Quorum sensing is a widespread mechanism of intercellular communication among bacteria controlling its gene expression as a function of cell density. Two major quorum sensing pathways, with characteristic signalling molecules, AI-1 and AI-2, have been identified[1]. LuxS is one of the principal components in the biosynthetic pathway of AI-2, the universal signal for bacterial inter-species communication. In some organisms, quorum sensing by LuxS has been shown to have a profound effect on pathogenicity by affecting toxin production or flagellar morphogenesis and hence motility and colonization. However, in some other species it has no direct pathogenic role and known to affect the metabolism. LuxS is a small metalloenzyme with Zn(II) ion at its active site, which was later predicted to be Fe(II) by Zhu et al[2]. Sequence alignment studies have indicated the presence of an invariant HXXEH motif[3] and the metal ion is tetra-coordinated by the side chains of the two histidines of this motif and cysteine, the fourth coordination site being occupied by a water molecule[4, 5]. LuxS was shown to exist as a homodimer in its biologically active form by dynamic light scattering studies[6], the two identical active sites being formed at the dimer interface[5]. This protein catalyzes the non-redox cleavage of a stable thioether bond in the SAM (S-adenosylmethionine) cycle to produce AI-2 as shown schematically in Figure 1. Such enzymatic reactions are rare in nature and these enzymes require redox-active cofactors. However, LuxS is unique as it does not require any redox-active cofactor to catalyze this chemically difficult reaction[7].

Protein association plays a dominant functional role such as specific recognition, signal transduction, regulation of gene expression and studies have shown that the symmetry in an oligomer plays an important role as it defines the overall architecture of the protein, and thus its function[8]. LuxS is a biologically active homodimer and extensive research has been carried out to explain the principles underlying protein association, with an aim to characterize the interface[9]. Characterization of the protein-ligand interface using network approach has given important insights into the effects of ligand binding and chain connectivity on network communication in the protein dihydrofolate reductase (DHFR)[10]. The dimerization domain was shown to be very important in relation to the structure and function of LuxS as is evident from the occurrence of the active site at the dimer interface. Yet another evidence for the importance of the homodimer is apparent from the studies of electrostatic potential and conserved residues on the monomer surface[5]. The crystal structures of LuxS from only four organisms are available in the PDB in their dimeric states, and it has been proposed to contribute a unique fold to the alpha-beta family of proteins[5, 11]. Interestingly the protein structures superpose very well at the backbone level (Additional file 1: Table S2) and this information does not provide any clue to the diversity in the function of LuxS. In the present study, we have highlighted the intricate details of side chain interactions at the dimeric interface, in order to distinguish LuxS structures from different sources.

Graph theoretical approaches have proved to be useful in capturing the side-chain interactions in a collective way through the representation of protein structures in the form of protein structure networks (PSN)[12]. For example, the analysis of the interface clusters in PSN has given insights regarding the sequence signature motifs responsible for the different types of quaternary association in legume lectins[13]. Also, structural domains and domain interface residues in multi-domain proteins were successfully identified by graph spectral analysis[14]. Protein network analysis has also provided insight into the domain-domain interactions in proteins in the context of allostery[15]. The importance of minor side-chain variations in the light of 'allostery without a backbone conformational change' has also been established[16]. Recently, crucial residues involved in the oligomerization of M. smegmatis Dps were predicted from graph based method and has been experimentally proven to be crucial for the functional oligomeric state of the protein[17]. The aim of the present study is to exploit the advantages of PSN graphs for the characterization of the interface in case of LuxS from bacteria of various species. Our study has revealed the potential of PSN graphs as an analytical tool in judging differences in side-chain interactions; the backbone level having almost no differences at all. The interfaces are characterized in several selected structures (X-ray structures as well as the modelled ones) and the results are extended to LuxS family of proteins in general. An indirect correlation between the dimeric interface organization and the biological function such as pathogenesis is brought out from these studies. Additionally, potential sites of mutation at the interfaces are also identified. Some of them are shown by previous studies to be structurally and functionally important residues in LuxS.

Bacteria have efficiently imbibed mechanisms for communication within themselves and the environment. Quorum sensing is a mode of such crosstalk as a function of cell density and has been divided into two categories (quorum sensing system 1 and 2) with their characteristic signalling molecules, autoinducer 1 and 2 respectively. LuxS is the precursor protein of autoinducer 2. Although the quorum sensing, in general, can directly influence pathogenicity by controlling bacterial gene expression as a function of cell density, it is the unique feature of LuxS induced quorum sensing system to control bacterial metabolism. So LuxS driven quorum sensing has the ability to control not only pathogenicity but also the 'overall well-being' of the organism. The effect of LuxS mutation manifests in different ways in different organisms. As an example, for the organisms like H. influenza, S. epidermis, and H. pylori, E. coli, it has been shown that mutation of LuxS affect biofilm development and metabolism/motility/flageller morphogenesis[20, 21] respectively (affecting metabolism), whereas for organisms like V. Cholera, C. perfringens, the effect was shown to be related to toxin production[21–23] (direct influence on pathogenicity).

In the present study we have analyzed the quaternary association features of the available and modelled structures of LuxS from 23 different organisms. Our study indicates the presence of distinct key dimer association features in the LuxS protein from different species related to the biological manifestation of the protein. This fact has enabled us to classify our dataset into six categories and perform a detailed analysis in order to characterize their interface. A simple analysis of the protein structures over our dataset at the backbone level is largely elusive and predicts the structures to be highly similar. However, finer differences exist at the side-chain level, which has become evident through our analysis of the interface clusters. A detailed description of cluster sizes, number of residues in the clusters is provided in the Additional file 1: Table S1. There is a wide variation of the interface cluster size and their orientation, although their backbone rmsds are comparable (Additional file 1: Table S2), which accounts for the range of finer diversities existent within the structure of the same protein from different organisms.

Our analysis has also enabled the prediction of signature motifs of completely/partially conserved sequentially apart but structurally proximal residues at the dimer interface for each class. Since these signature motifs appeared because of their spatial proximity, it would not have been possible to identify these motifs by sequence comparisons alone. Based on similar cluster analysis studies, it was predicted previously that the quaternary association of lectin from Lotus tetragonolobus did not correspond to any known type of association pattern[24] and indeed a new type of association was later discovered for this protein by crystallographic studies[25]. This type of analysis would at least enable, in this case, to predict, a priori, the probable physiological function of the LuxS protein from a given organism by superposition of interface clusters (described in details in the Methods section) and comparison of the motif obtained with the signature motifs from the different classes. These identified motifs can serve as possible epitopes for LuxS inhibitor design as they are highly vulnerable structurally conserved regions of the protein within a particular class. However, for organisms sharing the same niche, e.g. belonging to the enterobacterial group, like Helicobacter pylori, Escherischia coli and Salmonella typhimurium (results not shown), it becomes exceedingly difficult to target the LuxS of a particular organism without any effect on the other. On close inspection of the side chain interface clusters of these three organisms, it was found that they are mostly superposing implying a very large similarity in their structures even at the side chain level. So it is not possible in such cases to target one without influencing the other. However, as discussed later in the context of the mini-triad motif, some level of discrimination is possible.

Our study reveals the presence of a triangular pattern at the interface that is constituted by three most important clusters as shown in Figure 2. The probiotics and some extremophiles and class III members manifest an exception as this geometric pattern is distorted in them. Probiotics are beneficial bacteria and includes Lactobacillus, and Bifidobacterium species. They possess numerous potential therapeutic properties including anti-inflammatory and anti-cancer activities and other features of interest[26]. In recent years, studies with in vitro cell culture and animal models clearly demonstrated protective effects of probiotics for anti-tumor and anti-cancer effects[27]. It was shown previously that L. acidophilus secreted molecule(s) that either inhibited Quorum sensing signal or underwent direct interaction with bacterial transcriptional regulators, controlling the transcription of enterohemorrhagic E. coli O157 genes involved in colonization[28]. However, both the genomes have the luxS gene homologue. It was also proposed that AI-2 might interfere with the normal functioning of commensal organisms in the gut by competing with their communication and thus, taking colonization advantage[29]. So it is intriguing to note the difference of the geometric pattern of distribution of the active sites on the dimer interface as is captured by our interface cluster analysis. It is striking that the probiotics actually contained the LuxS protein that contributed to virulence and related phenomenon in several bacterial species. Thus, it is proposed that this geometric pattern at the dimer interface might play an important role in the behavioural outcome of LuxS in relation to its function; absence of which distinguishes the probiotics from the rest of the organisms considered in our study in terms of function.

Another important observation from our analysis is the presence of a mini-triad at the apex of the dimer which presented itself in the interface clusters throughout for all the structures in our dataset except for class III (LuxS from Bacillus sp.) in which it is absent from the interface cluster. However, this set of residues does not show up as conserved in multiple sequence alignment studies of the 23 proteins in our dataset. Its position in the structure at the apex of the dimer and also the repeated occurrence in the protein interface cluster indicate that it might be an important site for mutation; we hypothesize that their mutation might reorient or distort the dimerization interface. Also, we find that there is a marked difference in the chemical nature of the residues constituting the mini-triad motif between H. pylori (GXNXD) and that of S. typhimurium and E. coli (TXHXD), although they share a very high similarity even at the side-chain level. This mini-triad is a highly accessible target for small molecules as they are in the exposed regions of a loop at the apex of the protein dimer. These small molecules may be potential drugs leading to a perturbation at the dimer interface, targeting specifically E. coli and S. typhimurium LuxS but not H. pylori LuxS.

Graph theoretical analysis is helpful to visualize the actual global connectivity in the protein structure network at a given interaction strength. From this analysis, we predict that it is an ensemble of amino acid residues that dictates the function of LuxS. So we predict few potential sites of mutation in the LuxS protein. Some of our predictions of possible sites of mutation have been already experimentally justified; some would warrant experimental verification.

The results of hub analysis reveal a special characteristic feature in the extremophilic class (class IV); the recurrent presence of a core hub at a high interaction strength (I_min = 12%). Furthermore, the asymmetry of the interface discussed before also manifests within the core of the protein as shown by the presence of the "persistent hub" in one chain and not in the other. It was shown previously that hubs in real world networks provide robustness to such networks[30]. Previous studies have also demonstrated that the thermophilic proteins contained more hubs than their mesophilic analogues[31] and these hubs were found to be important for the overall integrity of the structure of such proteins. The "persistent hub" and its connecting residues are thus recognized as potential sites for de-stabilizing mutations in the LuxS proteins from extremophiles (class IV).

The LuxS family of proteins (there are 202 proteins with annotated sequences in UNIPROT database. In this regard our dataset is biased by the available sequence information) mostly share a high sequence similarity among various species, subsequently sharing a high level of structural similarity at the backbone level. Also there is over-representation of some species such as Streptococcus, Lactobacillus in our available dataset as shown in Additional file 3: Figure S2. However, among these 202 LuxS proteins we find a wide variety of bacterial species such as the pathogens, commensals, probiotics and the most prominent representatives of the human intestinal microbiota (bacteroidetes and firmicutes). In a nutshell, our analysis reveals the extent of finer structural diversity that can remain hidden within the structure of the same protein from different organisms with the same quaternary association and the same fold at the backbone level. Our method is robust enough to capture minor differences of side-chain orientation at the level of quaternary association, which gives rise to such diversity in function globally. However, LuxS orthologues are present in many more organisms. Metagenomic approaches, in which the genetic material collected from various environments is cloned directly into surrogate hosts, has given us a key to the huge repository of genetic diversity in bacteria. Such approaches will definitely give an even more descriptive idea about the structural conservation[32] of the AI-2 precursor protein in the human gut, bustling with a microbial agora.

Our analysis focuses on the structural properties of a single protein across a dataset to summarize its effect on the typical phenotypic/biological features. However the recent metagenomic data curations have suggested the presence of trillions of organisms in the human intestinal microbiota[33, 34]. But not all of these organisms contain the LuxS proteins and AI-2 does not have a signalling role in all the organisms. This definitely narrows down the complexity but then also the system is too complex to be predicted at the structural level of a single protein. A systems biology approach considering the whole network of events has to be employed to decipher the complex puzzle of the human intestinal microbiota. Our method is a contribution to this converging pipeline in terms of the fact that our classification at the structural level can be used to group similar categories for a larger dataset. But nevertheless to understand the overall machinery of LuxS activity and to target LuxS clinically, we need a more holistic view at the systems level.

The protein structure network graphs are constructed for the 23 LuxS proteins considered in our dataset. Our results indicate presence of completely/partially conserved signature motifs at the dimer interface of the LuxS protein for class (II–VI) in relation to their biological function. Our results show the presence of an ensemble of residues at the active site, all of which are required for the proper functioning of LuxS as S-ribosylhomocysteinelyase and thus potential sites of mutation are predicted. We also identify a definite geometric pattern in the form of an isosceles triangle at the dimer interface which we predict to be important for LuxS function. Also, a mini-triad formed by the residues of both the subunits, arising from a stretch of five residue amino acids is found at the apex of the dimer (for class I–II, and class IV–VI) and is predicted to have structural and thus, functional implications which would call for experimental verifications. This mini-triad on an exposed loop of the protein may be a potential epitope for the design of specific inhibitors of LuxS enzyme. Our characterization of the dimer interface for the probiotic LuxS also makes prediction of their notable characteristic structural features. Also the results of hub analysis reveal the presence of a core hub that is characteristic of the LuxS from extremophiles (class IV) and are proposed to contribute robustness to the system. So our analysis emphasizes the importance of the implications of differences in side-chain orientation on functional diversity. To conclude, our study is an attempt to bridge the structure of LuxS protein to its function with special emphasis on the interactions at the dimer interfaces; and hence towards the general phenomenon of quorum sensing in bacteria.