|Home | About | Journals | Submit | Contact Us | Français|
We synthesized a range of PQS (Pseudomonas quinolone signal; 2-heptyl-3-hydroxy-4(1H)-quinolone) analogues and tested them for their ability to stimulate MvfR-dependent pqsA transcription, MvfR-independent pyoverdine production, and membrane vesicle production. The structure-activity profile of the PQS analogues was different for each of these phenotypes. Certain inactive PQS analogues were also found to strongly synergize PQS-dependent pyoverdine production.
Pseudomonas aeruginosa is a major cause of chronic infections, especially among individuals with cystic fibrosis (CF) (12, 14). Virulence factor and biofilm production by P. aeruginosa is controlled by quorum sensing (QS). The QS system of P. aeruginosa involves the following two distinct types of signaling molecules: N-acylated homoserine lactones (AHLs) (23) and the Pseudomonas quinolone signal (PQS; a 4-alkyl-quinolone) (18). The biosynthetic precursor of PQS, 2-heptyl-4(1H)-quinolone (HHQ), has also been demonstrated to function as a signaling agent in P. aeruginosa (24). AHL-dependent signaling is widespread among Gram-negative bacteria (19). However, 2-alkyl-4-quinolone (AHQ) signaling is more restricted and has so far been detected only in P. aeruginosa and certain Burkholderia and Alteromonas spp. (7, 21). PQS has been reported in the sputa of CF patients (estimated minimal concentration of ca. 2 μM) (3), indicating that it may be an important factor in the establishment of chronic infections.
PQS has been shown to control production of multiple virulence determinants, including elastase, pyocyanin, rhamnolipids, and the siderophore pyoverdine (5, 9, 11, 18). PQS is extremely hydrophobic and is trafficked between cells in membrane vesicles (MVs) (17). MV formation is also stimulated by PQS, which has been shown to interact with the lipopolysaccharide (LPS) layer of the outer membrane (16). PQS biosynthesis is directed in 2 steps. The PqsABCD proteins catalyze the condensation of β-keto fatty acids with anthranilate to yield AHQs, prominent among which is 2-heptyl-4(1H)-quinolone (HHQ) (2, 6, 18). HHQ is a diffusible molecule that can be produced in one cell and then converted to PQS in another through the action of an unlinked monooxygenase, PqsH (6, 8). Synthesis of PQS is primarily under the control of the AHL-dependent QS system (22), although it can also be produced independently of it (9, 25). LasR stimulates transcription of pqsH and mvfR (a LysR-type regulator, also known as PqsR) (5, 9, 11, 24). MvfR activates transcription of pqsABCDE. PQS is not the only alkyl-quinolone made by P. aeruginosa (6, 18). Others include HHQ, 2-nonyl-4-quinolone (NHQ), and 2-heptyl-4-quinolone N-oxide (HHQNO). The signaling function(s), if any, of NHQ and HHQNO is not clear.
In spite of the important role(s) played by PQS in virulence, it has not yet been the subject of a systematic structure-activity relationship (SAR) analysis. Previous structural modifications of 4-alkyl quinolones have focused mostly on assessing the activity of alkyl chain congeners in stimulating transcription from the pqsA and lecA promoters (10) and MV production (15). In the current work, we generated a more diverse range of PQS analogues and tested them for their ability to stimulate (i) MvfR-dependent transcription of pqsA, (ii) MvfR-independent production of pyoverdine, and (iii) membrane vesicle formation. The analogues we made are shown in Fig. Fig.1.1. The synthesis of these analogues will be described elsewhere (J. Hodgkinson, M. Welch, and D. R. Spring, unpublished data).
The ability of the analogues to stimulate MvfR-dependent transcription from the pqsA promoter was assessed as previously described (4) in Escherichia coli strain DH5α containing pEAL08-2. This plasmid encodes MvfR (under the control of the tac promoter) and carries lacZ under the control of the pqsA promoter. Each analogue (or, as a control, dimethyl sulfoxide [DMSO]) was added at the start of the growth curve, and the β-galactosidase activity of the cells was measured after 8 h, as described previously (4). The results are shown in Fig. Fig.2.2. In this heterologous E. coli-based reporter system, the concentration of PQS required to stimulate PqsR-dependent pqsA transcription was ~1,000-fold lower (nM versus μM) than the concentration of PQS reported by Fletcher et al. (10) to do the same in P. aeruginosa. This suggests that in P. aeruginosa, additional factors (e.g., PQS efflux) may influence the measured response and that the E. coli-based system represents a more direct assay of PqsR activity. There was a strong alkyl chain length dependency on agonism; analogue 1, which carries only a short “stub” of an alkyl chain, was essentially inactive, while analogues 2 and 3 (containing alkyl chains with 5 and 6 carbon atoms, respectively) displayed progressively greater stimulatory activity. Analogues with alkyl chains longer than 7 carbon atoms retained substantial agonist activity, indicating that the PQS-binding pocket of MvfR can accommodate these variants. These data are consistent with those of Fletcher et al. (10), who noted similar alkyl chain length dependency using a pqsA::lux reporter system in a P. aeruginosa PAO1 background. However, the alkyl chain binding pocket of MvfR was apparently intolerant of nonalkyl substituents, since replacement of the PQS heptyl side chain with phenyl groups (Fig. (Fig.2,2, analogues 7 and 8) greatly reduced or abolished agonist activity. Introduction of chlorine atoms at each of the four available positions on the anthranilate ring (Fig. (Fig.2,2, analogues 9 to 12) led to a rather uniform diminution in agonist activity, whereas introduction of a fluorine atom at position 5 had no effect (cf. analogues 9 and 14). This may be due to steric effects of the substituents (F has a smaller van der Waals radius than Cl), although the introduction of electron donating −OH or −OMe groups at position 6 (Fig. (Fig.2,2, analogues 13 and 16) or 7 (analogue 15) led to a large reduction in activity, suggesting that the electron density of the aromatic ring may also play a part.
PQS can stimulate pyoverdine production in a MvfR-independent manner (1, 8). Since PQS is also an iron chelator, presumably due to the presence of the 3-hydroxy-pyridin-4-one moiety, this has led to the suggestion that PQS stimulates pyoverdine production indirectly by depleting the culture of utilizable iron (1, 8, 26). However, the dependence of pyoverdine production on PQS structure has not so far been systematically studied. Pyoverdine production was measured in P. aeruginosa strain DH125 (a PA14-derived pqsB::TnphoA strain) (4). Due to the disruption of pqsB, this strain is unable to generate endogenous alkyl quinolones. Dose-response curves showed that addition of 20 to 30 μM exogenous PQS stimulated a sharp increase in pyoverdine production by DH125, which saturated at around 60 μM PQS (data not shown). Therefore, we tested the PQS analogues at 30, 60, and 90 μM concentrations (Fig. (Fig.3).3). Since all our analogues contain the 3-hydroxy-pyridin-4-one moiety, we expected to see relatively little difference in their abilities to stimulate pyoverdine production. However, this was not the case. At 30 μM, only analogues PQS, 4, and 6 (and to a lesser extent, analogue 3) stimulated pyoverdine production to significant levels. Also, and unlike MvfR-dependent pqsA transcription, there was an abrupt abolition of pyoverdine-stimulatory activity when the alkyl chain exceeded 8 carbon atoms in length (Fig. (Fig.3,3, cf. analogues 4 and 5). Notably, introduction of an unsaturated bond at the end of the alkyl chain (Fig. (Fig.3,3, analogue 6) increased pyoverdine production compared with that of the saturated equivalog (analogue 4) and PQS. Substitution of the alkyl chain with a phenyl moiety (Fig. (Fig.3,3, analogues 7 and 8) essentially abolished activity. In contrast to the MvfR-dependent activity profile, introduction of a fluorine atom at anthranilate ring position 5 resulted in lower agonist activity, while chlorine atom substitutions at positions 5 to 8 displayed the following activity profile: position 5 > position 6 > position 7 > position 8. Modifications that introduced electron-donating groups (Fig. (Fig.3,3, analogues 13, 15, and 16) were inactive at all concentrations tested. We conclude that the stimulation of pyoverdine production shows a strong dependence on several aspects of PQS structure unrelated to the 3-hydroxy-pyridin-4-one moiety per se. In this regard, we note that analogue 1, which has been experimentally demonstrated to bind Fe3+ with high affinity (log β3 = 36.2) (8), was unable to induce pyoverdine production at any of the concentrations tested.
To further investigate the interplay between iron availability and AHQs, we examined whether iron deprivation or iron supplementation affected the pyoverdine-inducing activity of a selection of analogues (PQS and compounds 5, 6, and 8, all at a concentration of 90 μM). As shown in Fig. Fig.4,4, in the presence of excess iron in the medium (Fig. (Fig.4A),4A), pyoverdine production was low and unaffected by the addition of the AHQs. In LB alone (Fig. (Fig.4B),4B), the same pattern of induction as that shown in Fig. Fig.33 was observed. In deferrated medium (Fig. (Fig.4C),4C), pyoverdine production was uniformly high and unaffected by the AHQs. These data show that when iron is either limiting or in excess, it overrides the effect(s) of AHQs on pyoverdine production.
We wondered whether any of the “inactive” or partially active analogues might antagonize pyoverdine production. DH125 was grown in the presence of 25 μM PQS and 75 μM selected analogues (1, 2, 3, 5, 7, 8, 12, 13, 15, and 16), and pyoverdine was assayed. The results are shown in Fig. Fig.5.5. No antagonism was observed. Instead, we noted that some of the compounds strongly synergized the activity of PQS. Of particular note, analogues 2 and 5, which were themselves completely inactive at stimulating pyoverdine, and analogue 12, which very poorly stimulated siderophore production, yielded substantial induction ratios (Fig. (Fig.5).5). Compounds 15 and 16, which were also completely inactive (Fig. (Fig.3),3), displayed partial PQS synergism. Only compounds 1, 7, and 8 did not display synergism. These data raise the possibility that the nominally “nonactive” endogenous PQS analogues such as NHQ may function to “fine-tune” the activity of PQS itself. Speculatively, the molecular basis for the observed synergism might be through facilitation of PQS uptake or trafficking, perhaps linked to the ability of synergizing compounds to specifically interact with the cell envelope.
We next investigated whether the PQS analogues influenced membrane vesicle (MV) formation by P. aeruginosa. The results are shown in Fig. Fig.6.6. Consistent with previous reports (15, 17), we found that wild-type cultures produced more MVs than the PQS null mutant. However, in our hands, most of the compounds tested (including PQS itself) were able only to slightly stimulate MV production (<2-fold) when present at a 50 μM concentration. This contrasts with the findings of earlier works that reported nearly complete restoration of MV production at this concentration of PQS (Fig. (Fig.6,6, wild type [WT]) (17) and may reflect the different approaches used to quantify the MVs (total protein content measured in the previous work versus phospholipid content measured in the current work). This notwithstanding, compounds 9, 10, and 11 were able to restore MV production to much higher levels than PQS, indicating that the introduction of chlorine atoms at positions 5, 6, and 7 (respectively) of the anthranilate ring increases MV-stimulatory activity. Other modifications on the anthranilate ring (Fig. (Fig.6,6, compounds 12 to 16) did not substantially enhance the MV-stimulatory activity of the molecules.
Previous works have shown that most of the PQS made by P. aeruginosa is packaged into MVs (13, 17), raising the possibility that specific structural features of the molecule might affect interaction with the cell envelope by affecting its lipophilicity. A chemical informatics approach (www.molinspiration.com/cgi-bin/properties) was therefore used to calculate the octanol-water partition coefficient (logP value) for each analogue. In parallel, we also measured the solubility of each analogue in water. There was a good correlation between these parameters (Table (Table1).1). Notably, the least active compounds (1, 7, and 8) showed the greatest water solubility, consistent with there being a minimum lipophilicity required for PQS. However, lipophilicity alone does not account for the dramatic functional impact of subtle structural variations. For example, extension of the alkyl chain from 8 to 9 carbon atoms (Table (Table1,1, cf. analogues 4 and 5) increases the logP value but abolishes pyoverdine stimulation.
Overall, the correlation between MvfR binding and pyoverdine stimulation was moderate. When the induction ratios for each assay are plotted against one another, two distinct clusters are observed (Fig. (Fig.7).7). Compounds 4, 6, 9, 10, 11, 14, and PQS all showed good inducing activity in both assays (cluster I), indicating that these modifications to PQS are in structurally tolerant positions. Interestingly, some members of cluster I were also better at stimulating MV production than the compounds in cluster II. The latter were less active, indicating that if (i) the alkyl chain is too short (analogues 1 to 3) or too long (analogue 5), (ii) the alkyl group is substituted by a phenyl moiety (analogues 7 and 8), (iii) the anthranilate ring carries electron-donating groups (analogues 13, 15, and 16), or (iv) position 8 in the anthranilate ring carries an electronegative atom (analogue 12), the specific signaling or MV-promoting activity of the molecule becomes compromised. Therefore, almost every part of the PQS molecule contributes in some way toward its function as a signaling molecule. Alternatively, these specific structural factors may diminish the bioavailability of the molecules. We also note that, with the exception of the water-soluble analogues (1, 7, and 8), all members of cluster II retained PQS-synergistic activity, indicating that this function of the molecule is more dependent on its general properties (e.g., lipophilicity) rather than specific chemical groups.
Our results further support the notion that the structure of PQS is constrained through interaction with multiple factors (currently known to include LPS, MvfR, and iron). Furthermore, it is possible that additional PQS receptors remain to be identified. In this regard, both Burkholderia thailandensis and Burkholderia pseudomallei contain functional pqsABCDE homologues (7, 21), but neither of their genomes encodes a convincing MvfR homologue.
Research done in the laboratory of M.W. is generously supported by the BBSRC and MRC. Research done in the laboratory of D.R.S. is supported by the EPSRC, BBSRC, MRC, and the Wellcome and Newman Trusts. S.D.B. is a BBSRC-funded postdoctoral fellow. J.H. was supported by an MRC-funded Ph.D. studentship.
Published ahead of print on 21 May 2010.