The historical concept of microorganisms as single entities autonomously thriving in their environment has been clearly overcome by abundant evidence demonstrating that many, if not most, bacteria are able to live in organized communities requiring the exchange of some kind of information to coordinate their behavior. Part of this communication process, known as quorum sensing (QS), is defined as the awareness of bacteria to population densities and is based on the secretion and detection of different chemical molecules that are expressed in a density-dependent manner. Several cellular mechanisms are understood to be regulated by a QS mechanism in a variety of bacteria.

Plant-associated bacteria display a multiplicity of different lifestyles ranging from the commensal epiphyte and the endophytic symbiont up to a fully pathogenic existence. Many species are able to switch between different lifestyles depending on the environmental conditions met in planta and alter their population structure accordingly, which may range from a free-living unicellular state to communities of organisms living within extracellular matrices known as biofilms [1]. These processes, together with swarming behavior or the production of antibiotics and virulence factors, are often steered by the QS molecules, the so-called autoinducers (AI) [2–4].

The most widespread QS system in Gram-negative bacteria (QS-1) was first discovered in the marine symbiont Vibrio fischeri, where it controls bioluminescence. It consists of an N-acyl-L-homoserine lactone (AHL) synthase and a transcriptional activator encoded by the genes luxI and luxR, respectively [5,6]. In the related marine bacterium Vibrio harveyi, AHL is synthesized by the LuxM synthase, which shows no homology to LuxI-type AHL synthases although it is based on same biochemistry. In V. harveyi, the signal is detected by the LuxN histidine kinase [7]. Several AHLs (or type I autoinducers, AI-1), all sharing a common homoserine lactone core, but differing in their acyl side chain moieties have been described to date [2–4]. Various sets of AI-1 molecule and cognate receptor are employed (sometimes simultaneously [8,9]) among Gram-negative bacteria. Interspecies communication is basically limited to species sharing the same system, although a limited crosstalk is possible between bacteria that use chemically similar AHLs [10–12].

A second quorum sensing system (QS-2, [13]) was discovered in V. harveyi [14]. This system is controlled by the LuxS protein that catalyzes the production of (S)-4,5-dihydroxy-2,3-pentanedione (DPD). DPD is the direct precursor of a different autoinducer molecule (AI-2), (2S,4S)-2-methyl-2,3,3, 4-tetrahydroxytetrahydrofuryl borate (S-THMF-borate), a furanosyl borate diester [15] (Figure 1). In Vibrio spp., the AI-2 signal is detected by the two-component sensor kinase LuxPQ and is ultimately linked to the same transduction pathway used by the QS-1 sketched above and a third quorum sensing system, which is based on signal molecule (S)-3-hydroxytridecan-4-one (CAI-1) in Vibrio cholerae [16] and Vibrio harveyi [17]. All three signals are conveyed through the central signal relay protein LuxU and the terminal response regulator LuxO, the latter controlling gene expression together with sigma factor σ⁵⁴ [18], whereby the strengths of the different autoinducer signals are not equivalent and vary from species to species [19,20]. In bacterial orders other than the Vibrionaceae, the task of detecting a different form of the DPD-derived AI-2 signal, (2R,4S)-2-methyl-2,3,3,4-tetrahydroxytetrahydrofuran (R-THMF), is carried out by an ABC-transporter, the Lsr-receptor complex. Orthologs of this protein are prevalent mainly in Enterobacteriaceae [21,22], Pasteurellaceae and Bacillaceae, but were not detected in Vibrionaceae [23]. The presence of LuxS has been reported in several subgroups of the bacterial kingdom such as Actinobacteria, Bacilli, Bacteroidetes, Deinococci and Beta-, Gamma-, and Epsilonproteobacteria. Conversely, LuxS has not been described in Archaea nor in Eukarya [24,25]. The pervasive nature of the luxS gene, embracing both Gram-negative and Gram-positive bacteria, led to the presumption that QS-2 may be the foundation of a bacterial Esperanto; that is to say, a universal language spoken and understood by various bacterial species [26]. Unfortunately, this assumption has often neglected a crucial function of LuxS in bacterial cells. This protein has an enzymatic role in the activated methyl cycle (AMC) [25], which accounts for the regeneration of the major methyl donor S-adenosyl-L-methionine (SAM) and the recycling of methionine by detoxification of S-adenosyl-L-homocysteine (SAH) in the cell (Figure 1). This omission has repeatedly led to the misinterpretation as QS of metabolic effects in in vitro mutational experiments and to the incorrect interpretation of in silico genomic data (see Section 5) [23]. Further complicating data interpretation, also the presence of sequences encoding accessory elements like small regulatory RNAs (sRNAs) in the coding or the promoter region of the luxS gene has been shown be involved in QS-related behaviors. Thus, their involvement should carefully be considered when evaluating the phenotype of luxS mutants [27,28].

In this review we will critically review the latest publications on QS-2 to analyze whether LuxS is involved in signaling or if it may hold a mere metabolic role in the species studied. Furthermore, we will comb through the published genomes of bacteria and search for elements related to AI-2 signaling that may allow the formulation of bioinformatics-informed predictions on the importance of QS-2, with particular focus on plant-associated Enterobacteriaceae.

In those cases when the chemical complementation of luxS mutation by the means of exogenous AI-2 is successful even in the absence of genes coding for either LuxPQ or the Lsr-receptor complex, the presence of alternative receptors for AI-2 must be postulated; otherwise a sheer metabolic role of LuxS in the associated bacteria must be considered [23]. Different studies have hypothesized a number of such QS-2 related regulators and transporters.

Although the two-component sensor kinase LuxPQ is expected to be limited to marine bacteria of the genus Vibrio, there are indications that Lsr-mediated AI-2 sensing is a trait generally restricted to (but not a prerequisite for) bacteria that live in close association with specific groups of eukaryotes. In keeping with this idea, the lsr-locus is present in many species of the B. cereus group, but not in B. subtilis, a soil organism which is not considered to be an animal pathogen [23]. Among Salmonella spp., the lsr operon is only located in S. enterica subsp. enterica serovars that are associated with endotherms, whereas it is absent from the genome of S. bongori, a species usually associated with cold-blooded animals [78]. In the nematode symbiont P. luminescens the Lsr-receptor complex is apparently restricted to those strains that form a mutualistic consortium with Heterorhabditis bacteriophora, but the lsr genes show extensive deletions in strains associated with other hosts, suggesting that QS-2 is deeply involved in the association with this particular insect-pathogenic nematode [79]. Thus, the pattern of distribution of LsrB among the different bacterial species suggests that AI-2 may be a relevant signal for many animal pathogens and symbionts and that the Lsr-receptor complex may play a general role in host colonization by regulating processes, such as biofilm formation, for which a high cellular density is a critical prerequisite.

Among plant-associated bacteria, AI-2 specific receptors were only found on the pSymB-plasmid of the N₂-fixing Alphaproteobacterium S. meliloti, which forms an endosymbiotic relationship with legumes from the genera Medicago, Melilotus and Trigonella. Both S. meliloti (as well as related Alphaproteobacteria R. sphaeroides and R capsulatus) and the aforementioned Bacillus cereus group are thought to have acquired the genes coding for the Lsr receptor complex by horizontal gene transfer (HGT), probably from bacteria belonging to the Enterobacteriaceae and Pasteurellaceae, respectively, that share the same ecological niche [23,50]. This conclusion is supported by the unusual position of their lsrB genes in the corresponding phylogenetic tree (Figure 3). As Alphaproteobacteria themselves are unable to produce AI-2, it was speculated that they may use the Lsr-receptor complex to eavesdrop on the AI-2-mediated communication of other bacterial species in the same habitat [23,51].

Klebsiella spp. naturally occur as a free-living diazotrophs in soil or as endophytes in the roots of a wide variety of plant species. About 30% of strains can fix nitrogen under anaerobic conditions, but isolates that are close relatives of these diazotrophs are also associated with human diseases [80,81]. Isolates forming a clade that includes the type strain of Klebsiella variicola were also found within fungus gardens of the colonies leaf-cutter ants of the genera Atta and Acromyrmex where, together with bacteria belonging to the genus Pantoea, they participate in nitrogen fixation for the fungal symbiont [82]. Both bacterial species living within this consortium carry, beside the nif genes coding for the nitrogen fixation pathway, the lsr-operon needed for the AI-2 ABC transporter. Although prevalent in the genomes of different Klebsiella spp., none of the two gene clusters was previously reported in any sequenced Pantoea spp. [83–88] suggesting that these were acquired by Pantoea sp. At-9b through HGT during adaptation to this particular ecological niche [82]. However, the lsr-genes were not directly acquired from Klebsiella, as the significant distance between the two genera in the lsrB tree demonstrates (Figure 3). Just as in the symbiotic/pathogenic interactions involving gut colonization of insects and animals, the lifestyle needed for nitrogen fixation in root nodules or in fungus gardens entails high bacterial population densities within a closed system, thus providing a possible rationale for the presence of a QS system. A potential relationship between QS-2 and nitrogen fixation is confirmed by the presence of the Lsr receptor complex in another Alphaproteobacterium R. sphaeroides 2.4.1. This bacterium is similar to Sinorhizobium in that it does not carry the luxS gene, but, in contrast to the latter, it is a free-living diazotroph.

In plant-associated bacteria that are not N₂-fixing symbionts, the genes encoding the Lsr receptor complex could be identified only in the genomes of the mulberry-pathogen Enterobacter mori [89], in the plant growth-promoting endophyte Enterobacter sp. 638 [90], and in Enterobacter cancerogenus, which was originally isolated from poplars (Populus spp.) affected by a canker disease [91]. E. cancerogenus has also been found as clinical isolate [92]. Whether the AI-2 receptor plays a role in pathogenicity or plant association is unknown, but its presence could again reflect the importance of QS-2 in situations associated with bacterial populations thriving in a limited space such as in the case of bacterial cancers or in bacteria employing an endophytic lifestyle.

The major pitfall when analyzing the role of QS-2 in a microbial species has been the reliance on the mere presence of the luxS gene in the genome of the bacteria under investigation [23]. As the AI-2 precursor DPD is normally produced as a by-product of the AMC, even detection of AI-2 molecules in the culture supernatant, either by chemical analysis or through reporter strains such as V. harveyi BB170 [31], cannot be considered as an unmistakable proof of the existence of a functional QS-2 system. The same applies to traditional mutational analysis that, given the important role of LuxS in methionine recycling within the cell, is not able to discriminate between QS-2 and metabolism-related changes of the luxS mutant phenotype. This is because conventional in trans complementation will overcome defects induced in either pathway. Examples for this kind of “missed approach” are available in a wide variety of luxS positive bacteria [23], including plant pathogens such as Pectobacterium carotovorum [93,94] or Erwinia amylovora [95]. In the latter study, the effects of the luxS mutation were sweepingly ascribed to QS without considering any possible implication of the AMC, even though it was previously demonstrated that the mutant strain has impaired metabolic fitness, especially under sulfur-limiting conditions [96]. In the same vein, the failure to complement the phenotype of a ΔluxS variant by the luxS-positive wild-type strain in co-cultivation assays suggested the absence of an extracellular signal. A way to circumvent the complication of the dual role of LuxS in the cell is to shift the focus of the analysis from LuxS itself to the analysis of potential AI-2 receptors and the chemical complementation of the luxS mutant, including objective evidence of AI-2 binding or uptake [97,98]. Alternatively, it was proposed to replace the luxS gene not with an intact copy of itself, but with the sahH gene that encodes the SAH hydrolase, which is part of the one-step S-adenosylhomocysteine detoxification pathway (Figure 1). In this manner, it is possible to repair the function of the AMC without restoring AI-2 production: if the wild-type phenotype can be salvaged by an in trans complementation with sahH, then any difference observed in the luxS mutant has to be ascribed to metabolism, not to QS-2 [99]. Conversely, if it possible to fully complement the luxS mutant by adding exogenous AI-2 to the culture medium (chemical complementation), a role of the latter molecule in signaling must be considered.

A more promising tactic to determine whether QS-2 is really present in a bacterial species is to shift the focus of the investigation from AI-2 production to AI-2 recognition and detection. This may include bioinformatics approaches to scan for orthologs of known AI-2 receptors [23,38,50], which must withstand the reciprocal best-hit strategy to distinguish between orthologous and paralogous genes [100]. In this kind of approach, it is important that the actual receptor is employed in similarity searches, as the use of associated proteins or domains (e.g., transmembrane channel LsrCD or sensor kinase/phosphatase LuxQ) may yield multiple partial positive matches in several species, as these proteins may share an high degree of sequence identity among paralogous systems. A thorough analysis of the complete genomes of E. amylovora [101] and the closely related species E. pyrifoliae [102], E. piriflorinigrans [103], E. billingiae and E. tasmaniensis [104] revealed the complete absence of any specific receptor associated with QS-2, thus confirming the metabolic role of LuxS in the fully sequenced species of this genus. In addition to the bioinformatics approach, experimental evidence of a functional QS-2 system can be gathered by confirming AI-2 depletion from the culture media [97,98] or, as more recently proposed, the use of a DPD-based rhodamine-conjugated dendrimer as multivalent probe that allows specific imaging of cells bearing Lsr-type quorum-sensing receptors [105].

The fact that many of the alternative AI-2 receptors and QS-2 regulators are fairly common across different bacterial taxa and are not specific to luxS-positive bacteria suggests that they were recruited subsequently into their role in QS-2. This raises the question of whether their primary function is always true signalling or, especially in luxS-negative bacteria, whether it may also concern other types of interactions involving metabolism or activation of non-QS-2 related responses. In principle, because the PTS system initiates AI-2 import in E. coli [68] is virtually ubiquitous within the bacterial kingdom [106], almost every bacterium should at least theoretically be capable to internalize AI-2. The same is true for ribose ABC transporters, which were shown to be able to mediate AI-2 intake because of the structural similarity of the autoinducer molecule with their cognate substrate [56,63]. This could explain the successful chemical complementation of luxS mutants in those species that were found to be devoid of both major AI-2 receptors [23]. However, whether the transcriptional changes induced by the addition of exogenous AI-2 in the receiver organism are effectively related to QS-2 remains to be demonstrated. From an evolutionary point of view, for a secreted molecule to be a signal between different species, it must not only elicit a response in the receiving organism but also have evolved primarily to do so [107]. Interactions involving other types of non-concerted responses may be referred to as cues or coercion events [108]. A cue describes an emission which provides beneficial information to a receiver but which did not evolve explicitly for that purpose, and thus does not directly benefit the emitter. Conversely, coercion describes the situation in which a signal induces a detrimental response in a receiver to the benefit of the emitter [109]. Whatever names are used for these alternative activities, they illustrate the danger of concluding that a single prevailing environmental factor controls bacterial behavior in a manner that evolved specifically to carry out that function. In reality, both the ecological context of QS regulation and the process itself are complex and respond to a multiplicity of environmental stimuli. For this reason, the notion of ‘quorum sensing’ should perhaps be defined more narrowly to include only signals specifically used for communication between emitter and receiver cells. Otherwise, the concept of QS may be diluted to include environmental signals that affect bacterial metabolism and behavior in rather indirect ways that are not examples of intentional cell-to-cell- communication [110].