Biosynthetic Pathway of Pseudomonas aeruginosa 4-Hydroxy-2-Alkylquinolines

Bacterial strains and culture conditions.

Clinical P. aeruginosa strain SCV 20265, isolated from the respiratory tract of a CF patient who attended the cystic fibrosis clinic at Hanover Medical School, Hanover, Germany (12), was used in this study. This P. aeruginosa strain produced large amounts of HAQs and therefore was especially suitable for the feeding experiments with labeled precursors. P. aeruginosa was routinely cultured at 37°C on Columbia or Luria-Bertani (LB) agar. Transposon mutants of SCV 20265 with an insertion in the carB, pyrB, or pyrD gene that was generated using the transposon construction vector EZ:TN pMOD-2 (Epicentre) were grown in LB medium supplemented with 50 μg/ml gentamicin. The PAO1 wild-type strain and the rhlG:Tc mutant strain used in this study were kindly provided by G. Soberon-Chavez (Departmento de Microbiologia Molecular, Universidad Nacional Autonoma de Mexico) (3).

Broth cultures of P. aeruginosa SCV 20265 and transposon mutants affected in genes of the pyrimidine metabolic pathway were grown in LB medium containing in some cases, as indicated below, either 5 mM orotic acid or 5 mM UMP.

Extraction of extracellular P. aeruginosa HAQ metabolites and thin-layer chromatography (TLC).

To isolate HAQ metabolites, P. aeruginosa cultures grown on agar plates were harvested and suspended in methanol. Following a centrifugation step and evaporation of the methanol to completion, the dried residue was washed several times with distilled water. Alternatively, the secondary metabolites were extracted from P. aeruginosa broth cultures with dichloromethane. Briefly, the bacterial cultures were extracted with 1 volume of dichloromethane by vigorous shaking. After centrifugation at 2,000 × g for 15 min the lower organic layer was evaporated.

TLC was performed using a Silica Gel 60 F254 TLC plate. The extracted P. aeruginosa material was dissolved in methanol and separated by TLC using tert-butyl methyl ether-n-hexane (10:1) as the solvent. Fluorescent spots were visualized under UV light and photographed. 4-Hydroxy-2-heptylquinoline was chemically synthesized as described previously (22). Starting from 4-hydroxy-2-heptylquinoline, PQS was synthesized by the procedures described by Pesci et al. (20). The two HAQs and 4-hydroxy-2-heptylquinoline-N-oxide, the latter purchased from Sigma (St. Louis, MO), were used as standards in TLC.

Isotope labeling.

Isotope-labeled anthranilic acid (¹⁵N; 1 mg/ml) or [1,2-¹³C]acetate (2 mg/ml) was added to a bacterial culture that had been grown for 24 h on 0.2% agarose in modified Vogel-Bonner medium (3.3 mM MgSO₄, 10 mM citric acid, 28 mM NaNH₄HPO₄, 37 mM K₂HPO₄, 214 mM d-gluconic acid; pH 7.4). The plates were incubated for another 48 h before HAQs were methanol extracted. The anthranilic acid-labeled samples were subjected to GC-MS, whereas the ¹³C-labeled samples were subjected to one-dimensional (1D) ¹³C NMR spectroscopy.

To increase the ¹³C label of extracted HAQs, P. aeruginosa SCV 20265 was cultured by using 0.2% agarose in modified Vogel-Bonner medium without d-gluconic acid supplemented with 2 mg/ml [1,2-¹³C]acetate, [1-¹³C]acetate, or [2-¹³C]acetate with and without 5 mM orotic acid or 10 mg β-ketodecanoic fatty acid. In order to avoid ¹³C incorporation into the HAQ moiety originating from anthranilic acid, 1 mg/ml of unlabeled anthranilic acid was added to the medium in these experiments. The plates were incubated for 3 days, and the HAQs were methanol extracted and subjected to TLC and GC—MS-MS.

GC-MS analysis.

4-Hydroxy-2-alkylquinoline derivatives were analyzed after trimethylsilylation (50% pyridine—50% BSTFA [bistrimethylsilyltrifluoroacetamide] containing 1% TMC [trimethylchlorosilane]), (70°C, 1 h) with a Thermo-Finnigan GCQ ion trap mass spectrometer (Finnigan MAT Corp., San Jose, CA) running in the positive-ion electron impact (EI) mode equipped with a 30-m DB5 capillary column. For determination of the exact positions of ¹³C and ¹⁵N atoms after stable isotope labeling studies, MS-MS experiments were performed after isolation of the respective monoisotopic parent ions using excitation energies of ca. 2.0 eV.

NMR spectroscopy.

NMR spectra of the purified selectively ¹³C-labeled 4-hydroxy-2-alkylquinolines (1D ¹H and ¹³C) and of unlabeled synthetic material (1D ¹H and ¹³C and two-dimensional heteronuclear multiple bond correlation) were recorded at 300 K with Bruker DPX 300 and APX 400 NMR spectrometers locked to the major deuterium resonance of the solvent, CD₃OD.

Extraction of P. aeruginosa rhamnolipids.

For isolation of rhamnolipids, P. aeruginosa cultures were grown on soft agar (0.2% agarose [Difco]) in normal Vogel-Bonner medium or Vogel-Bonner medium in which the d-gluconic acid had been replaced by 25 mM sodium acetate. After incubation for 4 days at 37°C, 20-ml portions of the soft agar cultures were extracted with 10 ml dichloromethane by vigorous shaking. After centrifugation at 2,000 × g for 15 min, the lower organic layer was evaporated and subjected to GC-MS.

Analysis of isotope-labeled P. aeruginosa HAQs by GC-MS and NMR.

Previous studies on the biosynthesis of HAQs revealed that anthranilic acid is a direct precursor, and it has been speculated that HAQs are produced in a multiple-step reaction by the condensation of anthranilic acid and β-keto fatty acids. However, in a previous screening for HAQ-negative P. aeruginosa transposon mutants, we isolated mutants affected in the carB, pyrB, and pyrD genes. Although these genes are not neighbors on the chromosome, they all encode enzymes successively required for the biosynthesis of orotic acid in the pyrimidine biosynthetic pathway. In these mutants HAQ biosynthesis could be complemented by addition of 5 mM orotic acid to the growth medium, as demonstrated by TLC after 3 days of growth. Thus, it may be argued that orotic acid could be a direct precursor of HAQs. In an initial step anthranilic acid and orotic acid might react to form kynurenic acid, and the pqsABCDE operon could then be responsible for the attachment of an even-number carbon chain (C₆ or C₈) at the C-2 carbon of the quinolone ring to produce 4-hydroxy-2-heptyl- and 4-hydroxy-2-nonylquinoline, respectively (Fig. 1). To confirm either the previously proposed biosynthetic pathway for HAQs involving the condensation of anthranilic acid and a β-keto-fatty acid or the possibility that anthranilic acid and orotic acid are the precursors of HAQ synthesis via the formation of kynurenic acid, we performed feeding experiments with isotope-labeled substrates. First, P. aeruginosa SCV 20265 was grown in the presence of ¹⁵N-labeled anthranilic acid on modified Vogel-Bonner swimming agar plates with d-gluconic acid as the main carbon source. After growth for 3 days the bacterial mass was extracted with methanol, and the isolated secondary metabolites were subjected to GC-MS analysis. The upper panel of Fig. 2 shows a typical gas chromatogram of the trimethylsilylated HAQ derivatives, whereas the lower panel shows the EI mass spectrum of the unlabeled heptyl derivative and the fragmentation pattern assigned in the inserted scheme. Compared to the unlabeled control, all extracted 4-hydroxy-2-alkylquinolines contained approximately 66% ¹⁵N, as easily deduced from the corresponding EI mass spectrum of the trimethylsilylated derivatives by the characteristic shift of the molecular ions at m/z 315 (heptyl derivative) and m/z 343 (nonyl derivative) and the intense fragment ion generated by elimination of the alkyl side chain at m/z 231 by 1 Da (Fig. 2). Incorporation of the ¹⁵N from labeled anthranilic acid was also observed for all other HAQ metabolites, and this clearly indicated that anthranilic acid served as a common precursor for PQS, 4-hydroxy-2-heptylquinoline, 4-hydroxy-2-nonylquinoline, and 4-hydroxy-2-heptylquinoline-N-oxide. Moreover, since the only nitrogen of anthranilic acid was labeled, this indicates that N-1 of HAQ originates from anthranilic acid.

To shed light on the identity of the carbon substrate, especially at positions 2, 3, and 4 in the quinolone ring system, P. aeruginosa cultures were incubated for 3 days on modified Vogel-Bonner medium containing either [1-¹³C]acetate, [2-¹³C]acetate, or [1,2-¹³C]acetate in addition to unlabeled anthranilic acid as the main carbon source. According to the pattern of label incorporated, one can distinguish whether anthranilic acid and labeled β-keto acid (the pattern is dependent on the [¹³C]acetate source) condense “head to head” or whether an unusual polyketide synthesis mechanism extending activated anthranilic acid with labeled acetate moieties is responsible for PQS production. Substantial label incorporation into the HAQ metabolites was demonstrated. It is noteworthy that compared to cultures grown in minimal medium with d-gluconic acid as the main carbon source, it was 4-hydroxy-2-nonylquinoline and not 4-hydroxy-2-heptylquinoline that proved to be the main component in the HAQs. A characteristic mass shift of the molecular and fragment ions of the 4-hydroxy-2-alkylquinoline derivatives was observed for all labeling experiments by GC-MS after trimethylsilylation. Whereas the size of the molecular ion of the heptyl derivative increased by 4 Da after labeling with [1-¹³C]acetate, the size of the molecular ion of the corresponding nonyl derivative increased by 5 Da in accordance with elongation of the alkyl chain by an additional two-carbon unit. The intense fragment comprising the ring system at m/z 231 (Fig. 3, lower panel) incorporated only one ¹³C atom for both derivatives. Interestingly, the nonyl derivative exhibited in this and all other labeling experiments in which acetate was used as the major carbon source a clearly higher rate of ¹³C incorporation (ca. 30%) than the heptyl derivative. This behavior can be explained by a later onset of synthesis for the longer-chain derivative than for the shorter-chain derivative. Whereas before addition of acetate predominantly the short-chain variant was synthesized, the availability of large amounts of acetate caused increased synthesis of longer-chain variants. Thus, the chain size seems to depend on both the carbon source and the growth phase.

An identical experiment performed with [2-¹³C]acetate instead of [1-¹³C]acetate resulted in mass shifts of 5 and 6 Da for the molecular ions of the heptyl and nonyl derivatives, respectively, and a shift of 2 Da for the fragment ion at m/z 231. Therefore, the heterocyclic part of the quinoline ring system must also be synthesized from acetate units, since the labeled acetate could be clearly found not only in the alkyl side chain but also in the ring system. Furthermore, the difference between the two acetate labels can only be explained by the elimination of one CO₂ molecule originating from the carboxylate moiety of acetate during biosynthesis. As expected, labeling with [1,2-¹³C]acetate led to a mass increase of 9 or 11 Da for the molecular ions and a shift of 3 Da of the fragment ion at 231 Da. To confirm these results, MS-MS fragmentation was performed for the molecular ions of the [1-¹³C]acetate- and [2-¹³C]acetate-labeled 4-hydroxy-2-heptylquinolines at m/z 319 and 320, respectively, as shown in Fig. 3. The alternating presence of ¹³C and ¹²C carbons in the alkyl chain of the molecules can be directly deduced from the corresponding fragments and also the fragments at m/z 232 and 233, respectively, incorporating the quinoline ring system, indirectly supporting the hypothesis that all HAQ derivatives are synthesized by “head-to-head” condensation of anthranilic acid and β-ketodecanoic or β-ketododecanoic acid with elimination of the carboxylate group of the fatty acid as CO₂. A similar MS-MS experiment was performed for the molecular ion of the nonyl derivative, and analogous results were obtained (data not shown). The addition of unlabeled orotic acid in addition to anthranilic acid to the medium, which should have drastically reduced the incorporation of ¹³C into the ring system if the alternative biosynthetic pathway via kynurenic acid is correct, had absolutely no effect on this pattern. Addition of 3-oxo-decanoate, anthranilic acid, and [1,2-¹³C]acetate to the medium, however, resulted in labeling of the nonyl derivative (3 Da of the ring fragment and 11 Da of the molecular ion; compare with the data described above), whereas the labeling rate of the heptyl derivative was reduced by a factor of approximately 10, demonstrating that predominantly added unlabeled 3-oxo-decanoate is used for the biosynthesis of this derivative.

The biosynthetic pathway via β-keto-(do)decanoic acid could be independently demonstrated by ¹³C NMR spectroscopy. For this experiment, HAQ produced with relatively small amounts of [1,2-¹³C]acetate added to the medium was used, resulting in a degree of labeling of ca. 1%. Since the presence of two neighboring ¹³C atoms can be expected in the side chain and also in the heterocyclic ring system if our proposed biosynthetic model is correct, C-C coupling between intact acetate-derived units should be observed by 1D ¹³C NMR, whereas the probability of coupling between ¹³C atoms originating from different acetate units is too low to be observable. In this way, coupling between C-2 of the quinoline ring system and the first carbon of the side chain, C-9, with a characteristic coupling constant was indeed observed (Table 1), unequivocally demonstrating the presence of an intact acetate-derived unit at this position. The observation of doublet signals for the three carbons at the chain end confirmed the regular incorporation of acetate units in the fatty acid section of the molecule. No coupling was detected between C-3 and C-4 of the quinoline ring (Table 1), confirming the elimination of the carboxylate carbon of the final acetate unit as CO₂ used to synthesize HAQ derivatives via β-keto fatty acids. Therefore, the C-4 carbons of these molecules must originate from anthranilic acid.

Since we demonstrated that β-keto acids are direct precursors of HAQs, it might be speculated that the probable coenzyme A ligase encoded by pqsA is responsible for the activation of the β-keto acids. pqsD might then be involved in a transacetylation reaction, and finally, β-keto-acyl-acyl carrier protein might be condensed with anthranilic acid to form PQS. These processes are probably also dependent on enzymes of the primary fatty acid metabolism. Future studies will have to elucidate the enzymatic function of the proteins encoded by the pqsABCDE operon and to identify whether and which enzymes of the primary fatty acid metabolism are also required for PQS biosynthesis.

Growth characteristics of P. aeruginosa mutants affected in the pyrimidine metabolic pathway.

To further verify that orotic acid is not a direct precursor of HAQ production, we tested whether HAQ production by mutants with mutations in the pyrimidine metabolic pathway could be restored by not only orotic acid but also UMP. UMP is a direct precursor of DNA biosynthesis, and orotic acid is metabolized into UMP in the pyrimidine metabolic pathway. Indeed, both orotic acid and UMP could restore HAQ production in the carB, pyrB, and pyrD mutant strains. Since the lack of HAQ production in mutants affected in the pyrimidine metabolic pathway cannot be explained by the lack of orotic acid as a direct HAQ precursor, we analyzed the growth characteristics of the carB, pyrB, and pyrD mutant strains compared to those of the wild type. Despite comparable growth behavior in log-phase cultures, not only HAQ production but also growth in stationary-phase cultures was affected in all three mutants. The optical densities approached only about 30% of the optical densities of the wild-type cultures, unless the medium was supplemented with either orotic acid or UMP. These results imply that the absolute bacterial cell density in stationary-phase P. aeruginosa cultures and HAQ production are directly linked and that the mutants affected in the pyrimidine metabolic pathway lack HAQ production due to a general metabolic defect.

Common biosynthetic pathway for the fatty acid moiety of rhamnolipids and P. aeruginosa HAQs.

It is obvious that the majority of the HAQs of P. aeruginosa are either 4-hydroxy-2-heptylquinolines or 4-hydroxy-2-nonylquinolines. Accordingly, either β-ketodecanoic or β-ketododecanoic fatty acids are essential precursors of HAQ biosynthesis. In this context it is intriguing that rhamnolipids of P. aeruginosa are composed of rhamnose and β-hydroxy fatty acids that are the same chain length (namely, β-hydroxydecanoic and β-hydroxydodecanoic fatty acids). Recently, evidence has been provided that the rhlG gene of P. aeruginosa codes for a FabG homolog specifically involved in the synthesis of the β-hydroxy acid moiety required for rhamnolipid biosynthesis (3). While a mutation of the rhlG gene had no apparent effect on the growth rate and total lipid content of P. aeruginosa cells, the production of rhamnolipid was abrogated and the synthesis of poly-β-hydroxyalkanoate was delayed. This suggests that the biosynthetic pathway for the fatty acid moiety of rhamnolipids is separate from the general fatty acid pathway starting with a specific keto-acyl reductase step catalyzed by RhlG. In addition, the rhlG mutant strain was shown to produce significantly less pyocyanin, whose production is strongly dependent on PQS, a defect that could be complemented in the mutant strain after the parental gene was provided on a plasmid in trans (3). Since it was tempting to speculate that RhlG is required to provide a pool of β-hydroxy C₁₀ to C₁₂ fatty acids in P. aeruginosa that are essential precursors of not only rhamnolipid biosynthesis but also of the biosynthesis of HAQs, we tested whether the PAO1 rhlG:Tc mutant (provided by G. Soberon-Chavez) was deficient in HAQ biosynthesis. Indeed, total HAQ production was reduced in the PAO1 rhlG:Tc mutant strain after 4 h, 9 h, and 30 h of growth, as determined by TLC (data not shown). These findings suggest that RhlG plays a role not only in the biosynthesis of fatty acids used as substrates for rhamnolipid and poly-β-hydroxyalkanoate but also in HAQ production, albeit for the latter process an intact rhlG gene is not an absolute requirement. Moreover, assuming a common pathway for the fatty acid moiety of rhamnolipids and HAQs, one would expect that the addition of acetate to the growth medium would lead not only to a shift from 4-hydroxy-2-heptylquinoline to 4-hydroxy-2-nonylquinoline (as demonstrated in this study) but also to a shift in the C side chain length of P. aeruginosa rhamnolipids from C₁₀ to C₁₂. Indeed, acetate supplementation led to a shift in the biosynthesis of rhamnolipids with the longer C₁₂ side chain, as demonstrated by GC-MS. Whereas before addition of acetate the short-chain variants of the rhamnolipids were synthesized predominately, the availability of large amounts of acetate caused increased synthesis of 4-hydroxy-2-nonylquinoline and rhamnolipids with the longer C₁₂ side chain (Fig. 4). These results further strengthen the presumption that the fatty acid moiety of rhamnolipids and HAQs are derived at least in part from the same pool.

Conclusions.

In this study we elucidated the biosynthetic pathway of HAQs that are signal molecules involved in the regulation of virulence factor production in P. aeruginosa. Using labeled precursors and analysis of extracted HAQs by GC-MS and NMR spectroscopy, we confirmed the previously proposed biosynthetic pathway involving a “head-to-head” condensation of anthranilic acid and β-keto fatty acids. Moreover, the results of this study imply that the P. aeruginosa rhlG gene coding for an NADPH-dependent β-ketoacyl reductase provides a pool of C₁₀ and C₁₂ β-hydroxy fatty acids which not only are essential for rhamnolipid biosynthesis but also serve as substrates for HAQ production.

HAQ signal molecules are produced in the stationary phase of growth, and they seem to be important when the bacterial population is under stress. Under these conditions (for instance, during a chronic infection of a CF lung) maintenance of viability is of major importance for the bacterial population rather than cell division. The elucidation of the biosynthetic pathway and future research on the molecular mechanisms of HAQ signal transduction should provide valuable clues regarding the process of chronic infection and the pathogenesis of persistent disease. More insight could make the HAQ signal molecules a unique drug target for new therapies for treating P. aeruginosa infections.