|Home | About | Journals | Submit | Contact Us | Français|
Quorum sensing (QS) is under the control of N-acylated L-homoserine lactones (AHLs) and their cognate receptors (LuxR-type proteins) in Gram-negative bacteria, and plays a major role in mediating host-bacteria interactions by these species. Certain cyclic dipeptides (2,5-diketopiperazines, DKPs) have been isolated from bacteria and reported to activate or inhibit LuxR-type proteins in AHL biosensor strains, albeit at significantly higher concentrations than native lactones. These reports have prompted the proposal that DKPs represent a new class of QS signals, and potentially even interspecies or interkingdom signals; their mechanisms of action and physiological relevance, however, remain unknown. Here, we describe a library of synthetic DKPs that was designed to (1) determine the structural features necessary for LuxR-type protein activation and inhibition, and (2) probe their mechanisms of action. These DKPs, along with several previously reported natural DKPs, were screened in bacterial reporter gene assays. In contrast to previous reports, the native DKPs failed to exhibit either antagonistic or agonistic activities in these assays. However, non-natural halogenated cyclo(L-Pro–L-Phe) derivatives were capable of inhibiting luminescence in V. fischeri. Interestingly, additional experiments revealed that these DKPs do not compete with the natural lactone signal, OHHL, to inhibit luminescence. Together, these data suggest that DKPs are not QS signals in the bacteria examined in this study. Although these compounds can influence QS-regulated outcomes, we contend that they do not do so through direct interaction with LuxR-type proteins. This work serves to refine the lexicon of naturally occurring QS signals used by Gram-negative bacteria.
Quorum sensing (QS) has emerged as a prevalent cell-cell signaling pathway in chemical ecology. This population density sensing mechanism is used broadly by bacteria and is governed by a chemical “language” of small, diffusible signal molecules (or autoinducers) and their associated protein receptors (1–3). QS in Gram-negative bacteria is under the control of diffusible N-acylated L-homoserine lactone signals (AHLs) and the LuxR-type family of cytoplasmic receptors, and was first characterized in the luminescent marine symbiont Vibrio fischeri (3–6). The AHL ligands are most frequently generated by LuxI-type synthases, and their local concentration correlates with cell density (along with other environmental factors (7)). Above a threshold concentration, the AHLs bind to their cognate LuxR-type receptors, and these complexes most frequently activate the transcription of target genes required for bacterial group behavior. As many of these processes play crucial roles in both pathogenesis and symbiosis, there is significant interest in the development of non-native ligands that can block or mimic native autoinducer signals and attenuate QS (8, 9). Such molecules would represent new tools to study the molecular mechanisms of QS and their roles in host-bacteria interactions (10, 11).
The structures of native AHLs from selected Gram-negative species are shown in Figure 1 (panel a): N-(3-oxo-hexanoyl)-L-homoserine lactone (OHHL, 1) that binds LuxR in V. fischeri and represents the canonical quorum sensing circuit, N-(3-oxo-octanoyl)-L-homoserine lactone (OOHL, 2) that binds TraR in the plant pathogen Agrobacterium tumefaciens, and N-(3-oxo-dodecanoyl)-L-homoserine lactone (OdDHL, 3) that binds both LasR and QscR in the opportunistic pathogen Pseudomonas aeruginosa. Interception of AHL:LuxR-type protein binding by synthetic ligands represents one strategy to control QS in Gram-negative bacteria, and considerable research efforts over the past 20 years have been directed in this area (12). The majority of this past work has focused on the design and synthesis of non-native AHL derivatives (5, 13, 14), and our laboratory has recently made contributions to this area for LuxR, TraR, LasR, and QscR based QS (e.g., the LuxR-type protein inhibitor 4-iodo phenylacetyl HL (4-I-PHL, 4); Figure 1, panel a) (15–20).
Looking beyond the AHL scaffold would certainly expand the range of possible ligands for use as LuxR-type protein modulators, and several research groups have reported strategies towards this end, including the screening of commercial small molecules libraries (21) and natural product isolates (22, 23), and computational pharmacophore modeling (24, 25). Recently, a set of cyclic dipeptides (2,5-diketopiperazines, or DKPs) were isolated from a range of Gram-negative bacteria and reported to modulate LuxR, TraR, or LasR activity in sensitive AHL “biosensor” strains previously considered specific for AHLs (DKPs 5–12; Figure 1, panel b) (26–28). These DKPs have been isolated either individually or as mixtures from culture supernatants of P. aeruginosa, P. fluorescens, P. putida, P. alcaligenes, Proteus mirabilis, Enterrobacter agglomerans, Vibrio vulnificus, and Citrobacter freundii. As these compounds appear to be common to a broad number of bacteria, they have been suggested to represent both a new class of naturally occurring QS signals, and potential interspecies signals (4, 6, 8, 10, 29–31). Further, as DKPs are common in fungi and elicit a range of effects in higher animals and plants (32), these compounds have also been proposed to play roles in interkingdom signaling. Questions remain with regards to their actual physiological function in bacteria, however, as DKPs elicit their activities against LuxR-proteins at significantly higher concentrations than AHLs (up to 106 times higher). To date, a critical examination of the DKP structure class as QS modulators is yet to be reported, and would provide insights into their role as putative cell-cell and cell-host signals (33).
Here, we report the design and synthesis of a library of non-native DKPs, an evaluation of these compounds as antagonists and agonists of LuxR, TraR, and LasR, and a comparison of their activities to the naturally occurring DKPs 5–10. Surprisingly, the previously reported DKPs failed to exhibit either antagonistic or agonistic activities in the reporter strains examined in this study. However, several non-native DKPs were identified that are capable of inhibiting (but not activating) luminescence in V. fischeri. Interestingly, these compounds exhibited structural features reminiscent of both the naturally occurring DKPs 8 and 9 and several of our recently reported, non-native AHL antagonists of LuxR, TraR, and LasR (15–18). Additional experiments indicated that luminescence inhibition by these non-native DKPs does not occur through LuxR in V. fischeri. Collectively, these data suggest that DKPs can influence QS regulated outcomes, yet may not do so through direct interactions with LuxR-type proteins. These findings have important implications for the alleged role of DKPs in bacterial QS (34).
We designed a focused library of 23 DKPs around the structures of native DKPs 5–10 (Figure 1, panel b), as these six DKPs had the most closely related structures of the eight (i.e., each contained L-Pro). We maintained the Pro unit in each library member, and then systematically altered the other amino acid side chain and stereochemistry at each chiral center to generate sub-libraries of cyclo(Pro-Xxx) 13–16 (shown in Figure 2). Sub-library 13 represented the simplest analogs, containing only one chiral center (L-Pro). DKPs 14 were derived from L-amino acids and designed to display various aliphatic and aromatic functionality (both natural and unnatural) on their non-Pro side chains. We included halogenated- and nitro-Phe derivatives because we have found that AHLs containing halogenated- and nitro-phenylacetyl groups (PHLs) exhibited both strong antagonistic and agonistic activity against TraR, LasR, and LuxR (e.g., 4-I-PHL 4) (15–18, 20). Structures 15 were stereoisomers of control DKPs 6–10, containing L-Pro and D-amino acids. Lastly, DKPs 16 were enantiomers of sub-library 15, containing D-Pro and L-amino acids.
We developed an efficient solution-phase synthetic route to DKPs 5–10 and sub-libraries 13–16 (Scheme 1). In brief, L- or D-Pro-OMe was coupled to N-Boc-protected amino acids using standard carbodiimide-mediated conditions. After cleavage of the N-Boc group under acidic conditions, intramolecular cyclization proceeded smoothly at room temperature upon treatment with piperidine to generate DKPs in sufficient yields for biological evaluation (38% average overall yield; 60–750 mg scale). All of the DKPs were purified to homogeneity by silica gel column chromatography (>97% purity, 95:5 d.r.; Supplementary Table 1).
Note, we did not utilize routine reverse-phase HPLC methods for DKP purification, as our HPLC instruments are utilized for the analysis of AHLs in related research (15–20), and we sought to minimize the risk of contamination by these highly active compounds. Such contamination, in part due to the similar physicochemical properties of DKPs and AHLs, has impacted research in this area previously. For example, Degrassi et al. fractionated P. putida WCS358 supernatant extracts using HPLC, tested each fraction against a number of AHL biosensor strains, and isolated DKPs 6, 8, 9, and 12 as the major products in each active fraction (Figure 1, panel b) (27). However, upon re-synthesis of these four DKPs, they found that the synthetic DKPs failed to activate the biosensor strain used for their initial detection, although they did activate other reporter strains. The researchers attributed the initial activities to AHLs that co-purified with their DKPs. For these reasons, we took precautions throughout this study to ensure the purity of our synthetic DKPs.
We examined the abilities of DKP controls 5–10 and sub-libraries 13–16 to modulate TraR, LasR, and LuxR activity in the same biosensor strains used in previous DKP reports (26–28) to allow for direct comparisons. We note that these biosensor strains can produce varying levels of LuxR-type protein (most frequently, substantially higher than native levels), and this can directly affect their sensitivity for exogenous ligands (i.e., higher protein levels correlate with higher sensitivity (35)). We sought to obtain data using biosensor strains with both high and native protein levels for each protein in this study to increase the stringency of the assays. Therefore, we also investigated DKPs 5–10 and 13–16 in bacterial reporter strains containing native LuxR-type protein levels (34). All of these strains lack AHL synthases, but contain LuxR-type proteins and cognate promoter sequences that control reporter gene expression. As a result, LuxR-type protein activity, and consequently exogenous ligand activity, can be measured using standard reporter gene read-outs. These assays can be performed in solution in multititer plates where activity is assessed using a plate reader, or the compounds can be overlaid with bacteria in warm agar and, following incubation, colorimetric reagents allow for visualization of activity.
All of the primary antagonism and agonism assays performed in liquid culture were tested at 500 μM DKP concentrations. Competitive antagonism assays were performed with DKP in the presence of native AHL ligand at its EC50, while agonism assays were performed with DKP alone. Negative controls reported the activity of media and DMSO only.
We began our investigation of DKPs 5–10 and 13–16 by overlaying them with a TraR overproducing strain (A. tumefaciens NTL4 (pZLR4), a second-generation strain of NT1(pDCI41E33) (36)) analogous to that described by Holden et al. and Degrassi et al. (26, 27, 37). Although we did not observe activation from cyclo(L-Leu-L-Pro) (6) as previously reported by both groups, its isomer cyclo(L-Ile-L-Pro) (14a) was active in this overlay assay, along with DKPs 13a and 15b (see Supplementary Figure 2). In contrast to both previous reports, DKP 9 did not activate TraR in our hands. We next sought to test the activity of the DKPs in a β-galactosidase reporter strain that produces TraR at native levels (A. tumefaciens WCF47 (pCF372) (38)). In this liquid culture assay, β-galactosidase, and therefore TraR, activity was measured using routine Miller absorbance assays (15, 17, 18). Neither the control DKPs 5–10 nor the DKP sub-libraries 13–16 were observed to activate TraR in this strain beyond the level of the negative control (Supplementary Figure 3). In turn, we did not observe inhibitory activity for any of the DKPs (5–10 and 13–16) against OOHL (2, at 100 nM) in competitive antagonism assays in this A. tumefaciens strain.
We next screened our DKP controls 5–10 and libraries 13–16 for agonistic activity against LasR in the heterologous E. coli biosensor strain previously used to examine DKP activity (pSB1075) (39). Controls 5–10 were inactive in this LasR overproducing strain. This result for control cyclo(L-Met-L-Pro) (7) contrasted with that of Holden et al., who reported that 7 could activate LasR (albeit at concentrations above 1 mM (26)). We did, however, observe partial activation (20%) of this strain with DKPs 14a, 14c, and 14h at 500 μM (Supplementary Figure 4). We sought to compare these data with reporter strain E. coli (NH5α (pJN105L pSC11)) (40). This strain contains LasR under the control of an inducible promoter and reports LasR activity via β-galactosidase production; LasR protein levels were induced to approximately native levels for P. aeruginosa with arabinose (17). Similar to the TraR assay data above, neither the control DKPs 5–10 nor the DKP sub-libraries 13–16 were capable of activating or inhibiting LasR (vs. 7.5 nM OdDHL, 3) in this native protein level reporter strain (see Supplementary Figure 5).
We evaluated the activities of the DKPs against LuxR using the E. coli JM109 (pSB401) biosensor (34, 39). This heterologous strain contains luxR and the luxI promoter from V. fischeri MJ-1, and the lux operon (luxCDABE) from Photorhabdus luminescens; LuxR is overproduced and protein activity is reported as luminescence (16). Control DKPs 5–10 failed to activate this strain, but 5 and 9 were able to inhibit luminescence by 20% (vs. 20 nM OHHL, Supplementary Figure 6). These data conflict with previous reports of controls 7–9 activating LuxR and all six controls 5–10 inhibiting LuxR in this same strain (26–28). However, several non-native DKPs from sub-libraries 13–16 were able to weakly inhibit (but not significantly activate) luminescence in this strain, most notably 14e, 14f, and 15f (~25% inhibition, Supplementary Figure 6).
We next examined control DKPs 5–10 and sub-libraries 13–16 for LuxR agonistic and antagonistic activities (vs. 3 μM OHHL, 1) in a ΔluxI derivative of V. fischeri ES114 (41, 42). In this strain, LuxR is produced at native levels, and the native V. fischeri lux operon behaves as the luminescent reporter (43). Although we identified no LuxR agonists in these screens (Supplementary Figure 7), several non-native DKPs were antagonists at a 100:1 ratio relative to OHHL (1). Compounds 14a, 14e, 14i, and 16c gave moderate inhibitory activities (33–40%, Figures 2 and and3).3). More notably, cyclo(L-Pro-L-4-Cl-Phe) 14b (the “natural” diastereomer of 16c) and cyclo(L-Pro-L-4-I-Phe) 14c were capable of inhibiting luminescence by 76% and 95% at 500 μM, respectively. These active DKPs (14b and 14c) exhibit (L, L) stereochemistry and structural features reminiscent of the naturally occurring control DKPs 8 and 9, yet contain non-native side chains. Interestingly, DKP 14c shares the 4-iodo phenyl motif with our previously reported, AHL-derived LuxR inhibitor, 4-I-PHL (4; Figure 1, panel b).
We performed dose response analyses on DKPs 14b and 14c against 5 μM OHHL (1) in V. fischeri ES114 (Δ–luxI), and obtained IC50 values of 208 μM and 116 μM for luminescence inhibition, respectively (Figure 4, panels a and b). These IC50 values are 2–3 orders of magnitude higher than those previously reported for AHL-derived LuxR inhibitors (such as 4) (16, 17).
To obtain further insight into the mechanism of luminescence inhibition by non-native DKPs, we performed competitive dose response analyses of the native ligand OHHL (1) in the presence of increasing amounts of the most active DKP, 14c, in V. fischeri ES114 (Δ–luxI). Such Schild analyses can reveal whether a compound behaves via a competitive or non-competitive antagonism mechanism, either directly interfering with the binding of an agonist or not, respectively (44). In earlier work, Holden et al. had presumed that DKPs bind at, or near, the OHHL (1) binding site on LuxR because their activators displayed weak competitive inhibition against OHHL (26). However, competitive dose response experiments with 14c showed that this compound reduces the maximal luminescence induction level of OHHL (1) without affecting the EC50 value of OHHL (1) (Figure 4, panel c), suggesting that 14c does not directly compete with OHHL (1) for LuxR binding. In addition, this experiment indicates that DKP 14c does not affect production of LuxR in V. fischeri, as this outcome would also alter the EC50 for OHHL (1) at different concentrations of 14c. Later experiments revealed that the luminescence inhibitory activity of DKP 14c is also reversible in V. fischeri ES114 (Δ–luxI), as luminescence could be induced by adding fresh OHHL (1) after washing cells that had been treated with DKP 14c (data not shown).
In order to further probe the mechanism of action of DKP 14c, we performed dose response analyses in several additional mutant strains of V. fischeri. These strains lacked either proteins that have been implicated in DKP activity in other Vibrio spp. (OmpU and ToxR) (28, 31, 45, 46), or proteins that have been recently reported to regulate luminescence in V. fischeri (CheV, FlrC, and YehT) (47, 48). We found, however, that the luminescence inhibitory activity of 14c was not affected by the loss of these proteins, indicating that this DKP does not directly interact with them (see ref. 45 and 48 for additional details).
We next turned our focus toward the luciferase enzyme itself. If DKP 14c interfered with luciferase production in V. fischeri, or if it directly inhibited the enzyme, we would expect to observe a reduction in luminescence. To test these possibilities, we measured the levels of active luciferase in cells cultured with and without 14c. We used our previously reported synthetic LuxR inhibitor 4-I-PHL (4) as a control in these experiments (16). We pre-cultured V. fischeri ES114 (Δ–luxI) in the presence of 3 μM OHHL, DMSO alone, 3 μM OHHL (1) with 1 mM 14c, or 3 μM OHHL (1) with 10 μM 4-I-PHL (4) to the same optical density to normalize the number of cells in the culture (~0.4; 14c and 4-I-PHL were added to the pre-culture at 10x their IC50 values). We then pelleted the cells and tested luminescence levels according to the procedure described by Lei and Becvar (49). Briefly, this procedure utilizes an excess of the reactants (decanal and flavin mononucleotide (FMNH2)) to ensure that luciferase is the rate-limiting component in the reaction, and the initial maximum intensity of light emission is proportional to luciferase activity (see Supporting Information for details). The results from this assay are shown in Figure 5. As expected, we observed that OHHL (1) is required for luciferase production during the pre-culture (luciferase output (LO) = 72.4 ± 7.6). Addition of 4-I-PHL (4) fully inhibited the luminescence of the pre-cultured cells and also decreased the amount of active luciferase produced during the pre-culture (LO = 11.0 ±1.0). When added directly to the enzymatic assay reaction, 4-I-PHL (4) had no effect on the levels of luciferase activity (LO = 73.0±7.0). Similarly, the amount of active luciferase was reduced by >80% when the cells were pre-cultured with DKP 14c (LO = 13.2±4.4). However, in contrast to 4-I-PHL (4), 14c affected the activity of luciferase in the enzymatic assay when added directly to the enzymatic assay buffer (LO = 24.4±1.4), although not to the same degree as in pre-cultured cells. We were unable to test DKP 14c at higher concentrations in this assay, as the compound is not soluble in the absence of DMSO and greater than one percent DMSO in the assay buffer (v/v) greatly affected luciferase activity (LO ~ 49 ± 6). Investigations are ongoing to further characterize the effects of 14c on luciferase in V. fischeri. Nevertheless, these data suggest that DKP 14c affects the levels of active luciferase in V. fischeri, potentially by inhibiting the enzymatic activity of luciferase itself.
Few molecules other than AHLs are known to modulate QS in Gram-negative bacteria, and DKPs have attracted considerable interest as a potential new class of QS signals, and even interspecies or interkingdom signals (50). There have been no reports investigating either their mechanism or the structural features of these molecules that are required for activity. We designed and synthesized focused libraries of non-native DKPs (13–16) based on the structures of naturally occurring DKPs 5–10, and examined their abilities to agonize and antagonize well-characterized LuxR-type proteins using both sensitive biosensor strains and reporter strains with native protein levels. The results of these assays suggest, in contrast to previous reports, that DKPs do not interact with LuxR-type proteins to affect QS. DKPs 5–10 and all of the DKPs in libraries 13–16 derived from natural α-amino acids failed to activate or inhibit the native protein level QS reporter strains utilized in this study based on TraR, LasR, and LuxR. The only DKPs capable of significantly modulating LuxR QS outputs (i.e., luminescence) were derived from non-natural amino acids, DKPs 14b and 14c. Further experiments showed that these DKPs do not compete with the natural ligand OHHL (1) for LuxR. However, DKP 14c is able to reduce the level of active luciferase in V. fischeri when pre-cultured in the presence of OHHL (1). As natural DKPs 5–12 were discovered largely using AHL biosensors in which the LuxR-type proteins are overproduced, our results also highlight the need to use caution when interpreting such biosensor data for the identification of small molecule LuxR-type protein modulators.
Therefore, the question remains: Do DKPs function in nature as true QS signal molecules? Considerable further experimentation is required to provide a conclusive answer to this question, but the data outlined herein challenge the hypothesis that they do so through interaction with LuxR-type proteins in Gram-negative bacteria. It is important to note that the DKPs in this study were only active in V. fischeri, and that naturally occurring DKPs, while present in other Vibrio species, have yet to be reported in isolates from this bacterium. Nonetheless, the present work has provided evidence that DKPs do not act as LuxR-mediated QS signals, and further underscores the value of synthetic chemistry to probe the language of bacterial QS.
All chemical reagents were purchased from commercial sources (Sigma Aldrich, EMD Biosciences, Advanced ChemTech) and used without further purification. Solvents were purchased from commercial sources (Fisher Scientific, Mallinckrodt Baker) and used as obtained, with the exception of dichloromethane (CH2Cl2), which was distilled over calcium hydride prior to use. AHL controls 1–3 and 4-I- PHL (4) were synthesized according to our previously reported procedure (15–18).
DKP controls 5–10 and sub-libraries 13–16 were synthesized as follows: N-Boc-α-amino acids were coupled to H-Pro-OMe with 1 equiv. EDC and 1 equiv. Et3N in CH2Cl2 at 4 °C overnight (51). The resulting dipeptides were dissolved in a minimal amount of CH2Cl2 and cooled to 0 °C. A solution of AcCl (~5 equiv.) in MeOH (1:2.4 v/v) was added drop-wise to affect amine deprotection, and the reaction was stirred for 3 h. The solvent was removed in vacuo, and the deprotected dipeptides were dissolved in a minimal amount of DMF. Approximately 2 equiv. of piperidine were added to these solutions to facilitate cyclization, and the reactions were stirred at room temperature for 1 h. The solvent was removed, and the resulting solids were purified by flash silica gel chromatography (1–4% (v/v) gradient of MeOH in CH2Cl2) to give DKPs in 15–86% yields and >97% purities with 95:5 d.r.
References to the strains can be found throughout the text. Additional information about these strains and culture conditions is given in Supplementary Table 2.
Stock solutions of synthetic compounds were prepared in DMSO and stored in sealed vials. For agonism assays, an appropriate amount of DKP stock solution was added into wells of a 96-well multititer plate to give a 500 μM final concentration. For competitive antagonism assays, an appropriate amount of AHL stock solution was also added to each well to give a final AHL concentration equal to the EC50 value in that strain. The concentration of DMSO was normalized in all wells (<2% (v/v)). Absorbance assays for A. tumefaciens WCF47 (pCF372) and E. coli DH5α (pJN105L pSC11), and luminescence assays for E. coli JM109 (pSB401) and V. fischeri strains were performed according to our previously reported methods (15–18, 20). All assays were performed in triplicate. GraphPad Prism software (version 4.0c) was used to calculate IC50 and EC50 values.
We thank the NIH (AI063326-01), Research Corporation, Burroughs Wellcome Foundation, Greater Milwaukee Foundation Shaw Scientist Program, Johnson & Johnson, and DuPont for financial support of this work. H.E.B. is an Alfred P. Sloan Foundation Fellow. J.C.O. was supported by a Novartis Graduate Fellowship in Organic Chemistry. G.D.G was supported by an ACS Division of Medicinal Chemistry graduate fellowship. We thank Professors J. Handelsman, E. Ruby, S. Winans, and K. Visick for generous donations of bacterial strains, technical assistance, and contributive discussions. B. Carlson is acknowledged for assistance in the synthesis and characterization of several DKPs.
Supporting Information Available: Full characterization data for DKPs 5–10 and 13–16, details of all bacterial reporter strains tested, assay procedures, and assay data. This material is available free of charge via the Internet.