High levels of AI-2 activity arises from ribulose-5-phosphate in vitro
A solution of 5 mM Rul-5-P was incubated at 37°C for 24 h, based on the methods previously described for MHF formation [12, 13], and assayed for bioluminescence-inducing activity in V. harveyi BB170, a bioreporter for detection of AI-2 activity [17]. Addition of this solution to the reporter strain induced high levels of bioluminescence, much higher than those observed for a positive control of 5 mM MHF. By comparison, the isomers Rib-5-P and xylulose-5-phosphate (Xyl-5-P) stimulated little or no bioluminescence (Fig. 2A). Since bioluminescence in V. harveyi is also controlled by the signal molecule N-3-hydroxybutanoyl-L-homoserine lactone (AI-1), each compound was also tested with the AI-1 responsive bioreporter, V. harveyi BB886 [17]. No bioluminescence was induced in this strain (Fig. 2A) indicating Rul-5-P-mediated bioluminescence to be induced specifically via the AI-2 detection system.
To demonstrate that the observed AI-2 activity was not caused by impurities present in the commercial preparations, Rul-5-P was also produced enzymatically. Incubation of 5 mM ribose-5-phosphate in the presence of active spinach phosphoribose isomerase (10 U/ml), which isomerised the substrate to Rul-5-P, resulted in the formation of high AI-2 activity, whereas no significant activity was observed after incubation with heat-inactivated enzyme (Fig. 2B). Furthermore, time course experiments performed with freshly prepared 0.5 mM Rul-5-P incubated at 37°C showed that AI-2 activity gradually increased over time, typically displaying a maximum after 5 hours (Fig. 2C). This was consistent with a conversion of Rul-5-P firstly to DPD, and then to the less-active MHF, in agreement with previously observed spectrophotometric data [11, 13]. Fig. 2C shows that approximately 20-fold induction of bioluminescence was observed for the sample removed at the beginning of the time course experiment (0 h). This activity most likely originated from Rul-5-P conversion during the several hours of incubation in the V. harveyi BB170 bioassay rather than an intrinsic activity of the compound itself.
The concentration range of Rul-5-P used in these experiments was chosen because it matches the intracellular concentrations reported for some organisms in the literature, such as yeast (approximately 0.5 mM; [18]). However, much lower values in the range of 1–10 μM can be estimated from metabolite determinations in animal and human tissues, including liver, brain, blood samples [19–22].
Rul-5-P has been previously reported to give rise to MHF with approximately 1.3% of the sugar-phosphate converting to the furanone after incubation at pH 7.5 and 35°C for 15 h [13]. Liquid chromatography/tandem mass spectrometry (LC-MS/MS) confirmed that MHF was also formed under the conditions used in this study (data not shown), with conversion rates below 1% after 24 h of incubation at pH 7.7 and 37°C. Taking into account the low conversion rates, much higher AI-2 activities were observed in Rul-5-P solutions than could be expected solely due to formation of MHF (Fig. 2A and data not shown). This demonstrates that another compound derived from Rul-5-P, but different from MHF, was responsible for the observed activities. Since DPD is known to be generated in Rul-5-P solutions [13] and was also identified after o-phenylenediamine derivatisation under the incubation conditions used in this study (data not shown), this was most likely S-THMF-borate (see Fig. 1), a DPD-derivative known to activate the BB170 bioreporter [9].
Modification of the AI-2 bioassay
To determine whether LuxS-independent production of AI-2 also occurs in vivo, E. coli DH5α (which contains a frameshift mutation in luxS) and Staphylococcus aureus Newman luxS were each cultured in LB medium in the presence of 0.5% glucose. In E. coli, glucose at this concentration had previously been shown to prevent AI-2 uptake via catabolite repression [23]. In each case, cell-free culture supernatants from mid- to late exponential phase onwards appeared to display 3–6 fold greater induction of bioluminescence in V. harveyi BB170 than controls of sterile control medium (not shown). However, following careful scrutiny it became apparent that this induction was not caused by true AI-2 activity. Metabolism of glucose led to acidification of the culture environment, most prominently between 3–5 hours of growth, during which time pH values typically declined from pH 7 to pH 5. Addition of culture-supernatants from this period onwards, to the V. harveyi BB170 bioassay, led to a subsequent decrease in pH of the bioassay medium (not shown). This coincided with enhanced recovery of endogenously-induced bioluminescence in V. harveyi BB170, following dilution into the assay medium (Fig. 3A). Acidification has been previously reported by DeKeersmaecker and Vanderleyden [24] to impact on the AI-2 bioassay, therefore, the assay medium was modified to include a final concentration of 25 mM HEPES buffer to increase the total buffer strength. This prevented the more acidic culture-fluids from influencing the pH of the assay medium and simultaneously eliminated the induction of bioluminescence observed for the luxS mutant culture supernatants (Fig. 3B). Incorporation of HEPES did not affect the detection of true AI-2 activity in culture-fluid samples as shown by the unaffected response to positive control samples. Overall, the data demonstrated that culture-supernatants of the two luxS mutants grown in LB medium did not contain detectable AI-2 activity.
AI-2 activity is detectable in an E. coli luxS mutant with altered carbon flow
AI-2 activity was not apparent in the culture-fluids of luxS mutants investigated above. However, in E. coli at least, just 25% of exogenously supplied glucose is believed to be catabolised via the oxidative pentose phosphate pathway (OPPP) to yield the intermediate Rul-5-P [25]. It is therefore possible that LuxS-independent generation of AI-2 from Rul-5-P may be limited under the growth condition used in this study. It is also possible, that Rul-5-P concentrations are generally very low in E. coli and related organisms. In order to maximise the potential for detectable AI-2 production from the sugar-phosphate in vivo, an E. coli MG1655 pgi, edd, eda triple mutant (E. coli pgi-EDP; [25]) was utilised. In this strain, glucose catabolism occurs exclusively via the OPPP as entry of the sugar into the glycolytic and Entner-Doudoroff pathways are blocked ([25]; pgi encodes phosphoglucose isomerase; edd and eda encode phosphogluconate dehydratase and 2-keto-3-deoxy-6-phosphogluconate aldolase, respectively). Production of LuxS-derived AI-2 activity by E. coli pgi-EDP was eliminated through the introduction of a luxS mutation. This was achieved by P1 bacteriophage transduction of the disrupted luxS gene from E. coli BL21 luxS [9] into E. coli pgi-EDP. The mutation was also introduced into the E. coli MG1655 parent strain.
E. coli MG1655 luxS and pgi-EDP luxS were grown in LB containing 0.5% glucose. Cell-free culture-supernatants were harvested and tested for bioluminescence-inducing ability using the optimised (HEPES-containing) V. harveyi BB170 bioassay. E. coli pgi-EDP luxS culture-supernatants induced a response in V. harveyi BB170 that, whilst modest, was clearly not observed for E. coli MG1655 luxS (Fig. 3C). Typically a peak in bioluminescence-inducing activity was observed in E. coli pgi-EDP luxS culture-fluids following 5 hours of culture and displayed 4–6 fold greater bioluminescence induction than sterile LB alone. However, in the absence of glucose, AI-2 activity could not be detected (data not shown).
Several potential sources of false-positive results were investigated to determine whether the observed effects were a result of some non-AI-2-specific influence as witnessed for the unmodified bioassay. Assay medium pH was recorded following addition of all test samples and verified as being unaffected, thus confirming that elevated bioluminescence-induction by E. coli pgi-EDP luxS culture-fluids was not an artefact of acidification of the culture medium. In addition, viable counts were performed for the V. harveyi BB170 assay samples used to generate the data in Fig. 3C to ensure that differences in bioluminescence-induction between supernatant samples were not caused by effects on proliferation of the reporter strain (not shown). Furthermore, the possibility of obtaining erroneous results due to the repressive effects of glucose upon bioluminescence, as described by [24], was eliminated through the use of AI-2-negative media control samples with and without added glucose. These data therefore indicated that E. coli pgi-EDP luxS culture-fluids contained very low but detectable levels of AI-2 activity. This suggests that increased flux through the OPPP may indeed lead to increased cellular generation of AI-2-like molecules via the DPD intermediate. However, whether the OPPP represents one of the alternative major AI-2 production pathways predicted to exist in E. coli by stochastic modelling [26] remains doubtful, as only very low amounts of AI-2 could be detected for a metabolically crippled E. coli MG1655 pgi-edd-eda-luxS quadruple mutant but not E. coli MG1655 luxS. Future work will require the development of more stringent physical or chemical methods to unequivocally detect and quantify AI-2 molecules in complex biological samples, particularly when present at low concentrations. Once developed, such methods may allow more robust quantitative detection of, and distinction between, different DPD-derivatives. This would allow conclusive confirmation as to whether the Rul-5-P-dependent mechanism of AI-2 production is operational in vivo. An overview of the different pathways leading to DPD and AI-2 formation is given in Fig. 1
Implications
Our finding that the DPD levels formed during the spontaneous conversion of Rul-5-P are sufficient to give rise to measurable AI-2 activity has several important implications. First, given the ubiquitous presence of Rul-5-P in metabolically active cells [27], it seems possible that molecules with AI-2 activity are intrinsically formed as by-products of pentose phosphate metabolism in all organisms and independently of the LuxS enzyme. Indeed, AI-2 activity has been reported in stationary phase culture-supernatants of Streptococcus pyogenes luxS mutants [28]. Furthermore, some microorganisms, plants, and animals are known to produce MHF [29–33], a compound which, at least in the case of certain yeasts, is believed to be derived from Rul-5-P [12, 13].
Production of AI-2-like activity by higher organisms has also been demonstrated for algae of the genera Chlamydomonas and Chlorella [34] and evidence for the formation of DPD from Rul-5-P in tomato fruits has been provided [13]. Thus, organisms other than bacteria may have developed the machinery necessary to either metabolise or exclude DPD-derived by-products such as AI-2. This may also explain the presence of lsr-type AI-2 uptake systems in bacteria that do not contain a luxS homologue [3]. Sinorhizobium meliloti, for instance, possesses a complete putative lsr AI-2 uptake system and an AI-2 kinase (locus tags SMb21016-21022). Systems like this may have been acquired to minimise the loss of intrinsically produced DPD-derived compounds, or alternatively to scavenge the molecules released from other organisms present in the same niche, either bacteria or plants.
Finally, given the relative ease with which DPD, and thus AI-2, appears to form from Rul-5-P, and the potential advantages currently believed to be gained by the utilisation of AI-2-dependent signalling, it would seem an interesting, but as yet unexplored, possibility that some organisms may have acquired enzymes to direct and enhance this process of AI-2 formation in a controlled fashion.