Quorum sensing in bacterial virulence

Introduction

Bacteria can communicate by producing and responding to small diffusible molecules that act as signals. These molecules have been termed autoinducers (AIs). AIs are produced at basal levels and their concentration increases with growth. Because the signals can diffuse through membranes, their concentration inside cells approximates the concentration in the environment. Upon reaching a critical concentration, the signal molecules can bind to and activate receptors inside bacterial cells. These receptors can then alter gene expression to activate behaviours that are beneficial under the particular condition encountered. As this phenomenon occurs in a cell-density-dependent manner, it has been termed quorum sensing (Antunes & Ferreira, 2009; Bassler & Losick, 2006; Fuqua et al., 1994; 2001; Fuqua & Greenberg, 2002).

Many classes of AIs have been described to date. The most intensely studied AIs are the N-acylhomoserine lactones (AHLs) of Gram-negative bacteria, the peptides of Gram-positive bacteria and a class of AIs termed AI-2, whose structures remain unknown in most cases (Antunes & Ferreira, 2009). AHLs are usually detected through binding to and activation of cytoplasmic receptor proteins, which dimerize upon signal detection and can bind to promoter regions of target genes to activate or repress their transcription (Fuqua et al., 1994). Peptides are usually detected through binding to membrane sensor proteins of the two-component system family, although some can also be transported to the cytoplasm before interacting with their receptors (Novick & Geisinger, 2008; Pottathil & Lazazzera, 2003). On the other hand, AI-2 binds a periplasmic protein and then interacts with either a two-component system or a transporter depending on the organism (Ng & Bassler, 2009; Taga et al., 2001). Binding to a membrane-associated sensor kinase causes the activation of a phosphorelay cascade, which results in the activation or repression of a response regulator, culminating in altered gene expression (Ng & Bassler, 2009; Novick & Geisinger, 2008).

Quorum sensing was originally described in the marine luminescent bacterium Vibrio fischeri, where it functions as the control mechanism of light production and numerous other traits (Antunes et al., 2007; Eberhard et al., 1981; Engebrecht et al., 1983; Engebrecht & Silverman, 1984). For years, it was thought that this phenomenon was limited to a few marine organisms but it is now widely recognized that many bacterial species utilize quorum sensing as part of their regulatory machinery (Antunes & Ferreira, 2009; Bassler & Losick, 2006; Lyon & Novick, 2004; Smith et al., 2006). Of interest, we now know that bacterial virulence is in many cases controlled by quorum sensing (Antunes & Ferreira, 2009). This has led to a burst in quorum sensing research and the role of quorum sensing in the virulence of multiple human pathogens has been studied in molecular detail. Here, we review the role of quorum sensing in bacterial pathogenicity using several model organisms. We discuss the quorum sensing mechanisms used to control virulence gene expression, the evidence that they are important during host infection and the available strategies to control bacterial virulence using quorum sensing inhibitors.

Quorum sensing in Staphylococcus aureus virulence

Many Gram-positive bacteria utilize peptide quorum sensing systems to control gene expression and S. aureus has served as a model to study bacterial peptide signalling (Novick & Geisinger, 2008). This organism is a member of the human microbiota and is found in approximately 30 % of the adult population (George & Muir, 2007; Roux et al., 2009). Despite its widespread prevalence in healthy subjects, S. aureus is also a very dangerous opportunistic pathogen which has been increasingly associated with antibiotic resistance (George & Muir, 2007). S. aureus has multiple virulence factors and can display very rapid transmission, aiding its importance as a human pathogen (Massey et al., 2006). Additionally, S. aureus forms biofilms on many surfaces, including indwelling devices such as urethral stents (Kehinde et al., 2004). These indwelling devices, and subsequent biofilms formed on them, pose a serious risk for Staphylococcus infection (Balaban et al., 2007).

One of the factors which contribute to S. aureus virulence is its peptide-based quorum sensing system, encoded by the accessory gene regulator (agr) locus (Novick & Geisinger, 2008) (Fig. 1⇓). The autoinducer in the agr system is an oligopeptide that has been termed the autoinducing peptide (AIP), encoded by agrD. AIP is trimmed and secreted by AgrB, a membrane-bound protein (Ji et al., 1995; Saenz et al., 2000; Zhang et al., 2004). The active AIP is 7–9 aa, with a 5-membered thiolactone ring (Roux et al., 2009). AgrC is a membrane-bound sensor kinase to which extracellular AIP binds, leading to AgrC autophosphorylation and activation of AgrA (Ji et al., 1995; Koenig et al., 2004; Lina et al., 1998). The agr system is intricately involved in the regulation of virulence genes, predominantly from two promoters, P2 and P3, which produce RNAII and RNAIII, respectively (Morfeldt et al., 1995; Novick & Geisinger, 2008). P2 promotes the transcription of the agr operon from the RNAII transcript, which includes agrA, agrB, agrC and agrD (Novick et al., 1995; Roux et al., 2009). Active AgrA may be a phosphorylated homodimer that induces transcription at the P2 and P3 promoters, with a higher affinity for P2 (George & Muir, 2007; Koenig et al., 2004). Transcription from P3 leads to the production of RNAIII, which is the effector molecule of the agr system (Roux et al., 2009). RNAIII is a 514 nt regulatory RNA, which also functions as the mRNA for the δ-toxin (Balaban & Novick, 1995; Benito et al., 2000; Kong et al., 2006; Novick et al., 1995). The 5′ end is thought to upregulate α-haemolysin, while the 3′ end is required for the repression of protein A synthesis (Kong et al., 2006; Morfeldt et al., 1995). RNAIII reduces the expression of surface adhesins, and increases the production of capsule, toxins and proteases (Novick & Geisinger, 2008; Roux et al., 2009).

The agr system is thought to regulate over 70 genes, 23 of which are known virulence factors (George & Muir, 2007). Of the virulence factors regulated by agr, there are two classes: the first class contains virulence factors involved in attachment to the host and immune evasion, while the second class contains genes involved in the production of exoproteins associated with invasion and toxin production (Bowden et al., 2005; Yao et al., 2006). It has been thought that the activation of the agr system essentially switches the bacterium from an adhesive, colonizing commensal to an invasive and aggressive pathogen (Roux et al., 2009).

Four distinct groups of agr polymorphism have been identified, these are categorized as I–IV (George & Muir, 2007; Ji et al., 1997). Each group has a distinct AIP, which is able to bind to the receptor from all groups. However, each AIP is only able to activate the receptor from the same group (Ji et al., 1997). Apart from groups I and IV, which are able to cross-activate, all other groups are cross-inhibiting, i.e. an AIP from group II can inhibit the receptor from a group III strain (George & Muir, 2007; Mayville et al., 1999; Otto et al., 1998). There does appear to be a correlation between the agr group and the relative fitness of the S. aureus strain in an insect model of infection (Fleming et al., 2006). Therefore, there is potential for therapies which are based upon cross-group inhibition (George & Muir, 2007). There remains no doubt that the agr system is involved in S. aureus virulence; however, exactly how it contributes remains controversial.

One of the ways in which agr is thought to impact virulence is through its role in biofilm formation. The formation of biofilms by bacteria is a multi-step developmental process that starts with adhesion to a surface. Attached bacteria divide and give rise to macrocolonies. These macrocolonies later develop into mature biofilms, which can assume multiple topographies. The last step in biofilm development is detachment, which may be important for dissemination during an infectious process (Parsek & Tolker-Nielsen, 2008). It is thought that S. aureus possesses two independent mechanisms of biofilm formation; the first involves an extracellular polysaccharide, polysaccharide intercellular adhesin (PIA), and the second is thought to be PIA-independent, possibly involving adhesive proteins and the sarA and agr global regulators (Lauderdale et al., 2009; Novick & Geisinger, 2008). Because biofilms are thought to play a critical role in S. aureus infection, the role of agr in biofilm formation has been explored. When agr is non-functional, S. aureus has enhanced adhesion abilities (Vuong et al., 2000). Therefore, when agr is not active, the bacteria remain in the first stage of biofilm formation, adhering to a surface. However, agr is also important for detachment of clusters of cells from the biofilm (Kong et al., 2006). An agr mutant strain has a detachment defect, and the detachment of bacterial cells from biofilms was found to coincide with agr expression (Kong et al., 2006; Yarwood et al., 2004). In addition, an agr mutant has increased adherence and more robust formation of static biofilms than its agr-containing counterparts (Novick & Geisinger, 2008). It is thought that this role of agr is brought about by the reduction in adhesin production and an increase in the production of both δ-haemolysin and proteases (Novick & Geisinger, 2008). Thus, an important role of quorum sensing in S. aureus is the regulation of biofilm formation, a central factor in S. aureus virulence.

The exact contribution of agr to disease remains somewhat controversial. Many studies have found that agr is required for virulence in models of septic arthritis, osteomyelitis, endophthalmitis and pulmonary infections (Abdelnour et al., 1993; Booth et al., 1997; Gillaspy et al., 1995; Heyer et al., 2002). However, other evidence seems to question the role of the agr system in virulence in other models (Cheung et al., 1994; Kielian et al., 2001). In cystic fibrosis (CF) patients with S. aureus infections, the agr system was inactive, suggesting that agr activation may not be necessary for infection (Goerke et al., 2000). Through a temporal analysis of agr expression, Wright et al. (2005) have found that agr expression varies during different steps of infection. This suggests that agr may play different roles during the course of infection and this may explain the discrepancy in some of the results regarding its role in bacterial virulence. Interestingly, specific agr groups have been associated with some diseases, such as toxic shock syndrome, while other diseases, such as infections of CF patients, are associated with all four agr groups (George & Muir, 2007).

RNAIII is the effector of the agr system and is involved in bacterial virulence, so mechanisms to inhibit RNAIII have received considerable attention. An RNAIII-inhibiting peptide (RIP) has been found to inhibit S. aureus biofilm formation and toxin production (Cirioni et al., 2007). It is thought that by inhibiting cell–cell communication, RIP is able to prevent adherence and virulence of S. aureus. Preliminary studies have shown that RIP has no adverse effects on laboratory animals (Balaban et al., 2007; Cirioni et al., 2007). Treatment with RIP and teicoplanin prevented S. aureus growth in a urethral stent model (Cirioni et al., 2007). To further support the efficacy of RIP, Balaban et al. (2007) found that when RIP was administered at the time of infection, a concentration-dependent reduction in bacterial load was observed. It was also found that treatment with RIP 2 days post-infection suppressed bacterial growth and was more efficacious with multiple doses.

Quorum sensing control of Pseudomonas aeruginosa virulence

P. aeruginosa is a Gram-negative bacterium capable of surviving in a wide range of environments. This organism is an opportunistic pathogen and it is commonly associated with nosocomial infections and infections of severely burned individuals, and is a leading cause of death in severe respiratory infections, such as chronic lung infections in CF patients (Bendiak & Ratjen, 2009; Bodey et al., 1983; Koch & Hoiby, 1993). Infections with P. aeruginosa are difficult to eradicate, due to their high levels of antibiotic resistance and growth in biofilms (Driscoll et al., 2007).

At least three intertwined quorum sensing systems and one orphan autoinducer receptor affect the ability of P. aeruginosa to cause disease (Fig. 2⇓). Two of these systems, las and rhl, rely on the production of AHLs as the signalling molecules (AIs) (De Kievit & Iglewski, 2000). In the las system, N-3-oxododecanoyl-homoserine lactone (3OC12-HSL) is produced by the enzyme encoded by the lasI gene. When P. aeruginosa reaches a certain threshold density, 3OC12-HSL binds to the transcriptional activator LasR. LasR, in turn, dimerizes and binds to target promoters to control gene expression (De Kievit & Iglewski, 2000). Similarly, in the rhl system, the rhlI gene encodes the enzyme involved in the production of N-butyryl-homoserine lactone (C4-HSL). As with 3OC12-HSL, C4-HSL binds to its cognate transcriptional regulator, RhlR, to control the activity of target promoters (De Kievit & Iglewski, 2000). The rhl system is controlled by the las system at both transcriptional and post-transcriptional levels (Latifi et al., 1996). Besides LasR and RhlR, P. aeruginosa encodes an orphan receptor protein, QscR, which can sense 3OC12-HSL to control its own regulon (Chugani et al., 2001; Fuqua, 2006; Schuster & Greenberg, 2006).

Multiple P. aeruginosa virulence factors are involved in the development of disease, including secreted factors (such as proteases) and cell-associated factors (such as lipopolysaccharide and flagella), as well as the ability to form biofilms (Lyczak et al., 2000). Quorum sensing regulates the production of several extracellular virulence factors, promotes biofilm maturation and regulates the expression of antibiotic efflux pumps, meaning that it has a key role in the pathogenesis of P. aeruginosa (Dekimpe & Déziel, 2009; Diggle et al., 2006; Fuqua et al., 2001; Schuster & Greenberg, 2006; Swift et al., 2001; Wagner et al., 2003; Whitehead et al., 2001). The las and rhl systems regulate the timing and production of multiple virulence factors, including elastase, alkaline protease, exotoxin A, rhamnolipids, pyocyanin, lectins and superoxidase dismutase (Schuster et al., 2003; Smith & Iglewski, 2003). The expression of these two quorum sensing systems has also been linked to the regulation of biofilm formation. Quorum sensing signalling may start in the early stages of biofilm development, which is characterized by microcolony formation, where lasI mutants are unable to form structurally normal biofilms (Davies et al., 1998). Expression of the lasI gene is maximal at day 4 of biofilm development, decreasing between days 6 and 8. The expression of rhlI fluctuates during biofilm development and phenotypes of biofilm development with a rhlI mutant vary, according to the media and model used, supposedly due to different iron levels present (Davies et al., 1998; De Kievit et al., 2001; Patriquin et al., 2008; Yoon et al., 2002).

Significant concentrations of the AI molecules 3OC12-HSL and C4-HSL have been detected in sputum of CF patients colonized with P. aeruginosa (Erickson et al., 2002; Singh et al., 2000). This may indicate that quorum sensing is active during P. aeruginosa colonization of CF patients; however, it is important to note that quorum-sensing-deficient P. aeruginosa strains are often isolated from CF patients (Erickson et al., 2002; Karatuna & Yagci, 2010; Le Berre et al., 2008; Schaber et al., 2004). This has spurred a major discussion in the scientific community about whether quorum sensing is really important during CF infections. It has been hypothesized that the maintenance of a functional quorum sensing system is a metabolic burden for P. aeruginosa and that co-colonization with quorum-sensing-proficient and -deficient strains is in the best interest of this community of pathogens (Heurlier et al., 2006; Kohler et al., 2009). Moreover, social exploitation in P. aeruginosa communities may provide an explanation for the emergence of quorum-sensing-deficient strains in human infections (Sandoz et al., 2007). Nevertheless, it has been shown that P. aeruginosa rhlI and lasI mutants cause less tissue destruction and decrease mortality when compared with wild-type strains in multiple animal models (Smith & Iglewski, 2003), indicating an important role for quorum sensing in P. aeruginosa pathogenesis.

Apart from regulating the expression of virulence factors, some of the AIs have been shown to directly interact with host cells. 3OC12-HSL induces interleukin (IL)-8 secretion from human bronchial epithelial cells (DiMango et al., 1995) and induces COX-2 production (Smith et al., 2002). It also inhibits lymphocyte proliferation, downregulates the production of tumour necrosis factor alpha (TNF-α) and IL-12, and activates T cells to produce gamma-interferon (Skindersoe et al., 2009; Smith et al., 2002). 3OC12-HSL also induces apoptosis in macrophages and neutrophils, suggesting that this molecule not only enhances the expression of virulence genes but also affects immune responses to infection (De Kievit & Iglewski, 2000; Smith et al., 2002; Smith & Iglewski, 2003; Tateda et al., 2003; Telford et al., 1998). Additionally, 3OC12-HSL has been shown to directly affect one of the major regulators of immune responses, NFκB (Kravchenko et al., 2008).

A third P. aeruginosa AI molecule, 2-heptyl-3-hydroxyl-4-quinolone (Pseudomonas quinolone signal; PQS) has also been identified (Déziel et al., 2004; Lepine et al., 2004; Pesci et al., 1999). PQS belongs to the 2-alkyl-4-quinolone (AQ) family and it is synthesized via the pqsABCD genes in the pqsABCDE operon, which are responsible for the synthesis of 2-heptyl-4-quinolone (HHQ), the immediate PQS precursor (Diggle et al., 2006). Both molecules have been shown to play a role in cell–cell communication in P. aeruginosa. PqsH, which is controlled by the las system (Gallagher et al., 2002), plays a role in converting HHQ to PQS (Déziel et al., 2004). PqsE is in an operon with pqsA–D, but is not involved in the synthesis of PQS. However, disruption of pqsE causes loss of signal transduction of several but not all of the processes believed to be regulated by the pqs system (Diggle et al., 2007b). The exact mechanism of action of this protein is still unknown. However, it has been shown recently that PqsE can activate the transcription of PQS-controlled genes in the absence of PqsR and PQS by enhancing the rhl system (Farrow et al., 2008). PqsR is a LysR transcriptional factor that is activated by HHQ and PQS, leading to the positive activation of many virulence factors, which include a large number of genes also controlled by las and rhl. In addition, PqsR also controls the pqs operon itself, generating a positive feedback loop (Cao et al., 2001; Déziel et al., 2004, 2005; Xiao et al., 2006). Interestingly, the expression of PqsR is positively regulated by the las system (Xiao et al., 2006). RlhR, in turn, affects the expression of the pqs system (Hazan et al., 2010; McGrath et al., 2004; Xiao et al., 2006).

Through transcriptome analysis, over 90 genes were found to be regulated by the pqs system (Bredenbruch et al., 2006; Déziel et al., 2005). PQS has been shown to affect biofilm formation and to regulate several virulence factors in P. aeruginosa, including elastase, pyocyanin and LecA lectin, and it is considered essential for full virulence in multiple hosts (Cao et al., 2001; Rahme et al., 1997, 2000). PQS has been found in sputum, bronchoalveolar fluid and mucopurulent fluid from CF patients, suggesting that it may play an important role during the infection process (Collier et al., 2002). Besides controlling bacterial virulence, PQS and HHQ have been shown to downregulate the host immune response through NFκB (Kim et al., 2009b). Mutants that do not secrete these molecules have been shown to elicit increased TNF-α and IL-6 expression in cultured cells (Kim et al., 2009b). In a mouse infection model, a pqsA mutant was unable to disseminate in the lung tissue as well as the wild-type strain, suggesting a role of these molecules in P. aeruginosa pathogenesis (Kim et al., 2009b). PQS can also act as an iron chelator, and both the synthesis of PQS and the activity of PqsR–PQS are involved in iron homeostasis, another indication of the global importance of quinolone signalling for P. aeruginosa (Bredenbruch et al., 2006; Oglesby et al., 2008).

Because of the significant role of quorum sensing in the regulation of hundreds of virulence factors in P. aeruginosa, significant efforts have been made to discover molecules that inhibit quorum sensing in this organism. Natural quorum sensing inhibitors include cyclic sulfur compounds (Persson et al., 2005), halogenated furanones (Givskov et al., 1996), patulin and penicillin acid (Rasmussen et al., 2005b), among others. Through the use of a screening system, garlic extracts and 4-nitro-pyridine-N-oxide were also identified as quorum sensing inhibitors (Rasmussen et al., 2005a). These molecules specifically inhibited the activation of virulence genes by quorum sensing, reduced biofilm tolerance to tobramycin and also decreased virulence in a Caenorhabditis elegans pathogenesis model. In a recent study, a screening of quorum-sensing-inhibiting molecules derived from the plant species Combretum albiflorum revealed that catechin has a negative impact on the production of quorum-sensing-dependent virulence factors (Vandeputte et al., 2010). Subinhibitory concentrations of macrolides have been shown to be effective in inhibiting quorum-sensing-dependent virulence both in vivo and in vitro (Imamura et al., 2004; Molinari et al., 1993; Sofer et al., 1999; Tateda et al., 2001). Furthermore, CF patients showed significant improvement in pulmonary function with the use of macrolides without any effect on the total Pseudomonas population (Fujii et al., 1995; Saiman et al., 2003). Despite its success in decreasing P. aeruginosa virulence, macrolides are not ideal because, like any other antibiotic, excessive usage is associated with bacterial resistance. Nevertheless, based on the results obtained with natural quorum sensing inhibitors, a number of new inhibitory molecules have been synthesized and used with some success (Kim et al., 2008, 2009a; Liu et al., 2010), demonstrating that quorum sensing inhibition is an attractive therapeutic target.

Quorum sensing in Escherichia coli virulence

Another bacterial species that uses quorum sensing to control virulence gene expression is E. coli. This organism produces a signal molecule termed AI-2 (Surette & Bassler, 1998). AI-2 was originally identified as one of the AIs controlling light production by the marine bacterium Vibrio harveyi (Bassler et al., 1993). By utilizing a V. harveyi reporter strain that does not produce AI-2 but can respond to it, Surette and colleagues have identified E. coli strains that could activate the V. harveyi AI-2 sensor (Surette & Bassler, 1998). Subsequently, the gene responsible for AI-2 production was identified and named luxS (Surette et al., 1999). Although the structure of AI-2 in E. coli is currently unknown, the molecule produced by the close relative Salmonella enterica serovar Typhimurium has been identified as (2R,4S)-2-methyl-2,3,3,4-tetrahydroxytetrahydrofuran (Miller et al., 2004).

Genetic studies in enterohaemorrhagic E. coli (EHEC) and enteropathogenic E. coli revealed that LuxS controls the expression of the type-3 secretion system encoded by the locus of enterocyte effacement (LEE) pathogenicity island (Sperandio et al., 1999). This important virulence determinant is required for the formation of the characteristic attaching and effacing (AE) lesions caused by these pathogens (Celli et al., 2000). Additionally, transcriptomic studies have revealed that LuxS is a global regulator in EHEC, controlling the expression of over 400 genes (DeLisa et al., 2001; Sperandio et al., 2001). Many of these genes have functions related to bacterial virulence such as flagellar motility, surface adhesion and Shiga toxin production.

Originally, it was believed that the phenotypes of luxS mutants could be attributed to the lack of AI-2 production. However, Sperandio et al. (2003) showed that the molecule responsible for the regulation of LEE and flagellar genes in E. coli is not AI-2 (Sperandio et al., 2003). Because luxS controls the expression of LEE, it was expected that a luxS mutant would be unable to produce AE lesions in cultured epithelial cells. However, the mutant was fully proficient in inducing the formation of such lesions (Sperandio et al., 2003). Through a series of elegant experiments, the authors showed that although the signal responsible for the regulation of LEE was dependent on luxS, it was not AI-2, but a new molecule. This new AI was then named AI-3 (Sperandio et al., 2003) (Fig. 3⇓).

The regulatory cascade linking AI-3 sensing and virulence gene expression in E. coli is extremely complex. Microarray studies of a luxS mutant have revealed that multiple regulatory genes are controlled by quorum sensing, and some of these genes have been characterized in further detail (DeLisa et al., 2001; Sperandio et al., 2001). QseBC is a bacterial two-component signalling system that is regulated by LuxS, with QseB being the response regulator and QseC the sensor kinase (Sperandio et al., 2002b). Similar to LuxS, it has been found that QseBC controls motility gene expression (Sperandio et al., 2002b). Mutant analyses have led to the conclusion that AI-3 affects gene expression through QseBC, since this AI activates qseBC expression and a qseC mutant is unable to regulate motility in response to AI-3 (Sperandio et al., 2003). Biochemical studies have revealed not only that the expression of QseC is activated by AI-3 but also that this sensor kinase is the receptor for the AI (Clarke et al., 2006). Interestingly, QseC also functions as the receptor for the host catecholamine hormones epinephrine and norepinephrine, indicating that small molecule signalling pathways in eukaryotes and bacteria can intertwine (Clarke et al., 2006). Addition of AI-3 or catecholamines to QseC-containing liposomes resulted in an increased phosphorylation of this protein. Additionally, the activation of QseC resulted in phosphorylation of QseB, showing that QseB acts as the cognate response regulator in this system (Clarke et al., 2006). Besides QseBC, multiple regulators are involved in the control of virulence gene expression. A second two-component regulatory system that controls the formation of AE lesions has been recently described (Reading et al., 2009). This system is composed of the histidine kinase QseE and the response regulator QseF. These two proteins are involved in the transcriptional control of the effector EspFu, which is translocated into host cells by EHEC. Interestingly, a third gene in the same operon, qseG, separates qseE and qseF. QseG is an outer-membrane protein and it is also required for the translocation of type III secretion system effectors into host cells (Reading et al., 2009). Besides these two-component systems, two LysR-type regulators, QseA and QseD, are also required for the control of LEE expression, revealing that a complex regulatory cascade links quorum sensing and virulence gene expression in E. coli (Sperandio et al., 2002a; Walters & Sperandio, 2006).

The observation that quorum sensing controls the expression of multiple genes involved in E. coli pathogenicity has raised the hypothesis that this signalling system must be important for virulence in vivo. To test this hypothesis, Clarke et al. (2006) used a rabbit model of infection to show that a qseC mutant is attenuated during infection of rabbits, confirming the role of quorum sensing in E. coli virulence. Additionally, qseC has proved important for bacterial virulence in both Salmonella enterica serovar Typhimurium and Francisella tularensis (Rasko et al., 2008).

AI-3 control of virulence in E. coli has prompted investigators to develop small molecules that could inhibit this signalling system. To this end, Rasko et al. (2008) have recently described the identification of one such molecule. Through a high-throughput screening of a large library of compounds, the authors identified N-phenyl-4-{[(phenylamino)thioxomethyl]amino}-benzenesulfonamide (LED209) as an inhibitor of QseC and bacterial virulence, both in vitro and in vivo. LED209 inhibits QseC autophosphorylation, virulence factor production and AE lesion formation by EHEC. Additionally, LED209 can inhibit virulence factor production and host colonization by Salmonella Typhimurium and F. tularensis (Rasko et al., 2008).

Additional roles of bacterial AIs

Although bacterial AIs have been studied for decades due to their signalling roles, these molecules can often perform other functions. Perhaps the clearest example of this comes from the fact that AI-2 is not only a signal molecule but also a metabolic by-product (Rezzonico & Duffy, 2008; Schauder et al., 2001). AI-2 is synthesized through reactions involving the enzymes Pfs and LuxS. S-Adenosylmethionine (SAM) acts as a methyl donor in bacterial cells, creating S-adenosylhomocysteine (SAH). Because SAH is toxic, bacteria have evolved mechanisms to recycle it. This is accomplished through the action of the enzyme Pfs, which converts SAH to S-ribosylhomocysteine (SRH). SRH is then converted to AI-2 and homocysteine by LuxS (Schauder et al., 2001). These findings indicate that the connection of AI-2 synthesis with bacterial metabolism is twofold. First, the precursor of AI-2 is a major methyl donor in bacterial cells, acting in multiple metabolic processes. Also, LuxS is involved in the detoxification of a metabolic intermediate. Therefore, the role of AI-2 as a bona fide signal has been questioned (Winzer et al., 2002). Although such a role is clear in organisms such as E. coli and V. harveyi, it is likely that in most cases, the production of AI-2 is a mere consequence of SAM utilization and SAH detoxification. This could explain why LuxS is widespread in bacteria whereas AI-2 receptors are not (Rezzonico & Duffy, 2008).

Besides the metabolic link described above, many other functions have been assigned to bacterial chemical signals. For instance, Bredenbruch et al. (2006) have demonstrated that the P. aeruginosa PQS signal has iron-chelating properties. Using transcriptome analyses of the P. aeruginosa response to PQS, the authors showed that a large group of differentially regulated genes were involved in iron metabolism. This prompted the authors to investigate a direct link between PQS and iron homeostasis, leading to the discovery that PQS has iron-chelating activity and can form iron-containing complexes. Also, some of the known effects of PQS on P. aeruginosa could be attributed to its direct action as an iron chelator, supporting the notion that bacterial signals may have many currently unidentified functions in bacterial physiology (Bredenbruch et al., 2006).

Another non-signalling function that has been assigned to bacterial signalling molecules is that of antimicrobials. Lantibiotics are small peptides produced by Gram-positive bacteria, and they have been studied for several decades due to their antibiotic activity (Asaduzzaman & Sonomoto, 2009). More recently, it has been found that in several cases, lantibiotics act as signalling molecules that regulate their own synthesis, much like the molecules originally described as AIs decades ago. A few examples of this involve the production of mersacidin and subtilin by Bacillus (Kleerebezem, 2004), nisin production by Lactococcus lactis (Kuipers et al., 1995) and streptin production by Streptococcus pyogenes (Wescombe & Tagg, 2003). It is worth noting that many other well-known antibiotics have also been shown to act as signalling molecules (Goh et al., 2002). At subinhibitory concentrations, antibiotics can have a substantial impact on bacterial gene expression, suggesting that they can act as signals as well as antimicrobials.

Bacterial AIs can also be used for competition between species or strains. In this case they do not act as true signalling molecules but rather as signalling inhibitors. As mentioned before, quorum sensing in S. aureus involves the production of a small peptide that can control gene expression in a cell-density-dependent manner (Novick & Geisinger, 2008). This AIP can control the expression of several S. aureus exoproducts. Ji et al. (1997) showed that different strains of S. aureus produced different peptides and that peptides from one strain could inhibit gene expression by other strains. This allowed the authors to divide S. aureus strains into three agr compatibility groups, based on the activity of their autoinducing peptides. Later on, a fourth agr compatibility group was described (Jarraud et al., 2000). These findings raised the possibility that different S. aureus clones could use agr inhibition to compete against each other during the colonization of hosts. Indeed, specific agr groups can be associated with different infections caused by S. aureus (Jarraud et al., 2000, 2002; Musser et al., 1990), suggesting that quorum sensing in S. aureus may function not only as a mechanism to control gene expression but also as a tool used to eliminate competitors and allow more efficient host colonization.

The discovery that bacterial AIs have many functions that are not directly related to signalling has implications for the study of the evolution of quorum sensing. Explaining social behaviours through the lens of evolution has been a major challenge for evolutionary biologists, and this is also true for the study of microbial group behaviours (Diggle et al., 2007a; West et al., 2006). Non-signalling functions of microbial small molecules could therefore represent non-social functions from which these molecules could be borrowed during the evolution of microbial signalling. Although this hypothesis remains to be experimentally investigated, the studies described above indicate that there may be many interesting properties of microbial signals that are yet to be recognized.

Concluding remarks and future directions

In the last few decades, much has been learned about the mechanisms used by bacteria to communicate and control virulence traits. Still, new molecules and their effects on microbial virulence continue to be discovered. It is clear that the interface between quorum sensing and bacterial virulence represents a promising area from which new, effective antivirulence drugs can emerge. The examples cited here demonstrate that inhibition of virulence through inhibition of quorum sensing is possible and somewhat practical. The challenge ahead lies in translating this knowledge into real therapeutics that could boost our nearly exhausted supply of effective antibiotics.