Cell, Vol.

109, 421424, May 17, 2002, Copyright 2002 by Cell Press

Small Talk: Cell-to-Cell Communication in Bacteria

Minireview

In a process called quorum sensing, groups of bacteria communicate with one another to coordinate their behavior and function like a multicellular organism. A diverse array of secreted chemical signal molecules and signal detection apparatuses facilitate highly productive intra- and interspecies relationships.

How Bacteria Talk to Each Other: Quorum Sensing One type of bacterial cell-cell communication is referred to as quorum sensing. Quorum sensing-controlled behaviors are those that only occur when bacteria are at high cell population densities. These behaviors are ones that are unproductive when undertaken by an individual bacterium but become effective by the simultaneous action of a group of cells. For example, quorum sensing regulates bioluminescence, virulence factor expression, biofilm formation, sporulation, and mating. Quorum sensing is achieved through the production, release, and subsequent detection of and response to threshold concentrations of signal molecules called autoinducers. The accumulation of a stimulatory concentration of an extracellular autoinducer can only occur when a sufficient number of cells, a quorum, is present. Thus, the process is proposed to be a mechanism for census taking. There are three archetypal quorum sensing systems (Figure 1; for a review see Miller and Bassler, 2001). A typical gram-negative bacterial quorum sensing circuit is shown in Figure 1A. In this type of system, the autoinducer is an acylated homoserine lactone (AHL) synthesized by a LuxI-type enzyme. Cytoplasmically synthesized autoinducer diffuses passively through the bacterial membrane and accumulates both intra- and extracellularly in proportion to cell density. When the stimulatory concentration of the AHL is achieved, a LuxR-type protein binds it. LuxR-AHL complexes bind to promoters of quorum sensing-regulated target genes and activate transcription. Over 50 species of gramnegative bacteria produce AHLs that differ only in the acyl side chain moiety, and each LuxR-type protein is highly selective for its cognate AHL signal molecule. Additional complexity exists in many of these circuits, such as the use of multiple AHL autoinducers and LuxR proteins that can act either in parallel or in series. Figure 1B shows the paradigm quorum sensing circuit of a gram-positive bacterium. The autoinducers are short, usually modified peptides processed from precursors. The signals are actively exported out of the cell, and they interact with the external domains of membrane bound sensor proteins. Signal transduction oc1

Correspondence: bbassler@[Link]

curs by a phosphorylation cascade that culminates in the activation of a DNA binding protein that controls transcription of target genes. Specificity exists because each sensor protein is highly selective for a given peptide signal. Similar to gram-negative bacteria, gram-positive bacteria can use multiple autoinducers and sensors. Some peptide autoinducers act exclusively from the outside, while others elicit a specific set of gene expression changes from the outside and are also transported back into the cell where they trigger a different set of behavioral changes. The final model system shown (Figure 1C) is that of the gram-negative bacterium Vibrio harveyi. This quorum sensing circuit controls bioluminescence. V. harveyi produces two autoinducers termed HAI-1 and AI-2. HAI-1 is a typical gram-negative-like AHL, although its synthesis is not dependent on a LuxI-like enzyme. The second autoinducer, AI-2, is unexpectedly a furanosyl borate diester. HAI-1 and AI-2 signal transduction occurs via a gram-positive-like phosphorylation cascade. Critical for AI-2 signal transduction is the soluble periplasmic AI-2 binding protein LuxP. Linguistic Terminology: Autoinducer Biosynthesis AHL autoinducers are derived from S-adenosyl methionine (SAM) (Figure 2). The LuxI-type enzymes couple the acyl group from a specific acylated acyl carrier protein to the methionine on SAM. Lactonization of the intermediate results in the production of the AHL autoinducer. The byproduct of the reaction, methylthioadenosine (MTA), is toxic. The nucleosidase enzyme, Pfs, cleaves MTA to two nontoxic products, adenine and methylthioribose (MTR). The subsequent fate of MTR is unknown. Some gram-negative bacteria have AHL synthases that are unrelated in sequence to the LuxI-enzymes although they are proposed to catalyze an identical reaction. These include LuxLM from V. harveyi (Figure 1C), AinS from Vibrio fischeri, and HdtS from Pseudomonas fluorescens. Examples of AHL autoinducers are shown in Figure 3A. The structure of the LuxI-type enzyme EsaI from the plant pathogen Pantoea stewartii was recently solved by X-ray crystallography (Watson et al., 2002). Alignment of the protein sequence of EsaI to other LuxI-type enzymes revealed that the most highly conserved amino acids are situated on one face of the protein in an apparent active-site cleft. Prior to this study, it was assumed that AHLs are synthesized through the formation of a covalent enzyme-substrate intermediate involving an active-site cysteine. Interestingly, the EsaI structure shows that no cysteine residue exists in the vicinity of the substrate. EsaI has structural similarity to phosphopantetheine binding N-acetyltransferase proteins (GNAT family), and they share a common catalytic core. Based on the GNAT structure and enzymatic mechanism, residues predicted to be key for catalysis in EsaI were identified and confirmed by site-directed mutagenesis. These studies suggest a new mechanism for AHL synthesis in which the first step is N-acylation, which occurs by nucleophilic attack on the 1-carbonyl of the acyl-ACP by the amine of SAM. Lactonization follows by nucleophilic

Cell 422

Figure 1. Three Types of Quorum Sensing Systems (A) Gram-negative bacteria. (B) Gram-positive bacteria. (C) The gram-negative bacterium V. harveyi. The open circle represents the AI-2 binding protein LuxP (see text). In (B) and (C), the P shows that signal relay is by phosphorylation. H and D denote histidine and aspartate, which are the phosphorylated residues on the signaling proteins.

attack on the carbon of SAM by the SAM carboxylate oxygen, and this results in production of the specific AHL and MTA. Oligopeptide autoinducers are synthesized by cleavage from longer precursor peptides. Representative oligopeptide autoinducers are shown in Figure 3B. Some peptide signals contain side chain modifications that include lactone or thio-lactone rings and other as yet undefined hydrophobic moieties. The machinery responsible for processing and modification is undefined, so it is not known how oligopeptides are created that contain biologically crucial but chemically minimal differences. A common feature in oligopeptide production is the requirement for a dedicated ATP binding cassette (ABC)

transporter for export of the signal. Additionally, the genes encoding the precursor peptide, the signal detection apparatus, and the transport machinery are usually located adjacent to one another. Similar to AHLs, AI-2 is produced from SAM. AI-2 production requires three enzymatic steps (Figure 2; Schauder et al., 2001). Utilization of SAM as a methyl donor produces S-adenosylhomocysteine (SAH). Like MTA, SAH is toxic. Pfs detoxifies SAH by hydrolyzing it to adenine and S-ribosylhomocysteine (SRH). SRH is the substrate on which LuxS acts to produce AI-2. LuxS converts SRH to 4,5-dihydroxy-2,3-pentanedione (DPD) and homocysteine. DPD cyclizes and undergoes additional rearrangements to yield the active AI-2 species. Until recently, the chemical structure of AI-2 remained unknown (discussed below). Understanding the Vernacular: Intra- and Interspecies Cell-Cell Communication In natural habitats, bacteria can exist in highly ordered communities composed of multiple species. These communities are constructed such that each species carries out a specific subset of functions that, collectively, allow the conglomerate to thrive. Successful associations of this type require effective intra- and interspecies cell-cell communication. AHLs and peptide autoinducers are highly specific and are used for intraspecies cell-cell communication. AI-2 and its synthase LuxS, on the other hand, exist in over 40 species of gram-negative and gram-positive bacteria, and AI-2 is proposed to act as a more universal interspecies chemical language. Each AHL autoinducer synthase manufactures a specific signal. The newly reported structure of the EsaI autoinducer synthase reveals clues as to how a series of AHL molecules can be generated that differ only in side chain length. The EsaI structure predicts the existence of a hydrophobic cavity into which the acyl side chain (3-oxo-hexanoyl) neatly packs (Watson et al., 2002). Modeling suggests that the corresponding cavity in LasI of Pseudomonas aeruginosa is larger, and this is consistent with LasI synthesizing the longer autoinducer, 3-oxo-dodecanoyl AHL. Selectivity of the LuxRtype AHL binding proteins for their cognate signal molecules has also been investigated. Deletion and mutation analysis showed that AHLs bind to the central region of the LuxR-type proteins, and the DNA binding domains reside at the carboxyl termini. The X-ray structure of the first LuxR-type protein, TraR of the plant pathogen Agrobacterium tumefaciens, bound to the cognate AHL (3-oxo-octanoyl AHL) has recently been solved. Surprisingly, the autoinducer is fully embedded within the protein with virtually no solvent contact (Zhang et al., 2002). This structure is consistent with earlier findings suggesting that folding of TraR requires autoinducer binding (Zhu and Winans, 2000). Additional structural studies of this and other LuxR-autoinducer-DNA ternary complexes should begin to unravel the mechanism by which the LuxR-type proteins recognize a specific ligand. Accurate oligopeptide autoinducer synthesis is guaranteed by the specific sequence of DNA encoding the oligopeptide precursors. Although the posttranslational processing and modification steps are not well understood, an in-depth analysis of the final structural features critical for oligopeptide signaling specificity exists

Minireview 423

Figure 2. AHL and AI-2 Biosynthesis AHLs and AI-2 are produced from SAM. The representative AHL shown is C4 AHL from P. aeruginosa.

for autoinducers from the pathogen Staphylococcus aureus (Mayville et al., 1999). Each S. aureus autoinducing oligopeptide initiates a virulence cascade in its own S. aureus group and inhibits the virulence cascade in other S. aureus groups. Presumably, the first invading S. aureus strain that successfully establishes its quorum sensing circuit out-competes other invading strains. The S. aureus peptide signals are 89 amino acids with thiolactone rings involving invariant cysteine residues located 5 amino acids from the C terminus (Figure 3B). Structure-activity studies show that a ring is essential for activation and inhibition, the linear tail region is required for activation but dispensable for inhibition, and a truncated peptide possessing the thiolactone ring but lacking the tail acts as a global inhibitor of S. aureus virulence (Lyon et al., 2000). Chen et al. (2002) determined the structure of the V. harveyi AI-2 by crystallizing and solving the structure of the V. harveyi sensor protein, LuxP, complexed to AI-2. AI-2 consists of two fused 5 membered rings with a boron atom bridging the diester (Figure 3C). The assign-

ment of boron was based on the crystallographic electron density, mass spectroscopic determination of the molecular weight of the ligand bound to the complex (194 D), and 11B-NMR. Finally, addition of borate to V. harveyi cultures strongly stimulated light production, the quorum sensing-controlled behavior in this organism. The authors proposed that cyclization of DPD and addition of naturally occurring borate generates active AI-2. This is especially intriguing because boron, although known to be essential for many organisms, previously had no defined biological function. The tight specificity inherent in AHL and oligopeptide communication circuits presumably results in noise reduction. However, this facet of AHL and oligopeptide circuits also renders them species-specific. In contrast, it appears that the LuxS enzymes from various species of bacteria generate identical intermediates in the AI-2 biosynthetic process. Although at present only one AI-2 structure is known, it is possible that the AI-2s synthesized by different bacteria are identical resulting in a species-nonspecific language.
Figure 3. Representative Autoinducers and the Enzymes that Produce Them (A) Gram-negative AHL autoinducers. (B) Gram-positive oligopeptide autoinducers. The asterisk denotes an unknown modification. (C) V. harveyi AI-2.

Cell 424

Many gram-negative bacteria use AHL autoinducers and also produce AI-2. Likewise, many gram-positive bacteria have oligopeptide signaling systems as well as AI-2. Making and responding to combinations of these and potentially other types of chemical signals could permit bacteria to take a census of their own population numbers and also the population density of other species in the vicinity. A distinct response to each signal, or a response that is based on a combinatorial sampling of a variety of signals, could enable bacteria to continuously modulate behavior depending on the species present in a consortium. Censorship of Free Speech: Anti-Quorum Sensing Strategies Interspecies cell-cell communication allows bacteria to exploit the diverse metabolic functions that exist in a mixed-species consortium. Several cases of interspecies bacterial communication have been documented, the most notable of which is V. harveyi detection of and response to AI-2 produced by other species of bacteria. AI-2 has only recently been discovered, so its signaling role in these bacteria is not fully clarified. To date, AI-2 has been shown to control specific target gene expression in V. harveyi, Vibrio cholerae, Escherichia coli, Salmonella typhimurium, Porphyromonas gingivalis, Shigella flexneri, Streptococcus pyogenes, Neisseria meningitidis, Actinobacillus actinomycetemcomitans, Borrelia burgdorferi, and Clostridium perfringens, suggesting that this molecule plays a critical cell-cell communication role in and possibly between a diverse range of bacteria. Some bacteria release AI-2 early in growth and internalize at later times. Signal turnover is not a novel feature restricted to AI-2, as it occurs in oligopeptide and AHL signaling systems as well. In these latter cases, elimination of the signal makes possible the initiation and termination of specific behaviors. This could also be the case for AI-2. Alternatively, because AI-2 allows communication between species, bacteria that consume AI-2 could be actively interfering with the signaling process. AI-2 production is integrally associated with SAM metabolism, and therefore, directly tied to cell growth. As such, AI-2 harbors information regarding how well a bacterial population is doing. Elimination of AI-2 from the extracellular environment could be a form of censorship that allows one species of bacteria to avoid alerting other species to its presence. Consumption of AI-2 could provide a competitive edge to a bacterial species by rendering other species of bacteria that rely on AI-2 as information at a disadvantage. This could be critical for disrupting the delicate balance that exists between mixed populations competing for colonization of a particular niche. Other examples of anti-quorum sensing censorship strategies exist. As mentioned, S. aureus groups use peptide quorum sensing to infect a host and to inhibit other S. aureus groups from doing so. The plant-associated bacterium Pseudomonas aureofaciens regulates antibiotic production by its own AHL and also by molecules secreted by other plant-associated bacterial species, suggesting that P. aureofaciens detects the presence of competitor bacteria and responds by attempting to kill them (Pierson et al., 1994). The soil bacterium Bacillus subtilis produces an enzyme (AiiA) that hydro-

lyzes the AHL of the commensal soil bacterium Erwinia carotovora. E. carotovora relies on its AHL for infecting plants, and AiiA renders E. carotovora avirulent (Dong et al., 2000, 2001). The soil bacterium Variovorax paradoxus uses AHLs as the sole source of carbon and nitrogen, suggesting that in its natural habitat, V. paradoxus gains a competitive edge by growing on other bacterias AHLs. Studying these forms of quorum sensing censorship is new, and many other mechanisms for inhibiting quorum sensing are likely exist. Conclusions Bacteria do not live solitary reclusive lives, but rather they communicate using diverse chemical languages. This review outlines communication by AHLs, oligopeptides, and AI-2. Bacteria also use butyrolactones, quinolones, amino acids, and possibly other molecules for chemical communication. Because these compounds are diffusible and because their accumulation correlates with cell density and species type, these compounds contain information about the metabolic potential of the environment and the constituent species present. The use of this diverse assortment of molecules as sources of information exemplifies the acuity with which bacteria sense, integrate, and modulate behavior in response to ever-fluctuating environments. Importantly, we now know that cell-cell communication controls virulence in many pathogenic bacteria, and we also know that successful anti-quorum sensing tactics have evolved between coexisting bacterial populations. These two pieces of knowledge have spurred growing interest in designing and implementing synthetic anti-microbial strategies that expressly target quorum sensing, suggesting that continued research aimed at deciphering these chemical languages could have enormous practical applications.
Selected Reading Chen, X., Schauder, S., Potier, N., Van Dorsselaer, A., Pelczer, I., Bassler, B.L., and Hughson, F.M. (2002). Nature 415, 545549. Dong, Y.H., Xu, J.L., Li, X.Z., and Zhang, L.H. (2000). Proc. Natl. Acad. Sci. USA 97, 35263531. Dong, Y.H., Wang, L.H., Xu, J.L., Zhang, H.B., Zhang, X.F., and Zhang, L.H. (2001). Nature 411, 813817. Lyon, G.J., Mayville, P., Muir, T.W., and Novick, R.P. (2000). Proc. Natl. Acad. Sci. USA 97, 1333013335. Mayville, P., Ji, G., Beavis, R., Yang, H., Goger, M., Novick, R.P., and Muir, T.W. (1999). Proc. Natl. Acad. Sci. USA 96, 12181223. Miller, M.B., and Bassler, B.L. (2001). Annu. Rev. Microbiol. 55, 165199. Pierson, L.S., 3rd, Keppenne, V.D., and Wood, D.W. (1994). J. Bacteriol. 176, 39663974. Schauder, S., Shokat, K., Surette, M.G., and Bassler, B.L. (2001). Mol. Microbiol. 41, 463476. Watson, W.T., Minogue, T.D., Val, D.L., von Bodman, S.B., and Churchill, M.E. (2002). Mol. Cell 9, 685694. Zhang, R., Pappas, T., Brace, J., Miller, P.C., Oulmassov, T., Molyneaux, J.M., Anderson, J.C., Bashkin, J.K., Winans, S.C., and Joachimiak, A. (2002). Nature, in press. Zhu, J., and Winans, S.C. (2000). Proc. Natl. Acad. Sci. USA 98, 15071512.