The impact of quorum sensing and swarming motility on Pseudomonas aeruginosa biofilm formation is nutritionally conditional

Authors


*E-mail parsem@u.washington.edu; Tel. (+1) 206 221 7871; Fax (+1) 206 543 8297.

Summary

The role of quorum sensing in Pseudomonas aeruginosa biofilm formation is unclear. Some researchers have shown that quorum sensing is important for biofilm development, while others have indicated it has little or no role. In this study, the contribution of quorum sensing to biofilm development was found to depend upon the nutritional environment. Depending upon the carbon source, quorum-sensing mutant strains (lasIrhlI and lasRrhlR) either exhibited a pronounced defect early in biofilm formation or formed biofilms identical to the wild-type strain. Quorum sensing was then shown to exert its nutritionally conditional control of biofilm development through regulation of swarming motility. Examination of pilA and fliM mutant strains further supported the role of swarming motility in biofilm formation. These data led to a model proposing that the prevailing nutritional conditions dictate the contributions of quorum sensing and swarming motility at a key juncture early in biofilm development.

Introduction

Pseudomonas aeruginosa is a common environmental bacterium that causes a range of opportunistic infections. Some are acute, such as burn wound infections. These are characterized by rapid bacterial growth, sepsis, and if untreated, death of the host. Some are chronic, such as cystic fibrosis (CF) airway infections, where the bacteria persist in the host for years. Two important processes central to P. aeruginosa pathogenesis are the formation of surface-associated communities called biofilms and intercellular signalling, or quorum sensing. These processes represent types of bacterial social behaviours.

Pseudomonas aeruginosa is a model organism for the study of bacterial social behaviour, or ‘sociomicrobiology’ (Parsek and Greenberg, 2005). P. aeruginosa possesses two quorum-sensing systems, las and rhl (Passador et al., 1993; Brint and Ohman, 1995). Each system has its own signal synthase, signal receptor and distinct acyl-homoserine lactone (AHL) signal. Quorum sensing-regulated functions are known to be critical for acute virulence (Passador et al., 1993). Biofilm formation has also been implicated in disease. Most chronic CF airway infections are caused by P. aeruginosa biofilm communities (Costerton et al., 1999).

These two types of social behaviour have been linked. In 1998, the role of quorum sensing in P. aeruginosa biofilm formation was first described (Davies et al., 1998). This study reported that a lasI mutant strain formed biofilms structurally distinct from the wild-type that were easily disrupted by addition of the detergent SDS (sodium dodecyl sulphate). Purevdorj et al. (Purevdorj et al., 2002) also reported that under high shear conditions, biofilms formed by a quorum-sensing mutant strain exhibited clear structural differences compared with the wild-type strain. However, other studies have indicated that quorum sensing may not be important for biofilm development. For example, Heydorn et al. (Heydorn et al., 2002) demonstrated that wild-type and quorum-sensing mutant strains formed flat, uniform biofilms indistinguishable from one another. Different culturing conditions and biofilm reactors have been suggested to explain some of the observed discrepancies, although the reason for the differences remains unclear. In order to develop biofilm- directed therapies targeting quorum sensing, it is important to understand these differences.

Another factor important for P. aeruginosa biofilm development is motility. Two cell surface appendages involved in motility and biofilm formation are flagella and type IV pili (Klausen et al., 2003a). The flagellum drives swimming motility, while type IV pili are involved in a type of surface translocation called twitching. Both the flagellum and type IV pili have been shown to mediate attachment to a surface (O'Toole and Kolter, 1998; Giltner et al., 2006). In addition, the expression of type IV pili is controlled by the catabolite repression control protein, linking its expression to the nutritional environment (O'Toole et al., 2000).

Pseudomonas aeruginosa employs another type of surface motility called swarming, and whether it contributes to biofilm formation is unknown. Swarming has been characterized as flagellar-assisted movement on a viscous surface, such as that found on solid medium containing low percentage agar. Previous studies have suggested that quorum sensing, rhamnolipid production, type IV pili and the flagellum all contribute to swarming (Köhler et al., 2000; Déziel et al., 2003). Quorum sensing control of swarming is thought to be mediated by RhlR, which activates expression of the rhlAB genes (Ochsner et al., 1994a). These genes encode enzymes required for production of the surfactant, rhamnolipid (Ochsner et al., 1994a,b).

In this study we provide evidence that swarming motility can contribute to early stages of P. aeruginosa biofilm formation. We further show that both the magnitude of swarming and its control by quorum sensing are carbon source-dependent. The biofilm and swarming motility phenotypes of the wild-type and quorum-sensing mutants were distinctly different when grown with succinate, but not when grown on either glutamate or glucose. This nutritionally conditional relationship between swarming motility and quorum sensing represents a way in which environmental conditions can dictate when quorum sensing is important for biofilm development.

Results

The nutritional environment influences the role quorum sensing plays in biofilm formation

The carbon source used to grow P. aeruginosa has been shown to impact biofilm structure (Klausen et al., 2003a). Klausen et al. reported that growth on either amino acids or citrate promotes the formation of flat, uniform biofilms, while growth on glucose promotes the formation of structured biofilms containing large cell aggregates. As quorum sensing has also been shown to impact biofilm structure, the relationship between quorum sensing and biofilm formation was examined with strains grown on different carbon sources.

A wild-type strain of P. aeruginosa and two isogenic quorum-sensing mutant strains, lasRrhlR and lasIrhlI, were cultured in biofilm flow cell chambers using a defined minimal medium supplemented with three different carbon sources: succinate, glutamate and glucose. The doubling times in liquid culture on these three different carbon sources were roughly equal (∼1 h). At 48 h, the wild-type strain exhibited different biofilm structure depending upon the carbon source (Fig. 1, and data not shown). Biofilms cultured with either glutamate or succinate were flat and fairly uniform. In contrast, glucose-grown biofilms consisted of a thin monolayer of cells punctuated with cell aggregates. The general structure of the biofilms remained similar at later time points, such as 4 days (data not shown). These results are consistent with those reported by Klausen et al. (Klausen et al., 2003a). Structural differences between flat biofilms grown on glutamate and succinate compared with the structured biofilms grown on glucose were measured using the image analysis software comstat (Table 1). The comstat parameter that best highlights the differences between flat, uniform biofilms and those containing cell aggregates is surface roughness.

Figure 1.

Flow cell biofilms for wild-type and lasIrhlI strains grown with different carbon sources. These images were acquired 48 h after inoculation of the system. SV, side view (x,z-plane); TD, top down view (x,y-plane). Gridlines are spaced 20.1 μm apart.

Table 1. comstat analysis of 48 h biofilm experiments.a
 Roughness coefficient (dimensionless)Average thickness (μm)
  • a. 

    Results are averages ± standard deviation for n = 10 samples.

  • b–e. 

    Roughness coefficients different for 99% confidence interval, P < 0.003, by t-test.

Glutamate
 wtb0.019 ± 0.00523.3 ± 1.8
 lasIrhlI0.045 ± 0.02224.9 ± 2.3
 lasRrhlR0.023 ± 0.00732.5 4.0
Glucose
 wtb,c0.664 ± 0.36018.7 ± 7.8
 lasIrhlI0.369 ± 0.35527.1 ± 18.3
 lasRrhlR0.367 ± 0.35126.6 ± 12.9
Succinate
 wtc,d,e0.033 ± 0.01119.8 ± 2.2
 lasIrhlId0.306 ± 0.19916.9 ± 6.0
 lasRrhlRe0.126 ± 0.08323.7 ± 2.7

When grown on succinate, quorum-sensing mutant biofilms were distinct from the wild-type strain (Fig. 1, Table 1). Unlike the flat, uniform biofilms formed on succinate by the wild-type strain, the quorum-sensing mutant strains formed biofilms containing cell aggregates. Complementation of the lasIrhlI mutant strain by addition of exogenous purified AHLs (5 μM butyryl homoserine lactone and 3-oxo-dodecanoyl homoserine lactone) to the succinate-containing growth medium restored wild-type biofilm architecture (Fig. S1). When grown on either glucose or glutamate, the quorum-sensing mutant strains formed biofilms similar to wild-type (Fig. 1, Table 1).

Predictive modelling suggests that differences in surface motility can explain differences in biofilm structure

Based upon previous work, we hypothesized that differences in surface motility could explain differences in biofilm structure at early stages of development (Landry et al., 2006). The motility hypothesis predicts that flat, uniform biofilms are characterized by significant surface-associated motility, while biofilms containing cell aggregates are the result of limited surface movement and clonal growth of attached cells. Mathematical simulations are an effective way to visualize this concept. We used a kinetic Monte Carlo method to illustrate the potential influence of surface motility on the architecture of a biofilm. In this method, the biofilm itself is represented as a collection of cells arranged in a three-dimensional lattice. The cells are then assigned values for reproduction, motility and attachment (i.e. cells that are no longer moving and become permanently immobilized on the surface or biofilm). Initially, we assumed an equivalent reproduction rate for motile and non-motile cells and only varied the surface motility rate; thus, simulations contain the same number of bacteria at all time points. For purposes of computing transition probability rates, cells were divided into two categories: surface cells and buried cells. It was assumed that the surface cells are at the periphery of the biofilm in direct contact with the overlying bulk liquid and thus have direct contact with nutrients. Therefore, surface cells are assumed to consume a majority of available nutrients and thus are the only cells able to reproduce. Similarly, only surface cells were able to move on the surface. However, a buried cell that is exposed by surface motility of an overlying cell is then considered a surface cell and may begin reproducing and moving. After cell division, daughter cells were placed in unoccupied sites in the lattice, which are adjacent to the parent cell.

The biofilm structure predicted by the simulation is shown in Fig. 2. Flat, uniform biofilms are observed when cells are actively moving on the surface (Fig. 2A), while cellular aggregates are observed under conditions of low surface motility (Fig. 2B). As the predicted biofilm structure was consistent with the biofilm structure observed in actual experiments, we examined the role of surface motility early in biofilm formation.

Figure 2.

Mathematical simulation predicting the effect of surface motility on biofilm structure.
A and B. Predicted biofilm structure where cells are exhibiting high surface motility and low motility. All cells are growing with equal reproduction rates for both simulations. The predicted biofilm structure is shown after 25 000 cell divisions of 10 initially adherent single cells.
C and D. Refined mathematical simulation predicting the effect of surface motility on biofilm structure. Experimental values derived from time-lapse microscopy series were used to refine the simulation. Predicted biofilm structure where cells are exhibiting high surface motility and where cells are exhibiting limited motility. The predicted biofilm structure is shown after growth of 10 initially adherent single cells. Green cells are motile, and red cells are non-motile.

Time-lapse microscopy supports the motility hypothesis

Pseudomonas aeruginosa strains were observed using time-lapse microscopy at early stages of biofilm development. Flow cell-grown biofilms were incubated in an environmental chamber at 30°C and a single field of view was examined for ∼20 min. Motility was monitored after attachment of bacteria to the surface during a period when single cells could still be clearly resolved by microscopy (∼4 h after inoculation).

Analysis of these time-lapse series showed patterns of surface motility consistent with the motility hypothesis (Fig. 3). Wild-type cells grown on succinate and glutamate had a greater velocity (0.58 ± 0.58 μm min−1 and 0.41 ± 0.48 μm min−1, respectively, n = 30), while forming a flat, uniform biofilm (Fig. 1). In contrast, bacteria grown on glucose were observed either to not move at all or to move slowly (0.11 ± 0.13 μm min−1, n = 30), and the development of biofilms containing cell aggregates was observed (Fig. 1). The paths of 30 randomly selected individual cells tracked over the course of the time period for the three carbon sources are shown in Fig. 3. (Time-lapse movie files are included in Supplemental materials.)

Figure 3.

Time-lapse microscopy of selected strains grown with different carbon sources. Individual cells were tracked for a period of 20 min for succinate-, glucose- and glutamate-grown wild-type and succinate-grown lasRrhlR. Thirty separate cells were individually tracked for each field of view, with different colours representing the tracking of a single cell for the given time period.

The quorum-sensing mutant strains grown with succinate formed biofilms with many cell aggregates, similar to the wild-type strain grown on glucose (Fig. 1). The motility hypothesis predicts that the mutant strains would exhibit reduced surface motility when grown on succinate. Time-lapse microscopy showed this to be the case. The lasRrhlR mutant grown on succinate showed greatly reduced surface velocity in flow-cells (0.07 ± 0.08 μm min−1, n = 30) (Fig. 3).

The data generated from time-lapse microscopy were used to revise the parameters of the mathematical simulation. New values were entered for key variables such as average surface speed, relative ratio of motile and non-motile cells, attachment rate (i.e. cells that do not move or stop moving) and reproduction rate (Table 2). In Fig. 2C and D, we show the predicted biofilm structure of revised simulations, which illustrate the impact of high and low surface motility rates on biofilm structure. Both experiments start with 10 individual cells randomly placed on the surface. High-motility conditions (e.g. succinate-grown wild-type) predict that the bacteria rapidly spread across the surface, forming a biofilm with a uniform mat of cells (Fig. 2C). For low-motility conditions (e.g. glucose-grown wild-type), growth is almost exclusively around the initial attached cells, leading to aggregates (Fig. 2D).

Table 2.  Surface cell transition probability rates for revised simulation.
 Wild-type succinate (high motility)Wild-type glucose (low motility)
Reproduction rate Motile daughter (divisions cell−1 min−1)7.07 × 10−32.58 × 10−3
Reproduction rate Non-motile daughter (divisions cell−1 min−1)5.05 × 10−41.10 × 10−3
Motility rate (Δx min−1)4.640.228
Attachment rate (attachment cell−1 min−1)0.0140.046

The biofilm structures predicted by these refined simulations are more representative of the actual biofilm structure observed in flow cell reactors. The simulation suggests that much of the observed biofilm structure observed at early stages of biofilm formation can be explained by surface motility.

Quorum-sensing control of swarming motility is nutritionally conditional

As wild-type and quorum-sensing mutant strains exhibited carbon source-dependent patterns of surface motility in flow cell reactors, known modes of surface motility were investigated. P. aeruginosa has two mechanisms for movement on a surface, twitching and swarming. In an attempt to relate our flow cell biofilm observations to either twitching or swarming, the wild-type and quorum-sensing mutant strains were assayed for these two types of motility when grown with different carbon sources.

Significant differences in swarm diameter were observed between strains on the tested carbon sources. The wild-type strain swarmed readily when grown with succinate or glutamate, but poorly on glucose (Fig. 4). This general trend was consistent over a range of surface viscosities (solid medium with 0.5–1.0% agar; see Fig. S2). Quorum-sensing mutant strains had swarming phenotypes that were carbon source-dependent. When grown with succinate, the quorum-sensing mutant strains were defective for swarming (Fig. 4, Fig. S2). As expected, addition of purified 5 μM 3-oxo-dodecanoyl homoserine lactone and butyryl homoserine lactone to the succinate plates restored wild-type swarming to the lasIrhlI strain, but not the lasRrhlR mutant strain (Fig. S1). In contrast, the mutants swarmed similar to wild-type for the other two tested carbon sources, showing little swarming on glucose and significant swarming on glutamate. To address whether quorum sensing itself was nutritionally conditional, signal production was measured for the wild-type strain grown on each of the carbon sources. Similar amounts of signal were produced for each culturing condition (data not shown).

Figure 4.

Swarm, swim and twitch plate motility assays for wild-type, lasIrhlI and lasRrhlR strains grown with different carbon sources. In this study 12 mM succinate, glucose, or glutamate was used as the sole carbon source in FAB medium. The agar percentage used for each type of assay is indicated on the figure.

All strains exhibited the same degree of twitching and swimming motility on the three carbon sources (Fig. 4). To ensure that our results were not strain-specific, the swarming motility patterns of another laboratory wild-type strain PAO1 and a PAO1 lasRrhlR mutant strain were tested and found to be similar to our quorum-sensing mutant strains (data not shown).

Rhamnolipids are not critical for swarming motility

Quorum-sensing control of swarming motility is thought to be exerted through regulation of rhamnolipid biosynthesis. We hypothesized that quorum sensing-regulated rhamnolipid production might be important for swarming on succinate and not for the other two carbon sources. Therefore, the dependence of swarming motility on rhamnolipid production was examined for all three carbon sources. Wild-type cultures produced rhamnolipid under all tested conditions, while lasIrhlI, lasRrhlR and the rhlAB mutant strains produced none that was detectable (Table 3). Interestingly, the wild-type strain produced the most rhamnolipid when grown on glucose, but showed the least amount of swarming on this carbon source (Fig. 4).

Table 3.  Total rhamnolipid (mono- and di-) produced by strains grown with different carbon sources.a
StrainGlutamatebGlucosebSuccinateb
  • a. 

    Cells cultured at 30°C in FAB medium with 50 mM glutamate, glucose, or succinate.

  • b. 

    Units are mg l−1.

  • ND, not detectable.

Wild-type8.668.418.9
lasIrhlINDNDND
lasRrhlRNDNDND
rhlABNDNDND
fliM9.767.123.4
pilA8.571.119.0

A precursor for rhamnolipid synthesis, C10-C10 3-(3-hydroxyalkanoyloxy)alkanoic acid (HAA), has also been implicated as a surfactant for swarming motility (Déziel et al., 2003). LC/MS analysis for HAA showed a compound with m/z of 357 was present in wild-type cultures (data not shown). While the amount of HAA produced could not be quantified (as no standard is commercially available), it was observed that the relative abundance of HAA produced by the wild-type strain was approximately threefold higher when grown on succinate than when grown on glucose or glutamate. The rhlAB and quorum-sensing mutant strains did not produce any detectable HAAs, which is consistent with Déziel et al. (Déziel et al., 2003).

The contribution of rhamnolipid to swarming motility was further analysed using a rhlAB mutant strain. The rhlAB mutant swarmed when grown on both succinate and glutamate, but this swarming was slightly reduced compared with the wild-type, in particular on succinate (Fig. 5). This strain did not have a distinguishing phenotype on twitching or swimming motility plates, and the production of rhamnolipid could be restored by providing the rhlAB genes in trans (data not shown). We also tested the common P. aeruginosa laboratory strains, PAO1 and PA14. Swarming of the PA14-rhlA and PAO1-rhlAB mutant strains was similar to those of our rhlAB mutant, verifying that our observations were not strain-specific (Fig. S3, data not shown). The same PA14-rhlA strain was shown to be swarming-deficient in a previous study using different culturing conditions (Caiazza et al., 2005).

Figure 5.

Swarm plate motility assays for wild-type, fliM, pilA and rhlAB strains. Plates contained 0.5% agar and FAB medium was supplemented with 12 mM succinate, glucose, or glutamate as the sole carbon source.

Swarming motility appears to be responsible for surface motility at early stages of biofilm development

Our data correlated swarming motility to the surface motility observed in flow cells. In order to determine if some of the movement might be due to twitching motility, a strain containing an unmarked deletion of the pilA gene was constructed. The pilA gene encodes the pilin subunit of type IV pili. Twitching motility was completely abolished in this strain and could be restored by providing the pilA gene in trans (Fig. 6). This strain was not impaired for swarming and actually displayed a hyper-swarming phenotype under some conditions (Fig. 5). The hyper-swarming phenotype was most pronounced on glucose, and when complemented in trans with a functional copy of pilA, wild-type levels of swarming were restored (Fig. 6). Swimming motility was not affected in this strain (data not shown). PA14 and PAO1 pilA mutant strains also exhibited wild-type swarming motility suggesting that this observation is not strain-specific (data not shown). Examination of rhamnolipid production revealed that the pilA mutant strain produced wild-type levels, suggesting that the hyper-swarming phenotype is not due to rhamnolipid overproduction (Table 3).

Figure 6.

Complementation of the pilA strain restores wild-type swarming and twitching. These strains were grown on FAB supplemented with 12 mM glucose at 0.5% agar for swarming motility, or LB at 1.0% agar for twitching motility. The control plasmid was pUCP18, while pDA2 was pUCP18 containing a functional copy of the pilA gene. Plates contained 300 μg ml−1 carbenicillin for plasmid selection and maintenance.

A fliM mutant strain, which does not produce a functional flagellum, was then constructed to serve as a negative control for swarming (Fig. 5). This mutant was completely defective for swarming and swimming motility under every condition tested, while rhamnolipid production was not altered (Table 3).

The pilA mutant strain was grown on glucose in flow cell reactors and examined by time-lapse microscopy. As described earlier, the wild-type strain showed little surface motility under these culturing conditions (Figs 3 and 7). However, the pilA strain displayed increased surface motility when grown on glucose (1.2 ± 1.6 μm min−1, n = 30) (Fig. 7), which is consistent with the strain's hyper-swarming phenotype on this carbon source. The fliM mutant strain was then grown on glutamate, a carbon source that promoted surface motility of the wild-type strain in time-lapse microscopy series (Figs 3 and 7). This strain failed to show surface motility under these conditions (0.05 ± 0.06 μm min−1, n = 30) (Fig. 7), consistent with this strain's defective swarming motility phenotype (Fig. 5).

Figure 7.

Time-lapse microscopy of selected strains. Individual cells were tracked for a period of 20 min for glucose- and glutamate-grown wild-type and glucose-grown pilA and glutamate-grown fliM strains. Thirty separate cells were individually tracked for each field of view, with different colours representing the tracking of a single cell for the given time period.

Discussion

Our data have led us to propose the following model relating surface motility to early steps in biofilm formation: under conditions promoting swarming motility, cells are continuously moving on the surface and as they multiply and continue moving, they form a flat, uniform biofilm (Fig. 8). Conditions resulting in limited swarming motility, such as growth of the wild-type strain on glucose, produce biofilms with cellular aggregates. Therefore, early in biofilm formation, the magnitude of swarming motility exhibited by cells may represent a juncture at which biofilm development proceeds to form a flat, uniform biofilm or a structured biofilm (Fig. 8).

Figure 8.

Model for nutritional contribution to early biofilm formation. The carbon source exerts control over surface motility and quorum sensing to influence the initial coverage of the substratum after attachment of cells in the early stages of biofilm formation. Cells with low motility form aggregates, leading to more structured biofilms, while cells with high motility spread across the surface, leading to a homogeneous and flat biofilm. Cells have been colour-coded for structured biofilms. Orange cells represent the immobile ‘stalks’ of structured biofilms, while light blue cells represent the motile subpopulation that produces the ‘cap’ as described by Klausen et al. (2003b). Three roles for quorum sensing have been proposed in biofilm development. These are highlighted by circled red letters: (A) the degree of swarming motility at early stages determines whether P. aeruginosa will proceed to form a flat or a structured biofilm; (B) the release of extracellular DNA contributes to the EPS matrix structural integrity; and (C) rhamnolipids maintain open channels between large cell aggregates.

In support of this model, the motility observed in time-lapse microscopy series correlated to swarming motility phenotypes. Strains grown under conditions promoting swarming motility in plate assays exhibited significant surface motility and formed flat, uniform biofilms in flow cell experiments, while strains or conditions showing limited swarming motility resulted in biofilms containing cell aggregates. Twitching and swimming motility levels did not differ when the strains were grown on the different carbon sources. In addition, time-lapse microscopy of the pilA and fliM strains further suggested that swarming motility is responsible for the observed surface movement. Taken together these data strongly suggest that the observed motility in time-lapse series is swarming, not twitching or swimming motility.

Previous work has suggested that type IV pili-driven twitching motility is a key component for the formation of flat biofilms. Our data are not necessarily inconsistent with this report, as twitching motility may be important in later stages beyond those at which the time-lapse microscopy series were taken. Twitching motility has also been shown to be important in later steps in the formation of a structured biofilm. Klausen et al. (Klausen et al., 2003b) demonstrated that in older glucose-grown biofilms (> 5 days), the bacterial migration that forms the ‘cap’ of structured biofilms is type IV pili-mediated. Cap formation is a process by which a motile subpopulation in older biofilms moves across the surface and accumulates on an immobile subpopulation of cells. This immobile subpopulation is an aggregate of cells and has been termed ‘the stalk’. The stalk subpopulation is depicted in Fig. 8 as the orange aggregate of cells.

The traditional assay for swarming motility is movement on semisolid nutritive agar plates. Microscopy of P. aeruginosa swarming colonies on these plates has led to the notion that swarming requires groups or rafts of cells in close physical proximity (Köhler et al., 2000). While this may be true on semisolid medium, our data suggest that on the surface of flow cell biofilm reactors single cells can effectively engage in this type of motility as well.

Mathematical simulations proved very useful for predicting the impact of surface motility on biofilm structure. The differences between our revised simulation (Fig. 2C and D) and the actual P. aeruginosa biofilm structures suggest there may be other key experimental variables that still need to be identified to further refine its predictive capabilities. For example, an important experimental observation not incorporated into the simulation is that under low-motility conditions only a fraction of initial attached cells give rise to cell aggregates. Why this happens is unclear; however, this may in part account for the thin carpet of cells lying between the aggregates seen in the actual biofilms (e.g. the wild-type biofilms grown on glucose, Fig. 1). This thin carpet of cells between aggregates is currently not predicted in our revised version of the simulation (Fig. 2D). Another potentially key parameter missing from the current model is exopolysaccharides (EPS). There is evidence in P. aeruginosa that expression of EPS is inversely regulated with motility (Garrett et al., 1999; Kirisits et al., 2005). EPS production has the potential to influence aggregation by facilitating cell–cell interactions. Ultimately, identification and incorporation of key parameters into the model may lead to a simulation that more accurately predicts biofilm structure.

The genetic determinants contributing to swarming motility were investigated. The roles of some factors were not influenced by nutritional conditions. For example, flagella were required for swarming under every condition tested. This supports previous studies and is consistent with the traditional definition of swarming as flagellar-assisted surface motility. Previous work had suggested that a pilA mutant strain was defective for swarming (Köhler et al., 2000). In our case a pilA mutant strain was not impaired for swarming motility, exhibiting a hyper-swarming phenotype under some culturing conditions. The explanation for this observation is not clear. One possibility is that type IV pili may physically impede flagellar rotation/function, because both appendages are polarly localized. Alternatively, the presence of type IV pili may exert physical drag on the cell in the viscous environment on the swarm plate surface. Another possibility is that the absence or presence of type IV pili may affect flagellar expression.

We also found that the nutritional environment can influence the functions required for swarming motility. Quorum sensing has been previously shown to control swarming motility (Köhler et al., 2000). Consistent with this, quorum-sensing mutants are defective for swarming when succinate is provided as the sole carbon source. However, quorum-sensing mutants grown on glutamate displayed wild-type levels of swarming motility. There is precedence for this observation, studies with Serratia liquefaciens have also indicated that quorum-sensing control over swarming motility is nutritionally conditional (Eberl et al., 1996). One possible explanation is that quorum sensing-regulated rhamnolipids are required for swarming on succinate, but not on glutamate. The rhlAB mutant strain did swarm less than the wild-type strain on succinate, although it swarmed to a greater degree than the quorum-sensing mutants. Therefore, some additional quorum sensing-regulated factor besides rhamnolipids may contribute to swarming on succinate. The quorum-sensing mutant strains swarmed as well as the wild-type on glutamate, even though these strains did not make rhamnolipid (Table 3). Complicating our analysis is the fact that rhamnolipid measurements were made in 24 h stationary-phase liquid cultures. Relating these data to what occurs in the flow cell and swarm plate culturing formats is difficult. Whether swarming on glutamate proceeds in a surfactant-independent fashion or requires an alternative surfactant is unclear. Initial water contact drop angle studies suggest that growth on glutamate does not produce an alternative surfactant (data not shown, ongoing work).

Our study has focused on the contribution of quorum sensing-regulated surface motility to early steps in biofilm development. It is difficult to compare our results with previous studies examining the role of quorum sensing in P. aeruginosa biofilm formation, because other quorum sensing-regulated functions have also been shown to be important in later stages of biofilm development (Fig. 8). Alleson-Holm et al. (Allesen-Holm et al., 2006), demonstrated that quorum sensing controls production of extracellular DNA and that this can impact biofilm structure when P. aeruginosa biofilms are grown on glucose. Quorum sensing-regulated rhamnolipid production has also been shown to maintain open channel formation in older (> 5 days), structured biofilms of P. aeruginosa (Davey et al., 2003). Therefore different quorum sensing-regulated functions may contribute to biofilm formation at different stages of development, and the contribution of these functions may also be influenced by environmental conditions.

The mechanism(s) behind the nutritional effects on swarming motility are unclear. Quorum sensing does not appear to be nutritionally conditional, as signal production levels were similar under all tested growth conditions. The effect of the nutritional environment on swarming appears to be very complex. In this study, we found that alternate carbon sources profoundly influence swarming and can also override functions previously thought to be required. Köhler et al. (Köhler et al., 2000) demonstrated that the nutritional environment exhibited significant control upon swarming motility, with certain amino acids inducing swarming, while Déziel et al. (Déziel et al., 2003) showed that iron availability can also influence swarming motility. Our study suggests that environmental/nutritional factors that influence swarming motility have the potential to influence biofilm structure. Why might it be advantageous to P. aeruginosa to alter biofilm structure in response to the nutritional environment? The answer to this is uncertain. However, biofilm structure influences the gradients within a community. Thus, altering structure may be a means of optimizing the gradient profiles of key chemical/nutritional species. An additional implication of this study relates to eradication of unwanted biofilms. Because biofilm structure can influence its antimicrobial resistance properties (Landry et al., 2006), understanding and controlling this process could have medical and industrial applications.

Experimental procedures

Bacterial strains and growth medium

Strains and plasmids utilized for this study are included in Table 4. Cells were grown using autoclaved FAB medium (Heydorn et al., 2000) with filter-sterilized glucose, glutamate, or succinate added as the sole source of carbon.

Table 4.  Strain list.
Strain or plasmidRelevant characteristicsSource or reference
P. aeruginosa
 Wild-typeATCC 15692 wild-type 1CATCC
 lasIrhlI15692 ΔlasI ΔrhlI; Gmr, TcrThis study
 lasRrhlR15692 ΔlasR ΔrhlR; Gmr, TcrThis study
 wt-gfp15692 miniTn7 gfp2; Cmr, GmrThis study
 lasIrhlI-gfplasIrhlI miniTn7 gfp3; Gmr, Tcr, Kmr, SmrThis study
 lasRrhlR-gfplasRrhlR miniTn7 gfp3; Gmr, Tcr, Kmr, SmrThis study
 fliM-gfp15692 fliM, miniTn7 gfp3; Kmr, Smr, TcrThis study
 rhlAB15692 ΔrhlAB; GmrThis study
 pilA15692 ΔpilA; markerlessThis study
 pilA-gfppilA miniTn7 gfp3; Kmr, SmrThis study
 PA14Wild-typeRahme et al. (1995)
 PA14-rhlAPA14 ΔrhlA; GmrPukatzki et al. (2002)
 PA14-pilAPA14 pilA::Tn5-B30; TcrO'Toole
 PAO1A wound isolate commonly used by many laboratoriesHolloway et al. (1979)
 PAO-JP3ΔlasR ΔrhlR; Tcr, HgrPearson et al. (1997)
Plasmids
 pEX18ApAllelic replacement vector, AprHoang et al. (1998)
 pRK600Mobilization plasmid, CmrKessler et al. (1992)
 pUX-BF13Conjugation helper plasmid, AprBao et al. (1991)
 pBK-miniTn7-gfp2mini Tn7 gfp2; Cmr, Apr,GmrKoch et al. (2001)
 pBK-miniTn7-gfp3mini Tn7 gfp3::Cmr, Apr, Kmr, SmrKoch et al. (2001)
 pTTN61fliM::Tcr used for allelic replacementKlausen et al. (2003a)
 rhlAB-KOrhlA::Gmr::rhlB used for allelic replacementBoles et al. (2005)
 pJDS101pilA allelic replacement vector in pEX18ApThis study
 pUCP18Expression vector, AprWest et al. (1994)
 pEC03rhlAB expression plasmid, AprBoles et al. (2005)
 pDA1rhlR + lasR expression plasmid in pUCP18An et al. (2006)
 pDA2pilA expression plasmid in pUCP18An et al. (2006)

Biofilm flow cell experiments

Biofilms were cultured in sterilized flow-cells at 30°C for 48 h in FAB with 0.6 μM glucose, glutamate, or succinate. The medium flow was approximately 3.75 ml h−1. Flow cell chambers were inoculated at the start of the experiment with 100 μl of log-phase growing cells (OD600 ≈ 0.7) diluted with fresh medium to OD600 = 0.05. P. aeruginosa strains used for biofilm experiments contained a mini-Tn7 chromosomal, constitutive, GFP-expressing insertions that allowed visualization of cells using fluorescent microscopy (Koch et al., 2001). Images were acquired using a Bio-Rad confocal laser and Nikon microscope equipped a 40× or 60× objective and collected with Bio-Rad LaserSharp software. Image series were compiled using Volocity (Improvision, Lexington, MA) software. Characteristics of image series (n = 5 × 2 duplicate channels = 10) were quantified using comstat analysis software (Heydorn et al., 2000).

Mathematical simulations

We conducted a numerical experiment to illustrate the influence of surface motility on the biofilm architecture. We used the kinetic Monte Carlo method to simulate growth of a biofilm (Bortz et al., 1975; Binder, 1986). In this method, the biofilm itself is represented as a collection of cells arranged in a rectilinear lattice. The cells are then assigned transition probability rates for reproduction, motility and attachment.

The boundary conditions are taken to be periodic, and the system is assumed to be saturated with substrate, so there is no preferential growth at the top of the biofilm versus the bottom. Cell transition probability rates were determined by dividing cells into two categories: surface cells and buried cells. It is assumed that the surface cells consume the majority of the available substrate, so surface cells are the only cells able to reproduce. Similarly, only surface cells are able to move along the surface. However, a buried cell that is subsequently exposed by surface motility is then considered a surface cell and may begin reproducing and moving. Reproduction probability rates were set at one cell division per unit time for both low and high motility cells while motility probability rates were set at one cell length per unit time and 1000 cell lengths per unit time for low and high motility respectively. For cell reproduction, daughter cells are placed in unoccupied sites in the lattice, which are adjacent to the parent cell. Moving cells are not permitted to cause the biofilm to become detached from the substratum. For the first simulation, the time period was set so that the biofilm will occupy approximately 50% of the computational domain at the end of the simulation.

Transition probability rates for the revised simulation were determined from data taken from experimental observations. The transition probability rate of reproduction was computed by r = ln(N(T)/N(0))/T, where T is the elapsed time of the experiment, and N(t) is the population count at time t. The fraction of daughter cells that were motile was taken to be proportion of the whole population that was motile. Reproduction rates for motile and non-motile cells were taken to be the same. The attachment transition probability rate was computed using a similar formula, r = ln(M(0)/M(T))/T, where M(t) is the number of motile cells at time t. The transition probability rate for diffusion of motile cells was drawn from a standard random walk formulation, r = <d2>/(TΔx2), where <d2> is the observed mean square displacement of a motile cell after time T, and Δx is the length of a lattice segment. The lattice sites were taken to be cubic, with Δx = 1.65 μm. The boundary conditions are taken to be periodic in the lateral directions. We assume the substrate is saturating so that all cells with available neighbour lattice sites are permitted to reproduce. All cells are identified as motile or non-motile. Motile cells may move and/or attach to the substratum or other non-motile cells. Non-motile cells are not permitted to detach and become motile.

Motility plate assays

Plate assays were used to determine differences in motility upon different carbon sources. FAB medium with 12 mM glucose, glutamate, or succinate was solidified with Noble agar (0.4–1.0%). Plates were inoculated using a sterilized platinum wire with log-phase cells (OD600 ≈ 0.7) grown in the respective carbon source approximately 16 h after pouring the plates and incubated at 30°C for 60 h.

LC/MS analyses of rhamnolipid and HAA production

Quantification of rhamnolipid was performed using liquid chromatography/mass spectroscopy methodology similar to that described by Déziel et al. (Déziel et al., 1999). Liquid samples were prepared from filtered (0.2 μm) supernatant of stationary-phase cultures (OD600 ≈ 2.0). The analyses were performed with a single quadrupole mass spectrometer Agilent 1100 series LC/MSD (Agilent Technologies Canada, Montreal) using electrospray ionization in negative mode. Capillary voltage was set at 3000 V both positive and negative. The drying gas flow was kept at 12 L min−1 and the drying gas temperature was set at 350°C. The nebulizer pressure was constant at 35 psi. The mass spectrometers fragmentor was set at 70 and the gain at 1.0. Dwell time was 580 ms and per cent relative dwell was 100. Molecular weight was set for each of the three single ion monitoring experiments at 357, 509 and 649 m/z corresponding to the predominant di-rhamnolipid compound (Rha-Rha-C10-C10) made by P. aeruginosa and two precursors, mono-rhamnolipid and HAA (Déziel et al., 1999; 2003; Lépine et al., 2002). The instrument was interfaced to an Agilent 1100 HPLC (Agilent Technologies Canada, Montreal) equipped with an Agilent G1313A autosampler and an Agilent G1311A quaternary pump using a 150 mm × 3.0 mm Agilent Zorbax SB-C8 C8 column (particle size 3.5 μm). An isocratic method consisting of 70% acetonitrile, 30% water was used with a flow rate of 300 μl min−1. Peak areas for m/z 509 and 649 were compared with a known mixture of Rha-C10-C10 mono- and Rha-Rha-C10-C10 di-rhamnolipid provided by the Jeneil Biosurfactant Company (Saukville, Wisconsin) to determine mass.

Plasmid and strain construction

DNA manipulations were performed using standard methods. Genomic DNA and plasmid DNA were prepared using the QIAmp Blood Mini Prep Kit and the QIAprep Spin Miniprep Kit (QIAGEN, Valencia, CA), respectively, and DNA fragments were excised and purified from agarose gels by GFX polymerase chain reaction (PCR) kit (Amersham Biosciences, Piscataway, NJ). To construct the pilA mutant strain, a PCR product upstream of the pilA gene (using primers with the sequences: GTCCGAATTCACCTTACCGCTCAGTTGAT and GCGTGGATCCAGGTTGTGATAACTAAGGT) was ligated into the pEX18Ap vector between the EcoRI and BamHI sites and a PCR product downstream of the pilA gene (using primers with the sequences: GACTTGATCCGCCTGAACCTCAACAAGCAC and GCAGCATGCATCCTCTTGGGTGGACTTG) was ligated into the pEX18Ap + upstream DNA vector between the BamHI and SphI sites to create plasmid pJDS101. Allelic replacement vectors bearing mutations were mobilized from Escherichia coli SM10 into P. aeruginosa by conjugational mating. For pilA, single-cross-over recombinants were selected by plating on LB containing 500 mg carbenicillin per millilitre. Double-cross-over recombinants were then selected by plating on LB containing 5% sucrose. Knockouts of rhlAB and fliM were acquired similarly using selection of 200 mg ml−1 gentamycin and 500 mg ml−1 tetracycline respectively (Klausen et al., 2003a; Boles et al., 2005). All strain constructions were confirmed by phenotypic assay and PCR analysis.

Acknowledgements

Support was provided by NIH Grant No. 1R01-GM67248-01, NSF Grant MCB 0133-833 and NIH Training Grant No. T32 AI07511 (J.D.S.). Strains P. aeruginonsa PA14, PA14-rhlA and PA14-pilA were provided by George O'Toole of Dartmouth Medical School. Plasmid TTN61 and strains for transposon mutagenesis were provided by Tim Tolker-Nielsen, Paula Ragas Lotte Lambertsen, and Soren Molin of The Technical University of Denmark. Purified rhamnolipids were provided by the Jeneil Biosurfactant Company (Saukville, Wisconsin). We would also like to thank D. D'Argenio, D. Kearns, L. Hoffman, L.L. McCarter, P.K. Singh, C.S. Harwood and A.L. Schaefer for helpful comments and discussion.

Ancillary