Chemical synthesis of (S)-4,5-dihydroxy-2,3-pentanedione, a bacterial signal molecule precursor, and validation of its activity in Salmonella typhimurium.

We describe an original, short, and convenient chemical synthesis of enantiopure (S)-4,5-dihydroxy-2,3-pentanedione (DPD), starting from commercial methyl (S)-(-)-2,2-dimethyl-1,3-dioxolane-4-carboxylate. DPD is the precursor of autoinducer (AI)-2, the proposed signal for bacterial interspecies communication. AI-2 is synthesized by many bacterial species in three enzymatic steps. The last step, a LuxS-catalyzed reaction, leads to the formation of DPD, which spontaneously cyclizes into AI-2. AI-2-like activity of the synthesized molecule was ascertained by the Vibrio harveyi bioassay. To further validate the biological activity of synthetic DPD and to explore its potential in studying DPD (AI-2)-mediated signaling, a Salmonella typhimurium luxS mutant was constructed. Expression of the AI-2 regulated lsr operon can be rescued in this luxS mutant by addition of synthetic DPD or genetic complementation. Biofilm formation by S. typhimurium has been reported to be defective in a luxS mutant, and this was confirmed in this study to test DPD for chemical complementation. However, biofilm formation of the luxS mutant cannot be restored by addition of DPD. In contrast, introduction of luxS under control of its own promoter complemented biofilm formation. Further results demonstrated that biofilm formation of the luxS mutant cannot be restored with luxS under control of the strong nptII promoter. This indicates that altering the intrinsic promoter activity of luxS affects Salmonella biofilm formation. Conclusively, we synthesized biologically active DPD. Using this chemical compound in combination with genetic approaches opens new avenues in studying AI-2-mediated signaling.

Bacteria possess an arsenal of chemical signal molecules that enable them to communicate within and between species. As such, these unicellular organisms are able to behave in a multicellular way. When a few bacteria release such signal molecules into the environment, their concentration remains below detection limits. However, when the signal molecules accumulate and the concentration reaches a threshold level, they induce the population to cooperate in diverse behaviors such as bioluminescence, virulence, and biofilm formation, by activating or repressing target genes (1)(2)(3)(4)(5). This phenomenon is referred to as quorum sensing.
A broad variety of quorum-sensing signal molecules has been identified in the past 20 years that can roughly be divided into (i) acyl-homoserinelactones in Gram-negatives (6), (ii) processed oligopeptides in Gram-positives (7), and (iii) AI-2 1 in both Gram-negatives and Gram-positives (8). In contrast to the first two types of signal molecules, which are usually involved in intraspecies communication, the AI-2 molecule has been proposed to serve as a "universal" signal for interspecies communication (9,10). The synthase for AI-2 production, LuxS, is widely conserved among Gram-negative and Gram-positive bacteria (9). Many genera of bacteria, including Salmonella typhimurium, produce AI-2 (11).
In S. typhimurium, AI-2 is produced and released during exponential growth. It subsequently is reinternalized by the bacteria via the Lsr (LuxS-regulated) ABC transporter (15,16). AI-2 induces transcription of the lsrACDBFGE operon, of which the first four genes encode the Lsr transport apparatus. Additionally, a Salmonella luxS mutant has been reported to be defective in biofilm formation on gallstones (17). However, whether this is the result of a signaling role of AI-2 is a current topic of debate (8,18). Winzer et al. (19) advocated that AI-2 is not a signal molecule in organisms other than V. harveyi but rather is a discarded by-product of S-ribosylhomocysteine recycling. Changes in gene expression due to inactivation of luxS could be due to defects in the methionine metabolism, rather than to the role of AI-2 as a signal. If, however, wild-type phenotypes in luxS mutants can be restored by the addition of exogenous AI-2, a role of luxS as signal molecule synthase could be justified. Therefore, there is a need for chemically well defined AI-2 signal molecules.
In this study we present an original, short, and convenient chemical synthesis of enantiopure DPD, starting from commercial methyl (S)-(Ϫ)-2,2-dimethyl-1,3-dioxolane-4-carboxylate. In addition to demonstrating the biological activity of the chemically synthesized DPD by means of the Vibrio bioassay, we succeeded in rescuing the expression of an AI-2 regulated target gene in a Salmonella luxS mutant by adding chemically synthesized DPD. The availability of synthetic DPD in combination with a genetic approach, allowed us to reveal the complexity of Salmonella biofilm formation.

Synthesis of DPD
The synthesis route to DPD, starting from methyl (S)-(Ϫ)-2,2-dimethyl-1,3-dioxolane-4-carboxylate (Fluka), is depicted in Fig. 1A (2). The key step is the ozonolysis (iv) of (S)-1,2-dihydroxy-4-methyl-4penten-3-one (see Fig. 1A, 5). 0.78 mg (6 mol) of the enone 5 was dissolved in 2 ml of methanol and cooled to Ϫ78°C. A stream of 0.75 g/h (15.7 mmol/h) O 3 in O 2 was bubbled through the solution for 0.5 min with a total flow of 10.8 dm 3 /h. Next, the excess of ozone was removed from the solution with a stream of nitrogen gas and dimethyl sulfide was added (5-10 eq). The reaction mixture was allowed to warm up and kept at room temperature for 24 h. After the addition of 3 ml of water, the volatiles (methanol, dimethyl sulfide) were removed by evaporation. NMR showed the signals of DPD and its anomers, accompanied by minor traces of remaining formaldehyde hydrate, which is split off during ozonolysis, methanol, dimethyl sulfide, and Me 2 SO or dimethyl sulfone. Details can be found in the supplemental data.

AI-2 Activity Bioluminescence Assay
The activity of DPD was assayed using the V. harveyi BB170 (luxN::Tn5) reporter strain, which responds only to AI-2-like molecules (11). The luminescence bioassays were performed as described previously (11,22) by applying different concentrations of synthetic DPD, MHF (4-hydroxy-5-methyl-3-(2H)-furanone) (Sigma), and all intermediates of the DPD chemical synthesis (see Fig. 1A, [2][3][4][5]. Light production was measured using a CCD camera (Berthold Night Owl, PerkinElmer Life Science) or a luminescence reader (Fluoroskan Ascent FL, Labsystems). Optical density was determined at 620 nm with a microtiter plate reader (VERSAmax, Molecular Devices). Sterile AB medium and a mixture of Me 2 SO, methanol, and water (ratio v/v 1:1:8) served as negative controls. Cell-free AB broth (22) of an overnight culture of the acylhomoserine lactone-defective V. harveyi BB152 was used as the AI-2-producing positive control (prepared as reported (11)). Results are reported as a percentage of the induction level produced by the positive control. Assays were performed at least in triplicate.

DPD Complementation Assay Measuring lsrA Expression
A transcriptional fusion between the lsr operon promoter and the reporter luxCDABE (pCMPG5638) was constructed as described in the supplemental data. The lsrA::luxCDABE fusion was used to measure promoter activity of lsrACDBFGE as counts/second of light with a CCD camera, as reported previously (15) and briefly described in the supplemental data. Gene expression was normalized per cell by dividing the luminescence value by the A 620 value of each sample. All experiments were performed in triplicate.

Biofilm Assay
Biofilms were grown essentially by the method of Stepanovic et al. (23) with modifications specified in the supplemental data. One of these modifications is the device used for biofilm formation, i.e. a platform carrying 96 polystyrene pegs (Nunc number 445497) that fits as a microtiter plate lid with a peg hanging into each microtiter plate well (Nunc number 269787), as described previously (24,25). Synthetic DPD was added to the growth medium at the time of inoculation and when changing the medium. Alternatively, a fed batch experiment was set up during a period of 48 h. Synthetic DPD concentration was gradually increased by the addition of a certain amount of DPD (i.e. 0, 0.1, 1, 5, 10, 20, 30, 50, 72 M) to the growth medium every 6 h. Methionine (Sigma), cysteine (Sigma), and S-adenosylmethionine (Sigma) were used at a final concentration of 1 mM. Subsequent quantification of biofilm production was performed as previously described (23), with modifications specified in the supplemental data. Each strain and/or condition was tested 8-fold. Each experiment was performed at least in triplicate.
To determine the presence of AI-2 at different time points during biofilm formation (i.e. 6, 24, 30, 48 h), pegs were sonicated for 5 min at 40 kHz (Branson 5210) in 200 l of 1:20 trypticase soy broth. The cell-free supernatant was prepared and assayed in the V. harveyi bioassay, as described previously (11).

Chemical Synthesis of DPD
Like its structure suggests, DPD is a highly reactive molecule, and, in solution, it exists in a dynamic equilibrium with its cyclic anomers (Fig. 1A, 1) (26). Consequently, the product as such is hard to purify. Hence the last step in the chemical synthesis of DPD should be a clean transformation, which makes further purification redundant. Therefore, a short and convenient chemical synthesis of DPD was designed with the ozonolysis of a double bond as the final step (Fig. 1A, 1).
Starting from commercial methyl (S)-(Ϫ)-2,2-dimethyl-1,3dioxolane-4-carboxylate (Fig. 1A, 2), DPD was obtained in four steps (Fig. 1A). First, the methyl ester, 2, was transformed into the amide, 3, which was subsequently transformed to olefin, 4, with isopropenylmagnesium bromide in a Grignard reaction. Hydrolysis of the dioxolane ring in 4 was performed on an acid Dowex resin to yield the enone 5. Ozone treatment of the double bond in 5 and in situ reductive cleavage of the ozonide with dimethyl sulfide yielded DPD. A simple workup, with addition of water and evaporation of volatiles resulted in an aqueous solution of DPD. In principle, it would be possible to reverse the order of the deprotection and ozonolysis (Fig. 1A, iii and iv), but the deprotection turned out to be much less clean when performed as the last step.

Characterization of DPD
Besides its biological activity (see below), evidence for the formation of DPD was collected in different ways. Based on 1 H NMR spectroscopic measurements (see supplemental data), the aqueous product solution contains DPD, 1, along with its two cyclic equilibrium anomers (Fig. 1A, 1b and 1c), which result . It is possible that in the aqueous solution the ketone group in the anomers 1b and 1c is further hydrated to the corresponding geminal diols (Fig. 1A, 1d and 1e); such a hydration has been reported for the structurally related cyclized 3-deoxyglycosones (27). Note that a hydrated form of DPD has been prepared by Semmelhack et al. (28); it has been proposed to be of biological significance in S. typhimurium (14).
Further evidence for the formation of DPD and the presence of its equilibrium products was obtained by using 1,2-phenylenediamine ( Fig. 1B) (29,30). Addition of this diamine to the aqueous solution resulted in the disappearance of DPD and its equilibrium products, and led to the formation a single, stable quinoxaline derivative 6 through a Maillard reaction, with only a few other signals remaining in the spectrum. This proves that at least 80% of the products of the aqueous equilibrium mixture are derived from DPD 1. NMR data on this derivatization can be found in the supplemental data.

Biological Activity of Synthetic DPD in the Vibrio Bioassay
The biological activity of different concentrations of synthetic DPD was determined in the V. harveyi bioassay for AI-2 (31) (Fig. 2). As a reference, MHF, which is structurally related to AI-2, was included in the test (Fig. 2). The bioluminescenceinducing capacity of synthetic DPD clearly exceeds that of MHF in the same concentration range. The maximal activity of synthetic DPD was observed at a concentration of ϳ15 M. The EC 50 value (concentration producing half-maximal stimulation) of synthetic DPD correlated to ϳ125 nM, compared with ϳ30 M MHF. Synthetic DPD thus possesses roughly a 240-fold higher activity than MHF. All intermediates of the chemical synthesis of DPD (Fig. 1A, 2-5) were assayed in the V. harveyi bioreporter, but none of the tested compounds induced bioluminescence (not shown).
DPD has been reported as a rather unstable and reactive molecule (12,19), especially at concentrations above 0.1 M where it looses biological activity (26,32). To test the stability of DPD, bioluminescence induction assays were repeated with synthetic DPD kept at room temperature during 7 days or stored at 4 or Ϫ20°C for a period of at least two months. Similar results as described above were obtained (not shown).

Synthetic DPD Complements lsrA Expression in S. typhimurium luxS Mutant
In S. typhimurium, the active uptake and modification system for AI-2, encoded by the lsr operon, is regulated by AI-2 (15,16). To prove that synthetic DPD is sufficient for lsr regulation, we investigated whether exogenously supplied synthetic DPD could induce the lsrA::lux fusion in S. typhimurium containing a null mutation in luxS. S. typhimurium lsrA::lux fusion strain, both in the wild-type and luxS null (CMPG5602) background, was grown to the late exponential phase in LB medium supplemented with different concentrations of synthetic DPD. Results obtained with 72 M DPD are shown (Fig.  3). As previously reported (15), and confirmed by our experiments (not shown), this is approximately twice the AI-2 concentration estimated to be present in S. typhimurium cell-free supernatant. Nevertheless, in the tested range, maximal expression of the lsrA::lux fusion in the Salmonella luxS mutant occurred at a concentration of ϳ360 M added DPD. In the absence of DPD, expression of the lsrA::lux fusion in the luxS mutant (luxS null), was much lower than in the luxS wild-type background (wild-type luxS). Addition of 72 M DPD restores lsrA expression in the luxS mutant.
For comparison with genetic complementation, we transferred the luxS gene to the luxS mutant fusion strain. pC-MPG5664 carries the luxS gene driven by its own, weak promoter (33). Introduction of the parent vector without luxS did not affect lsrA expression in a luxS null background (data not shown). Introduction of pCMPG5664 (luxS null ϩ luxS) induced the expression of lsrA::lux fusions to a level comparable with that observed when wild-type luxS is present on the chromosome (Fig. 3). Supplying synthetic DPD to the S. typhimurium lsrA::lux strains that were wild-type for luxS on the chromosome or that were luxS null with luxS complemented in trans, showed an increased level of lsrA::lux expression, as compared with endogenous levels of AI-2. These results corroborate those previously obtained by Taga et al. (15) with enzymatically synthesized AI-2. They showed that the lsr operon, under the conditions applied, was not fully induced by the endogenously produced AI-2.

The Effect of Synthetic DPD on S. typhimurium Biofilm Formation
To investigate the role of luxS and DPD in Salmonella biofilm formation, we tested S. typhimurium wild type and the luxS mutant (CMPG5602) for their ability to form biofilms on polystyrene pegs. The Salmonella luxS mutant is impaired in forming mature biofilms. Introduction of a functional luxS gene driven by its own promoter (pCMPG5664) in the luxS mutant restores biofilm formation to the wild-type level (Fig. 4A). However, synthetic DPD cannot rescue the biofilm formation defect of the Salmonella luxS mutant at 72 M concentration, which is sufficient to restore lsrA expression in a luxS mutant (Figs. 3  and 4A). Alternative ways of adding DPD to the biofilm assay were applied in an attempt to rescue biofilm formation in the luxS mutant. DPD concentrations were increased, the incubation time of biofilm formation was extended to 7 days, and the incubation conditions of biofilm formation were adapted to enhance contact with DPD (slightly rocking). Alternatively, complementation with synthetic DPD was performed in a fed batch experiment, in which the concentration of DPD was increased stepwise during 48 h in the assay for biofilm formation. Eventually, DPD supplementation was restricted to the time of inoculation. In none of these cases was restoration of biofilm formation in the luxS mutant observed (not shown). This latter condition was inspired by our observations that after 6 h no AI-2 could be detected in the biofilm (not shown).
To exclude an indirect effect of the activated methyl cycle, in which LuxS plays a role (12,19), as a possible cause of the lack of restoration of biofilm formation by DPD, the biofilm medium was supplemented with methionine, cysteine, or S-adenosylmethionine. However, no rescue of biofilm formation in the Salmonella luxS mutant was observed. Addition of S-adenosylmethionine even completely abrogated biofilm formation in Salmonella wild type (not shown).
To explore the hypothesis that controlled synthesis of AI-2 at a given time point, and therefore tight control of LuxS activity, is required for S. typhimurium biofilm formation as suggested from the previous experiments, we tested whether introducing luxS under the control of the strong constitutive nptII promoter (34) (pCMPG5643), resulted in restoration of Salmonella bio-film formation in the luxS mutant. Interestingly, introduction of luxS driven by this nptII promoter (pCMPG5643) cannot rescue biofilm formation in the luxS mutant (Fig. 4B). On the other hand, wild-type Salmonella containing the same construct, i.e. pCMPG5643, is able to form a biofilm in the same amount as wild type not containing this plasmid (Fig. 4B). As such, it can be further stated that the chromosomal wild-type luxS allele, i.e. driven by its own promoter, is dominant in the luxS merodiploid S. typhimurium strain containing two luxS copies, one driven by its own promoter and one by the nptII promoter. This experiment also allows excluding possible interference of a particular antibiotic added in the biofilm assay, because introduction of pCMPG5643 in wild-type Salmonella and consequently adding of the corresponding antibiotic, does not affect biofilm formation. DISCUSSION A most exciting development in recent years is the discovery that bacteria communicate thereby mimicking a multicellular organization. The AI-2 signal molecule is an exciting chemical signal, because it is produced and interpreted by a variety of bacteria (31). In many cases, luxS orthologs have been shown to be essential for AI-2 synthesis (9). The precursor of AI-2, DPD, is ubiquitous (9,12,14,19). However, because of the additional metabolic role of LuxS (19), it is not clear whether all the bacteria that produce AI-2 actually use it as a signal molecule. To elucidate this, a chemically defined AI-2 is required. We designed a chemical synthesis of DPD, because the lack of FIG. 2. Activity of synthetic DPD and structural analog MHF. Induction of bioluminescence in V. harveyi BB170 by different concentrations of synthetic DPD (1) (diamonds) and MHF (squares). Results are reported as a percentage of the induction level produced by the positive control BB152 (10% v/v), after background correction (negative control). Depicted are the results of a single experiment representative of at least three independent experiments. The concentration of synthetic DPD was determined based on the initial concentration of starting product (5) prior to ozonolysis by assuming a complete conversion to DPD. We noticed that DPD concentration of 300 M affected the growth of V. harveyi as shown by optical density measurements. MHF has been shown to inhibit growth of various bacteria at a concentration of 5 mM (18). chemically defined AI-2 (or its precursor DPD) hampered further studies on AI-2 signaling. Indeed, at the time we initiated our DPD synthesis, only partial purification of AI-2 by methanol-dichloromethane extraction from conditioned medium had been described (35). Recently, an affinity column with a borate resin showed effective in providing a method for concentrating and purifying V. harveyi AI-2 from the biosynthetic product (32). In vitro enzymatic production of AI-2 starting from Sadenosylhomocysteine is limited in concentration of end product by the low solubility of S-adenosylhomocysteine and results in a mixture of AI-2, MHF, homocysteine, and adenine (12,18). During the preparation and handling of this manuscript, two synthetic procedures for DPD synthesis have been published (26,28).
We here have described a unique synthesis of enantiopure DPD, starting from the commercially available methyl (S)-(Ϫ)-2,2-dimethyl-1,3-dioxolane-4-carboxylate in four straightforward steps. The end product of the synthesis is a solution containing DPD as an equilibrium mixture with its cyclization products, among which are 2 anomeric furanones and possibly hydration products (Fig. 1A), as shown by NMR and 1,2-phenylenediamine derivatization. A convincing proof for DPD synthesis is that our NMR spectrum fully matches that of the DPD, recently prepared via a completely different chemical route (26). The synthesis route of Meijler et al. (26) and Semmelhack et al. (28) comprises 6 or 7 steps, including a potentially aselective oxidation of an alkyne. In contrast, the route presented in this work comprises only 4 steps from a commercial precursor, with satisfactory yields in all steps.
The biological activity of our synthetic DPD was shown using the V. harveyi bioreporter. Biosynthetic AI-2 has been reported to possess EC 50 values ranging from 80 nM (12)  We compared the biological activity of synthetic DPD to that of MHF, a compound similar in structure to cyclized DPD that has been identified as a by-product of the in vitro degradation of S-ribosylhomocysteine (12,19). Our NMR analyses demonstrated that no MHF is formed during the chemical synthesis of DPD. Accordingly, DPD and MHF are not spontaneously interconvertible under these conditions, as previously put forward as a possibility (19). MHF has light-inducing activity, but a higher concentration than that of AI-2 is reported to be required to stimulate the reporter (12,19). Meijler et al. (26) observed a roughly 500-fold higher activity for synthetic DPD than for MHF. In our work, MHF showed a half-maximal activation in the V. harveyi bioassay at the concentration of 30 M. As such, the EC 50 of synthetic DPD in our assay is roughly 240-fold lower than that of MHF. In the literature, the EC 50 values for MHF of 100 M (12), 125 M (26), and 1 mM (19) have been reported. Variability of the AI-2 bioassay (31, 36, 37) could be responsible for the observed differences in the respective data.
The role of AI-2 in S. typhimurium has not yet been completely elucidated. Until now, only the lsr operon has been found to directly respond to AI-2 in S. typhimurium (15,16,38). This lsr operon has recently been shown to function analogously in Escherichia coli (39). At an approximate concentration of 72 M, chemically synthesized DPD, in the condition applied, is sufficient to restore expression of the lsr operon in a Salmonella luxS mutant (Fig. 3), indicating that Salmonella can indeed detect the presence of AI-2 and that DPD does not require LuxS to direct it into biologically active AI-2. This observation further validated the biological activity of our synthetic DPD.
The report on Salmonella luxS mutants being impaired in biofilm formation on gallstones (17) offered an interesting, more complex phenotype to evaluate biological activity of our DPD. The suggested role of LuxS (AI-2) in bacterial biofilm development is ambiguous. For instance, one study found a role for the LuxS system in Streptococcus mutans, but two other studies indicated that LuxS is not required for biofilm formation in S. mutans (40 -42). Our results showed that Salmonella Lack of complementation was observed in the luxS mutant strain when introducing luxS under the control of a strong constitutive promoter (pCMPG5643; luxS null, black bars), in contrast to introduction of luxS under the control of its own promoter (pCMPG5664; luxS null, gray bars), which restored biofilm formation in luxS null. Introduction of pCMPG5664 (wild-type luxS, gray bars) or pCMPG5643 (wild-type luxS, black bars) did not affect biofilm formation by wild-type Salmonella. Error bars depict S.D. of 8 independent measurements. Means that were found to be significantly different (p Ͻ 0.05) by the Tukey test are indicated by different letters. Depicted results are representative for at least three independent experiments. biofilm formation on polystyrene is dependent on an intrinsically regulated luxS, because a Salmonella luxS mutant, which is impaired in biofilm formation, could be rescued by the introduction of a plasmid containing wild-type luxS but not by luxS driven by the strong nptII promoter. In line with this, addition of synthetic DPD cannot rescue the biofilm formation defect of the Salmonella luxS mutant. Although in planktonic state, S. typhimurium luxS expression was reported to be weak and constitutive (33), during biofilm formation this might not be the case. The sessile lifestyle associated with persistence within a biofilm is distinct from the lifestyles of planktonic culture, resulting in different gene expression (43,44). Our results indicate that not AI-2 as such, but altering the intrinsic promoter activity of luxS, changes expression in time and/or intensity of Salmonella genes, among which are genes crucial for biofilm formation. The role of LuxS in the activated methyl cycle seems not to be a major player in biofilm formation, because supplementation with methionine, cysteine, or S-adenosylmethionine could not restore biofilm formation in a Salmonella luxS mutant. The observation that wild-type luxS is dominant in a merodiploid strain containing a wild-type luxS allele and a luxS allele engineered for strong constitutive expression further suggests a complex mechanism of LuxS-mediated biofilm formation. We will follow a genome-wide approach to disentangle the role of luxS regulation and its influence on the different levels of the biofilm formation cascade. Synthetic DPD offers a convenient, powerful tool to assist in this endeavor.