Phylogenetically Novel LuxI/LuxR-Type Quorum Sensing Systems Isolated Using a Metagenomic Approach

Abstract

A great deal of research has been done to understand bacterial cell-to-cell signaling systems, but there is still a large gap in our current knowledge because the majority of microorganisms in natural environments do not have cultivated representatives. Metagenomics is one approach to identify novel quorum sensing (QS) systems from uncultured bacteria in environmental samples. In this study, fosmid metagenomic libraries were constructed from a forest soil and an activated sludge from a coke plant, and the target genes were detected using a green fluorescent protein (GFP)-based Escherichia coli biosensor strain whose fluorescence was screened by spectrophotometry. DNA sequence analysis revealed two pairs of new LuxI family N-acyl-l-homoserine lactone (AHL) synthases and LuxR family transcriptional regulators (clones N16 and N52, designated AubI/AubR and AusI/AusR, respectively). AubI and AusI each produced an identical AHL, N-dodecanoyl-l-homoserine lactone (C₁₂-HSL), as determined by nuclear magnetic resonance (NMR) and mass spectrometry. Phylogenetic analysis based on amino acid sequences suggested that AusI/AusR was from an uncultured member of the Betaproteobacteria and AubI/AubR was very deeply branched from previously described LuxI/LuxR homologues in isolates of the Proteobacteria. The phylogenetic position of AubI/AubR indicates that they represent a QS system not acquired recently from the Proteobacteria by horizontal gene transfer but share a more ancient ancestry. We demonstrated that metagenomic screening is useful to provide further insight into the phylogenetic diversity of bacterial QS systems by describing two new LuxI/LuxR-type QS systems from uncultured bacteria.

INTRODUCTION

Bacteria interact with one another using chemical molecules as sensing signals. Detection of the molecules allows bacteria to distinguish between low and high cell population densities and to control gene expression in response to changes in cell number (46) and local environment (12). This process, referred to as quorum sensing (QS), allows a population of bacteria to coordinately control gene expression. Several types of QS signals have been found, including N-acyl-l-homoserine lactone (AHL) in the Proteobacteria (5, 45, 46).

AHL-producing bacteria have been identified in over 37 genera within the Alphaproteobacteria, Betaproteobacteria, and Gammaproteobacteria (9, 11, 14, 32) and the Cyanobacteria (10). In these bacteria, AHL-dependent QS systems have been shown to regulate many bacterial behaviors, such as virulence (8, 48) and biofilm formation (24, 29), mainly in response to cell densities. Therefore, AHL-dependent QS systems are now recognized for playing important roles in the regulation of bacterial behavior.

The AHL-based QS systems usually contain a luxI gene homologue, responsible for the synthesis of AHLs, and a luxR gene homologue, an AHL-dependent transcriptional regulator (19). The LuxI/LuxR-type QS systems have been experimentally identified and studied in more than 70 different species in the phylum Proteobacteria (6, 15). In addition, genome sequencing of a number of cultured Deltaproteobacteria (30, 39) and a yet-to-be cultured bacterium belonging to the phylum Nitrospirae (40) indicates that they may harbor putative LuxI/LuxR-type QS systems. It has been speculated that half the bacterial phyla (26 candidate phyla) do not have cultivated representatives, although at least 52 bacterial phyla have been identified from 16S rRNA gene sequences in environmental samples (33). These results suggest that not only Alphaproteobacteria, Betaproteobacteria, and Gammaproteobacteria but a diverse array of bacteria, including as-yet-uncultivated bacterial phyla, likely possess the LuxI/LuxR-type QS systems, but these possible QS systems have not been shown to be functional. More recently, a bacterial LuxI/LuxR-like QS system found in a methanogenic archaeon, Methanosaeta harundinacea 6Ac, was shown to be involved in regulating cell assembly and carbon metabolic flux (49).

A better understanding of QS systems will provide us with greater insight into the complex interaction mechanisms used widely among the Bacteria and even the Archaea in the environment. This research has been limited by the lack of information on community members without cultivated representatives. However, by using a non-cultivation-based metagenomic approach, new AHL synthase genes have been identified from uncultured organisms (21, 47). Screening of metagenomic libraries constructed from Alaskan soil using the reporter activity of green fluorescence protein (GFP) led to the discovery of a novel LuxI/LuxR-type QS system with low similarity to the known homologues found in Gammaproteobacteria (47). Moreover, metagenomic libraries constructed from activated sludge and soil resulted in the isolation of three new LuxI/LuxR-type QS systems producing previously unknown AHLs most closely related to those previously found in Alphaproteobacteria, Betaproteobacteria, and Gammaproteobacteria (21). These results demonstrated that metagenomic approaches are useful for the discovery of novel QS systems from uncultured bacteria.

In this study, we used a metagenomic approach to find novel LuxI/LuxR-type QS systems in uncultured bacteria belonging to classes other than the previously studied Proteobacteria, such as the phylum Nitrospirae. The metagenomic libraries were constructed from activated sludge of a coke plant wastewater treatment system containing various organic pollutants (22, 44) and forest soil as a source of phylogenetically diverse microbes. An intercellular screening method was used, and GFP fluorescence of the biosensor was detected using a spectrofluorometer. Structural analyses using ¹H nuclear magnetic resonance (NMR), ¹³C NMR, and fast atom bombardment mass spectra (FAB-MS) were used to identify the QS signals produced, and the novelty of QS systems was determined by phylogenetic analysis.

MATERIALS AND METHODS

Bacterial strains and culture conditions.

Metagenomic DNA was cloned into the fosmid CopyControl pCC1FOS (a high-copy-inducible fosmid cloning vector [Epicentre, Madison, WI]) and transferred into Escherichia coli TransforMAX EPI300 cells (Epicentre). E. coli DH5α served as the host for subcloning using plasmids pUC19 and pUC118. E. coli EPI300 and DH5α were cultured in Luria-Bertani (LB) medium at 37°C, and E. coli strain JB525-MT102 (pJBA132) (2) was cultured at 30°C. When necessary, antibiotics were supplied in the following concentrations: chloramphenicol (Cm), 12.5 μg ml⁻¹; tetracycline (Tc), 20 μg ml⁻¹.

Soil and activated sludge samples.

Activated sludge was collected from the aeration tank of a coke plant wastewater treatment facility in Japan (41), and the soil sample was collected from a forest on the grounds of the National Institute of Advanced Industrial Science and Technology (AIST) (Tsukuba city, Japan), located at 36°3′48.7614″N and 140°7′46.9878″E, on 22 November 2007 (soil temperature, 15°C at 5-cm depth). The soil sample was sieved (2-mm mesh size) to remove fine roots, leaves, and other organic debris. After sampling, the soil and the activated sludge were frozen and stored at −80°C until DNA could be extracted.

Construction of metagenomic libraries.

Metagenomic DNA from soil and sludge was extracted as described previously using sodium dodecyl sulfate and proteinase K (23). Extracted DNA was purified, fractionated (around 40 kb), and ligated into pCC1FOS for fosmid cloning following the methods of Rondon et al. (35). Fosmid libraries from sludge and soil were constructed using the CopyControl fosmid library production kit (Epicentre) according to the manufacturer's protocols. The transformants were selected on LB agar plates containing 12.5 μg ml⁻¹ chloramphenicol (LB/Cm) (22). Recombinant fosmids and polymorphisms of the insert DNA were examined by agarose gel electrophoresis of EcoRI-digested purified fosmids from randomly selected E. coli transformants.

Approximately 103,700 colonies from the LB plates were inoculated into 500 μl of LB/Cm (using ∼1,080 96-well plates) and grown at 37°C with shaking overnight. Then, overnight cultures, in groups of 48 clones per well, were dispensed into 96-well microtiter plates (Becton, Dickinson) (total of 23 96-well plates, 1 plate for every 4,608 clones). These plates were stored at −80°C and called the 96-well format libraries.

Screening for biosensor-inducing clones from metagenomic libraries.

To isolate metagenomic clones containing putative synthase genes for QS signals, the luxI luxR GFP indicator system E. coli strain JB525-MT102 harboring the gfp plasmid pJBA132 was used as the biosensor strain. Strain E. coli JB525-MT102 produces an unstable GFP in response to N-butyryl-l-homoserine lactone (C₄-HSL) at a concentration of 1,000 nM, whereas lower levels are needed for N-3-oxohexanoyl-l-homoserine lactone (3-oxo-C₆-HSL) at 1 nM, N-hexanoyl-l-homoserine lactone (C₆-HSL) at 10 nM, N-octanoyl-l-homoserine lactone (C₈-HSL) at 10 nM, and N-3-oxododecanoyl-l-homoserine lactone (3-oxo-C₁₂-HSL) at 10 nM (2).

Clone mixtures (∼0.1 μl) from each well of the 96-well format libraries were inoculated into a set of 96-well plates containing 100 μl of LB/Cm and incubated at 37°C for 18 h without agitation. Then, 400 μl of fresh LB/Cm was added to each well and the plates were incubated at 37°C for 30 min with vigorous agitation at 230 rpm. Subsequently, 0.5 μl of CopyControl induction solution (Epicentre) was added to each well. The solution increased the fosmid copy numbers in E. coli to help promote protein expression in the cells. Cultures were grown with agitation (230 rpm) at 37°C for 5 h, and then cells were removed by centrifugation (1,109 × g, 10 min, 4°C). Supernatant (100 μl) from each well was transferred into fresh 96-well microtiter plates. Then, 100 μl of a 5-fold-diluted overnight culture of the reporter strain, JB525-MT102, was added in equivalent numbers into each well of the 96-well microtiter plates and mixed with the supernatants (average pH 6.6). These plates were incubated at 30°C for 4 h without agitation to induce detectable GFP expression from the reporter cells. After incubation, reporter cells were collected by centrifugation (1,109 × g, 10 min, 4°C) and resuspended in 50 μl of sterile water to eliminate background autofluorescence of the LB medium. A SpectraMax Gemini XS Microplate spectrofluorometer (Molecular Devices, Sunnyvale, CA) was used to measure GFP fluorescence of the cell suspension at an excitation wavelength of 485 nm and an emission wavelength of 538 nm. Between-well comparisons were made by normalizing relative fluorescence unit (RFU) measurements to the optical density at 600 nm (OD₆₀₀) of the reporter strain E. coli JB525-MT102 added to each well. The wells with fluorescence intensities greater than that of the negative control based on a t test (P < 0.01) were selected for further analyses. An E. coli strain harboring fosmid pCC1FOS without any inserted DNA was used as the negative control in this screening. To isolate individual biosensor-activating clones from wells with fluorescence, the 48 clone mixtures were diluted and cultured on LB/Cm agar plates at 37°C. After incubation, 96 colonies were randomly picked, cultured, and screened for GFP fluorescence intensity greater than that of the control.

Sequence analysis.

The genes responsible for production of the biosensor-activating molecules were subcloned using a series of standard recombinant DNA techniques that included BamHI or KpnI restriction enzyme digestion, cloning, and transformation (37). Clones were sequenced using a 454 Life Sciences pyrosequencer. Sequences were assembled using the Roche Newbler program. The program ORF Finder from the National Center for Biotechnology Information (NCBI) (http://www.ncbi.nlm.nih.gov) was used to identify open reading frames (ORFs) that started with ATG, CTG, or TTG and terminated with TAG, TGA, or TAA. To annotate the fosmids, ORFs with putative genes longer than 1,000 bp were translated and compared to the NCBI database using the BLASTP program (1).

Phylogenetic analysis.

The phylogenetic positions of two pairs of LuxI/LuxR homologues, AubI/AubR and AusI/AusR, were determined through alignment against amino acid sequences of LuxI and LuxR families (73 sequences in total) whose functions have been experimentally determined, together with homologous but functionally undetermined sequences (27 sequences of LuxI and LuxR homologues) of the top BLASTP hits (E values of <1e−11) against the NCBI nonredundant protein sequence database. In total, 102 LuxI family sequences, including AubI and AusI, and 102 LuxR family sequences, including AubR and AusR, were independently aligned and used to construct phylogenetic trees. Amino acid sequences were aligned using ClustalW (43), and any gap-containing columns in the alignment were removed. Construction of phylogenetic trees using the neighbor-joining (36) and maximum likelihood methods was performed in the MEGA program (42), and bootstrap analyses were performed with 1,000 replicates in each algorithm.

Extraction, purification, and structure analysis of the QS signals in clones N16 and N52.

Clones N16 and N52 were cultured in LB medium (20 liters) at 30°C for 3 days with shaking (180 rpm). The active molecule was solid-phase extracted from the culture broth using a column packed with Diaion HP-20 (Mitsubishi Chemical Co., Tokyo, Japan) and then eluted from the resin with ethanol. Concentrated extract was subjected to solvent partitioning between water and CH₂Cl₂, and the CH₂Cl₂ fraction was chromatographed on a silica gel column (CH₂Cl₂/MeOH = 20:1). Fractions containing QS signal, confirmed using the biosensor assay, were combined and further purified by silica gel column chromatography, resulting in a pure QS signal compound. The purified QS signals were subjected to reversed-phase analytical high-performance liquid chromatography (HPLC) using a Cosmosil5 C₁₈-MS-II column (4.6 by 250 mm; NacalaiTesque, Kyoto, Japan) and an elution gradient of 30 to 100% MeOH at a 0.6-ml min⁻¹ flow rate for 30 min. Detection was made at 220 nm with a retention time of 27.1 min. Fast atom bombardment (FAB) mass spectra were measured on a JEOL JMS-700 MStation mass spectrometer using glycerol as a matrix. Optical rotation was determined on a Jasco DIP-1000 digital polarimeter in MeOH. Nuclear magnetic resonance (NMR) spectra were recorded on a JEOL JNM-A500 NMR spectrometer at 500 MHz for ¹H and 125 MHz for ¹³C in CDCl₃. Chemical shifts of ¹H and ¹³C in NMR spectra were determined relative to solvent reference peaks δ_H 7.24 and δ_C 79.0 for CDCl₃.

Nucleotide sequence accession numbers.

The nucleotide sequences of inserts from clones N16 and N52 reported in this study were deposited in the GenBank database with the accession numbers AB649138 and AB649139, respectively.

RESULTS AND DISCUSSION

Screening of biosensor-inducing clones.

Metagenomic libraries constructed using DNA from activated sludge and soil samples were composed of 4 × 10⁴ and 6 × 10⁴ clones, respectively. The average insert size was estimated to be 40 kb by restriction enzyme analysis of randomly selected clones. The predicted lengths of total cloned DNA were 1.6 Gb in the activated sludge library and 2.4 Gb in the soil library. Approximately 103,700 metagenomic fosmid clones in 96-well format libraries were inspected using a microplate spectrofluorometer to identify clones that induced GFP expression of the biosensor strain E. coli JB525-MT102. Of all investigated clones, three clones induced GFP expression significantly more than the biosensor strain after 4 h of incubation (Fig. 1). These clones were isolated from the soil metagenomic library (clone N16) (P = 0.005) and the activated sludge metagenomic library (clone N43 and N52) (P = 0.001 and P = 0.009, respectively). Sequence analysis of clone N43 indicated there were no homologues of AHL synthases; therefore, further analyses were not performed at this time.

Fig 1.

GFP fluorescence intensity of the biosensor induced by QS signal-producing clones. The biosensor detects small diffusible signal molecules that induce QS. The extracellular QS signals bind and form a complex with the LuxR transcriptional regulator. The complex binds to the promoter and activates the expression of GFP. Error bars represent the standard deviations of the means for three replicates.

Sequence analysis of clone N16.

Sequence analysis revealed that fosmid pN16 contained a 34,076-bp insert that was assembled using over 16,000 sequences, resulting in 121× coverage. The insert was comprised of 28 ORFs (Fig. 2A; see Table S1 in the supplemental material). A 5-kb region at the 3′ end of the insert contained luxI and luxR homologues (orf26 and orf25, respectively) oriented in convergent transcriptional directions (Fig. 2A). The highest amino acid sequence similarity to the LuxI homologue (ORF26), designated AubI (237 amino acids), was 42%, to a putative LuxI in Geobacter uraniireducens Rf4 (39) (Fig. 2A; see Table S1). The whole genome of a member of the Deltaproteobacteria, G. uraniireducens Rf4, has been sequenced, but the function of luxI and luxR homologues present in its genome has not been experimentally tested. The AubI protein also had 40% similarity to the QS signal synthase protein found in the nearly completed genome sequence of uncultivated Leptospirillum group II bacteria (40). The highest protein sequence similarity to a characterized LuxI homologue was 38%, to CviI in Chromobacterium violaceum ATCC 31532 (28), belonging to the Betaproteobacteria. The transcriptional regulatory gene (aubR) was adjacent to the aubI gene. There was 33% amino acid similarity between AubR and the LuxR family transcriptional regulator in the genome sequence of Geobacter sp. strain FRC-32. The deduced amino acid sequence of aubR (267 amino acids) had low similarity to CviR (31%), one of the most studied LuxR family transcriptional regulators, from C. violaceum (28). The low similarities of the AubI and AubR to any previously described LuxI/LuxR homologues found in the Proteobacteria suggest that they represent a new evolutionary branch of QS systems.

Fig 2.

Physical map and gene organization of fosmids pN16 (A) and pN52 (B). The arrows in the physical map indicate the sizes, locations, and directions of transcription of the ORFs. Black arrows represent luxI family genes; gray arrows indicate luxR family genes. Predicted protein functions of ORFs are listed below the map.

Furthermore, genes related to two-component sensor histidine kinase (orf27) and signal transduction histidine kinase (orf28) were found next to the aubI and aubR genes. The presence of these histidine kinases may indicate the presence of a second QS system in this microbe. A membrane-bound hybrid sensor kinase protein, LuxN, used instead of LuxIR to recognize AHL was first described from the luminous marine bacterium Vibrio harveyi (18). ORF27 and ORF28 have no amino acid sequence similarities to LuxN homologues, but a LuxI synthase annotated as a signal transduction histidine kinase has been found in the methanogenic archaeon (49). Therefore, the presence of histidine kinases is being noted for potential future studies of alternative or additional QS systems.

There were no phylogenetic marker genes in the pN16 insert DNA that could be used to identify the potential source of the DNA. However, 23 of 28 ORFs (from orf1 to orf16 and orf18 to orf24) showed high amino acid sequence similarity (56% to 89%) and gene arrangement similarity to uncultivated “Candidatus Nitrospira defluvii” (27), and the remaining five ORFs containing AubI and AubR had low similarity (31% to 42%) to previously described proteins from the phyla Chloroflexi, Actinobacteria, and Proteobacteria. It is possible that the DNA fragment of fosmid pN16 originated from a bacterium belonging to the phylum Nitrospirae, including “Ca. Nitrospira defluvii” or Leptospirillum group II bacteria. In addition, a new LuxIR-type QS system has been recently identified in a methanogenic archaeon (49). These facts strongly indicate that bacterial phyla other than the Proteobacteria and even other archaea may possess LuxIR-type QS systems that are similar to the system in Proteobacteria.

Sequence analysis of clone N52.

The nucleotide sequence of fosmid pN52 was 37,553 bp long and was assembled using over 11,000 sequences resulting in 78× coverage (Fig. 2B; see Table S2 in the supplemental material). The insert was composed of at least 35 intact ORFs and a truncated ORF. The deduced gene product of ausI (orf4, 202 amino acids) from clone N52 was 57% similar to the LuxI family protein LuxI_QS6-1 of a metagenomic clone from activated sludge (21) (Fig. 2B; see Table S2). The highest amino acid sequence similarity to a characterized LuxI homologue was to PpuI (40%) from Pseudomonas putida. The transcriptional regulatory gene, designated ausR (orf3), was adjacent to the ausI gene in the same transcriptional orientation. The deduced gene product AusR (255 amino acids) was 37% similar to the LuxR family transcriptional regulatory protein LuxR_QS10-1 in a metagenomic clone from forest soil (21). Among experimentally characterized LuxR homologues, BraR in Burkholderia kururiensis had the highest similarity (35%) to AusR.

Of the 36 ORFs on pN52, 32 had significant similarity to proteins from Betaproteobacteria, such as Alcaligenes faecalis and a Delftia sp. (see Table S2 in the supplemental material). A large gene cluster (orf11 to orf30) upstream of ausI and ausR was highly similar to genes responsible for the complete metabolism of phenol to tricarboxylic acid (TCA) cycle intermediates via the meta-cleavage pathway (4, 50). This suggests that these sequences were potentially obtained from a phenol-degrading bacterium, which is highly possible because of the organic pollutants in the coke plant wastewater treatment system where it originated. The effect of AHLs on phenol degradation rates in an activated sludge sample has been reported (44). Genes related to recombination (orf1) and transposition (orf34-36) were found flanking the phenol catabolic gene cluster and ausI and ausR. These mobile genetic elements indicate that this region may have been acquired by horizontal gene transfer, making it difficult to confirm the bacterial source of this fosmid.

Comparison between luxI and lux box-like element sequences in clones N16 and N52.

Comparison of sequences of the two metagenomic fragments from pN16 and pN52 indicated that these DNA fragments were derived from different organisms. The deduced gene product of aubI from fosmid pN16 was 30% similar to the deduced gene product of ausI from fosmid pN52. The G+C content in pN16 (56%) was lower than in pN52 (65%).

Promoter prediction software BPROM (Softberry, Mt. Kisco, NY) identified possible σ⁷⁰ promoters in the upstream regions of both AHL-synthesizing enzyme genes, aubI and ausI (Fig. 3). In addition, a 20-bp length inverted repeat sequence was found close to the −35 box of each σ⁷⁰ promoter. This inverted repeat is typically considered to be a lux box-like element where a LuxR regulator binds to the LuxI/LuxR-type QS systems (3, 13, 38).

Fig 3.

Analysis of lux box-like elements. (A) Location of lux box-like elements upstream of aubI and ausI. (B) Comparison of the elements with known lux box-like elements from Pseudomonas aeruginosa RhlI and LasI (7, 31), Ralstonia solanacearum SolI (17), Aliivibrio fischeri LuxI (16), Acidithiobacillus ferrooxidans AfeI (34), and Burkholderia cepacia CepI (26). Sequences with more than 60% identity are shaded.

The arrangement of the structural and regulatory genes on the two-fosmid clones was similar, but the orientation of transcription differed. The luxI genes (ausI and aubI) were adjacent to their putative luxR regulatory genes (ausR and aubR). However, the aubI and aubR genes were in convergent orientation whereas ausI and ausR were tandem. Adjacent luxI and luxR genes have been frequently observed not only in cultivated bacteria but also in uncultured bacteria (7, 19, 21, 47). The putative LuxR homologues found in this study would likely form a complex with the AHL produced by the LuxI homologues. The complexes would bind to the lux box-like element found upstream of luxI to regulate its expression in the same manner as in previously studied LuxI/LuxR-type QS systems.

Phylogenetic analysis of the AubI/AubR and AusI/AusR.

Phylogenetic analysis using neighbor-joining trees generated from the LuxI and LuxR sequences revealed that AusI and AusR sequences from clone N52 clustered with two metagenomic LuxIR sequences (LuxIR_QS6-1 and LuxIR_QS10-1) within the Betaproteobacteria and Gammaproteobacteria (Fig. 4). In contrast, AubI and AubR sequences from clone N16 were found to be distant from any LuxI or LuxR homologues whose function to synthesize AHL has been demonstrated. Putative LuxI/LuxR homologues found in the Deltaproteobacteria were also distant from other well-studied Proteobacteria and formed a distinct clade. However, AubI and AubR were more deeply branched than the deltaproteobacterial clade and also clearly distinct from a putative homologue found in the Nitrospirae, Leptospirillum group II bacteria (Fig. 4). The LuxI and LuxR trees also indicated that the majority of proteins from the Alphaproteobacteria, Betaproteobacteria, and Gammaproteobacteria did not make distinct clades (Fig. 4) that reflect their phylogeny based on 16S rRNA gene sequences. The phylogenetic positions of AusI/AusR and AubI/AubR were similar in neighbor-joining trees and maximum likelihood trees (see Fig. S1 in the supplemental material).

Fig 4.

Phylogenetic trees of (A) LuxI and (B) LuxR protein sequence homologues obtained from Proteobacteria and Nitrospirae. Sequences obtained from metagenomic clones are indicated in bold type. AubI, AubR, AusI, and AusR, characterized in this study, are enclosed in red boxes. Species belonging to the Alphaproteobacteria, Betaproteobacteria, Gammaproteobacteria, and Deltaproteobacteria are grouped within brackets. The bold brackets represent two groups of phyla. The Leptospirillum sequence was used as the outgroup. Trees were constructed using the neighbor-joining algorithm with bootstrap values obtained from 1,000 replicates. Nodes with bootstrap values of more than 50% are labeled. The scale bars represent 0.1 (A) and 0.2 (B) substitutions per amino acid site. Red branches represent sequences that have been experimentally tested. Abbreviations for bacterial genus names: Ab, Acidithiobacillus; Ae, Aeromonas; Ag, Agrobacterium; Al, Aliivibrio; Az, Azospirillum; B, Burkholderia; C, Chromobacterium; D, Desulfovibrio; E, Erwinia; G, Geobacter; Ga, Gallionella; M, Mesorhizobium; Mb, Methylobacterium; Ni, Nitrosospira; Nc, Nitrococcus; P, Pseudomonas; Po, Polymorphum; R, Rhizobium; Ra, Ralstonia; Rb, Rhodobacter; Rp, Rhodopseudomonas; S, Sinorhizobium; Se, Serratia; V, Vibrio; Y, Yersinia. Accession numbers for all LuxI and LuxR homologues used in this analysis are listed in Table S3 in the supplemental material.

The phylogenetic analysis based on amino acid sequences indicated that AubI/AubR was clearly distant from those in the Proteobacteria and deeply branched in the phylogenetic tree. There has been a hypothesis that luxI and luxR genes frequently spread by horizontal gene transfer within the Alphaproteobacteria, Betaproteobacteria, and Gammaproteobacteria (20, 25). The remote phylogenetic position of AubI/AubR suggests that the genes were not recently transferred from the Proteobacteria to the Nitrospirae. This critical finding also indicates that the QS systems did not originate within the Proteobacteria but likely evolved prior to the divergence of the phyla Proteobacteria and Nitrospirae. However, the ausI and ausR genes for the other positive clone, N52, were also novel but fell within the Betaproteobacteria.

Characterization of quorum sensing signals from clones N16 and N52.

Structural analyses of the QS signal identified from fosmid clones N16 and N52 using ¹H NMR, ¹³C NMR, and FAB-MS spectra revealed that both of them direct the synthesis of an AHL signal, C₁₂-HSL (see Fig. S2 and S3 in the supplemental material). The proton NMR spectrum of the active molecule showed typical AHL signals (Fig. S2A and S3A). There were no olefinic protons observed in the spectrum, suggesting that the acyl chain was saturated. In the ¹³C NMR spectrum there were only two signals around the carbonyl carbon region, indicating the absence of a keto group on the acyl chain (Fig. S2B and S3B). This speculation was supported by two-dimensional (2D) NMR data, including correlated spectroscopy (COSY), heteronuclear multiple quantum coherence (HMQC), and heteronuclear multiple-bond correlation (HMBC) (data not shown). FAB-MS analysis clearly determined the chain length to be C₁₂ based on ion peaks at m/z 282.2 (negative mode) (Fig. S2C and S3C). Both ¹H and ¹³C NMR spectra of the synthetic C₁₂-HSL could be superimposed with data of the compound from these metagenomic clones. Additionally, analytical reversed-phase HPLC and NMR analysis confirmed its identity with authentic C₁₂-HSL with a retention time of 27.1 min. Finally, comparison of the optical rotation value with that of a synthetic standard confirmed that its absolute stereochemistry was a 2S configuration. The aubI gene from clone N16 and ausI from clone N52 have been subcloned, and structural analyses of each subclone showed similar production of C₁₂-HSL (data not shown). Although sequence similarity was low between AubI and AusI, their QS signals were identical to that of C₁₂-HSL (Fig. S2 and S3).

The deduced amino acid sequences from clone N16 supported our hypothesis that it originated from an uncultured bacterium belonging to the phylum Nitrospirae (40). However, the regulatory gene arrangements and structure of the QS signal produced were similar to those of the LuxIR-type QS system mediated by an AHL (C₁₂-HSL). In conclusion, these results demonstrate that metagenomics can be used to discover LuxI/LuxR homologues from yet-to-be cultivated bacterial phyla in order to gain a greater understanding of bacterial interactions in natural ecosystems.