Quorum sensing provides the basis for coordinating community-wide, microbial behaviors in many mesophilic bacteria. However, little attention has been directed toward the possibility that such phenomena occur in extremely thermal microbial environments. Despite the absence of luxS in hyperthermophile genomes, autoinducer-2 (AI-2), a boronated furanone and proposed ‘universal’ interspecies mesophilic bacterial communication signal, could be formed by Thermotoga maritima and Pyrococcus furiosus through a combination of biotic and abiotic reaction steps. AI-2 did not, however, induce any detectable quorum-sensing phenotypes in these organisms, although transcriptome-based evidence of an AI-2-induced stress response was observed in T. maritima. The significance, if any, of AI-2 in hydrothermal habitats is not yet clear. Nevertheless, these results show the importance of considering environmental factors, in this case high temperatures, as abiotic causative agents of biochemical and microbiological phenomena.
In mesophilic, microbial habitats, many community-wide behaviors, including biofilm formation, virulence, swarming, competence, sporulation, and luminescence, are coordinated through a class of extracellular small molecules termed autoinducers by a process known as ‘quorum sensing’ (Waters & Bassler, 2005). Autoinducer-2 (AI-2), a boron-chelated, furanone derivative of 4,5-dihydroxy-2,3-pentanedione (DPD), has been postulated to be central to a universal, interspecies mode of cell–cell communication in many mesophilic marine and terrestrial microorganisms, living either independently or in a host-associated (parasitic or symbiotic) manner (Sun et al., 2004).
Considering the ubiquity of quorum sensing in mesophilic niches, the question arises as to whether similar mechanisms would exist for conserved coordinated behaviors in extreme environments. However, the existence and/or the nature of quorum sensing, universal or species-specific, in extreme microbial environments, is largely unknown (Paggi et al., 2003; Johnson et al., 2005). Indeed, biofilm formation and symbiotic interactions have been documented in both hyperthermophilic bacteria and archaea (LaPaglia & Hartzell, 1997; Huber et al., 2002; Pysz et al., 2004a; Johnson et al., 2005, 2006). In fact, species-specific, peptide-based quorum sensing in hyperthermophilic syntrophic cultures of fermentative anaerobes and methanogens has been observed (Johnson et al., 2005). However, to date, no evidence exists for the presence of AI-2 in hydrothermal environments. Here, we demonstrate that two model hyperthermophiles, a bacterium, Thermotoga maritima (Topt of 80 °C), and an archaeon, Pyrococcus furiosus (Topt of 100 °C), mediate AI-2 formation, despite lacking the canonical set of biosynthetic enzymes, Pfs and LuxS.
Materials and methods
Growth of microorganisms and preparation of cell-free extracts
Thermotoga maritima (DSM strain MSB8) and P. furiosus (DSM strain 3638) were cultivated anaerobically at 80 and 85 °C, respectively, in BSMII media, following the protocols described previously (Johnson et al., 2005). Cells were harvested by centrifugation, washed with 100 mL of isotonic buffer (40 g L−1 sea salts, 3 g L−1 PIPES, and 1.25 g L−1 NaCO2CH3, pH 7.0), lysed by sonication in 25 mL of 20 mM Tris-HCl, pH 7.5, containing 50 mM NaCl and 2 mM dithiothreitol, clarified by centrifugation, and dialyzed against lysis buffer at 4 °C.
Preparation of enzymes
Thermotoga maritima nucleoside phosphorylases (TM1596, TM1737) and phosphosugar mutase (TM0167) were obtained from Dr Scott Lesley at the Joint Center for Structural Genomics (La Jolla, CA, http://www.jcsg.org) as His6-tagged constructs in the pBAD derivatives pMH1 (TM1737) and pMH4a (TM0167 and TM1596) (DiDonato et al., 2004). TM0167 and TM1596 were expressed in Escherichia coli HK100 (Invitrogen), while TM1737 was expressed in E. coli Rosetta (DE3) (Novagen). One-liter cultures with Luria–Bertani (LB) (TM0167) or Terrific Broth (TB) media (TM1596 and TM1737) were initiated with a 50-mL inoculum and grown at 37 °C with appropriate antibiotics to an OD600 nm of c. 0.6. Cells were induced with l-arabinose (0.2% w/v final) and maintained at 37 °C for an additional 3–4 h. Pyrococcus furiosus 5′-methylthioadenosine phosphorylase II (PF0016) was cloned by PCR and ligated into pET15b (Novagen). Pyrococcus furiosus phospho-sugar mutase (PF0588) was obtained from Dr Michael Adams at the University of Georgia (http://www.secsg.org). Both constructs were transformed into Rosetta (DE3) for expression. One-liter cultures with LB containing 50 μM isopropyl-1-thio-β-d-galactopyranoside (IPTG) and appropriate antibiotics were initiated with a 20-mL inoculum and grown for 20 h at 30 °C. Escherichia coli pfs (S-adenosylhomocysteine nucleosidase) and luxS (S-ribosylhomocysteinase) were cloned from E. coli strain RK4353 (Stewart & MacGregor, 1982), ligated into pET28b (Novagen), and transformed into BL21 (DE3) (for LuxS) or BL21 (DE3) Tuner (for Pfs) (Novagen) for expression. One-liter cultures with LB (Pfs) or TB (LuxS) media containing appropriate antibiotics were inoculated with 20 mL of an overnight culture and grown at 37 °C to an OD600 nm of c. 0.6. Cells were induced by the addition of IPTG to 250 μM (Pfs) or 500 μM (LuxS) and grown for a further 6 h at 30 °C.
All recombinant cells were harvested by centrifugation and washed with phosphate-buffered saline (PBS). Cells were lysed by sonication in 30 mL PBS and, except for Pfs and LuxS, heat treated at 65 °C for 20–45 min to precipitate E. coli proteins. Proteins were purified by Ni2+-affinity chromatography using a 5-mL HisTrap FPLC column using the method described by the manufacturer (GE Healthcare). Further purification was performed as necessary using a Q-sepharose anion exchange column (in 50 mM Tris-HCl 8.0 with a 0–1 M NaCl gradient). Purified proteins were dialyzed against PBS. Before use, LuxS was incubated with 100 mM EDTA for 30 min at 4 °C. EDTA was removed by dialfiltration, following which the protein was incubated with 5 mM Fe2SO4 for 1 h at 4 °C. Residual Fe2+ was removed by gel filtration using a PD10 desalting column (GE Healthcare) equilibrated with 20 mM Tris-HCl, pH 7.5, and 100 mM NaCl.
AI-2 reporter assay and AI-2 production
In vitro AI-2 synthesis reactions were performed at 80 °C in a buffer of 20 mM potassium phosphate, pH 7.4, unless otherwise stated (see Figs 1 and 2). Thermotoga maritima and P. furiosus cell extracts were buffer exchanged into 20 mM potassium phosphate, pH 7.4, or 10 mM Tris-HCl, pH 7.5, with a PD10 column before use. S-ribosylhomocysteine (SRH) was generated by incubating 2 mM S-adenosylhomocysteine (SAH) with E. coli Pfs (5 μM) in 20 mM Tris-HCl, pH 7.5, for 2 h at 30 °C, followed by filtration through a YM-10 ultrafiltration membrane (Millipore Inc.). The production of AI-2 was monitored utilizing the Vibrio harveyi reporter strain MM32 (ATCC #BAA-1121, genotype luxN∷Cm, luxS∷Tn5kan), as described elsewhere (Miller et al., 2004). Where necessary, strain BB152 (luxM∷tn5kan) (Bassler et al., 1997) was used as a positive control for AI-2 production. All V. harveyi strains were provided by Dr Bonnie Bassler, Department of Molecular Biology, Princeton University, Princeton, NJ. All were grown in autoinducer bioassay (AB) medium, with the respective antibiotics (chloramphenicol 20 μg mL−1 and kanamycin 50 μg mL−1) added when necessary.
In general, assays combined 20-μL aliquots of each reaction with 180-μL AB medium and 5 μL of MM32 overnight culture, and were carried out in dark-walled, 96-well plates (Corning) incubated for 12–15 h at 30 °C. To limit the slow degradation of ribose-5-phosphate, these reactions (and controls) were incubated for only 4 h with a 1 : 4 dilution of an overnight culture of the reporter strain. Luminescence was quantified with a PerkinElmer HTS 7000 Plus microplate reader using an integration time of 150 ms, and a gain of 150. Experimental and control samples were always assayed in parallel using the same overnight culture of MM32.
Reactions (1 mL) were prepared based on the methods described in the text. Positive controls utilized AI-2 derived from SAH using both Pfs/LuxS as well as cell-free extracts of E. coli BL21 (DE3); 10 μM Pfs was incubated with 2 mM SAH in 20 mM Tris-HCl, pH 7.5, in a volume of 500 μL for 2 h at room temperature. The reaction was then diluted to 1 mL with LuxS (to 20 μM) in 20 mM Tris-HCl, pH 7.5, and incubated for a further 6 h at 37 °C. SAH (1 mM) was also incubated with 1 mg mL−1 BL21 cell-free extracts in 20 mM Tris-HCl, pH 7.5, for 6 h at 37 °C. Before use, 500 μL of BL21 extract was buffer exchanged into 20 mM Tris-HCl, pH 7.5, and 100 mM NaCl. TM0167 and TM1737 (5 μM each) were incubated with 1 mM adenosine in 20 mM potassium phosphate, pH 7.4, for 90 min at 80 °C. Ribose-5-phosphate and ribose (1 mM) were each incubated for 3 h at 90 °C in 10 mM NaHCO3, pH 8.0. All reactions also contained 5 mM borate, and the presence of AI-2 was verified by bioluminescence induction. Following incubation, all samples were filtered through a YM-10 ultrafiltration membrane, and the filtrates were dried under vacuum. Samples were subsequently extracted with 1 mL methanol and stored at −70 °C. Samples (1 μL) were injected (splitless) at 300 °C on an HP Model 5890, Series II GC, with a DB-5 MS column (Agilent) and a GCD-EI detector and chromatographed in a mobile phase of He at 1 mL min−1 over a 40–250 °C thermal gradient (10 °C min−1).
Transcriptional response of T. maritima to abiotically produced AI-2
Transcriptional response analysis of T. maritima to AI-2 was performed using a continuous culture bioreactor in Rinker defined media with 3 g L−1 maltose using a 1-L 5-neck round-bottom flask at a dilution rate of 0.32 h−1, following previously described methods (Pysz et al., 2004a). Cultures were dosed with two different concentrations of AI-2 derived from ribose-5-phosphate, or a control sample containing equimolar concentrations of ribose and phosphate. Experiments were designed to yield theoretical maximum concentrations of 10 and 100 μM AI-2 or 25 μM ribose (control). Samples were prepared as 100 × stocks and incubated for 4 h at 90 °C in the presence of a fivefold excess of borate (and equimolar potassium phosphate for the ribose control) in 10 mM NaHCO3, pH 8. Both the culture and 1 L of culture feedstock were dosed after c. five volume changes, at a cell density of c. 1 × 108 cells mL−1, with 10 mL of concentrated sample each. The concentration of inorganic phosphate released was calculated with malachite green and used to estimate the final concentrations of AI-2 in the cultures; these were c. 4 and 38 μM for the low- and high-dose experiments, respectively. The concentration of ribose and phosphate in the control sample was 25 μM (each) in the culture. Cells were harvested 1 h postexposure. RNA extraction, purification, and subsequent transcriptional analysis were performed as described previously (Pysz et al., 2004a, b; Conners et al., 2005; Johnson et al., 2005).
Results and discussion
The apparent ubiquity of quorum sensing in environmentally moderate habitats raises the following question: are the same behaviors observed in extreme environments, and if so, are they triggered and controlled by similar mechanisms? Previous work suggesting the possibility of a peptide-based quorum-sensing system in T. maritima (Johnson et al., 2005) led us to screen for the existence of other possible quorum-sensing systems among hyperthermophiles. Because of inherent thermolability, the possibility of acyl-homoserine lactone-based signaling in hydrothermal systems appears unlikely. In fact, exhaustive blast searches revealed no apparent orthologs to mesophilic acyl-homoserine lactone synthetases in (hyper)thermophile genome sequences. Orthologs to several genes of the E. coli AI-2 transporter (lsr) could be found in the T. maritima genome (Sun et al., 2004), although these are also known pentose transporters (Conners et al., 2005; Nanavati et al., 2006). To date, the only extreme thermophile whose genome encodes an ortholog to luxS is Thermus thermophilus (Topt of 70 °C), suggesting that AI-2 is predominantly a mesophilic phenomenon.
Although genome sequence analyses suggested that T. maritima and P. furiosus would utilize neither acyl-homoserine lactone- nor AI-2 based-signaling, a response to the V. harveyi AI-2 reporter assay was noted in spent culture media from each of the two hyperthermophiles (Fig. 1). This response could have arisen from abiotically generated AI-2 or from other closely related 4-hydroxy-3-(2H)-furanone(s). Indeed, previous reports have demonstrated that a positive signal in the V. harveyi reporter assay can also be induced by related furanones, such as 5-methyl-4-hydroxy-3(2H)-furanone (MHF) and 2,5-dimethyl-4-hydroxy-3(2H)-furanone (Schauder & Bassler, 2001; Schauder et al., 2001). Subsequent experiments were, therefore, directed at determining whether or not either T. maritima or P. furiosus was synthesizing AI-2.
As described above, neither organism's genome encodes orthologs to LuxS, the enzyme that catalyzes the terminal step in the canonical biosynthesis pathway for AI-2 – the cleavage of SRH. However, the T. maritima genome does have an ortholog to the gene coding for Pfs – the enzyme responsible for the conversion of SAH to SRH. Furthermore, both organisms contain SAH hydrolases, the enzyme that catalyzes the cleavage of SAH to adenosine and homocysteine, thereby bypassing the canonical AI-2 biosynthesis pathway. Previous reports suggest that bacteria contain either a gene for LuxS or the gene for SAH hydrolase (Sun et al., 2004). Taken together, it was not surprising that experiments with soluble cell extracts indicated that SAH was not involved as a precursor during the synthesis of AI-2 by either T. maritima or P. furiosus. However, experiments with SRH raised the possibility of a LuxS isoenzyme (Fig. 2a). Addition of borate, found in oceanic waters at a mean concentration of c. 0.4 mM (Coughlin, 1998), amplified the signal, although this also significantly boosted abiotic SRH conversion to detectable levels of AI-2 (Fig. 2b). Unfortunately, repeated attempts at fractionating cell extracts in order to isolate the enzyme(s) responsible were unsuccessful.
Taking a different tact, incubation of adenosine with soluble cell extracts from either organism produced a strong, temperature-dependent signal in the V. harveyi reporter assay (Fig. 3). This is a departure from the canonical AI-2 biosynthetic pathway from SAH, which is a two-step process involving the initial depurination of SAH to form SRH, followed by release of homocysteine from the ribose 5-carbon and concomitant rearrangement of ribose to DPD. DPD spontaneously rearranges to yield AI-2 (Henke & Bassler, 2004). Observations that adenine was being released in our assays, and that inorganic phosphate enhanced apparent AI-2 production (as judged by an increased signal in the reporter assay), suggested nucleoside phosphorylase activity – known to catalyze the cleavage of nucleosides to a free base and ribose-1-phosphate (Tozzi et al., 2006) (data not shown). It was hypothesized that the observed formation of AI-2 in these assays might be the result of thermally induced rearrangement of phosphorylated ribose. MS analysis of in vitro reactions, involving cell-free extracts, suggested the presence of DPD, adenine, and phosphorylated ribose. However, the terminal step in canonical AI-2 formation from SAH occurs via a C5-substituted ribose derivative (Henke & Bassler, 2004), and multiple reports have indicated that DPD formation from phosphorylated pentoses requires the sugar to be substituted at the fifth carbon (Crout & Hadfield, 1987; Blank & Fay, 1996; Hauck et al., 2003). Thus, it was concluded that the observed conversion of adenosine to AI-2 in these assays might also involve the activity of a phosphosugar mutase, converting ribose-1-phosphate to ribose-5-phosphate, which would then spontaneously convert to DPD in the presence of heat.
To test this, genes encoding a nucleoside phosphorylase (TM1737 and PF0016) and phosphosugar mutase (TM0167 and PF0588) were cloned from T. maritima and P. furiosus, respectively, and tested using assays similar to those performed with cell-free extracts. Figure 4 shows that, as with cell-free extracts, in the presence of heat, adenosine can be converted to AI-2 or a related furanone. GC-MS analysis of in vitro reactions using T. maritima nucleoside phosphorylase and phosphosugar mutase showed a predominant peak, with a similar fragmentation pattern, at the same position as that obtained from AI-2 produced using purified E. coli LuxS and Pfs (Fig. 5). It was interesting that the observed dose response in our experiments with the V. harveyi AI-2 reporter assay is consistent with that reported for MHF (Schauder & Bassler, 2001; Winzer et al., 2002a). It is possible that MHF is the predominant furanone formed in these in vitro assays – both AI-2 and MHF can result from DPD formation, and AI-2 is likely formed in the presence of borate simply by being ‘trapped’ by the borate (Winzer et al., 2002a). Indeed, DPD can yield a variety of final compounds, depending on the environment (Blank & Fay, 1996). While MHF/AI-2 are favored in this assay, in thermophilic habitats a range of furanones could be formed through a variety of mechanisms.
The observed positive signal in the V. harveyi AI-2 reporter assay was likely a consequence of furanones produced from the thermal degradation of pentose phosphates. The in vitro results here can be attributed to thermal degradation of ribose-5-phosphate to DPD. Ribose-5-phosphate is a common, even central, metabolic intermediate, and the enzymes associated with it are by no means novel – nucleoside phosphorylases and phosphosugar mutases have been well studied and are common to mesophiles and thermophiles alike. Certainly, this is not the only way furanones could be produced in thermophilic habitats – the incubation times for in vitro assays are much shorter than what is experienced by growing cells, and there would be more than one in vivo route to furanone formation both inside and outside the cell (Blank & Fay, 1996). Furanones are abundant in nature and their spontaneous formation, generally through Maillard adducts of sugars with amino acids, is well documented (Slaughter, 1999). There are, indeed, many potential substrates. High-temperature bacteria, such as T. maritima, produce a number of glycoside hydrolases and a series of sugar-specific ABC transporters, allowing it to grow on and utilize a plethora of carbohydrates, and suggesting that a range of sugars are available in its environment (Conners et al., 2006; van de Werken et al., 2008).
The chemistry of furanones, such as the AI-2 molecule, is consistent with abiotic formation in hydrothermal environments. However, the potential significance of furanones in these environments is not clear. Furanones can be innocuous, beneficial, or deleterious (Slaughter, 1999; Gonzalez & Keshavan, 2006). When T. maritima cultures were exposed to ribose-5-phosphate that had been incubated at a high temperature of 90 °C for 4 h, to mimic biological processes using a prime source of DPD (and ultimately furanone) formation, no obvious phenotypic changes were observed compared with control cultures (i.e. no apparent biofilm formation or flocculation). The T. maritima transcriptome from the chemostat culture was only impacted slightly; only a few genes responded to the stimuli with a greater than twofold change in transcription compared with the control (high dose=17 up/4 down, low dose=9 up/4 down) (Supporting Information, Table S1). This furanone challenge did, however, appear to induce a stress response, as several known stress genes were upregulated twofold or more, including dnaK (TM0373), hsp20 (TM0374), groES (TM0505), groEL (TM0506), dnaJ (TM0849), grpE (TM0850), and hrcA (TM0851) (Fig. 6).
Here we report that two hyperthermophiles produce AI-2 using a noncanonical abiotic/biotic reaction pathway (see Fig. 7). However, AI-2 did not induce the classical quorum-sensing response noted in certain mesophilic bacteria. Thus, the role of AI-2 (and other furanones) in hydrothermal environments, if any, is not clear. The fact that AI-2 does not seem to impact these hyperthermophiles suggests that this form of cell–cell signaling may be confined to thermally moderate biotopes.
We acknowledge the NASA Exobiology, DOE Energy Biosciences, and NSF Biotechnology Programs for support of this research. We also thank William Miatke for his help with this project. We are grateful to Scott Lesley, Joint Center for Structural Genomics, La Jolla, CA, and Michael Adams, University of Georgia, for providing the T. maritima and P. furiosus genes, and to the Analytical Instrument Group (Raleigh, NC) for MS analysis. Vibrio harveyi strains were graciously provided by Bonnie Bassler, Princeton University.