Characterization of a Novel Type of Oxidative Decarboxylase Involved in the Biosynthesis of the Styryl Moiety of Chondrochloren from an Acylated Tyrosine

Abstract

Myxobacteria are soil-dwelling bacteria notable for several unique behavioral features, such as cellular movement by gliding and the formation of multicellular fruiting bodies. More recently they have gained recognition as producers of several unique polyketide and nonribosomal polypeptide metabolites with potential therapeutic value. The biosynthesis of these compounds often involves highly unusual mechanisms including the formation of the chloro-hydroxy-styryl moiety of the chondrochloren antibiotic produced by Chondromyces crocatus Cm c5. Here it is shown that the final product of the chondrochloren megasynthetase is the novel natural product pre-chondrochloren, a carboxylated and saturated derivative of chondrochloren. This compound was isolated from strains harboring mutants of a hypothetical oxidative decarboxylase (CndG) identified in the chondrochloren gene cluster. CndG was heterologously expressed in Escherichia coli and shown to be an FAD-dependent oxidative decarboxylase. Biochemical characterization of the protein was achieved using the intermediate described above as the substrate and yielded chondrochloren by oxidative decarboxylation. It was also demonstrated that the CndG post-assembly line modification of pre-chondrochloren is essential for the biological activity of chondrochloren.

Introduction

Studies on several species of myxobacteria have resulted in the isolation and characterization of multiple interesting secondary metabolites that possess promising antibiotic and cytotoxic activities (1). The majority of known compounds produced by these bacteria are the products of biosynthetic mega enzymes termed polyketide synthases (PKSs)² and nonribosomal peptide synthetases (NRPSs). The assembly of monomeric organic and amino acid building blocks via mixed PKS-NRPS systems results in the formation of hybrid PKS-NRPS compounds exhibiting enomerous structural complexity. End products of PKS and/or NRPS megasynthases are usually modified by dedicated tailoring enzymes that attain internal conformation, eventually increase the degree of compound diversity, and often establish biological activity. These reactions may occur on a framework of growing acyl-S-enzyme intermediates and are usually mediated by enzymes that are integrated into the biosynthetic megasynthetases themselves (2, 3). In addition, the natural products may undergo post-assembly modifications, catalyzed in trans by dedicated enzymes to yield the final structures (4–7).

Enzymatic decarboxylations are widespread in nature. The reaction is known to occur through a variety of mechanisms that modify specific substrates (8, 9). Dependent on their catalytic cofactor, decarboxylases are categorized in two major classes. The first class utilizes organic cofactors such as biotin, flavin, and NAD⁺/NADP⁺ (10–12), and the second class of decarboxylases requires inorganic cofactors (13). Recently, a third enzyme class was identified that performs decarboxylation without a known cofactor (8, 14).

Molecular and biochemical studies of natural product biosynthesis have revealed the essential role of decarboxylases in generating structural diversity, in some cases during the maturation of secondary metabolites. One example of these enzymes is the decarboxylase that is encoded in the biosynthetic pathway of the potent lipopeptide antibiotic barbamide in the marine cyanobacterium Lyngbya majuscula. According to the proposed biosynthetic scheme, a carrier protein-bound 4-carboxy-thiazoline intermediate is presumed to be released by the thioesterase domain of BarG and then undergo a final decarboxylation to give the thiazole system of the end product. Interestingly, no oxidative domain is found in the respective module of the terminal NRPS protein, BarG. Therefore, it was postulated that BarJ, which is encoded downstream of the gene cluster, might play a role in the oxidative decarboxylation step to create the thiazole ring of the end product (15, 16).

Moreover, the biologically active long chain N-acyl aromatic amino acids, eneamides, and their enol esters have been isolated after heterologous expression of environmental DNA (feeA–feeM) in Escherichia coli (17). Formation of the amide is thought to be catalyzed by an N-acyl synthase that couples free tyrosine to an activated long chain fatty acid, delivered by an acyl carrier protein of the Fee assembly system. It has been shown that inactivation of a putative flavoprotein oxidase (FeeG) of the fee gene cluster mediates accumulation of an N-acyltyrosine intermediate. From this evidence, FeeG was proposed to catalyze styryl moiety formation by oxidative decarboxylation (18).

The chondrochloren biosynthetic pathway possesses several highly unusual and intriguing features, including biochemical transformations that are rarely found in other bacterial secondary metabolite biosyntheses. This includes an early chlorination of the tyrosine moiety and the formation of an ethoxy moiety in chondrochloren B, which is performed by a radical S-adenosylmethionine-dependent protein methylating a nonreactive carbon center in chondrochloren A. Another unusual feature in this compound class is the unusual hydroxyl-styryl functionality at the terminus of the structure. We hypothesized that the final step in the biosynthesis of chondrochloren, like in eneamide biosynthesis, would be an oxidative decarboxylation of a functionalized tyrosine, to yield both chondrochlorens A and B (Fig. 1).

FIGURE 1.

Oxidative decarboxylation steps proposed for the biosynthesis of barbamide catalyzed by BarJ (a), fatty acid enol ester of the fee gene cluster by FeeG (b), and chondrochloren (c). CndG is shown here to decarboxylate the pre-chondrochlorens A ((3)) and pre-chondrochloren B ((4)) to yield the mature chondrochloren A ((1)) and chondrochloren B ((2)), respectively.

Here, we report the inactivation of cndG in the chondrochloren gene cluster, which resulted in the accumulation of carboxylated chondrochlorens A and B (pre-chondrochlorens), biosynthetic intermediates that are shown to be the substrates of CndG. We cloned and biochemically characterized CndG and show this enzyme to be a novel catalyst for the oxidative decarboxylation of biologically inactive pre-chondrochloren. These results provide insights into the post-assembly modulation of pre-chondrochloren and its maturation into the final chondrochloren antibiotic.

EXPERIMENTAL PROCEDURES

General Molecular Biological Methods

Standard methods for DNA isolation and manipulation were used (19, 20). DNA fragments were isolated from agarose gels using the NucleoSpin Extract gel extraction kit (Machery-Nagel, Düren, Germany). PCRs were performed with Taq DNA polymerase (Fermentas) to generate DNA fragments for gene inactivation or Pfu polymerase (Stratagene) for generation of DNA cloned for heterologous expression. Conditions for amplification using an Eppendorf Mastercycler were as follows: denaturation, 30 s at 95 °C; annealing, 30 s at 48–60 °C; extension, 45 s at 72 °C; 30 cycles and a final extension for 10 min at 72 °C. PCR products were purified using the High Pure PCR product purification kit (Roche Applied Science). Ligations were performed using T4 ligase.

Inactivation of cndG

An internal fragment of the cndG gene containing a frameshift at the 5′-end was amplified using oligonucleotides Cnd-decar-frame-up (5′-GAT CTT CTG ACT TCC GCC TC-3′) (the mutagenic base pair is indicated in italics) and Cnd-decar-dn (5′-CAG CTC TCG GTC GTA CAT-3′). The PCR product was cloned into pCR2.1TOPO for sequencing creating pTOPO-CndG. After sequence verification, the insert was excised as an EcoRV/HindIII fragment and subcloned into vector pSUPHyg to generate plasmid pSR13. pSR13 was introduced into cells of E. coli ET12567 carrying pUB307 for biparental mating with Chondromyces crocatus Cm c5 by conjugation as described previously (21). The mutants were selected on Pol03 agar supplemented with 100 μg ml⁻¹ hygromycin and 120 μg ml⁻¹ tobramycin. Correct integration of the vector into genome and thus disruption of the gene were confirmed by PCR analysis using primers PSUP-EV and FAD-XhoI-dn (5′-CCT CGA GTC AGT TGT CCG CGG GCG-3′) or FAD-EcoRI-up (5′-GGA ATT CAT GAA CAC ACA GCC CCT GGA-3′) and FAD-XhoI-dn, using genomic DNA of three isogenic Cmc-cndG⁻ mutants in comparison with the C. crocatus Cmc5 wild type. The binding site of primer FAD-XhoI-dn is not located on plasmid pSR13. Cmc-cndG⁻ mutants were grown in Pol03 medium supplemented with hygromycin and 1% adsorber resin (XAD-16) at 30 °C for 7 days. Methanolic extracts of the cultures were prepared and subjected to analysis by HPLC-MS.

Purification of the Pre-chondrochloren B

Mutant Cmc-cndG⁻ was grown in Pol03 medium supplemented with 100 μg ml⁻¹ hygromycin in the presence of 1% XAD 16 resin. Methanolic extracts of the resin were applied to a Sephadex LH-20 column (GE Healthcare) with methanol as solvent. Fractions (7 ml) containing pre-chondrochlorens A and B were identified by HPLC-MS and combined. The pre-chondrochloren B was obtained at high purity by HPLC in two steps using a Zorbax C8 column. Step 1 was solvent: water (A)/acetonitrile (B) containing 0.1% formic acid; gradient 45% B for 7.5 min, 2.5 min to 60% B, 6 min to 95% B; flow rate: 6 ml min⁻¹. Step 2 was solvent: water (A)/acetonitrile (B), gradient: 20–50% B for 8.5 min, then to 95% B in 1.5 min; flow rate: 3.6 ml min⁻¹. After purification, the biological activity of the compound in comparison with chondrochloren B was analyzed by using an antibiogram agar diffusion assay employing 6-mm filter paper discs saturated with 30 μg of the compounds and subsequent growth inhibition assaying of Micrococcus luteus and Bacillus subtilis.

Expression and Purification of Recombinant CndG

cndG was amplified from chromosomal DNA of C. crocatus Cm c5 using primers FAD-EcoRI-up and FAD-XhoI-dn. The resulting product (1639 bp) was cloned into pCR2.1TOPO for sequencing and then subcloned into the expression vector pGEX-6P-1 (Invitrogen). The resulting construct, pGEX-CndG, was transformed into E. coli BL21 cells for protein expression. The cells were grown in LB medium containing 100 μg ml⁻¹ ampicillin at 37 °C until an A₆₀₀ level of 0.6 was reached, and then protein expression was induced with 0.1 mm isopropyl β-d-thiogalactopyranoside, and growth was continued at 16 °C overnight. The cells were harvested by centrifugation and disrupted by French Press at 4 °C in 1× phosphate-buffered saline. Cell debris was removed by centrifugation at 25,000 × g. The supernatant was bound to a MicroSpin^TM GST purification module, washed twice with 10 volume 1× phosphate-buffered saline buffer, and then the pure CndG was released by cleavage with PreScission Protease (Amersham Biosciences) at 4 °C in cleavage buffer, according to the manufacturer's instructions. The molecular mass of the protein obtained was confirmed by MALDI mass spectrometry.

Enzymatic Activity of the CndG

Activity assays of the CndG (0.2 μm) was carried out in buffer (50 mm Tris-HCl, pH 6.0, containing 10 μm pre-chondrochloren and 100 μm FAD). Control reaction was performed in parallel without the addition of the enzyme. The reactions were incubated at 30 °C, for various time points (5, 10, 30 min), followed by quenching with 1 μl of trifluoroacetic acid. Methanol (80 μl) was added to each aliquot and centrifuged, and then the samples were analyzed using a Bruker HCT plus mass spectrometer. For analysis of cofactor binding, CndG (50 μg; 0.83 nmol) was denaturated at 100 °C for 10 min to free the flavin cofactor, and then the mixture was centrifuged at 16,060 × g for 20 min at 4 °C. The supernatant was lyophilized, and the pellet was resuspended in 20% methanol and then subjected to HPLC-MS relative to a commercial sample of FAD. The effect of FMN as alternative cofactor was tested either by the addition of both FAD and FMN into the reaction or by the addition of FMN alone. To study the effect of the tag on the enzyme activity, the reactions were done using purified NusA-tagged and GST-tagged protein. To study the effect of metals, Zn²⁺, Mg²⁺, Mn²⁺, or Fe²⁺ at 0.1 mm was added into the enzyme reaction. The effect of metal chelators on CndG activity was determined by the addition of 2.5 mm EDTA or/and EGTA into to the reaction, and the control reactions were carried out in parallel under standard conditions. Identification of the pH optimum of the enzyme reaction was done using Tris-HCl buffer (50 mm, pH 4.0–9.0) under standard conditions. All of the reactions were performed in triplicate. The CndG activity was estimated according to the amount of the resulted product (chondrochloren B). The concentration of the chondrochloren B was quantified based on the analysis of the characteristic fragment ion (retention time (R_t) = 12.0 min, m/z [M + H]⁺ = 434.1), using a Zorbax C8 column (solvent: water (A)/acetonitrile (B) containing 0.1% formic acid; 0–0.5 min 45% B; 0.5–2 min linear from 45% to 55% B; 2–11 min isocratic at 55% B; 11–12 min linear from 55% to 95% B; 12–13 min isocratic at 95% B, and then 13–14 min linear from 95% to 45% B). A calibration curve based on the chondrochloren B reference standard was built and used to compare the concentration of the product in the samples. All of the reactions were performed in triplicate. For comparison of CndG activity under aerobic and anaerobic conditions, buffer (50 mm Tris-HCl) was purged of oxygen using oxygen-free nitrogen, overnight at 30 °C. Buffer (100 μl) containing pre-chondrochlorens B (10 μm) or a mixture of pre-chondrochloren A and B, 100 μm FAD, and 0.2 μm CndG was then added. The “anaerobic” reaction was covered with mineral oil (100 μl), incubated for 2 h at 30 °C, and then quenched by the addition of methanol (100 μl). A control reaction was carried out in parallel under standard conditions. Each assay was repeated in triplicate, and the results were analyzed by HPLC-MS. To test substrate specificity of the enzyme, in addition to l-tyrosine and 3-chloro-tyrosine, various N-acyl-l-tyrosine and meta-chloro-tyrosine derivates (N-acetyl-l-tyrosine, N-butyryl-l-tyrosine, N-cinnamoyl-l-tyrosine, N-benzoyl-l-tyrosine, and N-octanoyl-l-tyrosine) were chemically synthesized (supplemental “Methods”) and used as analogue substrate.

Kinetic Characterization of CndG

CndG (0.033 μm) was assayed in buffer (50 mm Tris-HCl, pH 6.0), containing 1 mm FAD and varying concentrations of pre-chondrochloren B (10.73, 16.1, 21.4, 26.8, 32.2, 64, and 96.6 μm). The reactions were incubated at 30 °C, and 40 μl of each reaction were removed at various time points (4, 6, 8, 10, 12, 14, and 20 min), followed by quenching with 1 μl of trifluoroacetic acid. Methanol (80 μl) was added to each aliquot, and then the samples were analyzed using the Bruker HCT plus mass spectrometer by tandem MS. The amount of product at each time point was quantified based on the analysis of the characteristic fragment ion (R_t = 12 min, m/z [M + H]⁺ = 434.1) as described previously.

RESULTS

CndG is encoded within the chondrochloren biosynthetic gene cluster and shows homology to FAD-dependent oxidases. Sequence analysis of the protein shows significant homology (30–45%) to several putative FAD-linked oxidases identified from bacterial genomes, including the putative oxidase enzyme from Cyanotheca sp. PCC7822 (accession number ZP_03154108), p-cresol methyl hydroxylase from Pseudomonas putida, and a putative oxidase (FeeG) from an environmental DNA (accession number AAM97300) (supplemental Fig. S1). Alignment with conserved FAD-binding motif of other flavoprotein sequences showed a putative FAD-binding region at the N-terminal sequence of the CndG (supplemental Fig. S1).

CndG Is Essential for Chondrochloren Biosynthesis

To evaluate our proposal for generation of the styryl moiety and involvement of CndG as a key enzyme for maturation of the metabolite, we aimed to inactivate the cndG gene in the Cm c5 strain. The experiment was achieved via homologous recombination of the cndG inactivation plasmid into the bacterial genome (Fig. 2a). Verification of the resulting mutants Cmc-cndG⁻ was achieved by PCR amplification of the inserted fragment. A 1016-bp PCR product was generated by the use of the oligonucleotides PSUP-EV and FAD-XhoI-dn and the genomic DNA of the mutants Cmc-cndG⁻, whereas no PCR fragment was generated from the wild type DNA under these conditions. In addition, a 1639-bp PCR product covering the whole cndG sequence was generated by the use of FAD-EcoRI-up and FAD-XhoI-dn primers from the genomic DNA of the wild type only, confirming disruption of the cndG (Fig. 2b). This result, in combination with hygromycin resistance of the mutants, clearly shows plasmid integration into the correct position of the bacterial genome. Consequently, a frameshift mutation was introduced into the first copy of cndG, and the second copy is promoterless (Fig. 2a). After fermentation of the mutants and the wild type in Pol03 medium with XAD resin, no difference in growth of the mutants was observed in comparison with the wild type. Methanolic extracts from cultures of the wild type and the Cmc-cndG⁻ mutants were analyzed by HPLC-MS, showing abolishment of chondrochloren production in the mutants (Fig. 2, c and d). By careful search for chondrochloren production in the mutant extracts, we could, however, identify two new products differing from chondrochlorens A or B by 46 mass units (Fig. 3). High resolution mass spectrometry data, as well as MS/MS fragmentation data of the new compounds, suggested the compounds to be two isomers of pre-chondrochloren A C₂₈H₄₂ClNO₉ [M + Na]⁺ = 594.2 (3 and 5 represent isomers): R_t = 17.7 and 25.2 min, respectively, and two isomers of chondrochloren B C₂₉H₄₄ClNO₉ [M + Na]⁺ = 608.2 (4 and 6 represent isomers): R_t = 22.0 and 27.3 min, respectively. Low amounts of the same isomers were also detected in the methanolic extract of the wild type strain (Fig. 3a).

FIGURE 2.

Inactivation of the cndG gene in Cm c5 by insertion of the pSR13 plasmid into the bacterial genome via single cross-over, creating Cmc-cndG⁻. a, schematic representation of the pSR13 plasmid, the genomic chondrochloren region in the wild type and in the Cmc-cndG⁻ mutant. b, verification of the resulting mutants, showing amplification of a 1016-bp PCR fragment (using the oligonucleotides PSUP-EV and FAD-XhoI-dn) in three isogenic mutants (lanes 3–5), in comparison with the wild type (lane 2). Lane 7 shows amplification of a 1639-bp PCR fragment of the intact cndG (using the oligonucleotides FAD-EcoRI-up and FAD-XhoI-dn) in the wild type in comparison with the mutants (lanes 8–10). Lanes 1 and 6, 1-kb DNA marker (Fermentas). c, HPLC chromatogram of chondrochlorens A (peak 1) and B (peak 2) production by Cm c5. d, HPLC analysis of the Cmc-cndG⁻ mutant, showing disappearance of chondrochloren peaks.

FIGURE 3.

a and b, HPLC-MS analysis of cell-free extracts of the Cm c5 wild type (a) and the Cmc-cndG⁻ mutant (b), showing peaks 3 and 5 (pre-chondrochloren A isomers) and peaks 4 and 6 (pre-chondrochloren B isomers) in Cm c5 and in the mutant. c, mass spectrum of the pre-chondrochloren A (peak 3) (R_t = 17.8 min, m/z [M + Na]⁺ = 594.3). d, mass spectrum of carboxylated chondrochloren B (peak 4) (R_t = 22.0 min, m/z [M + Na]⁺ = 608.3).

Structure Elucidation of Pre-chondrochloren B

To confirm our hypothesis concerning the chondrochloren biosynthetic intermediates, the most abundant pre-chondrochloren B compound 4 was purified from 10 liters of culture of the Cmc-cndG⁻ mutant, and its identity was confirmed by two-dimensional NMR (Fig. 4). The spectra of the purified compound were recorded on a Bruker Avance 500 instrument at 500 MHz. The spectra were assigned by comparison with the published data for chondrochloren B (22), which showed significant differences in the signals corresponding to 1′-H and 2′-H, as well as the appearance of a new carbon signal (δ_C = 170.8) (Table 1). ¹H,¹³C-heteronuclear multiple bond correlation analysis revealed that the new carbon is located adjacent to 1′-H (Fig. 4). High resolution mass spectrometry data showed (m/z): [M + Na]⁺ calculated for C₂₉H₄₄ClNO₉, 608.2602; found, 608.2588; Δ = −1.4 ppm, and UV-visible: λ_max 190, 233 nm. An antibiogram agar diffusion assay using 30 μg of purified isomer 4 showed that the compound lacks activity (no inhibition zone) against growth of M. luteus and B. subtilis in contrast to the purified chondrochloren B, which showed an inhibition zone of 17 and 14 mm, respectively.

FIGURE 4.

NMR analysis and structure elucidation of the carboxylated chondrochloren B and heteronuclear multiple bond correlations found for compound 4.

TABLE 1.

¹³C and ¹H NMR spectral data of the carboxylated chondrochloren B recorded at 500 and 125 MHz

d, doublet; dd, doublet of doublets; ddd, doublet of doublets of doublets; s, singlet; m, multiplet.

Position	13C	1H (J)	m
	ppm	ppm (Hz)
1	170.8
2	81.9	3.86 (3)	d
3	74	3.92 (3/7)	dd
4	76.8	4.34 (7/9)	dd
4OMe	56.7	3.28	s
5	137.2	6.38	d
6	142.2
6Me	12.3	1.81	m
7	206.5
8	42	1.23–1.28	m
8Me	14.7	3.52	m
9	76.5	3.78 (3/9)	dd
10	35	1.60–1.70	m
10Me	13	1.23–1.28	m
11	33.8	1.02	m
12	29.2	1.23–1.28	m
13	22.5	0.88	m
14	14.4	0.88	m
1′	54.4	4.76 (3/5/8)	ddd
2′A	36.9	3.20 (5/14)	dd
2′B	36.9	3.01 (8/14)	dd
3′	129.1
4′	129.5	7.19 (2)	d
5′	119.9
6′	150.3
7′	119.5	6.94 (8)	d
8′	128.9	7.02 (2)	dd
1″	67.8
2″	15.8	1.23–1.28	m
COOH	172.6

Heterologous Expression and Purification of CndG

To examine the catalytic function of the enzyme, cndG was cloned and overexpressed in E. coli. Because purification of the CndG protein was not successful by heterologous expression of the protein as NusA-His-tagged version (data not shown), we performed heterologous expression of the protein as GST-tagged protein. The expression resulted in a 86.5-kDa protein, which was then purified by affinity chromatography. PreScission Protease (Amersham Biosciences) did effectively remove the N-terminal GST tag (Fig. 5a). The molecular mass of the resulting protein was determined by MALDI mass spectrometry (60,123 kDa; calculated molecular mass; 60,157 kDa found).

FIGURE 5.

Purification of recombinant CndG. a, lane 1, protein marker. Lane 2, whole cell extract of induced E. coli BL21, containing pGEX-CndG. Lane 3, purification of GST-tagged CndG (asterisk) by glutathione affinity chromatography. Lane 4, purified CndG (asterisk), following treatment of the GST fusion protein with PreScission Protease. b, conversion of pre-chondrochlorens 3 and 4 to the mature chondrochlorens 1 and 2, respectively, catalyzed by in vitro action of CndG, monitored by HLPC analysis over a 30-min incubation period.

Characterization of CndG

The activity of CndG in vitro was investigated by incubating with pre-chondrochlorens A and B at 30 °C. Product formation was monitored by HPLC analysis over 30 min of incubation time (Fig. 5b). The conversion of the chondrochloren intermediates 3 and 4 to the final products 1 and 2 by the oxidative decarboxylase function of CndG was observed in comparison with the negative control reaction where products 1 and 2 were not obtained. HPLC-MS analysis of the supernatant from denatured CndG showed the presence of a substance with masses m/z [M + H]⁺ = 786.2, and m/z [M + Na]⁺ = 808.2 at a retention time that was consistent with that of authentic FAD.

The addition of external FAD (100 μm) into the reaction resulted in a 30% increase in CndG activity. Introduction of FMN into the CndG reaction as cofactor did not change the enzyme activity. However, the addition of equal amounts of FMN and FAD increased the enzyme activity only 24%. NusA-tagged and GST-tagged CndG proteins were also found to accomplish the oxidative decarboxylation. However, activity was reduced 3- and 2-fold, respectively. Maximum catalytic activity of the enzyme was found in 50 mm Tris-HCl buffer, pH 6.0. The addition of EDTA or/and EGTA chelators for trapping ions, as well as the effect of the external metals like Zn²⁺, Mg²⁺, Mn²⁺, and Fe²⁺, resulted in no significant change in the observed enzyme activity.

We also directly compared decarboxylation by CndG under aerobic and anaerobic conditions. Under low oxygen conditions, production of chondrochloren B was essentially undetectable, whereas decarboxylation to chondrochloren A was reduced by ∼95% (Fig. 6, a and b). To test the substrate specificity of CndG various N-acyl derivatives of l-tyrosine 7a–e and 3-chloro-l-tyrosine 9a–e were synthesized (supplemental “Methods”). The selective N-acylation of l-tyrosine was easily achieved through refluxing with stoichiometric amounts of the corresponding acid chloride (supplemental Scheme S1). The synthesis of the N-acyl-3-chloro-l-Tyrosine derivatives 9a–e commenced with the selective monochlorination of l-tyrosine with sulfuryl chloride, delivering the hydrochloride salt 8 in good yield (23). Subsequent N-acylation of 8 was accomplished by two different procedures (supplemental Scheme S2). In the first procedure an excess of N-ethyldiisopropylamine and acid chloride and hydrolysis of possible O-acylation under basic conditions afterward delivered 9a and 9c in moderate yields. In the second procedure an excess of hexamethyldisilazane for neutralization of the hydrochloride 8 and in situ protection of the phenol and acid functionality and silyl deprotection under acidic conditions afterward afforded 9b, 9c, and 9e in excellent yield.

FIGURE 6.

a and b, oxidative decarboxylation by CndG under aerobic (a) and anaerobic (b) conditions. Under anaerobic conditions, production of compound 2 from its carboxylated counterpart was essentially undetectable, whereas the yield of compound 1 was reduced by 95% (compounds 3 and 4 represent pre-chondrochloren peaks). c, characterization of the decarboxylase CndG. Determination of CndG kinetics was performed using purified compound 4 as substrate. Each point represents the average of three measurements, and the error bars indicate the standard deviation. The data were fit to the Michaelis-Menten equation by nonlinear regression.

The substrate specificity of the enzyme was probed using 7a–e and 9a–e analogues. HPLC-MS analysis of the reactions showed that CndG failed to catalyze conversion of the analogue compounds in comparison with the native substrate. Furthermore, no oxidative decarboxylation of either free l-tyrosine or 3-chlro-tyrosine was seen when they were added into the reaction as analogue substrates.

Kinetic studies of the enzyme under optimal catalytic conditions with various substrate concentrations were conducted by monitoring the amount of the characteristic fragment ion of the product at each time point as described formerly. The kinetic parameters (k_cat = 15.5 ± 0.45 min⁻¹; K_m = 21.5 ± 1.65 μm) (Fig. 6c) were calculated from Lineweaver-Burk plots using a substrate concentration range of 5.36 μm up to 96.6 μm.

DISCUSSION

The biosynthetic pathways of secondary metabolites from myxobacteria are of significant interest, because they frequently challenge established paradigms for chain assembly on modular multienzymes, as well as revealing novel and intriguing functional group transformations (24–29). Chondrochloren biosynthesis proceeds in a progressive fashion via the multimodular PKS/NRPS megasynthetase enzymes CndA-CndH. Several modification enzymes are also involved and include the halogenase CndH, which catalyzes the chlorination of tyrosine, and a radical methyltransferase that furnishes the ethoxy moiety at C-2 of the compound (24). Our previous adenylation domain specificity experiments as well as in silico analysis suggested the addition of tyrosine to the penultimate chondrochloren intermediate by the A domain of the terminal NRPS module of the gene cluster (24). However, the formation of the functionalized styryl group of the mature compound remained elusive. We suggested a post-assembly step in the chondrochloren biosynthesis for modification of the tyrosyl group via oxidative decarboxylation of the enzyme free intermediate (Fig. 1c).

The cndG gene found at the downstream margin of the biosynthetic gene cluster shows significant homology to known flavoprotein oxidases. Indeed, CndG shares sequence homology (31% identity) with FeeG. This protein is encoded in the biosynthetic cluster of a fatty acid enol ester and is predicted to be involved in the decarboxylation of a long chain N-acyltyrosine to an eneamide intermediate (17, 18). Interestingly the CndG protein shows no significant sequence homology to any of the decarboxylases involved in ribosomaly biosynthesized polypeptides such as mersacidin or epidermin (30–32). Nor does it display homology to enzymes implicated in the oxidative decarboxylation of barbamide (33).

To investigate the function of CndG in the biosynthetic process, a cndG mutant in the native Cm c5 strain was generated. Because cndG inactivation resulted in no bacterial growth limitation in comparison with the wild type, we can exclude the possibility of pleiotropic effects resulting from integration of the inactivation plasmid. HPLC analysis of the cndG mutant in comparison with the wild type showed a complete loss of chondrochlorens (Fig. 2, c and d), and confirmed the requirement CndG for biosynthesis of the compound. Because CndG inactivation hampered the formation of chondrochlorens, new compounds (compounds 3–6) with a +46 mass shift (Fig. 3, a–d) appeared in the mutant extracts. Structure elucidation of purified compound 4 by two-dimensional NMR indicated the presence of a new carboxyl group, suggesting that compound 4 is a carboxylated biosynthetic intermediate of chondrochloren B.

This finding shows that the decarboxylation and the oxidation reactions are carried out on the pre-chondrochloren as final steps in their biosynthesis (Fig. 1c). However, it was unclear whether both catalytic steps are performed via CndG. To address this question, CndG was purified and characterized in vitro. This experiment showed that the enzyme successfully decarboxylated and oxidized pre-chondrochloren, thus producing the styryl moiety of the mature chondrochlorens (Fig. 5).

Sequence analysis of CndG revealed several conserved residues at the N-terminal region of the protein that are characteristic of FAD-binding sites (supplemental Fig. S1). This theory was supported by the identification of FAD in the supernatant of denatured CndG. Our data show decarboxylation of the chondrochloren intermediate 4 directly after introduction of the CndG enzyme to the assay, even without the addition of external FAD. In addition, we observed an increase of CndG catalytic activity upon the addition of FAD to the reaction mixture. This increase in enzyme activity could by explained by a partial loss of FAD during purification of the enzyme. Similar results were described for the decarboxylase of the ribosomaly synthesized antibiotic, mersacidin; the authors suggested loose binding of the cofactor to the enzyme cavity (10). Moreover, a slight inhibition in the CndG reaction resulted from the addition of FMN into the assay. This may indicate a weak binding competition between FAD and the nonspecific FMN cofactor within the cofactor cavity site of the enzyme. No inhibition of CndG activity was observed in the presence of ion chelators, and the addition of external metal ions into the reaction also had no effect. This suggests that metal ions are not required for catalytic activity of the enzyme. Based on these results, CndG can be categorized within the “organic cofactor” class of decarboxylases.

Our data show that the CndG reaction requires molecular oxygen for oxidative decarboxylation of the pre-chondrochloren. The weak decarboxylase activity observed for chondrochloren A under the same conditions might result from imperfect “anaerobic” conditions (Fig. 6a). Although the measured k_cat for CndG is relatively low (15.5 min⁻¹), the rates for the biosynthesis by PKS (34, 35) and NRPS (36–38) in vitro (k_cat values typically in the 0.1–10 min⁻¹ range are nearly of the same magnitude, making it unlikely that this step would be rate-limiting for the complete biosynthesis. In addition, we assume that the enzyme has been adapted to the slow growth of the bacteria, because generation times tend to be longer than 7 h.

Our data also reveal that the enzyme is unable to decarboxylate a number of chlorinated and nonchlorinated analogue substrates containing variable N-acyl chains. Similarly, no decarboxylation of free tyrosine or 3-chloro-tyrosine was catalyzed by CndG. These observations suggest that the acyl chain is crucial for substrate binding within the catalytic pocket and indicates a limitation for diverse substrate recognition. Lower catalytic activity of the NusA-tagged or GST-tagged-protein demonstrates that the specific tags might influence the structural conformation of the enzyme, which may negatively affect the degree of either substrate or cofactor binding. Future work toward the elucidation of the enzyme crystal structure should shed light on the molecular basis of substrate specificity and binding center of the enzyme. These data should then allow a more detailed understanding of the CndG catalytic mechanism, which is not only important for the biosynthesis of chondrochloren but also might shed light onto the unknown chemistry behind the formation of similar moieties in other natural products. Furthermore, an antimicrobial assay of the compound provides evidence that post-assembly modification by CndG displays a crucial role in the observed biological function of the compound, because pre-chondrochloren in contrast to chondrochloren caused no inhibition of growth of M. luteus or B. subtilis.

The analysis of the chondrochloren biosynthetic pathway suggested the involvement of a set of novel tailoring transformations. Our data provide evidence for the specific function of CndG, and demonstrate that the enzyme performs the successive decarboxylation and oxidation of the functionalized pre-chondrochloren to yield the chloro-hydroxyl-styryl moiety of the compound. Therefore, CndG operates post-PKS/NRPS assembly of chondrochloren and attains the biological activity to the compound.

To our knowledge, this is the first example of a biochemical study on a decarboxylase that is associated with structural variation of a PKS/NRPS-derived natural product. The availability of such an enzyme may enable new possibilities for combinatorial biosynthesis and enzymatic production of bioactive compounds. In addition, our data describe a new type of oxidative decarboxylation generating a styryl moiety in a natural product starting from an acylated tyrosine derivative. The finding that various genes encoding similar proteins are found in the data bases suggests that the many as yet uncharacterized members of this family may perform similar functions.