Volume 13, Issue 7 p. 1666-1681
Free Access

Key two-component regulatory systems that control biofilm formation in Pseudomonas aeruginosa

Helga Mikkelsen

Helga Mikkelsen

Imperial College London, Division of Cell and Molecular Biology, Centre for Molecular Microbiology and Infection, South Kensington Campus, Flowers Building, London SW7 2AZ, UK.

Search for more papers by this author
Melissa Sivaneson

Melissa Sivaneson

Imperial College London, Division of Cell and Molecular Biology, Centre for Molecular Microbiology and Infection, South Kensington Campus, Flowers Building, London SW7 2AZ, UK.

Search for more papers by this author
Alain Filloux

Corresponding Author

Alain Filloux

E-mail [email protected]; Tel. (+44) 20 75949651; Fax (+44) 20 75943069.Search for more papers by this author
First published: 09 May 2011
Citations: 170


Biofilm formation in P. aeruginosa is a highly regulated process that proceeds through a number of distinct stages. This development is controlled by a wide range of factors, of which two-component systems (TCSs) play a key role. In this review, we focus on some of the TCSs that regulate the switch from a motile to a sessile bacterial lifestyle, either via the production of extracellular appendages or by the production of exopolysaccharides. Extracellular appendages, such as flagella, type IV pili and Cup fimbriae are often involved in the initial attachment of bacteria to a surface. In P. aeruginosa, many of these surface structures are regulated by TCSs, and some systems regulate more than one type of appendage. Furthermore, the production of exopolysaccharides, such as Pel and Psl, is required for P. aeruginosa biofilm formation. The regulation of Pel and Psl is post-transcriptionally repressed by RsmA, the activity of which is controlled by a complex regulatory system involving several sensor kinases and accessory components. Furthermore, the Rsm system is a major control system that inversely regulates factors involved in motility and acute infection on one hand, and factors involved in biofilm formation and chronic infection on the other hand. Finally, a series of TCSs has recently been discovered that regulates biofilm development in a stage-specific manner. Taken together, these complex regulatory networks allow the bacterium to respond appropriately to diverse environmental stimuli, and increased knowledge of their mechanisms and signals could be of great importance in the design of novel antibacterial strategies.


Pseudomonas aeruginosa is a common inhabitant of soil and other natural environments. It is characterized by an immense ecological diversity and occupies perhaps the broadest range of ecological niches of any bacterial species (Kulasekara and Lory, 2004). It is also an opportunistic pathogen with a broad host range that spans from insects and plants to animals including humans. Accordingly, the P. aeruginosa genome has a high proportion of its coding capacity dedicated to regulation with 9.3% of the open reading frames encoding regulatory proteins, including two-component regulatory systems (Stover et al., 2000).

In nature, microorganisms are frequently found in surface-attached multicellular communities called biofilms. These sessile microbes are encased in a self-produced polymeric matrix, and this mode of life provides a high level of protection against environmental assaults and predation (Donlan and Costerton, 2002). In clinical settings, bacterial biofilms are increasingly recognized as a determinant for disease, and the high level of resistance to antimicrobials makes them extremely difficult to eradicate (Parsek and Singh, 2003). Pseudomonas aeruginosa is a model organism for biofilm formation (Tolker-Nielsen and Molin, 2004), which in vitro is a highly regulated process that proceeds through a number of distinct stages (Sauer et al., 2002). These stages can be divided into initial attachment, microcolony formation, maturation and finally dispersion, which releases planktonic cells back into the environment (Sauer et al., 2002; Stoodley et al., 2002).

Pseudomonas aeruginosa infections can be broadly classified into acute and chronic infections. Acute infections are often characterized by fast growth, motility, cytotoxicity and rapid progression of disease. Conversely, chronic infections are believed to be associated with slower growth, biofilm formation, antibiotic resistance and persistence (Furukawa et al., 2006). Examples of P.aeruginosa infections that have been associated with biofilm formation are colonization of urinary catheters and ventilators, infected burn wounds and the chronic pneumonia in cystic fibrosis (CF) patients (Bjarnsholt and Givskov, 2007).

The lifestyle change associated with the transition to a sessile lifestyle is reflected in diverse physiological changes, and a large number of factors have been shown to contribute to the different stages of biofilm formation. Some of the major players are extracellular appendages (e.g. flagella and pili), which are often associated with the initial attachment to a surface (O'Toole and Kolter, 1998; Vallet et al., 2001; Klausen et al., 2003; Shrout et al., 2006), and polysaccharides, such as Pel and Psl (Friedman and Kolter, 2004a; Vasseur et al., 2005), which play an important role in maintaining the integrity of the community (Branda et al., 2005). However, perhaps more important than individual components that contribute to the biofilm structure are the regulatory systems that control their production in response to environmental stimuli and eventually cause the bacterium to switch from one lifestyle to another. In P. aeruginosa, many of these regulators are two-component systems.

Two-component systems (TCSs) typically consist of a sensor kinase and a response regulator, and these systems are the predominant signalling mechanism in most bacteria. The sensor kinases are often membrane bound, but they can also be cytosolic. They enable bacteria to constantly monitor environmental conditions, such as nutrients, temperature, pH, osmolarity or the presence of toxic substances. The detection of environmental stimuli then elicits an appropriate response that ensures successful survival of the bacterium in a given niche. The prototypical sensor kinase consists of a variable N-terminal input domain, frequently located in the periplasm in the case of membrane bound sensors, and a conserved C-terminal transmitter domain located in the cytosol (Fig. 1) (Rodrigue et al., 2000). Response regulators have a conserved N-terminal receiver domain, which is most often followed by a C-terminal output domain with a helix–turn–helix (HTH) motif that is able to bind DNA (Stock et al., 2000). Upon detection of a signal, the sensor kinase autophosphorylates on a conserved histidine residue in the transmitter domain (Gao and Stock, 2009). The phosphoryl group is then transferred onto a conserved aspartate residue in the receiver domain of the response regulator, thereby activating it as a transcription factor.

Details are in the caption following the image

Domain organization of two-component regulatory systems in P. aeruginosa. Histidine sensor kinases identified upon P. aeruginosa genome mining can be divided into three groups: classical, hybrid and unorthodox (Rodrigue et al., 2000). A classical sensor (42 on the PAO1 genome) consists of an N-terminal input domain (brown), followed by a transmitter domain (light blue) with a conserved histidine (H) that can be auto-phosphorylated. The phosphoryl group (P) can then be transferred onto a conserved aspartate residue (D) in the receiver domain (purple) of the response regulator. This activates the output domain (green), which is commonly a DNA binding domain. Hybrid and unorthodox (12 and 5 on the PAO1 genome respectively) sensors contain an additional C-terminal receiver domain and therefore require an external phosphotransfer (Hpt) module (dark blue) in order to phosphorylate the response regulator. In the unorthodox sensor, the Hpt module is an integral part of the protein, while hybrid sensors require an independent Hpt protein in order to phosphorylate the response regulator (Rodrigue et al., 2000). The TCS classification presented here is not comprehensive and does not take into account alternative phosphorelay structures. For example, the TodS sensor in Pseudomonas putida uses a second transmitter domain as phosphorelay (Busch et al., 2009), and not a Hpt domain as shown in this figure.

The genome of P. aeruginosa PAO1 encodes 127 members of TCSs (Rodrigue et al., 2000). While the majority of these systems consist of a classical sensor kinase and a DNA-binding response regulator, several response regulators have been shown to have a variable C-terminal output domain that can carry out diverse enzymatic functions (Grebe and Stock, 1999). Furthermore, the sensor kinases can be divided into three groups termed (i) classical, (ii) unorthodox and (iii) hybrid sensors (Fig. 1). In the latter two groups, an additional receiver domain with a conserved aspartate is fused to the C-terminus of the transmitter domain. Hence, to ensure transfer of the phosphoryl group to the response regulator, an intermediate histidine phosphotransfer module (Hpt) is required. In unorthodox sensors, this Hpt module is an integral part of the protein, whereas hybrid sensors require an independent Hpt protein to transmit their signal (Gao and Stock, 2009). Three such Hpt modules are encoded on the genome of P. aeruginosa PAO1, called HptA, HptB and HptC (PA0991, PA3345 and PA0033 respectively). The unorthodox and hybrid sensors thus theoretically pass the signal to the response regulator via a multistep phosphorelay, which probably allows more modulatory inputs to fine-tune the regulation (Rodrigue et al., 2000).

The aim of this review is not to give a comprehensive overview of factors involved in biofilm formation or their regulation, as this has been done elsewhere (Karatan and Watnick, 2009). Instead, we will focus on a few key TCSs that control determinants that play a role at various stages of biofilm development. We will specifically focus on the TCSs that control assembly of extracellular appendages, such as Cup fimbriae and type IV pili (Fig. 2), and the production of exopolysaccharides (EPS), such as Pel and Psl. These determinants are often required at different stages of biofilm development. In a simplified model, we can consider cell surface appendages to be involved in the early stages, including during initial surface attachment, whereas EPS are involved at later stages to mature the biofilm and shape its structure. We will also describe TCSs that regulate the progression of biofilm development by sequentially modifying the transcriptional profile of the bacterial community. An improved understanding of these signalling systems could lead to the discovery of important drug targets to combat P. aeruginosa, as well as other bacterial pathogens (Matsushita and Janda, 2002; Watanabe et al., 2008).

Details are in the caption following the image

Key two-component regulatory systems that control the expression of extracellular appendages in Pseudomonas aeruginosa. Sensors are shown in blue, response regulators predicted to be DNA-binding are shown in green, and EAL-domain response regulators are shown in red. An unknown regulator proposed to be required for cupB gene induction is shown in dark brown. Arrows indicate a positive influence on protein activity or gene expression, whereas red lines indicate a repressing effect.

Cup fimbriae in P. aeruginosa

The first stage of biofilm development involves initial contact, which is followed by irreversible attachment of bacteria to a surface. This process can involve a multitude of different adhesins and extracellular appendages, the requirement of which is likely to depend on the surface and the environmental conditions. In P. aeruginosa, the flagellum is recognized as a central component in the biofilm formation process, since it provides the mobility needed to actively approach a surface (O'Toole and Kolter, 1998). Type IV pili that we will discuss in another paragraph are also central factors in surface attachment and colonization (O'Toole and Kolter, 1998). Furthermore, in 2001 we discovered a novel set of extracellular appendages in P. aeruginosa, which are assembled by the chaperone-usher pathway, and therefore were named Cup fimbriae (Vallet et al., 2001). These fimbrial gene clusters are conserved in Gram-negative bacteria, including many pathogens, and are frequently found in several copies on the same genome. They have been particularly well studied in uropathogenic Escherichia coli, where they are known as Fim or Pap fimbriae (Hung and Hultgren, 1998; Sauer et al., 2004). These fimbriae can serve a number of functions, such as promote attachment to inert surfaces, mediate adherence to host tissues through specific adhesins, or facilitate evasion of host defences, e.g. by shielding antigens that can be recognized by the immune system (Hasman et al. 1999; Sauer et al., 2004). There are three basic components in a chaperone-usher pathway: A major fimbrial subunit, a chaperone and an usher. The fimbrial subunits are synthesized in the cytoplasm and exported to the periplasm via the Sec system, where they are bound to a chaperone that ensures correct folding and protects from degradation. They are then transported to the usher, which forms a pore in the outer membrane through which the pilus fibre is assembled and exposed at the cell surface. In P. aeruginosa, five fimbrial (cup) gene clusters have been identified to date, termed cupA-E (Filloux et al., 2004). These clusters vary in composition as well as regulation, and they are therefore likely to serve specific functions. Whereas the cupC gene cluster only encodes the three basic components (pilin, chaperone and usher), the other clusters are predicted to encode additional fimbrial subunits or adhesins, and some also encode a second chaperone. Although the specific functions of these fimbriae are unknown, they have all been shown to contribute to biofilm formation or to alter the adhesive properties of the strain that expresses them. The expression of cup gene clusters is tightly controlled, and most of them are not expressed during growth in liquid culture. The cupA gene cluster is controlled by regulatory proteins encoded by the cgrABC genes (Vallet-Gely et al., 2007), the global regulator of anaerobic gene expression, ANR (Vallet-Gely et al., 2007), and the H-NS like transcriptional repressor MvaT (Vallet et al., 2004). However, all the remaining cup clusters in P. aeruginosa are controlled by TCSs.

The Roc systems promote cup gene expression

The first TCS controlling cup genes to be described in P. aeruginosa was the Roc1 system (regulator of cup). This locus was discovered in strain PAK by Kulasekara and colleagues in a transposon screen for mutants with active cupB or cupC gene clusters (Kulasekara et al., 2005). These two cup clusters were found to be co-regulated by the Roc1 locus, which consists of three genes encoding one sensor kinase and two response regulators. RocS1 is a membrane bound unorthodox sensor. It contains several sensory domains, including periplasmic domains similar to those found in solute binding proteins (Sbp3) and a cytosolic PAS domain (Kulasekara et al., 2005). RocA1 is a conventional response regulator with a HTH output domain, and RocR has an output domain with an EAL motif, the phosphodiesterase activity of which degrades the second messenger c-di-GMP (Rao et al., 2008). The same gene cluster was also identified by Kuchma and colleagues in strain PA14 as a regulator of biofilm maturation (named sadARS in this study) (Kuchma et al., 2005). The RocS1A1R system is homologous to the BvgSAR system, which regulates important virulence factors in Bordetella species. The Roc system has also been implicated in virulence, since a rocS1 mutant is attenuated in a Caenorhabditis elegans infection model (Gallagher and Manoil, 2001; Kulasekara et al., 2005).

Using genetic and biochemical methods, Kulasekara and colleagues showed that the RocS1 sensor signals through both RocA1 and RocR, which have opposite effects on cupC gene expression. On one hand, expression of rocA1 activates cupC gene expression, which leads to production of fimbriae and thereby increased attachment. On the other hand, expression of rocR reduces cupC gene expression by a currently unknown mechanism that is likely to involve degradation of c-di-GMP (Kulasekara et al., 2005). The mechanism by which the RocS1 sensor balances the activity of the two response regulators is currently unknown, but it could be controlled by their relative abundance, relative affinity for the sensor or by the relative efficiency of the RocS1 kinase or phosphatase activity.

In their transposon screen for cupB and cupC regulators, Kulasekara and colleagues also identified the Roc2 locus, which encodes paralogues of the Roc1 system. The Roc2 system consists of the unorthodox sensor kinase, RocS2, and the conventional response regulator, RocA2. In contrast to RocS1, the membrane bound RocS2 sensor has no recognizable sensory domain in the periplasm, although it contains the cytosolic PAS domain (Galperin and Nikolskaya, 2007). Despite having highly conserved components, TCSs are usually well insulated against cross-talk or cross-regulation (Laub and Goulian, 2007). Nevertheless, in a recent report, we showed that extensive cross-regulation occurs between the Roc1 and Roc2 systems (Sivaneson et al., 2011). Like RocS1, the RocS2 sensor kinase is also able to induce cupC gene expression. However, the induction by RocS2 is independent of the RocA2 response regulator, but instead requires RocA1. RocS1 and RocS2 can therefore both signal through the same response regulator to activate cupC gene expression. Additionally, both sensors can also induce cupB gene expression, and this induction requires neither RocA1, nor RocA2, but is likely to proceed via a currently unknown component (Fig. 2).

Bacterial two-hybrid analysis further showed that both RocS sensors also interact with RocR and RocA2. In the former case, the signalling results in repression of the cupB and cupC gene clusters, but no targets were known for the latter. Microarray analysis revealed that the primary action of RocA2 is repression of the mexAB–oprM gene cluster (Sivaneson et al., 2011). These genes encode a multidrug efflux pump that provides resistance to a range of different antibiotics including β-lactams. The repression of mexAB–oprM by RocA2 was shown to translate into an increased sensitivity to several antibiotics, and this sensitivity was reversed in a rocA2 deletion mutant.

As previously mentioned, one of the hallmarks of sessile bacteria is that they are many times more resistant to antibiotics than their isogenic planktonic counterparts (Parsek and Singh, 2003). It may therefore seem counterintuitive to have a signalling cascade that promotes biofilm formation via Cup fimbriae and at the same time increases sensitivity to antibiotics by repressing a multidrug efflux pump. However, the mechanism of antibiotic resistance in biofilms is still largely unknown, and multidrug efflux pumps do not seem to play a significant role in this process (Brooun et al., 2000; De Kievit et al., 2001). Furthermore, it has also been shown that many strains isolated from CF patients are hypersensitive to β-lactam antibiotics due to an inactivation of MexAB–OprM (Vettoretti et al., 2009). It is therefore conceivable that this pump carries a fitness cost in a biofilm environment and is therefore repressed or selected against during biofilm formation and chronic infection.

The Rcs/Pvr system controls cupD gene expression in PA14

Whereas the cupA, cupB, cupC and cupE gene clusters are present in all four sequenced strains of P. aeruginosa, the cupD gene cluster is only present in PA14 and PA7. In PA14, the five cupD genes are located on the PAPI-1 pathogenicity island (He et al., 2004), and adjacent to them are four genes encoding members of TCSs. These nine genes are flanked by direct repeats, suggesting that they have been acquired as one unit by horizontal gene transfer. Furthermore, mutants in several of these components have been shown to be attenuated in plant and animal infection models (He et al., 2004). The TCSs consist of two sensors and two response regulators. RcsC is an unorthodox sensor, and PvrS is a hybrid sensor; they are highly similar to each other, and also have strong homology to the Roc sensors. Both RcsC and PvrS are membrane bound sensors, which have no characteristic periplasmic sensory domain, whereas only RcsC has a recognizable cytosolic PAS domain (Mikkelsen et al., 2009). The resemblance to the Roc systems continues with the two response regulators, where RcsB has a HTH output domain (like RocA1 and RocA2), while the output domain of PvrR has an EAL motif (like RocR). PvrR was first identified as a phenotype variant regulator that, upon overexpression, could reverse antibiotic resistant small colony variants back to the wild-type phenotype (Drenkard and Ausubel, 2002). The specific mechanism of this has not been elucidated, but it is probably linked to the c-di-GMP phosphodiesterase activity, which has been experimentally demonstrated (Kulasakara et al., 2006).

Similar to the regulation of cupC by the Roc system, the Rcs/Pvr system has been shown to control the expression of the cupD gene cluster in that RcsB activates cupD gene expression, whereas PvrR has a repressing effect (Mikkelsen et al., 2009; Nicastro et al., 2009). Furthermore, production of CupD fimbriae leads to increased attachment, formation of small colony variants and reduced swimming and twitching motility. Conversely, overproduction of PvrR leads to loss of attachment and increased motility (Mikkelsen et al., 2009). The observation that the Rcs/Pvr system antagonistically controls biofilm formation, through cupD gene expression, and motility is one of many examples of regulatory switches that control the planktonic/sessile lifestyle choice or acute/chronic infection, as we will see later with the RetS/GacS/LadS network.

The P. aeruginosa type IV pili

Long before Cup fimbriae were discovered in P. aeruginosa, type IV pili were known as major extracellular appendages involved in attachment and motility of Pseudomonas species (de Groot et al., 1994; Mattick, 2002). Type IV pili are located at the cell pole and are involved in twitching motility, which is a unique type of movement across semi-solid surfaces. The movement results from the extension, tethering, and retraction of the pilus structure. Because type IV pili allow the bacteria to move onto biotic or abiotic supports, they have been proposed to be important for colonization of surfaces and formation of microcolonies. After surface colonization, clonal growth leads to the formation of microcolonies that can develop into mushroom-like pillars and give the mature biofilm a unique structure. Type IV pili also contribute to the formation of the mushroom-like structure, since type IV pili-dependent motility is required for the climbing of motile bacteria onto the stalk to form the mushroom cap (Klausen et al., 2003).

Formation of type IV pili involves the multimerization of a pilin subunit, PilA in P. aeruginosa. The pilin is synthesized as a precursor with a short N-terminal leader peptide (6–7 amino acids), which is cleaved by the PilD/XcpA prepilin peptidase prior to assembly (Nunn and Lory, 1991; Bally et al., 1992). The P. aeruginosa PilA is a type IVa pilin, while type IVb pilins, such as BfpA in enteropathogenic E. coli (EPEC), have significantly longer leader peptides (15–20 amino acids) (Donnenberg et al., 1997). Whereas the P. aeruginosa type IVa pili are involved in motility, the type IVb or bundle forming pili from EPEC are essential for adhesion. A subclass of type IVb pili, also known as Flp pili, is encoded on the P. aeruginosa genome. The leader sequence of the P. aeruginosa Flp pilin is characteristic of type IVb pilins but displays a tyrosine at position +6, which is the hallmark for Flp pilin subunits (Kachlany et al., 2001). The Flp pili are not involved in motility, but these long and bundled pili contribute to biofilm formation and attachment to respiratory epithelial cells (de Bentzmann et al., 2006). Finally, The PAPI-1 genomic island of P. aeruginosa PA14 carries a 10 gene cluster (pilM2-T2 and pilV2) that encodes a type IVb pilus. However, rather than contributing to motility and attachment, this pilus has been shown to be involved in bacterial conjugation and transfer of the PAPI-1 island between bacteria (Carter et al., 2010; Filloux, 2010).

The PilRS system controls spatial assembly of type IVa pili

In contrast to Cup fimbriae, type IV pili genes are expressed in laboratory conditions. However, the regulatory mechanisms involved in pil gene expression are highly complex. Expression of the pilA gene depends on the sigma factor RpoN, as well as on an additional transcriptional regulator, PilR (Hobbs et al., 1993). PilR is a TCS response regulator, and PilS is its cognate sensor (Fig. 2). In addition to this TCS, a chemotactic system, PilGHIJKL and ChpABC (Whitchurch et al., 2004) is likely to determine the direction or rate control of twitching motility in response to local environmental changes. Although PilRS appears to exert direct control on pil gene expression, other TCSs are also involved in modulating twitching motility, namely the response regulator AlgR and the sensor FimS (Whitchurch et al., 1996). Interestingly, the FimS sensor does not contain the highly conserved residues for nucleotide binding characteristic in autokinase catalytic domain, and is thus unlikely to phosphorylate AlgR. This further suggests that FimS could act by dephosphorylating AlgR.

The PilS sensor displays a number of notable features. It is an inner membrane protein that contains three distinct domains. The N-terminal domain is highly hydrophobic, and contains six trans-membrane helices that anchor the protein to the membrane. The central domain is cytoplasmic and is predicted to contain sensory PAS domains (Galperin and Nikolskaya, 2007), and the C-terminal region corresponds to the typical transmitter domain in classical sensor kinases (Fig. 1). Furthermore, as well as being inserted into the cytoplasmic membrane via its hydrophobic region, PilS has also been shown to be spatially localized. Using PilS-GFP fusions, the protein was shown to be targeted to the cell poles (Boyd, 2000). This localization was only observed in P. aeruginosa, while in E. coli the PilS-GFP chimera diffused laterally across the membrane. Although the hydrophobic region anchors the protein to the membrane, it is not required for its polar localization. Replacement of this region by the trans-membrane domain of the E. coli MalG protein does not change the polarity, which is only affected if the PAS-containing central region is missing. It has thus been suggested that the PAS-containing region senses or interacts with specific components located at the cell pole in P. aeruginosa. It is tempting to hypothesize that the sensing mechanism could be coupled to the presence of pili exclusively at the pole, but no further studies on PilS have confirmed this.

As previously mentioned, several proteins other than TCSs are required for the control of twitching motility, and one of these is FimX (Kazmierczak et al., 2006). Mutants in the fimX gene display very low levels of pili on their surface (Huang et al., 2003). FimX has an N-terminal receiver domain typical of response regulators, but it lacks the conserved aspartate residue, suggesting it is unlikely to be activated by a cognate sensor kinase (Huang et al., 2003). Furthermore, the receiver domain is coupled to a PAS domain, as well as an EAL domain. As for PvrR and RocR, FimX displays phosphodiesterase activity and degrades c-di-GMP. Importantly, like PilS, FimX is localized to the cell pole, but in this case it is unipolar and not bipolar. This is reminiscent of the situation in Caulobacter crescentus, where progression of the cell cycle requires proper localization of stalk and flagellum at opposite cell poles. In this case, the dynamic polar localization of the sensor kinases PleC and DivJ, and of the EAL-domain response regulator PleD, has been shown to be crucial for polar differentiation (Viollier et al., 2002; Paul et al., 2004).

The PprAB system controls both type IVb pili (Flp) and Cup fimbriae

As for the CupA, CupB, CupC and CupD fimbriae, the Flp pili are barely produced in laboratory conditions, although the production of Flp pilin can be observed in late stationary phase (Bernard et al., 2009). The flp gene is part of a large gene cluster that also contains the tad–rcp genes involved in the assembly of Flp subunits into a pilus structure. The TCS PprA-PprB is also part of this cluster, which suggests that these components are involved in the control of flp–tad–rcp gene expression. PprB is a NarL-type response regulator with a typical HTH DNA binding domain and PprA is a sensor kinase containing PAS and GAF sensory domains in addition to its transmitter domain. The flp–tad–rcp cluster is organized in five transcriptional units, all of which are under PprB control. Band shift assays also revealed that PprB binds directly to the DNA regions upstream of the flp–rcp, tadF–fppA, and pprB genes (Bernard et al., 2009).

Although the pprAB genes are part of the flp–tad–rcp gene cluster, it is now clear that they have a more global impact on regulation of gene expression. In particular, PprB induces expression of the cupE gene cluster, which is involved in the assembly of CupE fimbriae. The CupE fimbriae are produced by the multimerization of CupE1 subunits, and as for other Cup fimbriae, the assembly involves a chaperone (CupE4) and an usher protein (CupE5) (Giraud et al., 2011). In addition to CupE1, minor subunits are likely to be part of the fimbrial structure, namely CupE2, CupE3 and the putative adhesin CupE6. The pilin domain of all pilin and adhesin subunits of the CupE system is similar to the previously described fimbrial subunits encoded in the Acintetobacter baumannii csu locus (Tomaras et al., 2003). These pilin subunits are members of the σ-fimbrial clade and not of the γ4-fimbrial clade to which the other P. aeruginosa Cup systems belong (Nuccio and Baumler, 2007).

In the same way as CupB and CupC (Ruer et al., 2007), the CupE fimbriae have been proposed to be involved in cell–cell interaction, as well as in adhesion to inert surfaces. Furthermore, Giraud and collaborators showed that CupE fimbriae contribute to biofilm structure and the shaping of mushroom-like structures (Giraud et al., 2011). CupE fimbriae are produced when bacteria are grown in static condition or on solid medium, but they are not observed in liquid shaking cultures. These observations suggest that the cupE genes are induced during biofilm-like growth, and that the PprB response regulator is involved in this control. Overproduction of PprB readily induces cupE gene expression, and this occurs by direct binding of PprB on the cupE1 gene promoter (Giraud et al., 2011). The cupE gene cluster is therefore the second direct target identified for PprB (after the flp–tad–rcp genes), which indicates that the PprAB TCS controls a variety of extracellular appendages and is central to biofilm formation and maturation.

The P. aeruginosa polysaccharides

Maturation of biofilms involves production of the extracellular matrix. In P. aeruginosa, the matrix contains an array of components, such as DNA (Allesen-Holm et al., 2006; Ma et al., 2009) and proteinaceous adhesins, including CdrA (Borlee et al., 2010). However, exopolysaccharides have long been considered to be the most fundamental matrix components in P. aeruginosa (Ryder et al., 2007) and many other bacterial species.

The alginate polysaccharide is a polymer consisting of mannuronic and guluronic acid residues, and its overproduction gives bacterial colonies a characteristic mucoid appearance. Already in the early 60 s, mucoid P. aeruginosa was discovered as the first bacterial source of alginate, which is only loosely attached to the cellular envelope (Linker and Jones, 1966; Jain and Ohman, 2004). Alginate production was also found to be a hallmark of P. aeruginosa chronic infection in the lungs of CF patients (Linker and Jones, 1964). Alginate has been proposed to protect bacteria from adverse environments and contribute to biofilm development (Boyd and Chakrabarty, 1995). The mucoid conversion that typically occurs in the CF lung or in response to oxygen radical exposure is mainly associated with mutations in the gene encoding the anti-sigma factor MucA (mucA22) (Mathee et al., 1999). The relationship between alginate production and biofilm formation is unclear, since alginate is not essential for biofilm formation in vitro. However, alginate shapes the structure of biofilms and increases the resistance to antibiotics such as tobramycin (Hentzer et al., 2001). Regulation of alginate biosynthesis has been extensively investigated and involves a complex regulatory network. Among the many regulatory elements involved are also TCSs, such as the FimS/AlgR system, previously mentioned for its role in motility regulation, and KinB/AlgB (Ma et al., 1997). Both these systems positively regulate alginate biosynthesis, and the NtrC-type response regulator AlgB has been shown to bind directly to the algD promoter, which is the first gene in the alginate biosynthetic cluster (Jain and Ohman, 2004; Leech et al., 2008).

Non-mucoid P. aeruginosa strains are now known to be proficient in biofilm formation, and two gene clusters involved in polysaccharide biogenesis have been shown to be crucial for this process. The pel gene cluster was first discovered in strains PA14 and PAK (Friedman and Kolter, 2004b; Vasseur et al., 2005), and a functional psl gene cluster was then identified in P. aeruginosa strains ZK2870 and PAO1 (Jackson et al., 2004; Friedman and Kolter, 2004a). The structural composition of the Psl polysaccharide has recently been elucidated and consists of a repeating pentasaccharide containing d-mannose, d-glucose and l-rhamnose (Byrd et al., 2009). The assembly of Psl on the cell surface follows a remarkable helical distribution (Ma et al., 2009), and this organization has been proposed to provide a scaffold for other matrix components, as well as contributing to cell-cell interaction. Psl has also been proposed to facilitate attachment to eukaryotic cells (Byrd et al., 2010).

Regarding the Pel polysaccharide, the mechanism for biosynthesis and transport has also been extensively studied (Vasseur et al., 2007; Kowalska et al., 2010). However, the exact nature of the Pel polysaccharide has not yet been elucidated. It has previously been proposed to be a glucose-rich component (Friedman and Kolter, 2004b), but recent studies on carbohydrate components isolated from the P. aeruginosa matrix failed to identify any Pel-specific compounds (Coulon et al., 2010). Finally, a recent report showed that Pel has a structural as well as a protective role in biofilms of P. aeruginosa PA14, which does not have the psl gene cluster. This role is strain-specific, since it is not observed in PAO1, which is capable of producing both Pel and Psl (Colvin et al., 2011).

The RetS/LadS/GacS pathway is a central switch to pel and psl gene expression

Studies of the regulation of pel and psl gene expression have formed the basis for the discovery of a central regulatory network involved in the switch between planktonic and biofilm lifestyles, as well as between acute and chronic infections.

In order to identify factors involved in biofilm formation, Goodman and collaborators constructed 39 mutants in selected genes of P. aeruginosa PAK encoding non-classical sensors or response regulators of TCSs (Goodman et al., 2004). As expected, they observed that mutations in pilR, required for twitching motility, or fleR, involved in flagellar motility, resulted in decreased biofilm formation under the conditions tested. However, mutation of PA4856 resulted in a strain that displayed increased biofilm formation compared with the parental strain. This gene encodes a hybrid sensor, and transcriptional profiling revealed that deletion of PA4856 had pleiotropic effects. The mutant not only displayed highly increased expression of the pel and psl genes, hence the hyperbiofilm phenotype, it also had reduced expression of 40 genes involved in cytotoxicity and virulence, including the type III secretion system. The sensor was therefore named RetS (regulator of exopolysaccharides and type III secretion). Further studies showed that the reduced expression of virulence genes in the retS mutant translated into reduced cytotoxicity and cell death in CHO cells. P. aeruginosa virulence is known to be multi-factorial and involves other molecular devices in addition to the T3SS. Examples of this are other secretion systems and toxins, such as the type II secretion system (T2SS) (Filloux, 2004) and exotoxin A. Since the expression of such genes was also repressed in the retS mutant, this suggested that the RetS signalling pathway acted as a switch between biofilm formation and cytotoxicity (Goodman et al., 2004). Furthermore, based on the current knowledge on some of the virulence factors involved, this signalling could also correspond to a switch between acute and chronic infection.

In parallel with Goodman's study, we screened a transposon mutant library in PAKΔpilA, and selected strains that were reduced in their ability to form biofilms (Vallet et al., 2001). In addition to the identification of mutants affected in the cupA genes (Vallet et al., 2001) and in the pel genes (Vasseur et al., 2005), we identified a mutant in a second gene encoding a hybrid sensor. This gene was named ladS due to the lost adherence phenotype of the mutant (Ventre et al., 2006), and the LadS sensor was found to be very similar to the previously described RetS sensor. Interestingly, both hybrid sensors are membrane bound with eight transmembrane domains and contain a periplasmic detector domain of the type 7TMR-DISMED2. Transcriptional analysis of the LadS regulon highlighted some remarkable features regarding the targets of the RetS and LadS sensors. Whereas RetS significantly affects the expression of 397 genes, LadS affects only 79, but almost half of these were found to overlap with the RetS regulon (Ventre et al., 2006). Strikingly, gene modulation in the ladS mutant was found to be antagonistic to that of the retS mutant. Consequently, the pel genes, which are upregulated in the retS mutant, are downregulated in the ladS mutant explaining the loss of biofilm formation. Conversely, the T3SS genes, which are downregulated in the retS mutant, are all upregulated in the ladS mutant.

RetS and LadS are orphan hybrid sensors, which have not been assigned any cognate response regulators. Hybrid sensors are characterized by the presence of an additional receiver domain at their C terminus (Fig. 1). Furthermore, in the case of RetS, two receiver domains (REC) are found at the C terminus, whereas LadS has only one. It is generally assumed that phoshorylation cascades involving hybrid sensors require an intermediate phosphorelay (Hpt) to activate the cognate response regulator. However, RetS neither required the REC domains, nor an Hpt module to fulfil its function (Goodman et al., 2009). Instead, RetS was demonstrated to form heterodimers with another P. aeruginosa sensor, GacS. GacS is a membrane bound sensor, which contains a cytosolic HAMP domain, whereas the periplasmic domain has no similarity with the RetS sensory domain. As a consequence of the RetS/GacS interaction, GacS does not autophosporylate, and therefore cannot activate its cognate response regulator, GacA. This heterodimerization mechanism used by RetS for signalling is entirely novel and so far unique in bacterial signalling.

The GacS/GacA TCS has been known for a long time as a regulator of virulence and biofilm formation (Parkins et al., 2001). The response regulator GacA has only two targets, namely the promoters of the small RNAs RsmY and RsmZ, the activity of which it enhances (Lapouge et al., 2008; Brencic et al., 2009). Importantly, rsmZ transcription was also found to be a target of the RetS and LadS signalling pathways, and in agreement with the antagonistic activity of these two sensors, rsmZ gene expression was shown to be elevated in the retS mutant while it was reduced in the ladS mutant (Ventre et al., 2006). In P. aeruginosa and other bacteria (Lapouge et al., 2008), the role of these two regulatory small RNAs (sRNAs) is to titrate the translational repressor RsmA. RsmA prevents translation of a large set of transcripts by binding directly to target mRNAs and blocking accessibility to the ribosome-binding site. RsmA is also known to globally affect P. aeruginosa virulence (Pessi et al., 2001; Heurlier et al., 2004). Global transcriptomic and proteomic analyses comparing gene expression in an rsmA mutant to the isogenic parental strain indicated that genes involved in quorum sensing, multidrug efflux pumps, protein secretion, cytotoxicity and motility were affected (Burrowes et al., 2006; Mulcahy et al., 2006). Recent studies further characterized the RsmA regulon and showed that the pel and psl genes were among the direct targets of RsmA repression (Brencic and Lory, 2009; Irie et al., 2010). In contrast, other targets including the T3SS appeared to be positively regulated by RsmA, although this is likely to proceed through an indirect mechanism. It is thus now clear that a retS mutation results in upregulation of rsmY and rsmZ, which prevents RsmA activity and allows pel and psl gene expression, while at the same time positively influencing genes involved in the T3SS. RetS signalling intersects with the GacS/GacA pathway via direct interaction between RetS and GacS, and the phenotypes associated with the retS mutation can be suppressed with secondary mutations in either gacS, gacA or rsmZ (Goodman et al., 2004). Although the mechanism of LadS is currently unknown, it may also be hypothesized to intersect the GacS/GacA TCS to influence rsmZ (Fig. 3).

Details are in the caption following the image

The GacS/GacA/RsmA signalling network in P. aeruginosa.
Upper panel: Schematic overview of the signalling cascade that converges on the small RNAs RsmY and RsmZ, which act by sequestering the translational repressor RsmA. RsmA reciprocally regulates factors involved in acute infection (motility and the T3SS) and chronic infection (Pel, Psl and the T6SS). Arrows indicate a positive influence on gene expression or protein activity whereas red lines are indicative of a repressing effect. Dotted lines indicate that the connection has not been fully demonstrated or is not understood at the molecular level.
Lower panel: P. aeruginosa phenotypes characteristic for the motile lifestyle (left) or the sessile lifestyle (right). Images show swimming motility (left), biofilm formation in glass tubes stained by crystal violet (middle) and congo red binding, which is characteristic of elevated polysaccharide production (right).

Another interesting gene target of the RetS signalling pathway is the gene cluster encoding one of the type VI secretion systems (T6SS), H1-T6SS (Mougous et al., 2006). In contrast to the T3SS, which is important for P. aeruginosa acute infections, the T6SS has been proposed to be involved in chronic infections as shown in a rat model of chronic respiratory infection (Potvin et al., 2003). Furthermore, the T6SS is highly active in P. aeruginosa isolates from chronically infected CF patients (Mougous et al., 2006). The apparent inverse regulation between T3SS and T6SS provides novel insights into the regulation of potential disease determinants and shows that the RetS/LadS signalling pathway not only mediates the change from planktonic to biofilm lifestyle, but that it is an essential switch in the global control of genes associated either with chronic or acute infections (Fig. 3). A similar switch also exists in other species such as Pseudomonas syringae (Records and Gross, 2010).

The HptB signalling pathway

As previously mentioned, the hybrid sensor RetS does not act through an Hpt phosphorelay, but instead acts directly by interfering with GacS activity. However, an earlier report suggested that RetS participates in HptB-mediated phosphorelay (Hsu et al., 2008). In line with this observation, one of the two RetS REC domains has been shown to be required in some conditions (Laskowski and Kazmierczak, 2006). Hsu and colleagues (2008) also proposed that three additional hybrid sensors, PA1611, PA1976 and PA2824, were directly connected to HptB in a phosphorylation cascade. Interestingly, along with retS, a mutant in PA2824 was also reported by Goodman and collaborators to be a hyperbiofilm former (Goodman et al., 2004). We further investigated the hptB signalling cascade and showed that, as for RetS, the hptB mutant forms elevated levels of biofilms. However, in contrast to the retS mutant, the hptB mutant does not display increased levels of the T6SS genes (Bordi et al., 2010). Combined with the observation that RetS does not appear to autophosphorylate, this could suggest that RetS and HptB are not cognate partners and belong to two distinct regulatory cascades. We further showed that the HptB pathway intersects the GacS/GacA pathway and impacts expression of sRNA. However, an important observation was that the HptB pathway only alters the expression of the rsmY gene and has no effect on rsmZ gene expression (Fig. 3). The molecular mechanism of such differential control is unclear but the upstream regions of rsmY and rsmZ are quite different, and additional regulatory elements could therefore act either on the rsmY or rsmZ promoters (Brencic et al., 2009). Finally, HptB has been shown to not act directly on GacA, but via a more complex route that involves the response regulator PA3346. The output domain of PA3346 is a phosphatase 2C domain (PP2C), which is involved in the dephosphorylation of PA3347, a putative anti-anti-sigma factor (Hsu et al., 2008). The genes encoding PA3346, PA3347 and HptB (PA3345) are clustered on the chromosome, which further supports the hypothesis that they act in a coordinate manner. This pathway is thus reminiscent of RsbU/RsbV pathway in Bacillus subtilis (Delumeau et al., 2004), although the identity of the sigma and anti-sigma factors involved in the final branch of the HptB pathway is currently unknown (RsmW and σB in case of B. subtilis).

Two-component systems pace the progression of biofilm development

We have so far discussed a series of TCSs that are important in the control of biofilm determinants. This control translates into the production of a series of biofilm-related factors, such as extracellular appendages and exopolysaccharides. The role of the regulatory systems is to ensure the timely presence of all relevant components when appropriate during biofilm development.

Interestingly, most studies of TCSs in P. aeruginosa have been restricted to either planktonic culture or to biofilms at a single time-point. However, some studies have focused on temporal protein activation in P. aeruginosa biofilms, thereby dividing the development into a number of distinct stages. At least five stages have been defined and are referred to as reversible attachment, irreversible attachment, maturations 1 and 2 and finally dispersal (Sauer et al., 2002). In a recent study, Petrova and Sauer specifically investigated the temporal phosphoproteome in biofilms of P. aeruginosa PAO1 (Petrova and Sauer, 2009). Using a combination of proteomic techniques (2D-GE coupled with immunoblotting and ICAT coupled with LC-MS/MS), they showed that members of three TCSs are phosphorylated at distinct stages of biofilm formation, but not in planktonic cells (Fig. 4). They further showed that biofilms formed by mutants in these components are arrested in their development around the stage at which the component would have become phosphorylated. These signalling systems therefore appear to act as checkpoints for the progression of biofilm formation.

Details are in the caption following the image

P. aeruginosa biofilm life cycle and regulation.
Upper panel: Biofilm formation in vitro has been shown to proceed through a series of distinct stages. Initial attachment of motile bacteria to a surface is often mediated by extracellular appendages and is reversible. This is followed by irreversible attachment and microcolony formation, which are then encased in self-produced polysaccharides and can mature into mushroom-like structures. Finally, a combination of cell death (indicated in red) and matrix degradation dissolves the structures from the inside and releases planktonic cells back into the medium (Stoodley et al., 2002; Webb et al., 2003).
Lower panel: Stage-specific phosphorylation of partners of two-component regulatory systems during biofilm formation (Petrova and Sauer, 2009). See text for details.

The TCSs were named according to their role in biofilm development: BfiSR (biofilm initiation), BfmSR (biofilm maturation) and MifSR (microcolony formation). The three sensor kinases are all classical with typical N-terminal HisKA-HATPase domains. BfiS is predicted to be cytoplasmic with three N-terminal PAS domains that may be involved in signal sensing, BfmS is predicted to be membrane bound and has a HAMP domain, and MifS has no bioinformatically predicted input domain. The response regulators are all predicted to be DNA binding, BfiR via a C-terminal HTH domain, BfmR with a winged-helix domain, and MifR has an AAA-type ATPase domain between the receiver and DNA binding domains.

While no targets are known for the BfmSR and MifSR systems, further studies in P. aeruginosa PA14 indicated a link between the BfiSR system and the RetS/LadS/GacS signalling cascade (Petrova and Sauer, 2010). Importantly, it was observed that under planktonic growth conditions, the abundance of rsmYZ transcripts was unchanged in a bfiS mutant compared with the parental strain. However, in biofilms the levels of sRNAs were decreased several folds in the wild-type but not in the bfiS mutant. These observations suggested that the increase of sRNA that is observed in the retS mutant is required for initiating biofilm formation. However, once irreversible attachment has occurred, further progression of the biofilm development appears to require downregulation of these sRNAs. A direct target was identified for the BfiR response regulator, namely the promoter of cafA, which encodes a protein with ribonuclease G activity. In accordance with this, the biofilm phenotype of the ΔbfiS mutant could be complemented by overexpressing cafA. Furthermore, the RNase activity of CafA appears to be specific for RsmZ, but not for RsmY (Fig. 3). This is a striking example of post-transcriptional control and further highlights the distinct roles of the seemingly redundant sRNAs in biofilm formation, as previously mentioned in the HptB pathway (Bordi et al., 2010).

Concluding remarks

This review was aimed at giving an overview of key two-component regulatory systems involved in biofilm development in P. aeruginosa. The overview is by no means complete and may occasionally present conflicting information. However, the presentation of complex regulatory processes in simple diagrams is frequently a balance between the comprehensive and the comprehensible. This is further complicated by the fact that data obtained from different laboratories, from studies with different P. aeruginosa isolates and carried out in different growth conditions require extreme caution with respect to interpretation and extrapolation. However, a number of lessons can be learnt even from a simplified report, such as this one.

The process of biofilm formation involves a large and complex regulatory network that, in addition to TCSs, involves a multitude of other elements, such as quorum sensing, c-di-GMP signalling and sigma factors, which have purposely been omitted in our overview. Nevertheless, TCSs have been shown to play a central role in biofilm development, as discussed above. Furthermore, despite being outnumbered several times by the classical sensors (Barakat et al., 2009), the key role of the unorthodox and hybrid sensors is striking in this context. These sensors are part of phosphorelay pathways, which are likely to allow fine-tuning of the responses to complex environmental signals or rapidly fluctuating environments.

Another striking observation is that, despite extensive homology and high levels of specificity within the TCS family, a number of diverse signalling networks can be elucidated. In this review, two archetypes of branched signalling pathways have been discussed. The Roc system is a so-called one-to-many pathway (Laub and Goulian, 2007), in which a single sensor, such as RocS1, influences the activity of several response regulators. In this case, the same sensor can modulate gene expression in an antagonistic manner by activating response regulators that either activate or repress cup gene expression and thereby biofilm formation (RocA1 and RocR respectively) (Kulasekara et al., 2005). The other type of branched pathway is known as many-to-one. In this case, several sensors signal to the same response regulator, and multiple signals can thereby be integrated into one specific response. The RetS/LadS/GacS sensors are a special case of a many-to-one system, since they all appear to influence the activity of the GacA response regulator and subsequent levels of sRNAs (Ventre et al., 2006). A further extension of this is the combination of the two types of branched pathways. This is observed in the Roc regulatory network, in which two sensors, RocS1 and RocS2, converge on the same set of response regulators with each response regulator controlling a specific set of target genes (Sivaneson et al., 2011).

The presence of multiple and highly similar sensory systems, such as the Roc1 and Roc2 systems, poses a considerable challenge in terms of unwanted crosstalk or conflicting pathways. A number of solutions to avoid these undesirable effects have evolved, some of which have been presented above. In the case of the PilSR TCS, spatial localization is likely to prevent interaction with other TCSs (Boyd, 2000). Alternatively, temporal expression, or activation, of certain TCSs can occur during biofilm development, which could be a strategy to avoid cross-communication. Finally, the presence or absence of specific signals could restrict the activity of TCSs to particular environmental conditions.

In conclusion, much more work is required to elucidate the complex regulatory networks involved in biofilm development. The change from a planktonic to a sessile lifestyle is associated with extensive physiological changes, and bacteria have consequently developed global molecular switches to control these changes. The Gac/Rsm regulatory system is merely one of these switches. Importantly, the transition between planktonic growth and biofilm formation can also be correlated with the infection strategy, i.e. acute versus chronic, of P. aeruginosa. TCSs thus remain a target of choice for the development of new drugs that can interfere with bacterial pathogenesis (Gotoh et al., 2010).

The final question that remains unanswered for the vast majority of TCSs is which signal is detected by the sensory domain. The sensor kinases described in this review are no exception to this rule, and much could be learnt from further investigations into this area. In pursuit of such a signal, the structure of the RetS sensory domain has recently been solved and was found to be similar to carbohydrate binding modules found in other proteins (Jing et al., 2010; Vincent et al., 2010). This observation suggests that RetS, and probably also LadS, responds to carbohydrate-like structures. Future work is likely to reveal more about a wide range of such sensing domains, and this could provide an important lead for the development of novel anti-pseudomonal drugs.


A.F. is supported by the Royal Society, the BBSRC (Ref. No. BB/F019645/1) and European EST Marie Curie (Grant No. MEST-CT-2005-020278). H.M. is supported by the BBSRC (Grant No. BB/F019645/1), and M.S. is supported by a Marie Curie fellowship (Grant No. MEST-CT-2005-020278) and a grant from the ‘Fondation pour la Recherche Médicale’ (FRM: FDT20091217603).