Detailed knowledge of the developmental process from single cells scattered on a surface to complex multicellular biofilm structures is essential in order to create strategies to control biofilm development. In order to study bacterial migration patterns during Pseudomonas aeruginosa biofilm development, we have performed an investigation with time-lapse confocal laser scanning microscopy of biofilms formed by various combinations of colour-coded P. aeruginosa wild type and motility mutants. We show that mushroom-shaped multicellular structures in P. aeruginosa biofilms can form in a sequential process involving a non-motile bacterial subpopulation and a migrating bacterial subpopulation. The non-motile bacteria form the mushroom stalks by growth in certain foci of the biofilm. The migrating bacteria form the mushroom caps by climbing the stalks and aggregating on the tops in a process which is driven by type-IV pili. These results lead to a new model for biofilm formation by P. aeruginosa.
Investigations of living bacterial biofilms by the use of advanced microscopy have shown that these sessile communities often consist of mushroom-shaped complex multicellular structures separated by water-filled spaces (Lawrence et al., 1991; DeBeer et al., 1994). Because of their innate resistance to host immune systems, antibiotics and other biocides, biofilms in medical and industrial settings are difficult, if not impossible, to eradicate (Costerton et al., 1987, 1999). Biofilm formation therefore leads to various persistent and sometimes lethal infections in humans and animals and to a variety of problems in industry where solid–water interfaces occur. Detailed knowledge of the developmental process from single cells scattered on a surface to complex multicellular biofilm structures is essential in order to create strategies to control biofilm development.
The formation of complex multicellular structures in biofilms has been proposed to be the result of growth under nutrient-transfer limited conditions (Wimpenny and Colasanti, 1997; Picioreanu et al., 1998) or alternatively to be the result of bacterial differentiation and expression of genes that directly control the spatial organization of the organisms (Stoodley et al., 2002). The involvement of cell-to-cell signals in the development of mushroom-shaped structures in Pseudomonas aeruginosa biofilms suggested that the development of complex multicellular biofilm structures has similarities to fruiting body formation by the multicellular myxobacteria (Davies et al., 1998). Myxococcus fruiting body formation occurs through aggregation of bacteria via social gliding motility and requires type-IV pili, a set of chemotaxis related genes and cell-to-cell signalling molecules (Kaiser, 2001). Although cellular migration plays a key role in the formation of complex multicellular structures in Myxococcus (Kaiser, 2001), as well as in eukaryotic morphogenesis and organogenesis (Firtel and Meili, 2000; Locascio and Nieto, 2001; Moscoso, 2002), the role of bacterial migration in the formation of complex multicellular structures in the later phases of biofilm development has not been investigated.
We have previously shown that structural development of P. aeruginosa biofilms is dependent on the carbon source (Klausen et al., 2003). When citrate was used as a carbon source P. aeruginosa PAO1 formed a flat and very dynamic biofilm, whereas when glucose was used as a carbon source P. aeruginosa PAO1 formed a heterogeneous biofilm containing mushroom-shaped multicellular structures separated by water-filled channels. In the previous study (Klausen et al., 2003), we characterized the developmental steps in the formation of the flat citrate-grown P. aeruginosa biofilm using colour-coded bacteria and confocal laser scanning microscopy (CLSM). It appeared that the most important factor for the formation of the flat biofilm was extensive twitching motility exhibited by the cells past the initial phase of biofilm formation. Here we present a detailed characterization of the developmental steps occurring in the formation of the mushroom-shaped multicellular structures in P. aeruginosa biofilms. Our analysis with time-lapse CLSM of biofilms formed by various combinations of colour-coded P. aeruginosa wild-type and P. aeruginosa pilA mutants shows that the mushroom stalks are formed by proliferation of a non-motile subpopulation, and that the mushroom caps are formed by a motile subpopulation which use type-IV pili to climb the stalks and aggregate on the tops.
In order to investigate bacterial migration occurring during development of the heterogeneous P. aeruginosa biofilm, we inoculated a flow-chamber with a 1:1 mixture of Yfp-tagged and Cfp-tagged P. aeruginosa PAO1, irrigated the flow-chamber with glucose minimal medium and followed the structural development in the biofilm by time-lapse CLSM. After 1 day of development the biofilm consisted of microcolonies which contained either cyan fluorescent or yellow fluorescent cells and the surface between the microcolonies was covered with a layer of yellow and cyan fluorescent cells in a mixture (data not shown). This indicated that the initial microcolonies in P. aeruginosa biofilm form by clonal growth of non-motile bacteria and that the bacteria move in the layer between the microcolonies. After 4 days of development, the surface was only partially covered with bacteria and the biofilm consisted of mushroom-shaped structures with either cyan or yellow fluorescent stalks, and with caps composed of various mixtures of cyan and yellow fluorescent bacteria (Fig. 1). These locations of the colour-coded bacteria suggested that the mushroom stalks were formed by clonal growth of non-motile bacteria and that the mushroom caps were formed subsequently by aggregation of motile bacteria on the top of the existing microcolony stalks.
Because P. aeruginosa is able to move on surfaces by means of type-IV pili driven, so-called twitching, motility (Mattick, 2002), we speculated that mushroom formation in P. aeruginosa biofilms might occur in a sequential process of stalk formation by a subpopulation of non-motile bacteria and cap formation by a subpopulation of bacteria which use type-IV-pili to migrate towards, and climb up on, the stalks. In order to investigate a role of type-IV pili in the formation of mushroom structures in P. aeruginosa biofilms we used a mutant, pilA, which is deficient in the synthesis of type-IV pili. A flow-chamber was inoculated with a 1:1 mixture of Yfp-tagged and Cfp-tagged P. aeruginosa pilA bacteria and the structural development was followed in the biofilm irrigated with glucose minimal medium. The P. aeruginosa pilA biofilm did not undergo a stage with complete surface coverage, but consisted of irregularly shaped microcolonies which were either yellow or cyan fluorescent and grew bigger at fixed locations without the formation of caps (Fig. 2). This supported the proposed hypothesis, that the formation of regular mushrooms with caps in wild-type biofilms requires type-IV pili driven motility and that dispersal of the cells on the surface between the microcolonies also occur by twitching motility.
In order to study a role of differential type-IV pili driven motility in biofilm mushroom formation in more detail, we investigated the development in biofilms which were initiated with 1:1 mixtures of Yfp-tagged P. aeruginosa PAO1 wild-type bacteria and Cfp-tagged P. aeruginosa pilA mutants. According to the hypothesis of mushroom formation by non-motile and motile subpopulations, the wild-type population in glucose-grown biofilms should split in a subpopulation of motile bacteria exhibiting twitching motility and another subpopulation of non-motile stalk-forming bacteria, while the pilA bacteria would constitute a second subpopulation of non-motile stalk-forming bacteria. With the used colour coding we would then expect a wild-type/pilA mixed biofilm developing into mushroom structures with yellow fluorescent caps and with stalks which were either cyan fluorescent (the majority) or yellow fluorescent. A CLSM micrograph of a 4-day-old biofilm which was irrigated with glucose minimal medium and had been initiated with a 1:1 mixture of yellow fluorescent PAO1 wild type and cyan fluorescent pilA mutant is shown in Fig. 3. In accordance with the hypothesis of mushroom stalk formation by non-motile bacteria and mushroom cap formation by twitching bacteria, all the mushroom caps were formed by the yellow fluorescent wild-type bacteria, while most of the mushroom stalks were formed by the cyan fluorescent pilA bacteria and a minor fraction of the stalks were formed by the yellow fluorescent wild-type bacteria.
We have previously shown that when citrate is used as the carbon source P. aeruginosa wild-type bacteria exhibit extensive twitching motility and form a flat biofilm without mushroom structures (Klausen et al., 2003). It was of interest to investigate if the presence of a stalk-forming P. aeruginosa pilA population would lead to mushroom formation in a citrate-grown wild-type/pilA mixed biofilm. If so, the expected result should be even clearer than in the case with the glucose-grown wild-type/pilA mixed biofilm, as with citrate as the carbon source the motile population should be all the wild-type bacteria, while the non-motile population should be the pilA mutants only. In a citrate-grown biofilm consisting of Yfp-tagged P. aeruginosa wild type and Cfp-tagged P. aeruginosa pilA mutants all the mushrooms should therefore have cyan fluorescent stalks and yellow fluorescent caps. To test the hypothesis, we followed the structural development in a biofilm which was irrigated with citrate minimal medium and had been initiated with a 1:1 mixture of yellow fluorescent PAO1 wild type and cyan fluorescent pilA mutants. The yellow fluorescent PAO1 wild type initially formed a flat and expanding bacterial lawn, while the cyan fluorescent pilA mutants formed microcolonies at fixed positions (Fig. 4). In the later stages of biofilm development, mushroom-shaped structures formed, with the cyan fluorescent pilA mutant constituting the stalks and the yellow fluorescent PAO1 wild type constituting the caps (Fig. 5). The experiments thus showed that formation of mushroom-shaped structures may occur in citrate-grown P. aeruginosa biofilms if a stalk-forming subpopulation is present and suggested that the difference in structure between the citrate-grown and the glucose-grown P. aeruginosa wild-type biofilms is due to different capabilities to form the mushroom stalks.
Vertical sections of a time-lapse CLSM image series captured in a biofilm consisting of a 1:1 mixture of Yfp-tagged P. aeruginosa wild-type bacteria and Cfp-tagged P. aeruginosa pilA mutants showed directly that the yellow fluorescent wild-type cells climbed the existing cyan fluorescent pilA microcolonies (Fig. 6). The increase in biomass in the mushroom caps may, in addition to type-IV pili driven aggregation of the yellow fluorescent wild-type bacteria, to some extent be attributed to growth. It should be noted, however, that the cyan fluorescent pilA mutants did not grow significantly in the time period.
In order to show that the results obtained with the P. aeruginosa pilA mutant were not due to a secondary mutation in our strain, we repeated the experiments with the wild-type/pilA mixed biofilms using the strain P. aeruginosa PAKΔpilA (Kagami et al., 1998) and got the same results (data not shown).
The present work leads to the model for development of the heterogeneous P. aeruginosa biofilm which is shown in Fig. 7. The model suggests that the formation of mushroom-shaped structures in P. aeruginosa wild-type biofilms occurs through stalk formation by proliferation of bacteria which have downregulated twitching motility and cap formation by bacteria which climb the microcolony stalks by the use of type-IV pili and aggregate on the top. According to the model, type-IV pili driven bacterial migration plays a key role in structure formation in the late phase of biofilm development. A role for type-IV pili in the initial phase of P. aeruginosa biofilm development has been proposed in previous studies. O’Toole and Kolter (1998) studied the initial phase of P. aeruginosa biofilm formation in a static system with glucose medium and based on the observations (i) a pil mutant formed only a monolayer of scattered cells whereas the wild-type formed small microcolonies and (ii) the wild-type bacteria moved on the surface so that small cell clusters formed and dispersed, it was proposed that the initial microcolonies in P. aeruginosa biofilms form by aggregation of bacteria mediated by type-IV pili. Because fruiting body formation by Myxobacteria also occurs via type-IV pili driven aggregation, a parallel between P. aeruginosa microcolony formation and Myxococcus fruiting body formation was proposed in a review by O’Toole et al. (2000). Recently, however, it was shown that P. aeruginosa type-IV pili mutants are able to form protruding microcolonies in citrate-grown biofilms (Heydorn et al., 2002; Klausen et al., 2003) and in Tryptic Soy Broth-grown biofilms (Singh et al., 2002) and evidence was presented that the initial microcolonies in citrate-grown and Tryptic Soy Broth-grown P. aeruginosa wild-type biofilms do not form by cell aggregation but by clonal growth (Singh et al., 2002; Klausen et al., 2003). The present study with time-lapse CLSM of biofilm formation by colour-coded bacteria suggests that the initial microcolonies (mushroom stalks) in glucose-grown wild-type P. aeruginosa biofilms form by clonal growth and that P. aeruginosa pilA mutants with glucose as a carbon source are able to form biofilms consisting of irregularly shaped protruding microcolonies without mushroom caps.
Stalk formation may initiate in certain foci of a P. aeruginosa biofilm as a consequence of twitching motility suppression in response to local environmental cues. In agreement with a requirement of downregulation of twitching motility for initial microcolony formation, it was recently reported that a component of innate immunity inhibits the formation of large microcolonial structures in P. aeruginosa biofilms by preventing downregulation of twitching motility (Singh et al., 2002). However, it is also possible that stalk formation initiates because the cells in certain foci of the biofilms adhere strongly to each other creating an intercellular matrix, so that twitching motility becomes arrested. Because the ability to sense local population sizes by means of cell-to-cell signalling evidently has a role in the formation of P. aeruginosa biofilm mushroom structures (Davies et al., 1998), the possibility exists that the expression of cell-to-cell adherence factors is regulated by cell-to-cell communication. Because stalk formation is prevented by type-IV pili driven migration in P. aeruginosa wild-type biofilms grown on citrate minimal medium (Klausen et al., 2003), the process appears to be dependent on nutritional conditions. In this connection it is interesting that Vfr, the P. aeruginosa homologue of the Escherichia coli catabolite repressor protein (Crp), is involved both in the regulation of quorum sensing (Albus et al., 1997) and in the regulation of twitching motility (Beatson et al., 2002) and it has moreover been shown that downregulation of twitching motility is prevented by iron limitation (Singh et al., 2002).
Cap formation through climbing of P. aeruginosa bacteria on top of microcolonial stalks by means of type-IV pili is consistent with the current belief that these pili mediate cellular motility by an extension-grip-retraction mechanism (Skerker and Berg, 2001), although type-IV pili in P. aeruginosa were until now only known to be involved in bacterial translocation over solid surfaces (Mattick, 2002). Because chemotaxis-related genes are involved in twitching motility by P. aeruginosa (Darzins, 1994), and local consumption creates nutrient gradients in biofilms (DeBeer et al., 1994; Wimpenny and Colasanti, 1997; Picioreanu et al., 1998), it is possible that the bacteria climb on top of the existing microcolonies because more nutrients are available on the top. Twitching motility is not regulated by any of the known cell-to-cell signalling systems in P. aeruginosa (Beatson et al., 2002), indicating that migration of the cells on top of the stalks is not coordinated by a known cell-to-cell communication system. Because a role of quorum sensing in formation of the heterogeneous P. aeruginosa biofilm structures has been shown previously (Davies et al., 1998), there may, however, be parallel quorum sensing controlled pathways required for mushroom cap formation. It was recently shown that rhamnolipid surfactant production affects P. aeruginosa biofilm architecture and that a rhlA mutant did not maintain open channels in biofilms (Davey et al., 2003). Because rhamnolipid synthesis is quorum sensing regulated (Pearson et al., 1997), and surfactants may facilitate twitching motility, it is likely that quorum sensing plays a role in mushroom cap formation through these processes. Quorum sensing could also be involved in a process that makes the twitching bacteria settle on the top of the mushrooms. Because the presence of the stalk-forming P. aeruginosa pilA bacteria also resulted in mushroom formation in a citrate-grown wild-type/pilA mixed biofilm, where the wild-type bacteria exhibit extensive twitching motility and unaccompanied form a flat biofilm, it is likely that formation of some kind of cell-to-cell glue causes the twitching bacteria to settle on top of the mushrooms. A role of quorum sensing in the formation of mushroom structures is consistent with our previous finding that a P. aeruginosa wild type and lasI mutant formed similar flat biofilms in citrate medium (Heydorn et al. 2002), as citrate medium, unlike glucose medium, does not support mushroom formation by the P. aeruginosa wild type in the flow-chamber biofilms (Klausen et al., 2003).
Mushroom formation by P. aeruginosa during biofilm growth has similarities to fruiting body formation by Myxococcus which occur during the sporulation process through aggregation of bacteria via social gliding motility (Kaiser, 2001). Twitching motility and social gliding motility are essentially the same process (Semmler et al., 1999) and it appears that Myxococcus and P. aeruginosa by and large have recruited the same genes for the formation of fruiting bodies and biofilm mushrooms. Both examples demonstrate how development of complex morphogenetic structures may be directed by regulated cellular motility and cellular responses to the nutrient conditions. P. aeruginosa mushroom development in biofilms is evidently directed by two independent activities involving a first stage of stalk development influenced by the nutritional conditions and a second stage of cap formation dependent on twitching motility and possibly influenced by nutrient gradients and quorum sensing. In case of the true multicellular Myxococcus, the two activities seem to be under joint control resulting in a coordinate developmental programme for fruiting body formation. Multicellular fruiting body formation and sporulation confer several selective advantages to myxobacteria, one being that the fruiting body package may adhere to a small animal and be transported to a new place where nutrients are not scarce (Kaiser, 2001). The formation in biofilms of mushroom structures separated by water-filled spaces has been proposed to be a multicellular process, which creates a circulation system that ensures efficient nutrient supply to most cells and efficient removal of waste products (Costerton et al., 1995). It is intriguing that also the highly regulated events involved in eukaryotic organogenesis and morphogenesis are based on the interplay between cell proliferation, cell migration and the generation of an intercellular matrix (Firtel and Meili, 2000; Locascio and Nieto, 2001; Moscoso, 2002). The involvement of these factors in the development of complex multicellular structures in biofilms as in organogenesis and morphogenesis in multicellular organisms, suggests that they are common denominators to the formation of complex multicellular structures in nature.
Bacterial strains and growth conditions
The P. aeruginosa PAO1 (Holloway and Morgan, 1986) strain from John Mattick's laboratory and a ΔpilA derivative constructed by allelic displacement (Klausen et al., 2003) were used in the study. The P. aeruginosa PAK derivative PAKΔpilA (Kagami et al., 1998) was used in control experiments. The strains were fluorescently tagged at an intergenic neutral chromosomal locus with cfp or yfp in mini-Tn7 constructs (Klausen et al., 2003). Modified FAB medium (Heydorn et al., 2000) was used supplemented with 30 mM glucose or 10 mM citrate for batch overnight cultures and with 0.3 mM glucose or 0.1 mM citrate for biofilm cultivation. Biofilms and batch cultures were grown at 30°C.
Cultivation of biofilms
Biofilms were grown in flow chambers with individual channel dimensions of 1 × 4 × 40 mm. The flow system was assembled and prepared as described previously (Møller et al., 1998). The flow-chambers were inoculated by injecting 350 µl overnight culture diluted to an OD600 of 0.001 into each flow-channel with a small syringe. After inoculation flow-channels were left without flow for 1 h, after which medium flow (0.2 mm s−1) was started using a Watson Marlow 205S peristaltic pump.
Microscopy and image acquisition
All microscopic observations and image acquisitions were done with a Zeiss LSM 510 CLSM (Carl Zeiss, Jena, Germany) equipped with detectors and filter sets for monitoring of Cfp and Yfp fluorescence. Images were obtained using a 63×/1.4 objective or a 40×/1.3 objective. Simulated 3-D images and sections were generated using the IMARIS software package (Bitplane AG).
This work was supported by the Danish Biotechnological Research Program.