CloR, a Bifunctional Non-heme Iron Oxygenase Involved in Clorobiocin Biosynthesis*

The aminocoumarin antibiotics novobiocin and clorobiocin contain a 3-dimethylallyl-4-hydroxybenzoate (3DMA-4HB) moiety. The biosynthesis of this moiety has now been identified by biochemical and molecular bio-logical studies. CloQ from the clorobiocin biosynthetic gene cluster in Streptomyces roseochromogenes DS 12976 has recently been identified as a 4-hydroxyphenylpyru-vate-3-dimethylallyltransferase. In the present study, the enzyme CloR was overexpressed in Escherichia coli , purified, and identified as a bifunctional non-heme iron oxygenase, which converts 3-dimethylallyl-4-hydroxy-phenylpyruvate (3DMA-4HPP) via 3-dimethylallyl-4-hydroxymandelic acid (3DMA-4HMA) to 3DMA-4HB by two consecutive oxidative decarboxylation steps. In 18 O 2 labeling experiments we showed that two oxygen atoms are incorporated into the intermediate 3DMA-4HMA in the first reaction step, but only one further oxygen is incorporated into the final product 3DMA-4HB

The aminocoumarin antibiotics novobiocin, clorobiocin, and coumermycin A 1 (Fig. 1) are natural products from Streptomyces spheroides NCIMB 11891 (syn. S. caeruleus) (1), S. roseochromogenes DS 12.976, and S. rishiriensis DSM 40489, respectively (2). These antibiotics have been shown to target bacterial DNA gyrase, with binding constants (K d ) in the range of 10 nM (3,4). Novobiocin (Albamycin®, Pfizer) is licensed for the treatment of infections with Gram-positive bacteria (5) and has been shown to enhance the cytotoxic activities of the anti-tumor drugs etoposide and teniposide (6). So far, the therapeutic use of aminocoumarin antibiotics is limited due to their low solubility in water, toxicity in eukaryotes, and poor penetration in Gram-negative bacteria (7). Combinatorial biosynthesis may offer a chance to develop novel aminocoumarins with improved properties (8).
Novobiocin and clorobiocin contain a prenylated 4-hydroxybenzoate moiety (called Ring A) (Fig. 1A). Recently, Lafitte et al. (9) demonstrated that Ring A plays a role in the binding affinity of these antibiotics for gyrase. Knowledge of the biosynthetic pathway of Ring A may facilitate future efforts to create structural modifications of this moiety by combinatorial biosynthesis.
The dimethylallyl moiety of Ring A is derived from the methyl erythritol-4-phosphate (MEP) pathway (10). The 4-hydroxybenzoic acid (4HB) 1 moiety can be formed in nature by at least three well-established mechanisms: (a) the direct conversion of chorismic acid to 4HB by chorismate pyruvate-lyase (11); (b) the removal of acetyl-CoA from the side chain of 4-coumaroyl-CoA by an oxidative reaction mechanism, analogous to the ␤-oxidation of fatty acids, resulting in 4-hydroxybenzoyl-CoA (12,13); and (c) the removal of a C2 unit from the side chain of a ␤-hydroxy-␤phenylpropionate derivative via a retro-aldol reaction; the resulting benzaldehyde derivate can subsequently be oxidized to the corresponding acid (14,15).
In novobiocin biosynthesis, feeding experiments with isotope-labeled precursors showed that 4-hydroxyphenylpyruvate and tyrosine are efficiently incorporated into Ring A (10,16), ruling out the chorismate pyruvate-lyase reaction as the principal source of this compound. The aminocoumarin moiety of novobiocin (Ring B) is formed from tyrosine via the intermediate ␤-hydroxytyrosyl-S-NovH. Chen and Walsh (17) suggested that Ring A may also be formed from this intermediate, following the reaction mechanism shown in Fig. 2A, i.e. by a retroaldol reaction to 4-hydroxybenzaldehyde, oxidation to the acid and prenylation. 3-Prenylation of 4HB has a well-established precedent in ubiquinone biosynthesis (18).
Cloning and sequencing of the biosynthetic gene clusters of novobiocin, clorobiocin, and coumermycin A 1 (Fig. 1B) (19) first appeared to support this hypothesis. Three genes were identified in the novobiocin and clorobiocin clusters for which no homologues existed in the coumermycin cluster. We speculated that these genes might be involved in the biosynthesis of Ring A (which is absent in coumermycin A 1 ). These genes were: (a) cloR and novR, which showed sequence similarity to putative aldolases; CloR was indeed shown to be involved in Ring A biosynthesis by a gene inactivation experiment (19); (b) cloF and novF, which show sequence similarities to dehydrogenases; and (c) cloQ and novQ, which did not show sequence similarities to known genes in the data base. We recently * This work was supported by a grant from the Deutsche Forschungsgemeinschaft (to L. H. and S.-M. L.), by the Fonds der Chemischen Industrie, and by the BMBF-Modelprojekt Klinische Pharmakologie (Grant FKZ-01EC0001). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
In the present study, we investigated the conversion of 3DMA-4HPP to Ring A and the role of CloR in this reaction sequence. Unexpectedly, the results showed that the biosynthesis of Ring A follows neither of the two mechanisms depicted in Fig. 2. Rather, the conversion of 3DMA-4HPP to Ring A proceeds in two oxidative decarboxylation steps, via 3-dimethylallyl-4-hydroxymandelic acid (3DMA-4HMA) as intermediate, and is catalyzed by the bifunctional non-heme iron oxygenase CloR. This represents a new pathway for the formation of benzoic acids in nature.
DNA Manipulation-DNA manipulations and standard genetic techniques in Escherichia coli were carried out as described by Sambrook and Russell (23). DNA fragments were isolated from agarose gels using a NucleoSpin 2 in 1 extraction kit (Macherey-Nagel, Dü ren, Germany). Isolation of plasmids was carried out with ion-exchange columns (Nucleobond AX kit, Macherey-Nagel).
Overexpression and Purification of CloR Protein-E. coli BL21(DE3)/ cloR-pGEX4T1 cells were grown in 100 ml of LB medium containing 50 g/ml carbenicillin at 15°C until an A 600 of 0.6 was reached (ϳ20 h). IPTG was added to a final concentration of 250 M. After 24 h at 15°C, the cells were harvested by centrifugation (10 min at 5,000 ϫ g) and resuspended in phosphate-buffered saline (140 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , pH 7.3) containing 1 mg/ml lysozyme (76,344 units/mg, Fluka). Resuspended cells were broken by sonification (Branson sonifier, 10-min sonification). Cell debris was removed by centrifugation (30 min at 20,000 ϫ g). Purification of CloR as a GST fusion protein and subsequent cleavage of the GST tag by thrombin treatment were carried out according to the manufacturer's instructions (Amersham Biosciences) using glutathione-Sepharose 4B and the batch method.
Overexpression and Purification of CloQ-Construction of vector cloQ-pGEX4T1, overexpression in E. coli BL21(DE3), and purification of CloQ were carried out as described in Pojer et al. (20).
Isolation of 3-Dimethylallyl-4-hydroxymandelic Acid-The incubation mixture (1 ml) contained 500 l of CloQ reaction product (ϳ75 nmol of 3DMA-4HPP), 0.2 M Tris-HCl (pH 7.5), 80 mM potassium phosphate (pH 7.5), 25 mM ascorbic acid, and ϳ40 g of holo-CloR and was incubated for 1 h at 30°C. The reaction was stopped with 50 l of formic acid and extracted twice with 1 ml of ethyl acetate. After evaporation, the residue was dissolved in 150 l of ethanol. 3DMA-4HMA was isolated by HPLC as described above, and its identity was confirmed by mass spectrometry.
Reactions in the Presence of 18 O 2 -Two flasks, one containing enzymatically produced 3DMA-4HPP (see above) and one containing holo-CloR reaction mixture (see above) were degassed by application of a vacuum and flushed with argon for three times. The anaerobic holo-CloR solution was transferred to the substrate vial containing 3DMA-4HPP. The argon was removed by application of a vacuum, and finally 18 O 2 was allowed to enter into the flask. After incubation for 1 h at 30°C, the reaction was stopped with formic acid and analyzed by liquid chromatography (LC) coupled to an electrospray ionization mass spectrometer (TSQ Quantum, ThermoFinnigan, San Jose, CA) in negative ion mode. A linear gradient of acetonitrile (25-75%) in 0.1% aqueous formic acid was used. In a parallel 18 O 2 labeling experiment, 3DMA-4HMA (isolated as described above) was used as substrate instead of 3DMA-4HPP.

Sequence Analysis of CloR and NovR-
The gene cloR from the clorobiocin cluster and the corresponding gene novR from the novobiocin cluster (Fig. 1B) encode proteins of 277 and 270 amino acids, respectively, and show 95% identity between each other (GenBank TM accession numbers AAN65240 and AAF67511, respectively). A data base search revealed sequence similarity to class II aldolases, which are represented, e.g., by L-fuculose-1-phosphate aldolase (28) and L-rhamnose-1-phosphate aldolase (29), and to the structurally related L-ribulose-5-phosphate-4-epimerase (30). These enzymes contain a catalytic zinc residue in their active center and are involved in the catabolism of sugars in E. coli.
cloR is transcriptionally coupled to the gene cloQ (accession number: AF329398), which encodes the aromatic prenyltransferase involved in Ring A formation (20). The same situation is found in the novobiocin cluster for the corresponding genes novQ and novR (accession number: AF170880).
Expression and Purification of CloR-We decided to express CloR as a glutathione S-transferase (GST) fusion protein rather than as an His-tagged protein, because CloR showed sequence similarity to class II aldolases, and these enzymes require zinc for there activity (31). A purification using the His-tagged protein can lead to a complete loss of the aldolase activity due to interactions between the hexahistidyl tag and the metal ion (Zn 2ϩ ). E. coli cells harboring CloR expression constructs yielded only insoluble protein when grown at temperatures of 20°C or higher. To obtain soluble CloR-GST fusion protein, cells were cultured at 15°C and induced with 250 M IPTG. After purification, GST was cleaved from CloR by thrombin treatment and removed. This procedure resulted in apparently homogenous CloR protein as judged by SDS-PAGE (Fig.  3). The molecular mass observed in SDS corresponded to the calculated mass of the protein (30.5 kDa). A protein yield of 1 mg of pure CloR per liter of culture was obtained. By using gel chromatography, the molecular mass of native CloR was determined as 124.5 kDa showing that the protein was tetrameric in solution.
Characterization of the Reaction Products of the CloR Reaction-To investigate the catalytic activity of CloR, we first produced 3DMA-4HPP (Fig. 2B), the putative substrate of CloR, by incubation of 4-hydroxyphenylpyruvate (4HPP) with DMAPP and the prenyltransferase CloQ. Subsequently, CloR and different cofactors, e.g. Zn 2ϩ and NADH (32), were added to the reaction mixture. However, no formation of 4-hydroxybenzaldehyde (3DMA-4HBAL) or Ring A could be detected by HPLC.
A more sensitive analysis using a radioactive assay with 4HPP and [1-14 C]DMAPP as substrates, however, revealed the presence of a small amount of a new radioactive compound (termed product X), with a retention time of 11.1 min in HPLC (Ring A, 16.7 min; 3DMA-4HBAL, 20.7 min; and 3DMA-4HPP, 18.4 min) (data not shown). This product was absent if heatdenaturated CloR was used, or if 4HPP or DMAPP were omitted from the prenylation assay. This indicated that the new metabolite X was derived enzymatically from 3DMA-4HPP.
In the biosynthesis of chloroeremomycin in Amycolatopsis orientalis, 4HPP is converted to 4-hydroxymandelic acid under catalysis of the non-heme iron dioxygenase HmaS (ϭ ORF21) (26,27). As described by Hubbard et al. (26), HmaS and similar enzymes need to be activated by preincubation with an excess of Fe 2ϩ immediately before incubation and Fe 3ϩ , generated by oxidation, has to be reduced to Fe 2ϩ by ascorbic acid to maintain an active enzyme. Although CloR did not show sequence similarity to HmaS or other non-heme iron-dependent enzymes, we decided to test CloR under similar conditions. After preincubation of CloR for 20 min with 1 mM FeSO 4 , ascorbic acid and enzymatically generated 3DMA-4HPP were added. After incubation for 1 h, the products of the reaction were analyzed by HPLC (Fig. 4). Under these conditions, formation of the new product X was ϳ15-fold higher than in the absence of Fe 2ϩ and ascorbate. Furthermore, an additional product was CloR was found to be specific for 3DMA-4HPP and 3DMA-4HMA as substrates. No product formation was observed with the non-prenylated substrates 4-hydroxyphenylpyruvate or DL-4-hydroxymandelic acid, nor with D-mandelic acid, L-mandelic acid, ␤-hydroxytyrosine, or 4-hydroxyphenyllactic acid. Additional experiments confirmed that the prenyltransferase CloQ (20) specifically prenylated 4HPP and was not able to react with DL-4-hydroxymandelic acid or 4-hydroxybenzoic acid (data not shown).
Purified CloR was a colorless protein. UV-visible spectrometry showed an absorption maximum at 283 nm but no absorption in the visual range. Therefore, CloR is not a heme protein, as are, e.g., the cytochrome P 450 monooxygenases.
Investigation of the Reaction Mechanism of CloR-The requirement of CloR for Fe 2ϩ and ascorbate suggested that it belongs to the non-heme iron oxygenases (33)(34)(35). To confirm whether indeed molecular oxygen was the substrate of the CloR reaction, and whether one or both oxygen atoms of O 2 were incorporated into the product, we carried out isotope-labeling experiments with 18

O 2 . Incorporation of the label was analyzed by LC-MS-MS analysis
In the first experiment, CloR (after preincubation with Fe 2ϩ ) was incubated with 3DMA-4HPP and ascorbate in an 18 O 2 atmosphere. A control incubation was carried out in the usual 16 (Fig. 5A), resulting from the incorporation of a single 18 O atom into the product. This shows a certain dilution of the label and has been reported previously for HmaS as well as for other non-heme iron oxygenases (27). This dilution has been suggested to result from an exchange of a presumed Fe IV ϭO intermediate with water (27). The product mixture also contained some unlabeled 3DMA-4HMA ([M-H] Ϫ ϭ 235), most likely due to the presence of residual 16 O 2 in the incubation vial.
In addition to 3DMA-4HMA, the CloR reaction also produced 3DMA-4HB. Incorporation of two 18  comparison to [M-H] Ϫ ϭ 205 for the unlabeled compound (Fig.  5A). As expected, both these labeled oxygens were located in the carboxyl group (Fig. 5A).
For a second labeling experiment, the intermediate 3DMA-4HMA was first produced in unlabeled form and isolated by HPLC. This compound was then incubated in an 18 O 2 atmosphere with CloR (preincubated with Fe 2ϩ ) and ascorbate. The resulting 3DMA-4HB was analyzed by LC-MS-MS (Fig. 5B), and this clearly showed the incorporation of one 18 O atom into the carboxyl group of the product, as demonstrated by the molecular ion at [M-H] Ϫ ϭ 207, and the decarboxylation product at m/z ϭ 161 (Fig. 5B). Fig. 5C summarizes the results of the 18 O 2 incorporation experiments. Two 18 O atoms are incorporated in the first reaction step, and one 18 O atom is incorporated in the second reaction step. The other 18 O atom involved in the second step is most likely converted to water (see "Discussion"). DISCUSSION In the present study, CloR was identified as a bifunctional oxygenase that converts 3-dimethylallyl-4-hydroxyphenylpyruvate (3DMA-4HPP) in two consecutive reaction steps to 3-dimethylallyl-4-hydroxybenzoate (3DMA-4HB), i.e. to the Ring A moiety of clorobiocin. An 18 O 2 labeling experiment unequivocally confirmed that molecular oxygen is used as substrate by CloR. The purified CloR protein does not contain a heme prosthetic group, and its activation by Fe 2ϩ and ascorbate indicated that it belongs to the non-heme iron oxygenase.
The first reaction catalyzed by CloR, i.e. the conversion of 3DMA-4HPP to 3-dimethylallyl-4-hydroxymandelic acid (3DMA-4HMA), has a well-established precedent in the HmaS reaction in chloroeremomycin biosynthesis. HmaS belongs to the iron(II)-and ␣-ketoacid-dependent dioxygenases (reviewed in Refs. 33,35,36). These enzymes utilize O 2 and an ␣-ketoacid as cosubstrates. During the reaction, the ␣-ketoacid loses CO 2 , and the keto function is oxidized to a carboxyl group by introduction of one of the oxygen atoms of O 2 . The other oxygen may be used for a hydroxylation reaction, exemplified, e.g., by the prolyl 3-hydroxylase reaction. However, iron(II)-and ␣-ketoaciddependent oxygenases have been shown to catalyze not only hydroxylations but a wide range of diverse oxidative transformations, including epoxidations, desaturations, ring formation, and ring expansion reactions. Some of these enzymes are bifunctional (e.g. deacetoxy-/deacetylcephalosporin synthase) or even trifunctional (e.g. clavaminic acid synthase or thymine hydroxylase), catalyzing several consecutive oxidative transformations within a single biosynthetic pathway.
In the conversion of 4-hydroxyphenylpyruvate (4HPP) to homogentisate by 4HPP dioxygenase, 4HPP serves as a substrate for hydroxylation and at the same time as ␣-ketoacid cosubstrate. CloR (in its first reaction step) and HmaS carry out a very similar reaction as 4HPP dioxygenase but hydroxylate the benzylic position of the substrate instead of the phenyl ring. However, although HmaS shows obvious sequence similarity to 4HPP dioxygenase, CloR does not.
The second reaction step catalyzed by CloR is the conversion of 3DMA-4HMA to 3DMA-4HB. It is tempting to speculate that this reaction may involve hydroxylation of the ␣-position of 3DMA-4HMA, resulting in a ␣,␣-gem-diol, which eliminates water to give the corresponding ␣-ketoacid, i.e. 4-hydroxybenzoylformate. Oxidative decarboxylation of this compound would result in the final product, 3DMA-4HB. Thereby, the overall reaction mechanism would be similar to that of the first reaction and resemble that of other iron(II)-and ␣-ketoacid-dependent dioxygenases.
However, LC-MS-MS analysis did not confirm the presence of 4-hydroxybenzoylformate in the incubation mixture, and the exact mechanism of the CloR reaction remains speculative at present. It is believed that the reactions catalyzed by iron(II)and ␣-ketoacid-dependent oxygenases involve a reactive Fe IV ϭO species (35,37). Whether this is true for CloR remains to be shown. The different groups of iron(II)-and ␣-ketoacid-dependent oxygenases possess little overall sequence similarity to each other (35,36), and it is therefore not surprising that CloR does not show sequence similarity to known members of this family. However, a data base search reveals that CloR does show significant similarity to several proteins of so far unknown function, deduced from genome sequences of different microorganisms. These may possibly represent enzymes of similar function as CloR.
The common structural motif of iron(II)-and ␣-ketoacid-dependent enzymes is the so-called 2-His-1-carboxylate facial triad (36,38). It consists of two histidyl groups and one glutaryl or asparagyl residue, which together anchor the iron atom in the active site of the enzyme. Comparison of the primary sequence of CloR with that of NovR as well as with the sequences of six of the data base entries of unknown function with high sequence similarity to CloR (data not shown) identifies His-161, His-176, His-178, His-241, and Asp-170 of CloR as strictly conserved residues. These amino acids may be candidates for a potential 2-His-1-carboxylate facial triad of this oxygenase.
The involvement of CloR in the biosynthesis of the prenylated 4-hydroxybenzoate moiety of clorobiocin was proven in vivo by a gene inactivation experiment (19). The present results now allow the formulation of a detailed hypothesis for the formation of this moiety of clorobiocin, as shown in Fig. 6: CloQ is a prenyltransferase, which converts 4HPP to 3DMA-4HPP (20). CloR converts the CloQ reaction product to Ring A, as demonstrated in the present study. CloF shows sequence similarity to prephenate dehydrogenases and is therefore likely to produce 4HPP as the substrate for CloQ, similar to ORF1 of the chloroeremomycin biosynthetic gene cluster (39). Notably, the coumermycin cluster does not contain a CloF homologue (Fig. 1B).
Radioactive feeding experiments with the novobiocin producer (40) showed that [U-14 C]L-tyrosine was incorporated preferentially into Ring B, whereas [U-14 C]4HPP was incorporated preferentially in Ring A. This may suggest that a primary metabolic prephenate dehydrogenase provides tyrosine for the formation of Ring B, whereas CloF supplies 4HPP for Ring A biosynthesis, and that cross-talk exists between both pathways probably via a transaminase reaction (Fig. 6).
Chen and Walsh (17) have previously speculated that Ring A and Ring B of novobiocin may both be produced via ␤-hydroxytyrosyl-S-NovH as a common precursor. Our studies on CloQ and CloR (Refs. 19 and 20 and the present study), however, now establish that these two aromatic moieties of clorobiocin (and very likely those of novobiocin as well) are produced by two independent pathways (Fig. 6). The discovery of CloR adds a new interesting member to the diverse family of the nonheme iron oxygenases and demonstrates the existence of a new pathway for the formation of benzoic acids in nature.