Complete genome sequences of bacteria known to have a luxS gene were analyzed for the presence of AI-2 receptor sequences using the blastx and the tblastx programs with a cut-off E-value of 1E-80 by applying a reciprocal best hit strategy, which allows to distinguish between orthologous and paralogous genes 35. As starting queries the luxP gene of V. harveyi strain BB7 (accession number U07069) and the lsrB gene of E. coli strain K12 (accession number NC_000913, locus tag b1516) were used. The hits were manually checked to confirm that they had the same functional annotation as the query and re-submitted as novel queries until no new relevant positive match was found. In this second phase also luxS-negative strains were included in the analysis by performing a general blast search across the entire NCBI database http://www.ncbi.nlm.nih.gov. In bacteria found to carry a gene coding for an AI-2 receptor the presence of a complete QS-2 system was corroborated by verifying the existence of other genes associated to the respective AI-2 transduction pathway. Bacterial species were scored as positive for QS-2 receptors only when in addition to luxP, homologs of the luxQ (inner membrane sensor kinase), luxU (two-component phosphorelay protein) and luxO (response regulator) genes which encode the full transduction pathway 1213 were also found. Similarly, positive results obtained with lsrB were confirmed by verifying the presence of genes constituting a complete lsr-operon (lsrBDCARKG). These include lsrA (encoding the ATPase that provides energy for AI-2 transport), lsrC and lsrD (whose products form the heterodimeric membrane channel for the uptake of AI-2), lsrR (encoding the repressor of the lsr-operon active in the absence of phospho-AI-2), and lsrK (encoding the AI-2 kinase) 22.

Our investigation was extended to incomplete sequenced bacteria known to carry a luxS gene or for which a function of QS-2 was documented in the literature through an NCBI database search with relevant terms such as "luxS", "lsrB", "luxP", "AI-2" or "quorum sensing". Discovered pertinent sequences were used in the analysis.

A literature search for published studies on the effect of type-2 quorum sensing in the organisms carrying the luxS gene was performed using the NCBI database and the papers were critically read, with particular attention to those key experiments which are essential to ascertain the presence of a QS-2, such as the identification of AI-2 receptors, chemical complementation of the luxS mutant or detection of AI-2 depletion in the culture medium 31. In cases where these tests were omitted the possibility that a metabolic dysfunction in the AMC may have been misinterpreted as disruption of type-2 quorum sensing was proposed.

Phylogenetic analysis was performed in order to understand how AI-2 receptors may have (co-)evolved in different bacteria with respect to the presence of luxS. Phylogenetic trees were generated on the basis of complete luxS- and lrsB-sequences without choosing any outgroup. Housekeeping gene rpoB, encoding the RNA polymerase β-subunit, was used as reference. Retrieved DNA sequences were aligned with ClustalW 36. Sites presenting alignment gaps were excluded from analysis. The Molecular Evolutionary Genetics Analysis (MEGA) program version 4.0 37 was used to calculate evolutionary distances and to infer trees based on the neighbor-joining (NJ) method using the JC formula. Nodal robustness of the inferred trees was assessed by 1000 bootstrap replicates.

Complete genomes and key luxS and lsrB sequences were used.

The luxP gene of V. harveyi strain BB7 was used as a starting query for a search of LuxPQ-type receptors in bacterial genomes that was successful only for members of the orders of the Vibrionales and, to a much lesser extent, of the Oceanospirillales. This kind of receptor, along with the effector kinase LuxQ, was already known in several species of the genus Vibrio, i.e., V. anguillarum, V. cholerae, V. fischeri, V. harveyi and V. parahaemolyticus (see Table 1 and citations therein). In Vibrio species for which a response to AI-2 system is documented, but in which no receptor was yet described, i.e., V. vulnificus 38 and V. alginolyticus 39, we were able to retrieve the luxPQ genes by searching the relevant gene sequences in the corresponding published genomes. Conversely, the LuxPQ-receptor was absent from the genome of V. angustum, which does not respond to the AI-2 signal 39, even though the luxS gene coding for S-ribosylhomocysteinase is present. Besides that, we managed to identify the LuxPQ receptor complex also in other representatives of the genus for which no study about QS-2 related phenotype has been performed to date, i.e., V. splendidus, Vibrio sp. Ex25 and Vibrio sp. MED222, but not in closely related Photobacterium profundum SS9 (Fig. 2 and Table 3).

A LuxP homolog was found also in the genomes of some Oceanospirillales, albeit among the bacteria of this order no luxS gene has been sequenced or found to date. The similarity of these proteins to the LuxP-receptor present in V. harveyi was only moderate: 41.2% identity and 71.7% similarity for the Neptuniibacter caesariensis MED92 periplasmic protein (locus tag: MED92_04774) which carries a region similar to the RbsB-like ABC-type sugar transport system receptor locus and 24.5% identity and 61.8% similarity for the Marinomonas sp. MED121 LuxP protein precursor (locus tag: MED121_11334). However, both results withstood the reciprocal best hit strategy test and yielded a LuxP protein of Vibrio as best match in a blastx search. Furthermore, in Neptuniibacter caesariensis MED92 and Marinomonas sp. MED121 the genes encoding these putative LuxP homologs are arranged in an operon with presumed transcriptional regulators that likewise contain PAS domains (locus tags MED92_04779 and MED121_11339 respectively), but share otherwise little homology with each other. In Marinomonas sp. MED121, the transcriptional regulator showed moderate homology with LuxQ of V. cholerae and contains also AAA-type ATPase and DNA-binding domains and is followed by the pfs gene coding for a 5'-methylthioadenosine nucleosidase (locus tag: MED121_11344). A similar protein (locus tag Dace_1280) in Desulfuromonas acetoxidans (a δ-Proteobacterium) showed an equivalent identity to LuxP of V. harveyi (36.2%) and was analogously arranged in an operon with a putative transcriptional regulator containing PAS domains (locus tag Dace_1279), while in Desulfovibrio desulfuricans (also a δ-Proteobacterium) the matching protein (locus tag Dde_3311, identity 41.0%) also carries a putative receptor locus but the corresponding gene was not associated with either a luxQ homolog (or a similar transcriptional regulator) or the other usual genes of an ABC-transporter, indicating the probable absence of a signal transduction pathway toward the inner of the cell. However, also Dace_1280 and Dde_3311 pointed toward LuxP in the reciprocal best hit test. As for Oceanospirillales, both D. acetoxidans and D. desulfuricans (both δ-Proteobacteria) were found to lack the luxS gene. All Vibrionales carrying the luxPQ genes also carried open reading frames which encoded proteins homologous to the two component, σ⁵⁴-specific, transcriptional regulator LuxUO found in V. harveyi (accession number L26221). In Oceanospirillales a similar situation was discovered, but both luxP and luxO were flanked by other transcriptional regulator genes which displayed a less important resemblance with luxQ and luxU respectively. In all strains investigated the LuxO protein displayed the characteristic 8-AA glycine-rich consensus (GESGTGKE) which is needed for interaction with alternative σ⁵⁴ (data not shown) 40. Interestingly, in none of the strains were the genes encoding the single components of the QS-2 mechanism (luxS, luxPQ and luxUO) found in proximal positions on the genome (data not shown), thus they were probably acquired and/or developed independently.

Notably the tblastx search using luxP gene of V. harveyi strain BB7 resulted in local hits with the lsrB gene of Enterobacteriales (E-values > 1E-10). Similarity is limited to a small RbsB-like portion on the C-terminus of the LuxP protein, but is most probably indicative of the substrate (AI-2) binding domain of both receptors. Generally, the use of luxQ gene of V. harveyi strain BB7 as starting query yielded multiple partial positive matches in almost every strain tested, but these were limited to the cytoplasmic transduction domain of other sensor kinases and were not directly related to any known quorum sensing related gene (data not shown).

Our screening of completed genome projects for lsrB, the gene belonging to the lsr-operon encoding the periplasmic AI-2 binding component of the ABC-type sugar transport apparatus, revealed that this type of receptor is generally more widely spread than the LuxPQ-system. We found that lsrB is present both in Gram-negative and in Gram-positive bacteria, albeit its distribution was not homogeneous in and among the bacterial orders considered (Fig. 2 and Tables 1, 2, 3). An open reading frame coding for LsrB was found, for Gram-negative γ-Proteobacteria, essentially only in part of Enterobacteriales (i.e., in E. coli, Photorhabdus luminescens, Klebsiella pneumoniae, Yersinia spp., in Shigella dysenteriae and Shigella flexneri, and Salmonella spp. with the exception of S. bongori, but not in S. marcescens, S. glossinidius and in Erwinia spp.) and of Pasteurellales (i.e., in Haemophilus somnus, Pasteurella multocida and Actinobacillus actimycetemcomitans, but not in related species such as Haemophilus influenzae or Actinobacillus pleuropneumoniae). No evidence for lsrB was found in Vibrionales and in the luxS-negative families of Pseudomonadaceae and Xanthomonadaceae. No lsrB homolog was located either in the luxS-positive class of ε-Proteobacteria (Helicobacter, Campylobacter) or in Spirochaetales, in which the presence of luxS was reported to date only in Borrelia spp. 41.

For Gram-negative α-Proteobacteria, lsrB was found only on the pSymB-plasmid of Sinorhizobium meliloti and in three Rhodobacter species, i.e. R. capsulatus (contig hit, position on the genome unspecified) and on the high variable chromosome-2 of R. sphaeroides 2.4.1 and R. sphaeroides ATCC 17029, but was absent from R. sphaeroides ATCC 17026. These finding are of special interest since α-Proteobacteria are lacking the luxS gene coding for the S-ribosylhomocysteinase/AI-2 synthase.

In Gram-positive bacteria, lsrB was located solely in bacteria belonging to the Bacillus cereus group. B. anthracis, B. thuringensis, B. weihenstephanensis and B. cereus were found to carry the lsr-operon (see additional file 2, Figure 1s). However, in B. weihenstephanensis KBAB4 both lsrB and lsrD show frameshifts and are listed as pseudogenes, in a similar way to lsrB of B. thuringensis serovar konkukian 97-27 and to the associated ATPase-gene lsrA of non-pathogenic B. cereus ATCC 14579, which is apparently truncated. In B. thuringensis serovar israeliensis ATCC 35646 a pseudogene with very high homology to part of the lsrB-receptor gene is found well separated from the rest of the operon (lsrDCARK) suggesting that in the latter strain the ABC-transporter was inactivated by a rearrangement of the genome. The lsrB gene could not be identified in closely related Bacillus cereus ssp. cytotoxis NHV391-88 confirming a significant degree of divergence from typical B. cereus 42, which is corroborated also by our luxS-based phylogeny. More distant Bacillus species, including B. subtilis, were all found to be lsrB-negative (see additional file 2, Figure 1s) suggesting that in Gram-positive bacteria the Lsr-transporter was acquired or developed by an ancestor of the Bacillus cereus group. No lsrB homolog was found in the Actinobacteria Bifidobacterium longum 11 and Bifidobacterium adolescentis which are so far the only bacteria besides Escherichia blattae 43, to possess both known SAH-detoxification pathways (SAH hydrolase and Pfs/LuxS).

All strains which yielded a positive match for lsrB were further verified by demonstrating the presence of other genes (lsrDCARKG) encoding the Lsr-complex. However, in none of the positive matches the luxS gene and the lsr operon were found to lie in proximal regions of the genome, suggesting that (as for luxPQ and luxUO in Vibrionales) they were probably acquired and/or developed independently. On the other hand, in Enterobacteriaceae luxS is frequently found in association with another gene involved in sulfur metabolism such as gshA, coding for the γ-glutamate cysteine ligase (data not shown).

An ambiguous positive match to LsrB (67.1% similarity, 31.4% identity) was obtained for substrate binding protein Atu3487 (accession number NC_003305) on the linear chromosome of Agrobacterium tumefaciens C58. However, both the sequence of the receptor and the organization of the flanking genes in the operon displayed high similarity with the rhamnose catabolism gene locus of Rhizobium leguminosarum bv. trifolii plasmid pRleW14-2c (accession number AF085687) showing that these genes are probably not involved in quorum sensing. Similarly, a positive match (23.8% similarity, 19.1% identity) in Actinobacillus pleuropneumoniae to an uncharacterized conserved protein (locus tag Aple02001327) was discarded after additional analysis yielded that the retrieved sequences showed more similarity to E. coli xylF gene, coding for the membrane receptor protein of the ABC-type xylose transport system.

In order to understand how QS-2 developed in the different bacterial taxa during evolution, a phylogenetic analysis was performed by comparing, for those species for which a QS-2 has been previously described, the presence of the established AI-2 receptor genes luxP and lsrB to the tree based on the luxS-gene sequences (Fig. 2). Our results confirmed the conclusions of a previous work 19 in which a few species appeared be misplaced on the luxS tree if compared to species phylogeny and thus to have acquired the latter gene by an horizontal gene transfer (HGT) event: these bacteria are β-proteobacterium Neisseria meningiditis, ε-proteobacterium Helicobacter pylori, the spirochete Borrelia burgdorferi, and actinobacteria Bifidobacterium longum and Bifidobacterium adolescentis. Interestingly, however, in none of these bacteria the gene sequences encoding the LuxP- or LsrB-receptors have been found, indicating that most probably only the metabolic function of LuxS, and not a complete QS-2 system, was transferred.

More generally, as detailed above, the occurrence of luxP was restricted to Vibrionales, but the gene was not universally present in every species of this order, missing for instance from Vibrio angustum or Photobacterium profundum. Similarly, the presence of the lsrB-gene was generally limited to certain species belonging to the Gram-negative bacterial orders of Enterobacteriales. Pasteurellales, and to the Gram-negative positive classes of Bacilli and Clostridia and was not a collective feature of all the members of each group. Most importantly, several species for which an AI-2 dependent behavior was claimed (Table 3) were found to be devoid of either receptor genes.

Analysis of lsrB tree (Fig 3) shows that this gene is well distinct from related ABC-receptor genes such as rbsB or xylF. This result is supported by similar analyses performed on the other genes coding for the ABC-receptor complexes (data not shown) and indicates that the Lsr receptor is ortholog to those ABC-transporters and did not arose by gene duplication from related sugar transporters in each single species or genus. The lsrB-phylogeny of luxS-negative species R. sphaeroides,R. capsulatus and S. meliloti reveals that the Lsr-receptor complex of these α-proteobacteria was probably horizontally acquired from unrelated enterobacterial species. The HGT hypothesis is further supported by the fact that in these species the genes coding for the AI-2 receptor are not on the main chromosome, but are found on dynamic genetic elements such as a smaller chromosome as in the case of R. sphaeroides 2.4.1 or on a plasmid as in S. meliloti 1021.

Taking in consideration only lsrB-positive bacteria, we finally compared the topology of the phylogenetic trees based on luxS, lsrB and the housekeeping gene rpoB, encoding a subunit of the RNA polymerase and which is commonly used as a molecular marker for microbial evolution studies (Fig. 4). The luxS- and rpoB-trees displayed a branching pattern which was more similar to each other than to that of the lsrB-tree, however in the latter most of the differences were found at local level within single clades and (if present) they were not supported by high bootstrap values even when using more rigorous building methods such as minimum evolution or maximum parsimony (data not shown), indicating the lsrB may have undergone an HGT only in a limited number of cases. This result, together with the discrete distribution of lsrB among luxS-positive bacteria suggests that for luxS- and lsrB-positive bacteria the AI-2 receptor coevolved with the SAH-lyase/AI-2 synthase, but that for luxS-negative strains it was probably acquired by certain species (or their common ancestor) as adaptation to a specific niche.

Our search across fully published genomes clearly shows that the presence of this kind of AI-2 receptors is primarily limited to a narrow group of bacteria. Corroborating the results of Sun and coworkers 19 no complete QS-2 receptor complex based on the LuxPQ sensor kinase and the related signal cascade was found in bacteria not belonging to the order of Vibrionales. Within the latter order the presence of luxPQ genes was not ubiquitous, but was found to be largely in accordance with the responsiveness to AI-2 stimulation and the functionality of the QS-2 system that can be deduced from the literature already published on this topic (Table 1). No alternative receptor for AI-2 has been described so far in bacteria belonging to the Vibrionales. Even if an experimental verification is still pending, we predict therefore that other sequenced Vibrio species (i.e., V. splendidus, Vibrio sp. Ex25 and Vibrio sp. MED222) that carry both LuxS, LuxPQ and LuxUO will most probably have a functional QS-2 system, while Photobacterium profundum SS9 will not.

LuxP homologs were also found in luxS-negative Neptuniibacter caesariensis MED92, Marinomonas sp. MED121 and D. acetoxidans DSM684. Since previous evidence suggests that AI-2 may also be involved in interspecies gene regulation 4445 and mixed-species biofilm formation 46, it is therefore conceivable that some bacteria may be able to recognize the AI-2 signal even though they are unable to produce it. The evolutionary benefit would be the opportunity to sense the presence and abundance of AI-2 producers sharing the same niche. Both Neptuniibacter and Marinomonas are marine bacteria that share the same environment 47 or are associated with the same host as AI-2 producer Vibrio strains 48. However, given the less important similarity to the receptor found in Vibrionales and the different sensor kinase associated this hypothesis has in any case to be proven experimentally. In the case of D. desulfuricans, in which the luxP homolog is not physically associated to any gene indicating a signal transduction toward the inner of the cell, the prospect of an AI-2 related QS seems less realistic.

Although in recent years the activity of QS-2 was repeatedly hypothesized on the basis of the presence and functionality of the luxS gene in several bacterial species not belonging to the order of Vibrionales, few efforts have been made to verify the presence of a cognate AI-2 receptor, either experimentally by demonstrating successful chemical complementation of luxS mutant (using AI-2 containing supernatants) or genetically by showing the presence of the associated genes in the corresponding bacterial genomes. The only other clearly established AI-2 receptor, the Lsr ABC-transporter was identified so far only in S. typhimurium 20 and E. coli 21. For this reason we examined the genomes of completely sequenced bacteria with a reported quorum sensing activity ascribed to AI-2 and searched for the presence of DNA-sequences coding for receptor protein LsrB, the extracellular substrate-binding component of the Lsr-complex.

The outcome of our investigation was at first glance surprising, since a positive match was found only in part of Enterobacteriales, Pasteurellales and Bacillales (Fig. 2). However, a more detailed analysis of the results showed that these were largely congruent with the ability to complement the luxS mutation by the means of chemically pure AI-2 signal or AI-2 containing supernatants. The lsrB gene was retrieved exclusively in the genome of bacteria for which, provided that it was performed in the corresponding study, a chemical complementation yielded a successful result (Table 1). Conversely, our search for lsrB did not yield positive results in any of the species in which exogenous AI-2 was unable to restore the original wild-type phenotype in luxS mutants (Table 2). These results strongly corroborate the hypothesis that LsrB functions as receptor for AI-2 and the fact that in many bacteria the role of LuxS is restricted to the AMC.

In a few cases we were unable to identify either the Lsr-receptor complex or the LuxPQ sensor kinase in the genome of the corresponding bacteria even if a successful chemical complementation of the luxS mutant using AI-2 containing fluids was claimed in the literature (Tables 2 and 3), although it must be stressed that for some bacterial species at least one work came to contrasting conclusions and asserted a mere metabolic role for LuxS (Table 2). Strain-specific distribution or functionality of lsr-transporters are in principle possible explanations for this result. Although the lsr-operon seems to be ubiquitous in strains belonging to the Bacillus cereus group, in some non-pathogenic strains (e.g, B. cereus ATCC 14579) or insect-associated species like B. thuringiensis the sequences of the lsr-genes revealed frameshifts or rearrangements that will most likely lead to the production curtailed proteins and non-functional AI-2 receptors. Since in none of these strains the lsr-operon was completely absent from the genome, this may be the result of gene deterioration as a consequence of reduced selective pressure in those strains not tightly associated with warm-blooded animals.

Another option to explain successful chemical complementation of luxS-mutants in bacteria lacking both LuxPQ or the Lsr-receptor complex is to postulate the presence of alternative receptors for AI-2. Yoshida and coworkers 49 were able to complement the defect in biofilm formation of a Streptococcus mutans luxS mutant by culture media of other Streptococci or other luxS-positive bacteria which are normally associated with S. mutans in the human oral cavity such as Porphyromonas gingivalis and A. actinomycetemcomitans, while in a more recent study 59 genes responsive to AI-2 were identified in the same strain by global transcriptome analysis 50. In the latter study, the product of one of the strong AI-2 upregulated genes (SMU.408) was hypothesized to act as an uptake mechanism for the autoinducer molecule. The corresponding protein is a xanthine/uracil/vitamin C permease (COG2252) and belongs to the major facilitator superfamily (MFS) transporters. Homologs to this putative receptor are present in all Streptococcus species for which a QS-2 related phenotype was described, but also in many other species, including many for which a conventional receptor for AI-2 is already known, such as V. vulnificus, B. anthracis or P. multocida (complete data not shown).

A role as an alternative AI-2 receptor was also suggested for the ribose ABC-transporter encoded by the rbs-operon. Stevenson and coworkers speculated that RbsB may serve as the AI-2 receptor in Borrelia burgdorferi 51, while in Actinobacillus actionmycetemcomitans both the LsrB and RbsB proteins were shown to interact with AI-2 molecule 52. The AI-2 signal showed differential affinity to two receptors depending if found as THMF (like in S. typhimurium) or as borate diester (like in V. harveyi) (Fig. 1) leading to the hypothesis that the two receptors may be used together by the bacteria to scavange borate from the environment 25. The distribution of the ribose ABC-transporter is almost ubiquitous among most bacterial species regardless if the luxS- and lsrB-genes are present (data not shown), so that it is conceivable, although far to be proved, that RbsB may account for the internalization of AI-2 in a number of species lacking LsrB.

Interestingly, bacteria that possess a Lsr-receptor complex are mostly associated with warm-blooded animals. In fact, this ABC-transporter could not be located in the luxS-positive bacteria S. bongori, which unlike other Salmonella spp. is usually associated with cold-blooded organisms 53, or in B. subtilis, which is a soil organism not considered to be an animal pathogen 54. Exceptions are represented by the human pathogens S. boydii and S. sonnei in which the lsr-genes are absent 55 and insect pathogens B. thuringensis and P. luminescens which conversely feature the Lsr-complex. The case of P. luminescens is of particular significance since in this bacterium the presence of the lsr-locus seems restricted to those strains (i.e., TT01, Hb, and C1) forming a mutualistic consortium with entomopathogenic nematode Heterorhabditis bacteriophora, but shows ample gene deletions in strains associated with other nematode hosts or in related Xenorhabdus species suggesting that the lsr locus is strongly involved in the bacterial association with H. bacteriophora 56. Altogether this data strongly suggest that AI-2 may be a relevant signal for many invasive bacteria and that the Lsr-receptor complex may generally play an important role in host colonization.

An unexpected result of our study was the discovery of a region with high homology to the lsr operon (lsrBDCARKG) (Fig. 5) on chromosome II of two of the three sequenced strains of the free-living photosynthetic bacterium R. sphaeroides and on plasmid pSymB of nitrogen-fixing plant symbiont S. meliloti 1021. Both bacteria belong to the α-subdivision of Proteobacteria that regenerates homocysteine not by the means of the AMC, but directly from SAH using a one-step reaction catalyzed by SAH hydrolase. Since they are devoid of the LuxS enzyme they can not produce the AI-2 signal by themselves and are thus in principle dumb in the QS-2 language. In both bacteria the genes coding for the Lsr receptor complex are situated not on the main chromosome, but either on a megaplasmid mainly devoted at the codification of solute uptake systems in S. meliloti 57 or on a smaller chromosome with much lower information content density than the main chromosome, which is thought to provide highly adaptable organisms such as R. sphaeroides with genetic "safe zones" that allow DNA to change without affecting the organism itself 58. The phylogeny of the lsrB gene (Fig 3) suggest that the AI-2 receptor complex in R. sphaeroides and S. meliloti was eventually acquired horizontally from unrelated enterobacterial species, such as e.g. K. pneumoniae which occurs as free-living diazotrophs in soil, plant endophytes, and as opportunistic human pathogens 5960. It is conceivable that K. pneumoniae acquired the Lsr receptor from other endotherm associated Enterobacteria and passed it to cohabitant nitrogen fixing bacteria in soil. Whether the Lsr-complex in R. sphaeroides and S. meliloti really enables them to react to the AI-2 signal still has to be proven experimentally, but is tempting to speculate that these α-Proteobacteria may use this receptor to eavesdrop on QS-2 communication employed by K. pneumoniae in soil.

Contrasting with other Actinobacteria that bear only the SAH hydrolase for the production of homocysteine using SAH as substrate, both B. longum and B. adolescentis also possess luxS in their genome. These species are key commensals of the human gastrointestinal tract and appear to have acquired luxS by horizontal gene transfer in the gut from other Gram-positive endotherm-associated bacteria (Fig 2). The presence of the other genes coding for the complete AMC and absence of lsrB from the genome of Bifidobacterium suggest that the additional occurrence of LuxS in these strains is not aimed to AI-2 production (albeit an homolog of SMU.408 of S. mutans is present), but represents only a redundant metabolic pathway for the recycling of homocysteine as proposed previously 61. However, the possibility that an eventual AI-2 signal may be used for one-way interspecies communication can not be discounted.

Another non-pathogenic species, E. blattae, was also identified to have both pathways 19, but given its taxonomical classification among Enterobacteriaceae, here the horizontal gene transfer most probably involved sahH instead of luxS. As for Bifidobacterium, also in E. blattae the physiological significance of the double SAH detoxification pathway still awaits clarification.

Previous studies have assessed that at the level of bacterial phyla the phylogenetic tree based on luxS largely corresponds to the one constructed using 16S ribosomal RNA, demonstrating that the former gene is ancestral in groups like Proteobacteria and Firmicutes. ε-proteobacterium H. pylori, spirochete B. burgdorferi and actinobacterium B. longum, which clustered with Firmicutes in the luxS tree, represented the exception to this criterion indicating that horizontal gene transfer has occurred in these species 62. Our result confirm these findings and suggest that also B. adolescentis, or more probably a common Bifidobacterium ancestor, may have undergone an HGT (Fig. 2). In order to understand if and how the AI-2 receptors coevolved with the AI-2 synthesis gene, we first analyzed the distribution of the recognized receptor genes on the entire luxS phylogenetic tree (Fig. 2), then compared the tree based on the lsrB receptor with the luxS and rpoB trees of the matching strains (Fig. 4). As previously reported the presence of the LuxP receptor is fundamentally limited to Vibrionaceae 19 and its evolution corroborates the fact that this kind of receptor has been constantly present during the evolution of this family of marine bacteria (Fig. 2). On the other hand the Lsr complex is widespread (but not ubiquitous) in Enterobacteriaceae and Bacillaceae, as well as present in two species belonging to the family Pasteurellaceae (i.e., P. multocida and H. somnus). Despite the fact that these bacterial groups are fairly separate on an evolutionary level, the branching pattern of the lsrB tree is mostly concordant with the corresponding tree based on luxS and housekeeping gene rpoB (Fig. 4), implicating that this kind of AI-2 receptor genes are probably orthologs that have co-evolved with the AI-2 producer gene. An alternative hypothesis postulating the Lsr receptor complex being a paralog of closely related ABC transporter, such as ribose transporter Rbs, is not supported by the available data which demonstrate that rbs genes of distant related species are more similar to each other, than equivalent rbs and lsr genes in the same species (data not shown). As expected from previous studies on ABC systems 63, also shuffling of genes between different transporters is not supported by phylogenetic analysis (data not shown).

The fact that in no bacteria the components of the AI-2 receptor and signal transduction pathway are found clustered together with luxS in the genome further suggests that the QS-2 system was built upon a pre-existent general metabolic function by some bacteria only, which may help to explain why it is not an universal feature of all species.