PqsD Is Responsible for the Synthesis of 2,4-Dihydroxyquinoline, an Extracellular Metabolite Produced by Pseudomonas aeruginosa*

2,4-Dihydroxyquinoline (DHQ) is an abundant extracellular metabolite of the opportunistic pathogen Pseudomonas aeruginosa that is secreted into growth medium in stationary phase to concentrations comparable with those of the Pseudomonas quinolone signal. Using a combination of biochemical and genetic approaches, we show that PqsD, a condensing enzyme in the pqs operon that is essential for Pseudomonas quinolone signal synthesis, accounts for DHQ formation in vivo. First, the anthraniloyl moiety is transferred to the active-site Cys of PqsD to form an anthraniloyl-PqsD intermediate, which then condenses with either malonyl-CoA or malonyl-acyl carrier protein to produce 3-(2-aminophenyl)-3-oxopropanoyl-CoA. This short-lived intermediate undergoes an intramolecular rearrangement to form DHQ. DHQ was produced by Escherichia coli coexpressing PqsA and PqsD, illustrating that these two proteins are the only factors necessary for DHQ synthesis. Thus, PqsD is responsible for the production of DHQ in P. aeruginosa.

positive organisms (8). Anthranilic acid is the precursor to both PQS and HHQ and is produced by the phnAB operon (9). All of the gene products in the pqsABCD gene cluster are thought to be essential for the formation of PQS and HHQ (9,10). The prediction that PqsA is a CoA ligase is borne out by recent in vitro experiments illustrating that PqsA is an anthraniloyl-CoA synthetase (11). The details of HHQ biosynthesis are unknown, but the anthraniloyl-CoA intermediate is predicted to react with a ␤-keto fatty acid via a "head-to-head" condensation in a series of steps requiring PqsB, PqsC, and PqsD to form HHQ (12), which is converted to PQS by the hydroxylation of HHQ catalyzed by PqsH (3). DHQ is a recently identified secondary metabolite of P. aeruginosa. Although only pqsA in the pqs-ABCD gene cluster is reported to be required for DHQ biosynthesis (13), the addition of two carbons to the anthraniloyl moiety in the DHQ structure indicates that a condensing enzyme-like reaction is probably involved. Despite that both DHQ and HHQ contain an anthraniloyl moiety and require PqsA for the synthesis, there is no evidence of DHQ being the precursor of HHQ and PQS.
The condensing enzymes play a central role in bacterial fatty acid synthesis by elongating the growing acyl chains by two carbons (14). These enzymes catalyze the Claisen condensation reaction to form a carbon-carbon bond via a ping-pong kinetic mechanism involving an acyl-enzyme intermediate (15). In the first step, the acyl-enzyme is formed by the transfer of an acyl chain from either acyl-CoA or acyl-acyl carrier protein (ACP) to the conserved active-site cysteine residue to form a thioester linkage. In the second step, malonyl-ACP is decarboxylated, forming a carbanion, which attacks the acyl-enzyme intermediate to form the ␤-ketoacyl-ACP product (14,16). In addition to their essential functions in bacterial fatty acid synthesis, condensing enzymes are also important components for the formation of secondary metabolites such as polyketides (17). Although PqsB, PqsC, and PqsD are designated as condensing enzymes in the sequencing data bases, only PqsD (or FabH1) possesses the Cys-His-Asn triad of active-site residues present in bone fide ␤-ketoacyl-ACP synthase III (FabH) condensing enzymes. PqsC lacks the Asn, and PqsB lacks all three of the active-site resides, although it retains some overall similarity to the N-terminal domains in the FabH condensing enzyme superfamily. In this study, we used a combination of biochemical and genetic analyses to demonstrate that the condensing enzyme activity of PqsD is responsible for the synthesis of DHQ in P. aeruginosa.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Reagents-PAO1 (pqsA, pqsB, pqsC, and pqsD) strains were cultured at 37°C in LB or tryptic soy broth (BD Biosciences) unless specified otherwise. The reagents used in this study were purchased from the following suppliers: Sigma, anthranilic acid and [ring-UL-14 C]anthranilic acid (specific activity of 61.1 mCi/mmol), malonyl-CoA, ATP, and DHQ; Analtech, Silica Gel H plates; Promega, restriction enzymes; and BD Biosciences, Pseudomonas isolation agar. Malonyl-ACP was prepared using AcpS (ACP synthase) with apo-ACP and malonyl-CoA as described (18). Anthraniloyl-CoA was prepared by PqsA with anthranilic acid and CoA. Briefly, the reaction contained 82 M [ring-UL-14 C]anthranilic acid (specific activity of 61.1 mCi/mmol), 200 M CoA, 2 mM ATP, 10 mM MgCl 2 , and 12.5 g/ml purified PqsA protein in 100 mM Tris-HCl (pH 7.5). The reaction was incubated at 37°C for 30 min before being subjected to centrifugation in an Amicon Ultra concentration filter (10 kDa) to get rid of the PqsA protein. The complete conversion of anthranilic acid to anthraniloyl-CoA in the filtrate was confirmed by TLC.
Plasmid Construction and Mutant Generation-The pqsB, pqsC, and pqsD mutants generated by transposon insertions were obtained from the P. aeruginosa mutant collection at the University of Washington Genome Center. The DNA sequence located between the loxP sites in the transposon mutant was excised by Cre-mediated recombination by conjugation of SM10/pCre1 donors with the transposon mutant recipients, leaving a 189-bp insertion at the location of the original transposon. The pqsA mutant was created by gene replacement technology as described previously (19,20). Briefly, a DNA fragment containing the upstream sequence of pqsA and ϳ200 bp of pqsA 5Ј-end, a gentamicin cassette, and a DNA fragment containing downstream sequence of pqsA were cloned into pEX18ApGW (19). The resultant plasmid was conjugated from Escherichia coli SM10pir into strain PAO1 with selection of gentamicin and carbenicillin on Pseudomonas isolation agar plates (BD Biosciences). Merodiploids formed via a single crossover event were resolved through 5% sucrose selection in the presence of gentamicin. The gentamicin resistance cassette was subsequently removed by Flp recombinase, resulting in a 1.3-kb deletion in pqsA. The presence of the deletion was verified by PCR and sequencing.
Protein Purification-The gene fragment encoding PqsA or PqsD was subcloned into expression plasmid pET-28a (Novagen) between the NdeI and HindIII restriction sites. The resultant plasmids were transformed into the Rosetta(DE3) strain (Novagen) to express the proteins with an N-terminal His tag. Transformed cells were grown at 37°C in LB broth in the presence of 30 g/ml kanamycin and 34 g/ml chloramphenicol to an A 600 of 0.8, heat-shocked at 42°C for 15 min, and induced with 1 mM isopropyl ␤-D-thiogalactopyranoside at room tem-perature for 3 h. The cells were harvested by centrifugation; resuspended in 20 mM Tris-HCl (pH 8), 500 mM NaCl, and 10% glycerol containing 1 mM phenylmethylsulfonyl fluoride; and lysed with a French press. His-tagged protein in the cell-free extract was purified by nickel affinity chromatography as described previously (21). The fractions containing the purified protein were pooled, concentrated, and stored in 50% glycerol at Ϫ20°C.
PqsA Activity Assay-The conversion of anthranilic acid to anthraniloyl-CoA by PqsA was measured by TLC. Briefly, each 40-l reaction contained 82 M [ring-UL-14 C]anthranilic acid (specific activity of 61.1 mCi/mmol), 200 M CoA, 2 mM ATP, 10 mM MgCl 2 , and 250 ng of purified PqsA protein in 100 mM Tris-HCl (pH 7.5). The reaction was initiated by the addition of PqsA. After a 10-min incubation at 37°C, 20 l of the reaction mixture was spotted on an activated Silica Gel H plate, which was developed in butanol/acetic acid/water (5:2:4, volume ratio). The dried TLC plate was exposed to a storage Phosphor-Imager screen, and the autoradiograph was scanned using a Typhoon 9200. The radioactivity was quantitated using Image-Quant Version 5.1 software.
Activity Assay of PqsD-The activity of PqsD was analyzed by TLC using malonyl-CoA and [ 14 C]anthraniloyl-CoA produced in situ by PqsA or prepared in advance. The PqsD reaction contained 82 M [ 14 C]anthraniloyl-CoA, 125 M malonyl-CoA, and the indicated concentration of PqsD in 100 mM Tris-HCl (pH 7.5). After incubation at 37°C for 30 min, 20 l of the 40-l reaction mixture was spotted on an activated Silica Gel H plate. The TLC plate was developed in chloroform/methanol/acetic acid (95:5:1, volume ratio), and the product was quantitated and analyzed as described above.
Liquid Chromatography and Mass Spectrometry-The PqsA or PqsD reaction products with non-radioactive substrates were analyzed on a Waters Alliance HT liquid chromatography-mass spectrometry (LC-MS) system (Waters 2795 separation module linked to a Waters 2996 photodiode array detector), which was controlled by MassLynx Version 4.1 software. A 30-l aliquot was injected onto a Waters XBridge C18 column (3.5 m, 4.6 ϫ 50 mm) and eluted using a methanol gradient at a flow rate of 1 ml/min: 0% methanol for 1 min, linear gradient from 0 to 60% methanol over 1.5 min, linear gradient from 60 to 95% methanol over 4.5 min, and linear gradient from 95 to 0% methanol over 2 min. The samples from the separation module were injected onto a Waters Micromass ZQ mass spectrometer and analyzed using positive ion mode electrospray mass spectrometry with a cone voltage of 60 V, a source temperature of 130°C, and a desolvation temperature of 350°C.
Extraction of P. aeruginosa Medium Supernatant-Strains of P. aeruginosa were grown in 50 ml of tryptic soy broth to stationary phase (A 600 ϭ 4.0). The cells were removed by centrifugation and filtration. The resultant medium supernatant was extracted twice with an equal volume of acidified ethyl acetate in a separating funnel. The two extraction mixtures were combined, dried under nitrogen, and resuspended in 0.1 ml of ethyl acetate for LC-MS analysis.
[ 14 C]Anthranilic Acid Metabolic Labeling-Strains were grown in LB broth supplemented with 0.005% tryptophan and 33 M [ring-UL- 14 C]anthranilic acid (specific activity of 61.1 SCHEME 1 mCi/mmol) and incubated at 37°C for 16 h. The cells were removed by centrifugation, and the supernatant was collected. To detect DHQ levels,10 l of supernatant was loaded onto a Silica Gel H plate (pre-activated at 90°C for 1 h) and developed with chloroform/methanol/acetic acid (95:5:1, volume ratio). The dried TLC plate was exposed to a PhosphorImager screen, which was scanned with a Typhoon 9200.
Viability of MLE-12 Cells-MLE-12 cells were maintained in 2% fetal bovine serum supplemented HITES medium modified with transferrin (10 g/ml), HEPES (10 mM), and L-glutamine (2 mM in addition to that in the base medium). Monolayers of MLE-12 cells were treated with 0.25% trypsin/EDTA solution for 15 min and subcultured in 6-well plates at 5% density. After overnight incubation, fresh medium containing 150 M DHQ was added to DHQ-treated wells, whereas medium with an equal volume of solvent dimethyl sulfoxide (DMSO) was added to control wells. Each treatment was performed in triplicate. Total cells and viable cells excluding trypan blue stain were counted at 0, 24, and 48 h after DHQ or DMSO was added. To determine the effects of different concentrations of DHQ, the MLE-12 cells were treated with 10, 25, 75, and 150 M DHQ or DMSO for 48 h.

Functional Predictions in the pqs Operon of P. aeruginosa-
The addition of an acetate moiety to anthranilate is required to form the basic structure of DHQ. Such two-carbon additions are typically carried out by condensing enzymes that decarboxylate malonyl-CoA (or malonyl-ACP) in the process of forming the new carbon-carbon bond. The first protein encoded by the pqs operon PqsA is an anthraniloyl-CoA synthetase (11), suggesting that anthraniloyl-CoA is the primer for the condensing enzyme step. Condensation reactions that use acyl-CoA primers are carried out by the ␤-ketoacyl-ACP synthase III or FabH family of proteins (Pfam08545). Therefore, the predicted amino acid sequences of the pqs genes were compared with the E. coli initiation condensing enzyme (FabH) to identify potential gene products involved in the postulated condensation reaction in DHQ biosynthesis in P. aeruginosa. PqsB, PqsC, and PqsD all have some degree of overall sequence similarity to E. coli FabH. Multiple sequence alignment of these three proteins with E. coli FabH showed that they are most highly related in the three areas surrounding the location of the Cys-His-Asn active-site triad in E. coli FabH (Fig. 1). PqsD possesses all three predicted active-site residues and is the most highly related protein to E. coli FabH, leading to its annotation as FabH1 in the P. aeruginosa genome data base. These similarities strongly suggest that PqsD carries out a condensation reaction. PqsC is the second most related protein. The Cys and His residues are present; however, PqsC lacks the conserved Asn. All three active-site residues are missing in PqsB. Previous work suggests that all three active-site residues contribute to catalysis (22), indicating that PqsB and PqsC are not functional condensing enzymes, although their overall relatedness to E. coli FabH suggests that they may possess a similar overall fold. These functional predictions indicate that the most likely candidate condensing enzyme for a role in DHQ formation in the pqs operon is PqsD.
DHQ Production in the pqs Operon Mutants-To identify the functions of Pqs enzymes in DHQ production in vivo, pqsA, pqsB, pqsC, and pqsD mutant strains were tested for their abilities to accumulate DHQ in the growth medium by both LC-MS analysis and [ 14 C]anthranilic acid labeling. Each of these mutant alleles knocked out the function of the transcribed protein by the insertion or deletion of DNA sequences and did not introduce elements that could potentially create a polar effect on the downstream genes in the operon. The efficiency of DHQ extraction from the growth medium by acidified ethyl acetate was evaluated using a DHQ standard. LC-MS result showed that, at 0.15 mg/ml, Ͼ90% of DHQ was extracted into ethyl acetate. Thus, ethyl acetate was used to extract the extracellular  quinolines from the spent media of the wild-type and mutant strains. The wild-type PAO1 strain secreted DHQ into the medium in addition to HHQ, PQS, and 2-n-heptyl-4-hydroxyquinoline N-oxide species with different acyl chain modifications (Fig. 2, A and B). LC-MS analyses of the growth medium extracts showed that no detectable levels of DHQ or other quinolines were observed in the pqsA mutant strain, demonstrating a requirement of PqsA for the formation of DHQ and the quinoline signals ( Fig. 2A). Of the three FabH homologs, only the PqsD function was required for DHQ production because both the pqsB and pqsC mutant strains produced DHQ, whereas the pqsD mutant strain did not secrete detectable DHQ (Fig. 2B). Consistent with these results using the Pseudomonas mutant strains, expression of pqsA plus pqsD in E. coli, an organism that does not produce DHQ, led to robust DHQ formation from extracellular [ 14 C]anthranilic acid (Fig. 2C). These data show that PqsA and PqsD are both necessary and sufficient for DHQ biosynthesis. PqsA forms the anthraniloyl-CoA primer, which is converted to DHQ by PqsD.
DHQ Formation by PqsD in Vitro-On the basis of the structure of DHQ, we hypothesized that DHQ is formed by the condensation of anthraniloyl-CoA with an activated form of a three-carbon acid, such as malonyl-CoA. Analyses of the pqs mutant strains demonstrated that PqsD was involved in DHQ formation in vivo; therefore, we purified recombinant PqsD protein to test this conclusion in vitro. Anthraniloyl-CoA, which was produced in situ by PqsA, was condensed with malonyl-CoA by PqsD to form a new product as shown by TLC (Fig. 3A). The identity of the product formed in the coupled reaction was confirmed by LC-MS to be DHQ, which exhibited the same elution time (3.67 min) and mass spectrum as the DHQ standard (Fig. 3, C and D). The conversion of anthraniloyl-CoA to DHQ was nearly complete in these experiments using 1 g of each of the enzymes. As anticipated, anthraniloyl-CoA was the product in the presence of PqsA alone (Fig. 3B). Anthraniloyl-CoA was eluted at 3.43 min; the mass spectrum of the anthraniloyl-CoA product from the PqsA reaction revealed fragment peaks characteristic of both the anthraniloyl group and CoA (Fig. 3B). Malonyl-ACP was also a substrate for DHQ formation by PqsD, whereas malonic acid did not support DHQ formation (Fig. 3A).
We determined the kinetic parameters of PqsD for different substrates. The apparent K m values of PqsD for anthraniloyl-CoA, malonyl-CoA, and malonyl-ACP were 35 Ϯ 4, 104 Ϯ 37, and 18 Ϯ 2 M, respectively (Fig. 4, A-C). PqsD exhibited a specific activity of 0.22 Ϯ 0.03 pmol/min/ng for DHQ formation when assayed in the linear range. Consistent with the in vivo results, neither purified PqsB nor PqsC catalyzed DHQ synthesis in vitro.
Formation of Anthraniloyl-enzyme Intermediate-A common property of condensing enzymes is the presence of a conserved active-site Cys that forms a thioester bond with an acyl primer. Our results showed that anthraniloyl-CoA was used as  1). The increase in mass of 119 Da in peptide AQCSGLLYGLQ-MAR was consistent with the molecular mass of an attached anthraniloyl group (Fig. 5B). None of the other five Cys residues of PqsD was modified.
DHQ Inhibits MLE-12 Cell Viability-The lungs of cystic fibrosis patients are colonized by P. aeruginosa biofilms, where extracellular metabolites such as DHQ are accumulated. We tested the effect of DHQ on mouse lung epithelial MLE-12 cells in the presence of 150 M DHQ, the concentration found in the growth medium of P. aeruginosa. The addition of DHQ to the cell culture medium significantly inhibited the growth of MLE-12 cells, particularly after 48 h of treatment (Fig. 6A). In addition to the reduced total cell numbers, the fraction of viable cells treated with DHQ was reduced to 30% compared with 85% of the DMSO-treated control cells. The inhibition of MLE-12 cells by DHQ was evident at 10 M (Fig. 6B). The growth inhibitory effect of DHQ on the mouse lung epithelial cells suggests  , mass spectrum of a peptide from the anthraniloyl-PqsD sample that was eluted from an SDS gel and digested with trypsin protease. The deduced sequence of the peptide was AQCSGLLYGLQMAR; the cysteine in this peptide was predicted to be the activesite cysteine based on multiple sequence alignment (see Fig. 1). The mass difference between y 11 and y 12 is 222.1 Da, consistent with a cysteine residue (103.1 Da) covalently modified by an anthraniloyl moiety (119 Da).
that DHQ contributes to the pathogenicity of P. aeruginosa infection. Further experiments are needed to determine how DHQ reduces cell viability and what the outcomes are from long-term exposure to DHQ.
Conclusion-Our experiments led to the model for DHQ biosynthesis outlined in Fig. 7. The first step is the formation of anthraniloyl-CoA by PqsA. This intermediate acts as a primer to load the active-site cysteine of PqsD, forming an acyl-enzyme intermediate. Next, either malonyl-CoA or malonyl-ACP binds to the acylated enzyme, which catalyzes the condensation reaction with the release of bicarbonate and the formation of 3-(2aminophenyl)-3-oxopropanoyl-CoA (or ACP). PqsD exhibited a slightly higher affinity for malonyl-ACP than for malonyl-CoA (Fig. 4, B and C) in vitro. However, without the information on the endogenous concentrations of these two substrates in P. aeruginosa, we cannot rule out one or the other as the real substrate for PqsD in vivo. In E. coli, malonyl thioesters compose typically 5% of both the CoA and ACP pools, but the total CoA concentration is 10-fold higher compared with ACP (23)(24)(25)(26). These considerations suggest that malonyl-CoA provides the bulk of the two-carbon units to DHQ. The above-described intermediate then undergoes intramolecular ring closure to form DHQ. We postulate that this cyclization reaction may occur spontaneously, as is observed in the formation of triacetic acid lactone by the elongation condensing enzymes of fatty acid synthesis (18) and in the examples of spontaneous ring closure in polyketide synthases (27). However, the idea that the cyclization is promoted by the PqsD condensing enzyme cannot be  The model illustrated in Fig. 7 is consistent with both the in vitro biochemical analyses of PqsD and the requirement for only PqsA and PqsD to produce DHQ both in P. aeruginosa and in the heterologous E. coli system. Our conclusion that PqsD is the condensing enzyme in DHQ synthesis is different from that reached by Lépine et al. (13). On the basis of the production of DHQ by a pqsB insertion mutant that they surmised would have a polar effect that would eliminate the expression of pqsC and pqsD, these investigators concluded that none of the pqs genes other than pqsA was involved in DHQ formation. However, they did not describe their mutant in sufficient detail to allow a comparison with our work, and their assumption that the pqsB gene disruption used in their study also disrupted transcription/translation of the downstream pqsCD genes was not experimentally verified. Although our data definitively establish PqsD as a condensing enzyme, its role in PQS biosynthesis remains ambiguous. It may catalyze the postulated head-tohead condensation of anthraniloyl-CoA with a ␤-keto acid, but it is equally likely that the condensing enzyme is responsible for the formation of the ␤-keto acid precursor.