SEARCH

SEARCH BY CITATION

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. The Vibrio fischeri LuxI/LuxR paradigm
  5. Quorum quenching
  6. LuxI-type AHL synthases
  7. LuxR-type AHL receptors
  8. Vibrio harveyi and the discovery of a bacterial esperanto
  9. AI-2 as a universal signal
  10. Conclusions
  11. Note added in proof
  12. Acknowledgements
  13. References

Cell–cell communication via the production and detection of chemical signal molecules has been the focus of a great deal of research over the past decade. One class of chemical signals widely used by proteobacteria consists of N-acyl-homoserine lactones, which are synthesized by proteins related to LuxI of Vibrio fischeri and are detected by proteins related to the V. fischeri LuxR protein. A related marine bacterium, Vibrio harveyi, communicates using two chemical signals, one of which, autoinducer-2 (AI-2), is a furanone borate diester that is synthesized by the LuxS protein and detected by a periplasmic protein called LuxP. Evidence from a number of laboratories suggests that AI-2 may be used as a signal by diverse groups of bacteria, and might permit intergeneric signalling. These two families of signalling systems have been studied from the perspectives of physiology, ecology, biochemistry, and more recently, structural biology. Here, we review the biochemistry and structural biology of both acyl-homoserine-lactone-dependent and AI-2-dependent signalling systems.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. The Vibrio fischeri LuxI/LuxR paradigm
  5. Quorum quenching
  6. LuxI-type AHL synthases
  7. LuxR-type AHL receptors
  8. Vibrio harveyi and the discovery of a bacterial esperanto
  9. AI-2 as a universal signal
  10. Conclusions
  11. Note added in proof
  12. Acknowledgements
  13. References

It has long been known that some groups of bacteria exhibit complex patterns of coordinated behaviour, and diffusible chemical signals have been either demonstrated or suspected to coordinate the individuals in these populations. The past 10 years have witnessed an explosion of interest in bacterial signalling systems, to the point that scientists have now documented a lexicon of diverse signals in many groups of microorganisms. No one knows how prevalent chemical signalling is among prokaryotes, and it has been argued that signalling is still the exception rather than the rule (Manefield and Turner, 2002). Nevertheless, many types of bacteria release at least one or a small number of signals for intraspecific communication, and in addition, there is evidence that some bacteria may also engage in intergeneric communication. These chemical signals, sometimes referred to as bacterial pheromones, are required for diverse behaviours, including bioluminescence, the horizontal transfer of DNA, the formation of biofilms, and the production of pathogenetic factors, antibiotics and other secondary metabolites (Whitehead et al., 2001). Studies of these systems have progressed rapidly, and recently have included contributions from several structural biologists. These studies have led to the X-ray crystal structures of two proteins that synthesize chemical signals, and of two receptors for these signals. This review will focus primarily on the biochemistry and structural biology of signal production and signal detection.

Reviews of bacterial signalling often state that most of the known systems fall into two classes: Gram-positive bacteria typically communicate using oligopeptide signals that are detected by two-component phosphorelay proteins (Dunny and Leonard, 1997), while proteobacteria generally signal via acyl-homoserine lactones (AHLs). However, these generalizations are overly broad, and obscure the diversity of signalling systems that has been documented. For example, the actinomycetes use a variety of γ-butyrolactone signals to regulate the production of antibiotics and other secondary metabolites (Chater and Horinouchi, 2003). At least one proteobacterium, Xanthomonas campestris, communicates using the unsaturated fatty acid cis-11-methyl-2-dodecenoic acid, and many bacterial genera were reported to release similar or identical signal molecules (Slater et al., 2000; Vojnov et al., 2001; Dow et al., 2003; Wang et al., 2004). X. campestris also appears to use a second signal that resembles γ-butyrolactones (Poplawsky and Chun, 1997). Another proteobacterium, Ralstonia solanacearum, uses a different fatty acid, palmitic acid methyl ester as a diffusible signal, which happens also to be volatile (Flavier et al., 1997a,b). Pseudomonas aeruginosa, has been reported to use 2-heptyl-3-hydroxy-4(1H)-quinolone as a signal molecule (Diggle et al., 2003). Yet another proteobacterium, Myxococcus xanthus, uses at least five chemical signals to coordinate the developmental process that leads to fruiting body formation (Shimkets, 1999; Sogaard-Andersen et al., 2003). The cyanobacterium Anabaena utilizes a peptide signal to initiate the differentation of nitrogen-fixing heterocysts (Yoon and Golden, 1998; Golden and Yoon, 2003).

The marine microbe Vibrio harveyi uses two signals to regulate bioluminescence. One is an AHL (3-hydroxybutanoyl-HSL), while the second signal is a furanone borate diester (Chen et al., 2002). This signal, designated AI-2, is released by many groups of bacteria and may be used as a signal by each (Miller and Bassler, 2001; Xavier and Bassler, 2003). If this is so, these bacteria ought to be able to communicate intergenerically using what has been called a ‘bacterial esperanto’ (Winans, 2002). However, it is important to keep in mind that not all released metabolites that affect cell physiology should qualify as cell–cell signal molecules. Paul Williams and colleagues have described four criteria that should be satisfied for a molecule to be considered a genuine chemical signal: ‘First, the production of the signal should occur during specific stages of growth, under certain physiological conditions, or in response to changes in the environment; second, the signal should accumulate extracellularly and be recognized by a specific receptor; third, accumulation of the signal should generate a concerted response once a critical threshold concentration has been reached; and fourth, the cellular response should extend beyond physiological changes required to metabolize or detoxify the signal’ (Winzer et al., 2002a). Using these criteria, these authors expressed some doubt about the role of AI-2 in signalling in organisms other than V. harveyi (Winzer et al., 2002b).

Reviewers of bacterial signalling sometimes contend that the main goal of intercellular signalling is to estimate population densities, a phenomenon sometimes referred to as autoinduction or quorum sensing (Fuqua and Winans, 1994). The idea here is that certain bacterial behaviours are appropriate only if carried out simultaneously by large numbers of bacterial cells, sometimes referred to as a quorum. In order to estimate population size, bacteria are thought to send and receive chemical signals, and to use the concentration of these signals as an indication of population size. However, this view of signalling might be overly simplistic, a point nicely made previously (Dixon, 2003). One author conjectured that these molecules might be released simply to detect and measure diffusion barriers (Redfield, 2002). According to this idea, the accumulation of a signal molecule would indicate a barrier to diffusion, and the bacterium might use this information, for example, to release hydrolytic enzymes or a matrix needed for biofilm formation. It is equally plausible that bacteria might require a quorum to effectively send and receive chemical signals, but that the goal of signalling is not to measure the quorum, but rather to coordinate the behaviour of its members (Winans and Bassler, 2002). This could explain why several organisms have multiple signalling systems. If the function of these systems is to count cell numbers, then multiple systems would seem redundant. On the other hand, if these signals help coordinate behaviour, then each system could well serve a unique purpose.

In the past 3 years, several papers have appeared that describe the high-resolution crystal structures of proteins critical for chemical communication. For example, the structure of the EsaI protein of Pantoea stewartii was published in 2002 (Watson et al., 2002). This protein represents the family of LuxI-type proteins, which direct the synthesis of AHL signals. In the same year, two groups published the structure of the TraR protein of Agrobacterium tumefaciens, which is a member of the LuxR family of AHL receptors (Vannini et al., 2002; Zhang et al., 2002). Three laboratories independently described the structure of the LuxS protein, which directs the synthesis of a signal molecule called AI-2 (Hilgers and Ludwig, 2001; Lewis et al., 2001; Ruzheinikov et al., 2001). Finally, the structure of an AI-2 receptor (LuxP) was reported (Chen et al., 2002). Importantly, that study described the chemical structure of AI-2, which had previously been elusive. Here, we will briefly review recent literature on AHL-signalling and then describe in some detail the structures of these four proteins.

The Vibrio fischeri LuxI/LuxR paradigm

  1. Top of page
  2. Summary
  3. Introduction
  4. The Vibrio fischeri LuxI/LuxR paradigm
  5. Quorum quenching
  6. LuxI-type AHL synthases
  7. LuxR-type AHL receptors
  8. Vibrio harveyi and the discovery of a bacterial esperanto
  9. AI-2 as a universal signal
  10. Conclusions
  11. Note added in proof
  12. Acknowledgements
  13. References

One of the earliest described intraspecies communication systems was identified in the bioluminescent marine symbiotic bacterium Vibrio fischeri (Nealson and Hastings, 1979). These bacteria live in symbiotic association with various species of fish or marine invertebrates, which use the luminescence for a variety of purposes. A protein called LuxI synthesizes N-3-oxohexanoyl-l-homoserine lactone (OHHL), an AHL signal molecule (Eberhard et al., 1981), sometimes termed an autoinducer, while a protein called LuxR is the autoinducer sensor as well as an autoinducer-dependent transcriptional activator of the luciferase operon (Engebrecht et al., 1983; Engebrecht and Silverman, 1984; 1987). As a population of V. fischeri cells grows, the concentration of external AHL increases as a function of cell-population density. When the autoinducer concentration reaches the micromolar range, its efflux from the cells becomes balanced by an influx, so that it can interact with LuxR. The LuxR-autoinducer complexes are thought to bind the luciferase promoter and activate transcription (Fig. 1). Therefore, this quorum-sensing regulatory circuit allows luciferase to be expressed preferentially at high population densities. A puzzling aspect of this system is how each individual V. fischeri cell avoids detecting the AHL that it itself produces. The diffusion of AHL across the envelope must be sufficiently rapid to prevent this short-circuitry. The enormous surface-to-volume ratio of bacterial cells (approaching in some cases 100 000 cm2 per cm3) must suffice for rapid diffusion of these signals across the cell envelope, although some AHLs may be actively exported by antibiotic export pumps (Evans et al., 1998; Kohler et al., 2001).

image

Figure 1. The LuxR-LuxI quorum sensing system of Vibrio fischeri. Autoinducer synthase, encoded by luxI, synthesizes N-3-oxohexanoyl-l-homoserine lactone (OHHL) from S-adenosylmethionine and acyl-ACP. OHHL diffuses across the cell envelope and accumulates intracellularly only at high population density. It binds the LuxR receptor protein, and LuxR-OHHL complexes activate transcription of the luxICDABE operon, resulting in bioluminescence.

Download figure to PowerPoint

For over 10 years the V. fischeri LuxI/LuxR quorum-sensing circuit was considered to be an isolated example of bacterial communication. However, we now understand that many proteobacteria communicate using homologous signalling circuits, as quorum-sensing systems resembling that of V. fischeri control gene expression in many other species of proteobacteria (though they have not yet been found in other groups). In virtually every case, an AHL is synthesized by a protein resembling LuxI and is detected by a protein resembling LuxR. Although LuxR/LuxI-type systems are common, the current evidence suggests that the majority of bacteria do not contain these types of regulators (Manefield and Turner, 2002). Phylogenetic analysis suggests that these proteins arose early in the evolution of proteobacteria, with functional pairs of AI synthases and AI receptors coevolving as regulatory cassettes (Gray and Garey, 2001). In many cases, these cassettes appear to have been acquired by horizontal transmission.

A number of pathogenic bacteria use AHL signals during colonization of plants and animals. Probably the best-studied example of this is P. aeruginosa, which uses two LuxI/LuxR pairs (called LasI/LasR and RhlI/RhlR) that work in tandem to regulate expression of a battery of diverse virulence factors (Passador et al., 1993; Brint and Ohman, 1995; Parsek and Greenberg, 1999; de Kievit and Iglewski, 2000). Of these, the LasI/LasR system is thought to act first, inducing the transcription of a variety of proteases and other virulence factors (Jones et al., 1993; Passador et al., 1993; Davies et al., 1998). Among the target genes of LasR is the rhlR gene (Ochsner and Reiser, 1995), and active LasR therefore induces expression of the second system. The RhlI/RhlR system further activates genes that are under LasI/LasR control, as well as another set of specific target genes (Brint and Ohman, 1995; Latifi et al., 1996; Pearson et al., 1997; Pesci et al., 1997; Hassett et al., 1999; Parsek and Greenberg, 1999; Whiteley et al., 1999). Regulation of RhlI/RhlR by LasI/LasR causes the two P. aeruginosa quorum-sensing circuits to initiate sequentially and in the proper order. A third LuxR homologue, QscR, has recently been identified from the completed genome sequence of P. aeruginosa, although a corresponding LuxI-like autoinducer synthase was not identified (Chugani et al., 2001). Genes that are controlled by LasR or RhlR have been sought by Tn5lac mutagenesis and by transcriptional profiling (Schuster et al., 2003; Wagner et al., 2003). In both studies, hundreds of genes were identified as being induced by the two AHL signals, and dozens were repressed. The expression of these genes differed widely with respect to timing and AHL specificity, suggesting that these regulatory networks are quite complex.

Signalling by AHL was recently implicated in the maturation of biofilms formed by P. aeruginosa. Biofilms are highly ordered bacterial communities that adhere to surfaces. Biofilms can be made up of single or multiple species, and generally possess aqueous channels thought to help distribute nutrients and remove waste products. Differential patterns of gene expression can be observed in distinct locations within a biofilm, suggesting that the individual members of the community have specialized roles that contribute to the survival of the bacterial consortium (Costerton et al., 1994; Costerton et al., 1995). P. aeruginosa biofilms are found in the lungs of cystic fibrosis (CF) patients, and an intact quorum-sensing circuit is necessary for proper biofilm formation by P. aeruginosa inasmuch as LasI mutants make poorly differentiated biofilms that are easily disrupted by detergents (Davies et al., 1998; Parsek and Greenberg, 1999). In addition, autoinducers can be detected in the sputum of CF patients (Erickson et al., 2002). These findings imply that quorum-sensing could be vital for the virulence of P. aeruginosa in CF patients.

Signalling also plays an important role in the biology of the plant pathogen A. tumefaciens, which causes crown gall tumors. In this bacterium, an autoinducer synthase, TraI, and an autoinducer receptor, TraR, together control the vegetative replication and conjugation of the 200 kb Ti plasmid, which is essential for pathogenicity (Piper et al., 1993; Zhang et al., 1993; Fuqua and Winans, 1994; Zhu et al., 2000; Pappas and Winans, 2003). This system functions only in or close to crown gall tumors. This is because expression of TraR requires nutrients called opines that are produced only in tumors. Close relatives of A. tumefaciens are the Rhizobia, several species of which have been shown to use AHL signalling systems. Rhizobium leguminosarum contains no fewer than four LuxI-type proteins and six LuxR-type proteins (Lithgow et al., 2000), found both on the chromosome and on the symbiotic megaplasmid that is required for nodulation and symbiotic diazotrophy. Collectively, these systems regulate genes involved in biofilm formation, conjugation, and symbiosis.

In most LuxR/LuxI-type systems, the protein resembling LuxR acts as an autoinducer-dependent transcriptional activator, but a few LuxR homologues appear to act as repressors whose activity is inhibited by the cognate autoinducer. The best characterized of these systems is the EsaR/EsaI system of P. stewartii, which is a vascular pathogen of maize. The regulator EsaR represses its own transcription in the absence of the signal molecule produced by EsaI (3-oxo-C6-HSL), and expression becomes derepressed by the autoinducer (Minogue et al., 2002). This result indicates that the signal inactivates the EsaR protein rather than activating it, which is fundamentally different from most LuxR-type proteins. EsaR was recently shown to be capable of acting as an activator in a heterologous system, but even here, the autoinducer inhibited rather than stimulated its activity (von Bodman et al., 2003). The LuxR and TraR proteins are also able to act as repressors of artificial promoters containing their binding sites positioned near the transcription start site (Luo and Farrand, 1999; Egland and Greenberg, 2000). As might be predicted, repression requires AHLs.

Quorum quenching

  1. Top of page
  2. Summary
  3. Introduction
  4. The Vibrio fischeri LuxI/LuxR paradigm
  5. Quorum quenching
  6. LuxI-type AHL synthases
  7. LuxR-type AHL receptors
  8. Vibrio harveyi and the discovery of a bacterial esperanto
  9. AI-2 as a universal signal
  10. Conclusions
  11. Note added in proof
  12. Acknowledgements
  13. References

Given the importance of signalling to bacteria, it should not be surprising that host organisms or competing bacteria might benefit from disrupting the production or detection of these signals, and various laboratories have reported these activities. For example, an isolate of Bacillus was identified that produces an enzyme called AiiA, which hydrolyzes the lactone ring of AHLs, destroying their biological activity (Dong et al., 2001; 2002). The hypothesis that this enzyme could block signalling was tested using Erwinia carotovora, a plant pathogen that uses quorum sensing to synthesize pectinases and other hydrolytic enzymes and also to regulate antibiotic production. Expression of AiiA by this bacterium blocked the production of these enzymes and attenuated virulence (Dong et al., 2000). Transgenic tobacco plants expressing AiiA were protected against E. carotovora infection (Dong et al., 2001), and P. aeruginosa expressing AiiA was deficient in biofilm formation and showed attenuated pathogenicity. In a separate study, the soil bacterium Variovorax paradoxus was shown to degrade AHLs via an aminoacylase that hydrolyzes the fatty acyl group, and to use these compounds as sole sources of carbon and nitrogen (Leadbetter and Greenberg, 2000). The red alga Delisa pulchra releases halogenated furanones that can block AHL-mediated signalling by blocking association between AHLs and their receptors, and by accelerating the proteolysis of these receptors (Manefield et al., 1999; 2001; 2002; Rasmussen et al., 2000; Hentzer et al., 2002). Similarly, higher plants release compounds that can either stimulate or inhibit bacterial signalling (Bauer and Robinson, 2002).

LuxI-type AHL synthases

  1. Top of page
  2. Summary
  3. Introduction
  4. The Vibrio fischeri LuxI/LuxR paradigm
  5. Quorum quenching
  6. LuxI-type AHL synthases
  7. LuxR-type AHL receptors
  8. Vibrio harveyi and the discovery of a bacterial esperanto
  9. AI-2 as a universal signal
  10. Conclusions
  11. Note added in proof
  12. Acknowledgements
  13. References

Several LuxI-like enzymes have been studied at the biochemical level. These proteins produced AHLs from the substrates S-adenosyl-L-methionine (SAM) and acylated acyl carrier protein (acyl-ACP, see Fig. 2) (Moréet al., 1996; Val and Cronan, 1998; Parsek et al., 1999). AHL synthesis is proposed to occur via a ‘bi-ter’ (two-substrate, three-product) reaction mechanism (Parsek et al., 1999). Studies using reaction inhibitors suggested that the synthase first binds the SAM and subsequently binds acyl-ACP. At some point during catalysis, the acyl group is donated from acyl-ACP to the amine of SAM, and apo-ACP is released. The product acyl-HSL is then released, followed by the release of methylthioadenosine (MTA). During AHL synthesis, the acyl chain is provided as a thioester of the phosphopantetheine prosthetic group of ACP, whose 1-carbonyl carbon is attacked by the amine of SAM, resulting in a bound acyl-SAM intermediate. Formation of the homoserine lactone ring occurs by nucleophilic attack on the gamma carbon of SAM by its own carboxylate oxygen (Parsek et al., 1999). Both acyl-ACP and SAM are used in rather novel ways in these reactions, inasmuch as acyl-ACP is normally used in lipid biosynthesis, and SAM is normally used as a methyl donor, and less often as a homoserine donor (Moréet al., 1996). In a sense, the lactonization step resembles the use of SAM in this latter role, except that the nucleophile is provided intramolecularly. A second family of AHL synthases has been described whose members do not show any protein sequence similarity to the LuxI family (Hanzelka et al., 1999). Members of this family use the same substrates as LuxI-type proteins, although acyl-CoA can substitute for acyl-ACP (Hanzelka et al., 1999).

image

Figure 2. Structure of the active site of EsaI of Pantoea stewartii, an AHL synthase. A. With the 3-oxo-hexanoyl group modeled in the structure (in azure). B. Shows the key residues predicted to bind SAM and acyl-ACP, the predicted reaction intermediates, and the three products, OHHL, holo-ACP, and methyl-thioadenosine (MTA). Adapted with permission from Watson et al. (2002).

Download figure to PowerPoint

AHL synthases from different bacterial species produce AHLs that vary over a broad range in acyl chain length from four to 18 carbons, as well as in the oxidation state at C3 (either unsubstituted or substituted with hydroxy or oxo groups) (Fuqua and Eberhard, 1999; Teplitski et al., 2003). A small number of AHLs also have monounsaturated acyl groups (Kuo et al., 1994; Fuqua and Eberhard, 1999). The choice of fatty acyl substrates is partly due to substrate preference of the enzyme and to substrate availability (Fray et al., 1999; Fuqua and Eberhard, 1999). LuxI-type proteins share four blocks of conserved sequence (Fuqua and Eberhard, 1999). Within these blocks eight residues are absolutely conserved. When mutated, the most conserved residues severely impaired catalysis (Hanzelka et al., 1997; Parsek et al., 1997).

The structure of one LuxI-type protein, EsaI of P. stewartii, has been solved using X-ray crystallography (Watson et al., 2002). This protein, which synthesizes primarily 3-oxo-hexanoyl-HSL (OHHL), was confirmed to be monomeric and approximately spherical with a rather deep cleft that was implicated in substrate binding. A loop near the N-terminus was unstructured in the crystal, but has several highly conserved residues, suggesting that this loop plays a critical role in catalysis. This region may become more structured upon binding of substrate. All of the highly conserved residues lie either in the cleft or at the unstructured N-terminus.

The protein EsaI shows structural similarity (though no sequence homology) to the GCN5-related N-acetyl transferases (GNAT) family of proteins (Watson et al., 2002). In particular, the highly twisted beta sheet of EsaI resembles those of GNAT proteins, which bind acetyl-CoA and carry out the acetylation of lysine residues of diverse proteins. EsaI and GNAT proteins carry out similar reactions, since both cause the primary amine of one substrate (SAM or lysine residues) to attack the C1 carbon of the other substrate (acetyl-CoA or acyl-ACP). Based on this structural resemblance, the acyl group of the acyl-ACP was modeled into a cleft of EsaI (Fig. 2) (Watson et al., 2002). The pocket was long enough to accommodate a 6-carbon acyl chain, but not a longer one. It also contained hydroxylic residues predicted to contact the 3-oxo group of the 3-oxo-hexanoyl-ACP. Threonine residue 140 was predicted to hydrogen-bond with the 3-oxo group of the acyl chain. Mutation of this residue to alanine caused the protein to synthesize the unsubstituted hexanoyl-HSL rather than the 3-oxo-hexanoyl-HSL (Watson et al., 2002). This important result confirmed that this channel accommodates the acyl chain, and reveals a determinant of substrate specificity. Proteins that synthesize 3-oxo or 3-hydroxy AHLs generally have threonine or serine residues at this position, while proteins that synthesize 3-unsubstituted AHLs generally have small non-polar residues at that position.

LuxR-type AHL receptors

  1. Top of page
  2. Summary
  3. Introduction
  4. The Vibrio fischeri LuxI/LuxR paradigm
  5. Quorum quenching
  6. LuxI-type AHL synthases
  7. LuxR-type AHL receptors
  8. Vibrio harveyi and the discovery of a bacterial esperanto
  9. AI-2 as a universal signal
  10. Conclusions
  11. Note added in proof
  12. Acknowledgements
  13. References

Early efforts to purify LuxR-type proteins in an active form were hampered by low protein solubility. It was discovered that the TraR protein of A. tumefaciens, when overproduced in Escherichia coli, forms insoluble inclusion bodies, but that when the same strain was cultured in the presence of the cognate AHL 3-oxo-octanoyl-HSL (OOHL), most of the protein was recovered in a soluble form (Zhu and Winans, 1999; Pappas and Winans, 2003). This protein was fully able to bind to DNA with high specificity for suspected TraR-binding sites (Luo and Farrand, 1999; Zhu and Winans, 1999). The protein was also able to activate transcription of two divergent promoters that flank one binding site, but only when the DNA template was supercoiled (Zhu and Winans, 1999). Exogenous OOHL was not required for DNA-binding or for transcription reactions, indicating that the OOHL that copurified with the protein was bound virtually irreversibly.

TraR-OOHL complexes were eluted from a gel filtration column as dimers and bound DNA as dimers (Qin et al., 2000; Zhu and Winans, 2001). The N-terminal half of the protein is sufficient for OOHL-binding and for dimerization, since TraR fragments containing just this domain are able to bind and inactivate the full-length protein (Zhu and Winans, 1998; Qin et al., 2000; Chai et al., 2001; Luo et al., 2003). The C-terminal domain of TraR also contributes to dimerization (Qin et al., 2000). Several amino acid residues in both domains may interact directly with RNA polymerase in that they are required for transcription activation, but are not required for protein folding, OOHL binding, or DNA binding (Luo and Farrand, 1999; Luo et al., 2003). Another LuxR-type protein, the CarR protein of Erwinia carotovora, also bound two molecules of AHL per dimer, and AHL binding converted CarR from a dimer to a higher order oligomer (Welch et al., 2000).

LuxR was recently purified in a soluble form by adding its AHL (OHHL) to cultures overexpressing this protein (Urbanowski et al., 2004). The protein bound with high specificity to the predicted LuxR-binding site, and recruited RNA polymerase to the adjacent promoter. This required addition of OHHL, indicating that LuxR-OHHL complexes can be dissociated by dilution, and that bound OHHL is in equilibrium with the unbound form. This is quite unlike TraR-OOHL complexes, since OOHL binds TraR virtually irreversibly (Zhu and Winans, 1999; 2001). A third LuxR homologue, RhlR of P. aeruginosa, was reported to bind a target promoter in the presence or absence of the cognate AHL, although that promoter was activated only in the presence of the AHL (Medina et al., 2003). In a separate study, a RhlR-EGFP fusion protein was studied in vivo by fluorescence anisotropy (Ventre et al., 2003). This fusion protein was dimeric in the presence or absence of the cognate autoinducer, but was dissociated to monomers by a heterologous AHL.

In one study, evidence was provided that at least a fraction of apo-TraR is membrane associated, and it was argued that AHL may cause release of TraR from the membrane and allows it to dimerize (Qin et al., 2000). However, TraR was overexpressed in those studies, and it was never demonstrated directly that AHL causes release of membrane-bound TraR subunits. In the authors’ laboratory, it was shown that TraR is dramatically stabilized against proteolysis by OOHL. Apo-TraR has a half-life of only 2 min in vivo, while TraR-OOHL complexes has a half-life at least 20-fold longer (Zhu and Winans, 1999). Consequently, apo-TraR is not detectable by Western blots when expressed at native levels, though TraR-OOHL complexes are readily detected (Chai and Winans, 2004). Similar studies were done in E. coli with similar results, and it was found that both Clp and Lon proteases contribute to TraR turnover (Zhu and Winans, 2001). In pulse-chase experiments, OOHL protected TraR against proteolysis only when it was added before the radiolabel (Zhu and Winans, 2001). This indicates that TraR can productively bind OOHL only during synthesis on polysomes, and that full-length apo-TraR proteins are not functional OOHL receptors. Purified apo-TraR was rapidly degraded by trypsin to oligopeptides, whereas TraR-OOHL complexes were more resistant to trypsin and were cleaved at discrete interdomain linkers, indicating that TraR requires OOHL to attain its mature tertiary structure.

The structure of the TraR protein of A. tumefaciens was solved independently in two laboratories, which published extremely similar structures (Vannini et al., 2002; Zhang et al., 2002). Both structures were dimers of TraR, each subunit of which contained a single molecule of OOHL. The TraR dimer also bound a 20-nucleotide self-complementary DNA fragment containing an 18-nucleotide canonical TraR-binding site (Fig. 3). The bound DNA was slightly bent in the crystal, and this DNA bend was independently confirmed using gel retardation assays (Pappas and Winans, 2003). As predicted, the N-terminal domain of each TraR monomer binds the AHL molecule, while the C-terminal domain contains a bundle of four helices that binds to half of a tra box (Vannini et al., 2002; Zhang et al., 2002). The TraR DNA-binding domains in the dimer bears considerable structural homology to that of the E. coli NarL regulator (Maris et al., 2002), which also binds DNA as a dimer. The N-terminal and C-terminal domains of TraR are connected by a flexible 12-amino acid linker. The AHL-binding domains provide a significant dimerization interface, while the DNA-binding domains add additional dimer interactions.

image

Figure 3. Structure of the TraR protein of Agrobacterium tumefaciens, an AHL receptor. A. Shows the N-terminal domain consisting of an alpha-beta-alpha sandwich with the AHL molecule lying between the beta sheet and several amphipathic alpha helices. The AHL is therefore completely engulfed within the receptor with no exposure to bulk solvent, and contributes to the hydrophobic core of the protein. B. Shows the C-terminal domain of TraR, consisting of four helix bundle, complexed with a DNA fragment containing the tra box sequence. This DNA sequence is decoded primarily by residues Arg206 and Arg210.

Download figure to PowerPoint

Interestingly, the AHL molecule is fully embedded within the TraR protein and shows no significant contact with bulk solvent (Vannini et al., 2002; Zhang et al., 2002). The OOHL-binding domain has an alpha-beta-alpha structure with the AHL embedded between a layer of alpha helix and a beta sheet (Fig. 3). Earlier studies suggested that TraR requires the AHL molecule as a scaffold that helps the protein to acquire a protease-resistant tertiary structure (see above). The authors of one of these studies proposed that the autoinducer indirectly affects gene activation by increasing the stability of TraR and the formation of active dimers. The dimers are then predisposed to recognize specific TraR-binding sites and activate transcription. The recent structural studies seem to confirm the key role of the AHL in the correct folding of the nascent TraR protein (Vannini et al., 2002; Zhang et al., 2002). Another interesting finding of the structural work is the overall asymmetry of the dimer complex. The TraR dimer displays a two-fold axis of symmetry in each of its two domains, but the two axes of symmetry of these domains lie at a 90° angle to each other (Vannini et al., 2002; Zhang et al., 2002). It is possible that TraR shows the same static asymmetry in vivo, but it is equally likely that there is considerable flexibility between the two TraR domains in vivo, which might contribute to its ability to activate divergent promoters.

Vibrio harveyi and the discovery of a bacterial esperanto

  1. Top of page
  2. Summary
  3. Introduction
  4. The Vibrio fischeri LuxI/LuxR paradigm
  5. Quorum quenching
  6. LuxI-type AHL synthases
  7. LuxR-type AHL receptors
  8. Vibrio harveyi and the discovery of a bacterial esperanto
  9. AI-2 as a universal signal
  10. Conclusions
  11. Note added in proof
  12. Acknowledgements
  13. References

Although most known examples of bacterial signalling occur within a single species, Bassler and colleagues several years ago discovered that a chemical signal used by Vibrio harveyi, called AI-2, is released by many groups of eubacteria (Bassler et al., 1997). This raises the possibility that all these bacteria might use AI-2 as a signal, and that there is a universal chemical lexicon by which bacteria can communicate intergenerically.

AI-2 is only one of two chemical signals used by the luminescent marine bacterium V. harveyi (Bassler et al., 1993; 1994). The other signal, AI-1 (3-hydroxybutanoyl-HSL), is an AHL-type signal (Cao and Meighen, 1989). Until recently, the chemical structure of AI-2 was a mystery (see below). V. harveyi does not employ a typical LuxI/LuxR-type quorum-sensing cascade. Instead, each signal is detected by a membrane-spanning two-component kinase (LuxN for AI-1, and LuxQ for AI-2) (Bassler et al., 1993; 1994). Additionally, a soluble periplasmic protein, LuxP, which resembles several periplasmic ribose-binding proteins, is required for detection of AI-2 (Bassler et al., 1994). This homology seems appropriate, since AI-2 is now known to resemble ribose (see below) (Schauder et al., 2001). Both LuxN and LuxQ are hybrid proteins containing a sensor kinase domain and a response regulator domain that funnel phosphoryl groups via LuxU to LuxO, a σ54-dependent transcriptional activator that is hypothesized to control the expression of an unidentified repressor of luxR, whose product directly activates the luciferase structural operon (luxCDABE) (Bassler et al., 1994; Freeman and Bassler, 1999; Lilley and Bassler, 2000). Note that LuxR of V. harveyi does not resemble LuxR of V. fischeri. In the absence of signal, both proteins act as kinases, and the resulting phosphoryl-LuxO blocks luminescence. In the presence of signals, the proteins act as phosphatases, and the ultimate dephosphorylation of LuxO stimulates luminescence. Both signals are needed for bioluminescence, because in the absence of one signal, the cognate receptor acts as a kinase to block lux gene expression (Mok et al., 2003).

For a number of years, the synthesis and structure of AI-2 were mysterious, beyond the fact that the AI-2 receptor resembled the ribose-binding protein, suggesting that AI-2 might resemble ribose. In 1999, the luxS gene was identified in V. harveyi and a very similar gene was described in Salmonella typhimurium (Surette et al., 1999). Even then the chemical identity of AI-2 remained elusive. It was subsequently found that LuxS catalyzes a step in the turnover of SAM (Schauder and Bassler, 2001; Schauder et al., 2001). When SAM is used as a methyl donor, S-adenosyl-homocysteine (SAH) is generated as one of the products. The adenosyl group of SAH is removed by the Pfs protein, generating S-ribosyl-homocysteine (SRH). The LuxS protein converts SRH to homocysteine, which can be scavenged, and a molecule derived from ribose called 4,5-dihydroxy-2,3-pentanedione (DPD), which cyclizes spontaneously to and from a compound called pro-AI-2 (Fig. 4). It is striking that both pro-AI-2 and AHL-type signals are derived from SAM. The fact that pro-AI-2 is derived from ribose supports the earlier observation that the AI-2 receptor resembles the ribose receptor (see above). However, the precise structures of both pro-AI-2 and AI-2 still remained elusive.

image

Figure 4. Synthesis and detection of AI-2 by V. harveyi. AI-2 is synthesized from S-adenosyl-homocysteine, a byproduct of SAM metabolism. Pfs removes the adenine group, generating S-ribosyl-homocysteine, which is converted by LuxS to homocysteine and pro-AI-2. Extracellular pro-AI-2 binds borate, forming AI-2, which binds to the periplasmic LuxP protein, which then signals the LuxQ two-component kinase/phosphatase. This ultimately leads to expression of the Iux operon and to bioluminescence. In this simplified figure we have omitted LuxU (which transfers phosphoryl groups from LuxQ and LuxN to LuxO), and LuxR (the direct activator of the luxCDABEGH operon).

Download figure to PowerPoint

The structures of LuxS orthologues from Bacillus subtilis, Helicobacter pylori, Deinococcus radiodurans, and Hemophilus influenzae have been determined by X-ray crystallography (Hilgers and Ludwig, 2001; Lewis et al., 2001; Ruzheinikov et al., 2001). In one study, the structure of the B. subtilis protein was determined in three forms, (i) as an apoprotein, (ii) as a complex with the substrate S-ribosyl-homocysteine and (iii) as a complex with the product homocysteine (Ruzheinikov et al., 2001). All four LuxS-type proteins were homodimers with one zinc atom per subunit located at or near the active site. Zinc atoms were coordinated by two His residues and one Cys residue, all contributed by the same subunit (Fig. 5). The tri-coordinate binding and the position of the zinc atoms strongly suggested a role in catalysis. The zinc atom forms hydrogen bonds with the O2 and O3 oxygen atoms of SRH, quite distant from the homocysteine moiety, so it is not clear exactly what role this metal plays. However, in a separate study of the B. subtilis LuxS protein, it was found that iron-saturated LuxS was 10-fold more active than zinc-saturated LuxS (Zhu et al., 2003), suggesting that the zinc atoms found in the LuxS crystals could be artefacts of crystallography.

image

Figure 5. Structure of the LuxS protein of B. subtilis. A. The substrate S-ribosyl-homocysteine is modeled within a cavity with this dimeric protein, which binds one zinc atom per subunit (purple). Proposed hydrogen bonding between LuxS and its substrate are shown with dashed lines. B. Shows this substrate and zinc atom engulfed within the LuxS active site (reprinted with permission from Ruzheinikov et al., 2001 and Lewis et al., 2001).

Download figure to PowerPoint

The chemical structure of AI-2 was finally solved by X-ray crystallography of the ligand bound form of the AI-2 receptor (LuxP) (Chen et al., 2002). As expected, LuxP structurally resembles the family of periplasmic ribose-binding proteins, and possesses two domains linked by a three-stranded hinge and a deep cleft containing one molecule of AI-2 (Chen et al., 2002). Also as predicted, LuxP contains a cyclized carbohydrate that resembles ribose. Unexpectedly, the LuxP structure contained additional unassigned electron density, showing an atom that formed a diester with the carbohydrate moiety. This atom was proposed to be boron, indicating that AI-2 is a furanosyl borate diester (Fig. 6). Although boron is difficult to distinguish from carbon by electron density alone, a carbon atom at this position would be extremely unstable, while boron would be more stable. This surprising finding was confirmed by using 11B NMR, by electrospray ionization mass spectrometry, and by showing that addition of borate to whole cells stimulates this signalling pathway (Chen et al., 2002). AI-2 bears no resemblance to any previously characterized bacterial pheromone. Boron is an abundant component of seawater, and so it should be available to the bacteria. It remains to be determined whether pro-AI-2 and borate form a complex before binding LuxP, or whether pro-AI-2 and borate can bind LuxP independently and sequentially.

image

Figure 6. Structure of AI-2 bound to the LuxP receptor protein. The bottom ring of AI-2 contains five carbon atoms that are derived from the ribose moiety of SAM, while the top ring is a borate di-ester that was found unexpectedly in the LuxP-binding site (reprinted with permission from Chen et al. (2002).

Download figure to PowerPoint

AI-2 as a universal signal

  1. Top of page
  2. Summary
  3. Introduction
  4. The Vibrio fischeri LuxI/LuxR paradigm
  5. Quorum quenching
  6. LuxI-type AHL synthases
  7. LuxR-type AHL receptors
  8. Vibrio harveyi and the discovery of a bacterial esperanto
  9. AI-2 as a universal signal
  10. Conclusions
  11. Note added in proof
  12. Acknowledgements
  13. References

Perhaps the most striking finding about AI-2 is that it is synthesized by so many bacterial genera. The luxS gene is highly conserved in most groups of bacteria except the alpha proteobacteria (Surette et al., 1999; Miller and Bassler, 2001). Many of the bacteria that contain a luxS gene have been shown to produce AI-2, and in every case tested, mutagenesis of luxS eliminated AI-2 production (Sperandio et al., 1999; Joyce et al., 2000; Day and Maurelli, 2001; Lyon et al., 2001; Taga et al., 2001; Kim et al., 2003). Furthermore, purified Pfs and LuxS enzymes from diverse bacteria were used to synthesize AI-2 from SAH in vitro. Homocysteine and signalling molecules of identical specific activity were produced from each enzyme (Schauder et al., 2001), indicating that AI-2 molecules from diverse bacteria are identical.

AI-2 appears to have diverse roles in signalling. Pathogenic E. coli strains mutated in luxS show decreased expression of a Type III protein translocation system (Sperandio et al., 1999), and in transcriptional profiling experiments, luxS mutations alter the expression of large numbers of E. coli genes (Sperandio et al., 1999; DeLisa et al., 2001). These luxS mutations also affect hemin acquisition genes in Porphyromonas gingivalis (Chung et al., 2001), the expression of the VirB virulence factor in Shigella flexneri (Day and Maurelli, 2001), and the secretion of the SpeB cysteine protease virulence determinant of Streptococcus pyogenes (Lyon et al., 2001). However, in a recent study, addition of purified AI-2 to an EHEC E. coli strain did not restore expression of the Type III protein translocation system, and a new signalling molecule, AI-3, was described. It is important to keep in mind that these luxS mutations may significantly perturb cellular physiology, which could affect patterns of gene expression (Winzer et al., 2002a,b).

The most comprehensive genetic studies of AI-2 detection in bacteria other than V. harveyi come from the Bassler laboratory, working on S. typhimurium. Extensive screens for AI-2 inducible genes in S. typhimurium yielded just one operon, whose products direct AI-2 uptake and metabolism (Taga et al., 2001). This operon contains seven genes, denoted as lsrACDBFGE. Of these, the first four resemble the ABC-type uptake system for ribose transport, and the remaining genes are thought to play roles in AI-2 metabolism. Another linked gene encodes an AI-2 kinase (Taga et al., 2003). The products of the lsrACDBFGE operon allow carbon-limited bacteria to release very little AI-2 and to scavenge it from the growth medium, possibly for use as a nutrient. Expression of the lsr operon is induced by AI-2, and regulated by the divergently transcribed lsrR gene (Taga et al., 2003). Therefore, the detection of AI-2 by S. typhimurium is fundamentally different from its detection by V. harveyi.

Although many bacteria release AI-2 and at least some bacteria use it as a signal, it is unclear whether all the bacteria that make AI-2 actually use it for signalling. The available evidence suggests that AI-2 may be both an intercellular signal and a metabolic waste product, or to have properties of both. AI-2 synthesis is obligately and stoichiometrically linked to the utilization of SAM and so cannot be made solely in response to some environmental signal. However, even if AI-2 is partly a waste product, this does not preclude a role in signalling. Metabolites could well act as signals, since they accumulate at high population densities and provide information about the size and metabolic status of a population of cells. One argument that AI-2 is made for the purpose of signalling is that its production seems rather wasteful. Some groups of bacteria can process SAH to homocysteine and adenosine rather than homocysteine, adenine, and AI-2. The former reaction seems more efficient, and the latter reaction may therefore have evolved precisely for the purpose of releasing signals (Bassler, 1999).

Conclusions

  1. Top of page
  2. Summary
  3. Introduction
  4. The Vibrio fischeri LuxI/LuxR paradigm
  5. Quorum quenching
  6. LuxI-type AHL synthases
  7. LuxR-type AHL receptors
  8. Vibrio harveyi and the discovery of a bacterial esperanto
  9. AI-2 as a universal signal
  10. Conclusions
  11. Note added in proof
  12. Acknowledgements
  13. References

We are only beginning to glimpse the chemical polyphony of bacterial communication. Here we have reviewed pheromone-dependent signalling only in proteobacteria, although signalling in other groups of bacteria is at least as complex. While most bacterial signals are thought to act intraspecifically, others may act in intergeneric signalling, although it is far from clear what selective benefit bacteria would gain from intergeneric communication. Furthermore, bacteria and higher organisms seem capable of blocking these signals, either by metabolizing them or by releasing inhibitory molecules, though it is difficult to gauge whether these systems really evolved for that goal. Undoubtedly, the recorded examples of signalling and signal blocking are just the tip of the iceberg. This review has emphasized the biochemistry and structural biology of proteobacterial signalling. While the four structures described here are informative, they provide static images of these proteins rather than dynamic ones, and more will be learned by applying other methods of structural biology, including protein NMR. These sorts of studies are in their infancy, and will provide a far more robust picture of these proteins in the next five to 10 years. The number of AHL-type systems that have been intensively studied is extremely small, and biochemical and structural studies need to be done, as these proteins are sure to have diverse properties. For example, we already know that some LuxR-type proteins are antagonized rather than activated by their cognate AHLs, and more surprises such as this are bound to be uncovered. Studies of AI-2 signalling are also just beginning. Perhaps the most important unanswered question about AI-2 is whether all the bacteria that make it actually use it in signalling. For those that do, we know virtually nothing about how this signal is detected, beyond the work done in V. harveyi and S. typhimurium, which detect AI-2 by two completely different mechanisms. Clearly, more work will need to be done to unravel these questions.

Note added in proof

  1. Top of page
  2. Summary
  3. Introduction
  4. The Vibrio fischeri LuxI/LuxR paradigm
  5. Quorum quenching
  6. LuxI-type AHL synthases
  7. LuxR-type AHL receptors
  8. Vibrio harveyi and the discovery of a bacterial esperanto
  9. AI-2 as a universal signal
  10. Conclusions
  11. Note added in proof
  12. Acknowledgements
  13. References

The X-ray crystal structure CprB of Streptomyces coelicolor was recently described. CprB is composed of a ligand-binding domain and a DNA-binding domain. The regulatory domain contains a hydrophobic cavity, which probably serves as a binding site for gramma-butyrolactone-type signal molecules. Binding of these molecules is thought to trigger the release of this repressor protein from operator DNA (J Mol Biol336: 409–419).

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. The Vibrio fischeri LuxI/LuxR paradigm
  5. Quorum quenching
  6. LuxI-type AHL synthases
  7. LuxR-type AHL receptors
  8. Vibrio harveyi and the discovery of a bacterial esperanto
  9. AI-2 as a universal signal
  10. Conclusions
  11. Note added in proof
  12. Acknowledgements
  13. References

The authors thank the members of our laboratory for critical reading of this manuscript. K. M. P and C. L. W. contributed equally to this review. Research in the author's laboratory is supported by grants from the National Institutes of Health (GM42893) and from the National Science Foundation (MCB-9904917).

References

  1. Top of page
  2. Summary
  3. Introduction
  4. The Vibrio fischeri LuxI/LuxR paradigm
  5. Quorum quenching
  6. LuxI-type AHL synthases
  7. LuxR-type AHL receptors
  8. Vibrio harveyi and the discovery of a bacterial esperanto
  9. AI-2 as a universal signal
  10. Conclusions
  11. Note added in proof
  12. Acknowledgements
  13. References