolipids can mediate detachment of wild-type P. aeruginosa biofilms: elimination of rhamnolipid production reduced detachment, and the addition of rhamnolipids (by two different methods) increased detachment. Furthermore, these experiments suggest that the central hollowing pattern can be produced by different surfactants, even when the surfactant comes from outside of the biofilm.

Discussion

Much of the previous work on biofilm detachment has focused on factors that initiate the process (Webb et al., 2003; Sauer et al., 2004; Thormann et al., 2005). In this study we investigated a detachment effector mechanism using a P. aeruginosa variant that exhibits robust detachment under different biofilm growth conditions. Using several independent methods, we determined that this variant overproduces the biosurfactant rhamnolipid and that rhamnolipids are required for its hyper-detachment phenotype. The rhamnolipid-mediated detachment mechanism of the variant rapidly restores aspects of planktonic physiology to detached cells, including antibiotic sensitivity. Furthermore, we found that both exogenous addition of rhamnolipids and induced expression of rhamnolipid genes can produce detachment in wild-type P. aeruginosa biofilms.

Rhamnolipids produced by P. aeruginosa have many activities. Previous work indicates that rhamnolipids (or rhamnolipid precursors) can facilitate surface motility (swarming) (Kohler et al., 2000; Deziel et al., 2003), mediate the assimilation of hydrocarbons as nutrient sources (Beal and Betts, 2000), affect biofilm architecture (Davey et al., 2003) and alter cell surface polarity (Zhang and Miller, 1994; Al-Tahhan et al., 2000). Rhamnolipids also have antimicrobial activity against other bacteria (Haba et al., 2003), and may disrupt some host defences during infection (Read et al., 1992). Many of these disparate functions stem from the amphipathic properties of rhamnolipids. The presence of both hydrophobic and hydrophilic moieties allow rhamnolipids to concentrate at physical interfaces and change prevailing interactions (Ron and Rosenberg, 2001).

These amphipathic properties also likely mediate rhamnolipids’ detachment functions. Rhamnolipids could act directly on the biofilm matrix to disrupt and solubilize components, perhaps even incorporating the matrix into micelles. Rhamnolipids could also disrupt cell surface structures that act as adhesions. Previous work has shown that rhamnolipids can cause the release of lipopolysaccharide from P. aeruginosa (Al-Tahhan et al., 2000) and they may have similar effects on other surface appendages. If these structures act to tether bacteria to each other or to the matrix, their disruption could induce detachment. A similar mechanism may underlie rhamnolipids’ action in maintaining the fluid channels (open areas) that surround biofilm structures (Davey et al., 2003). However in this case, rhamnolipids may act on cell–surface rather than cell–cell or cell–matrix interactions. The ability of rhamnolipids to promote swarming motility and alter cell surface charge could also help to mediate detachment. Of note, rhamnolipids have been shown to disrupt Salmonella typhimurium biofilms, suggesting that they may have activity against a broad spectrum of organisms (Mireles et al., 2001).

An interesting finding from our study was that the rhamnolipid-mediated detachment mechanism of the variant involved the central hollowing pattern that is also seen in aged wild-type biofilms. Detachment from the biofilm interior has been observed in other organisms including Staphlococcus aureus (Yarwood et al., 2004) and in the oral pathogen Actinobacillus actinomycetemcomitans (Kaplan et al., 2003a,b). In A. actinomycetemcomitans detachment is mediated by the enzymatic action of a N-acetylglucosaminidase rather than a surfactant (Kaplan et al., 2003b), suggesting that the central hollowing detachment pattern may be common across species, even when different detachment mechanisms are operative.

What produces the central hollowing detachment pattern? Other investigators have proposed that central hollowing may occur because the signals that trigger detachment are strongest in the interior of biofilm structures (see model, Fig. 9A) (Allison et al., 1998; Hentzer et al., 2002; Hunt et al., 2004). This would result in preferential activation of detachment mechanisms in the biofilm centre. Two general types of triggers have been suggested that could show this kind of centralized activity: the accumulation of a signalling molecule or waste product, or the depletion of one or multiple metabolic substrates (Hunt et al., 2004).

Figure 9.

Models to explain the central hallowing detachment pattern.
A. Central triggering model. In this model, central detachment results from preferential triggering of detachment in the biofilm interior. The shaded region represents the intensity of a detachment trigger (such as starvation or signal accumulation). The graph depicts the relative concentration of the triggering signal along the biofilm cross-section. See text and the study by Hunt et al. (2004).
B. Central susceptibility model. Central hollowing detachment could also be explained by the presence of distinct subpopulations with different susceptibility to detachment effector mechanisms. Physiological differences in the susceptible central regions (white area) result in preferential detachment irrespective of where the detachment mechanism is triggered.

Interestingly, rhamnolipid expression can be regulated by both of these putative triggers. The rhlAB genes are regulated by cell density-dependent quorum sensing, in which the level of expression is determined in part by the concentration of asyl-homoserine lactone signalling molecules present in the local environment (Ochsner and Reiser, 1995). These (or other signals) could accumulate in the centre of biofilm structures. Rhamnolipids can also be induced by starvation of many nutrients including nitrogen, phosphate, magnesium, calcium, potassium, sodium, iron and trace elements (Desai and Banat, 1997). As with signal accumulation, nutrient limitation may be most severe in the biofilm interior. Furthermore, the multiplicity of starvation signals could provide redundancy for this important function. This could be of great advantage as nutrient limitation is a major threat to fixed, dense bacterial populations.

While it is certainly possible that either starvation or signal accumulation could induce rhamnolipid-mediated detachment, our experiments suggest that central triggering is not required to produce the central hollowing pattern. In all the experiments we conducted that involved increasing rhamnolipid levels in biofilms, the rhamnolipids originated from outside of or uniformly throughout the biofilm. In the case of the exogenous rhamnolipid addition experiments (Fig. 7E), and the coculture experiments in which the variants localized to the outer biofilm regions (Figs 3 and 4), rhamnolipids likely originated from the periphery. The system we used to induce rhamnolipid gene expression results in fairly uniform expression throughout the biofilm structures (P. K. Singh and B. R. Boles, unpubl. obs.; Lequette and Greenberg, 2005).

These findings suggest an alternative hypothesis to explain central hollowing detachment that does not invoke regional triggering: biofilm structures may consist of distinct subpopulations with differing susceptibility to detaching agents; the centre of biofilm structures are more susceptible while the peripheral biofilm regions form a more durable shell that is relatively resistant (see model, Fig. 9B). The fact that the exogenous addition of both rhamnolipids and SDS produced cavities in the centre of biofilm structures suggests that this region may be generally more susceptible to various soluble agents that produce detachment. Increased susceptibility of internal populations could be caused by regional diversity in matrix or adhesin composition, or other physiological differences. The existence of populations with differing susceptibilities may also explain how rhamnolipids act to maintain open areas (fluid channels) around biofilm structures without disrupting the adjacent biofilms (Davey et al., 2003; Espinosa-Urgel, 2003). If bacteria in the durable outer shell secrete moderate levels of rhamnolipids, this could detach bacteria in fluid channels without disrupting the biofilm itself. Of note, the two models are not mutually exclusive; some detachment mechanisms may involve both central susceptibility and central triggering.

Several of our findings lend support to the idea that biofilm detachment is under the control of a global regulator that might co-ordinate related detachment functions. First, whereas inactivation of rhamnolipid genes eliminated the detachment phenotype of the variant and rhamnolipids produced detachment in the wild type, neither rhamnolipid addition nor induced expression of rhamnolipids completely reproduced the pattern of detachment seen in the variant or aged wild-type biofilms. Induced expression brought about incomplete detachment (Fig. 7C), and neither method produced rapidly swimming bacteria within the central cavities (see Movies S1 and S2 in Supplementary material). This finding raises the possibility that the complete central hollowing pattern may require other co-ordinately expressed functions. Examples of such detachment cofactors could include motility functions or matrix degrading enzymes such as exopolysaccharide lyases (Boyd and Chakrabarty, 1995).

Second, although the variant overexpressed rhamnolipids in all the conditions tested, we found no mutation in the rhamnolipid gene sequence (including the promoter region). This suggests that some regulatory mechanism is responsible for rhamnolipid overexpression in the variant. Lastly, the variant's detachment function rapidly restored some aspects of planktonic physiology to the separating cells. This transition to planktonic physiology could also be regulated and serve to prepare bacteria for conditions outside the biofilm.

Biofilms that exist in natural environments or within infected patients are certain to be more complex than those studied here. These biofilms may differ physiologically, contain multiple bacterial species and may incorporate exogenous substances such as mucus and DNA from a host, or polysaccharides and proteins from the local environment. While it remains to be determined whether rhamnolipids would be effective against such biofilms, inducing detachment is an attractive treatment strategy especially given the difficulty in overcoming biofilm-induced resistance.

Acknowledgements

We thank P. Bontu, A. Turner, T. Moninger and M. Nevell for technical assistance and M. J. Welsh, M. R. Parsek, T. L. Yahr, E. P. Greenberg and H. R. Wilson for helpful discussions.

Ancillary