Properties and Substrate Specificity of RppA, a Chalcone Synthase-related Polyketide Synthase in Streptomyces griseus *

RppA, a chalcone synthase-related polyketide synthase (type III polyketide synthase) in the bacteriumStreptomyces griseus, catalyzes the formation of 1,3,6,8-tetrahydroxynaphthalene (THN) from five molecules of malonyl-CoA. The K m value for malonyl-CoA and thek cat value for THN synthesis were determined to be 0.93 ± 0.1 μm and 0.77 ± 0.04 min−1, respectively. RppA accepted aliphatic acyl-CoAs with the carbon lengths from C4 to C8as starter substrates and catalyzed sequential condensation of malonyl-CoA to yield α-pyrones and phloroglucinols. In addition, RppA yielded a hexaketide, 4-hydroxy-6-(2′,4′,6′-trioxotridecyl)-2-pyrone, from octanoyl-CoA and five molecules of malonyl-CoA, suggesting that the size of the active site cavity of RppA is larger than any other chalcone synthase-related enzymes found so far in plants and bacteria. RppA was also found to synthesize a C-methylated pyrone, 3,6-dimethyl-4-hydroxy-2-pyrone, by using acetoacetyl-CoA as the starter and methylmalonyl-CoA as an extender. Thus, the broad substrate specificity of RppA yields a wide variety of products.

Bacteria synthesize a number of polyketides, and their huge structural diversity reflects the variety of pharmacological and veterinary properties (1). The bacterial polyketide synthases (PKS) 1 are divided into three categories. The first category of PKS is type I PKS, a giant assembly of multifunctional polypeptides, which is mechanistically related to type I fatty acid synthase, typically found in yeast and mammals (2). Type II PKS is a large multi-enzyme complex of discrete enzymes with different functions (3), and its pivotal component responsible for the condensing activity resembles ␤-ketoacyl synthase II of type II fatty acid synthase found in bacteria and plants (4). The third type of PKS is a small homodimeric protein that possesses overall sequence homology to the chalcone synthase (CHS) family (5), which is related in structure and mechanism to ␤-ketoacyl synthase III of type II fatty acid synthase (6,7). RppA, which was found in the Gram-positive and soil-inhabiting bacterium Streptomyces griseus, is the first bacterial PKS identified as a member of the CHS superfamily and categorized in type III PKS (8). Phloroacetophenone, a precursor in the 2,4-diacetylphloroglucinol biosynthesis in Pseudomonas fluorescens, is synthesized by PhlD that shares 49% identity in amino acid sequence with RppA (9). DpgA, which is encoded in the balhimycin (a vancomycin derivative) biosynthetic gene cluster in Amycolatopsis mediterranei, possesses 22% similarity to RppA and catalyzes the 3,5-dihydroxyphenylacetic acid synthesis solely from malonyl-CoA (10). Together with the finding that RppA is concerned with melanin production in S. griseus (8), we suppose that RppA-type enzymes are involved in the biosynthesis of a wide variety of secondary metabolites not only in filamentous Streptomyces but also in single-cell bacteria.
RppA catalyzes polyketide synthesis by selecting malonyl-CoA as the starter, carrying out four successive extensions of malonyl-CoA, and cyclizing the resulting pentaketide to THN [1] (Fig. 1). 2 The final ring closure to THN [1] is accompanied presumably by decarboxylation of the carboxyl group of the malonyl-CoA used as the starter (8). CHSs found exclusively in plants catalyze sequential condensation of three malonyl-CoAs to p-coumaroyl-CoA to form an enzyme-bound tetraketide intermediate that is subsequently folded into chalcone. A wide variety of chemical structures of the CHS family-catalyzed products results from the differences in starter unit, extension unit, number of condensation, and position specificity of ring closure (11)(12)(13)(14)(15)(16)(17)(18)(19). In plants, CHS is a key enzyme to give the central intermediate, naringenin chalcone, for anthocyanins, used for pigmentation (20) and protection against UV irradiation (21). In S. griseus, however, RppA yields THN [1] as an intermediate for melanin biosynthesis (8). The most striking difference between RppA and CHS is that the former condenses malonyl-CoA four times to give a pentaketide and the latter condenses malonyl-CoA three times to give a tetraketide. Recent advances in three-dimensional structure studies of the CHS family have revealed that the size of the cavity in the active site determines starter molecule selectivity and the upper limit of the chain length in polyketide products (22,23). Although the crystal structure of RppA has not been solved, the difference in the volume of the active site cavity might reflect the product profile of RppA and CHS.
We thus expected that site-directed mutants of RppA might be useful in the synthesis and conversion of various compounds once the properties of RppA catalysis are understood. For this purpose, it is essential to determine its kinetic parameters, * This work was supported by grants from the "Research for the Future" Program of the Japan Society for the Promotion of Science, the Bio Design Program of the Ministry of Agriculture, Forestry, and Fisheries of Japan (BDP-01-VI-2-3), and the Industrial Technology Research Grant Program in 2000 of the New Energy and Industrial Technology Development Organization of Japan (00A03004). 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.
§ Supported by the Japan Society for the Promotion of Science. ʈ To whom correspondence should be addressed. Tel.: 81-3-5841-5123; Fax: 81-3-5841-8021; E-mail: asuhori@mail.ecc.u-tokyo.ac.jp. 1 The abbreviations used are: PKS, polyketide synthase; CHS, chalcone synthase; THN, 1,3,6,8-tetrahydroxynaphthalene; HPLC, highperformance liquid chromatography; LC/APCIMS, liquid chromatography/atmospheric pressure chemical ionization mass spectrometry; RppA, enzyme that causes S. griseus to produce a red-brown pigment. specificity of starters and extension units, and the number of condensation with starters different from malonyl-CoA. We optimized the reaction conditions of THN synthesis from malonyl-CoA and measured the K m value for malonyl-CoA and the k cat values for THN synthesis. Under these conditions, the substrate specificity and chemical structure of resultant products were determined. For structural determination of the products, we chemically synthesized several compounds, expected as products, and used them as authentic samples.

Polyketide Synthase Assay
Histidine-tagged RppA was purified as described (8). The standard reaction mixture contained 200 M starter-CoA, 20 M [2-14 C]malonyl-CoA (88,000 dpm) or 200 M dl-2-[methyl-14 C]methylmalonyl-CoA (88,000 dpm), 100 mM Tris-HCl (pH 7.5), 1 mM EDTA, and 17 g of RppA in a total volume of 600 l. For the reaction of methylmalonyl-CoA, 130 g of RppA was used. After the reaction mixture had been preincubated at 30°C for 3 min, the reaction was initiated by adding the substrate(s) and was continued for a further 10 min. When methylmalonyl-CoA was used as the substrate, the reaction period was 2.5 h. Reactions were stopped by adding 120 l of 6 M HCl, and the products were extracted with 600 l of ethyl acetate. The organic layer was dried up by N 2 flush, and the residual material was dissolved in 15 l of CH 3 OH for HPLC analysis. Reverse-phase HPLC conditions were as follows: ODS-80Ts (C18) column (4.6 ϫ 150 mm; Tosoh), maintained at 40°C; mobile phase for reactions A, B, C, D, G, and H (see Fig. 4), linear from 5% CH 3 CN in H 2 O (each contained 2% acetic acid) to 40% CH 3 CN in H 2 O over 30 min and then 100% CH 3 CN within 10 min with detection at 280 nm; flow rate, 0.8 ml/min. The mobile phase for reactions E and F (see Fig. 4) were: linear from 30% CH 3 CN in H 2 O (each contained 2% acetic acid) to 60% CH 3 CN in H 2 O over 20 min and then 100% CH 3 CN within 5 min with detection at 280 nm; flow rate, 1.0 ml/min. UV spectra were detected on a Waters 996 photodiode array detector. The eluate was collected every minute and measured directly by liquid scintillation counting. Liquid chromatography-atmospheric pressure chemical ionization mass spectrometry (LC/APCIMS) was performed on an LCQ (ThermoQuest). Samples prepared for LC/MS were scaled up 10-fold, and nonradioactive malonyl-CoA was used.

Determination of Kinetic Parameters
The standard reaction in a total volume of 100 l contained 100 mM Tris-HCl (pH 7.5), 1 mM EDTA, 1 mM 2-mercaptoethanol, [2-14 C]malonyl-CoA (the concentration was varied between 0.33 and 4.2 M), and 0.57 g of RppA. After the reaction mixture had been preincubated at 30°C for 3 min, the reaction was initiated by adding the substrate and was continued for 10 or 15 s. The reaction was stopped with 20 l of 6 M HCl, and the material in the mixture was extracted with 200 l of ethyl acetate. The organic layer was combined with nonlabeled authentic THN [1] and dried up by N 2 flush for HPLC analysis. The THN [1] produced was separated by HPLC and its amount was quantified by means of [2-14 C]malonyl-CoA incorporation.

Chemical Synthesis of Phloroglucinols
To a stirring solution of 1,3,5-trimethoxybenzene (1.65 g, 9.98 mmol) in 10 ml of dry diethylether, octanoyl chloride (12.9 g; 79 mmol) or hexanoyl chloride (10.8 g; 80 mmol) was added. The mixture was stirred for 10 min at room temperature, and 2 ml of concentrated H 2 SO 4 was added dropwise. The reaction was allowed to be stirred at room temperature for 18 h before being diluted with ice. The products were extracted with ethyl acetate, washed with a saturated solution of NaHCO 3 and brine, dried with Na 2 SO 4 , and concentrated under reduced pressure. The crude products were flash chromatographed (25-50% ethyl acetate in hexane as an eluant) to provide phloroglucinol trimethyl esters as orange oils. Phlorocaprylophenone trimethyl ether (2.5 g, 84% yield): 1  To a stirring solution of phloroglucinol trimethyl ethers (1.9 mmol) in 20 ml of dry CH 2 Cl 2 at Ϫ78°C, BBr 3 (2.65 g, 10.4 mmol) was added dropwise. The mixture was allowed to warm to room temperature for 5 h and was then recooled to 0°C. The reaction was quenched with water and extracted with ethyl acetate. The organic layer was washed with brine, dried with Na 2 SO 4 , and concentrated in vacuo. The crude products were flash chromatographed (30 -50% ethyl acetate in hexane as an eluant) to provide phloroglucinols as orange solids. Phlorocaprylophenone [5a] (286 mg, 60% yield): 1

RESULTS
Determination of Kinetic Parameters of RppA-In previous work (8) we constructed plasmid pET-RppA that directed the synthesis of histidine-tagged RppA with a structure of Met-Gly-His 10 -Ser 2 -Gly-His-Ile-Glu-Gly-Arg-His-RppA in the soluble fraction of Escherichia coli. We purified the His-tagged RppA with a nickel-nitrilotriacetic acid column to near homogeneity and used for the following assays. Incubation of RppA with [2-14 C]malonyl-CoA gave THN [1] as a product. However, the product THN [1] readily polymerized, giving structurally unknown compounds, and was auto-oxidized to flaviolin because of the chemical instability of THN [1] (8). This instability hampered quantitative analysis of the RppA catalysis. To determine the kinetic parameters of RppA, we improved the conditions of reaction by shortening the reaction period to avoid the oxidation and polymerization of THN [1]. When [2-14 C]malonyl-CoA was used as a substrate and the product was subjected to HPLC analysis, almost all of the radioactive substances were comigrated with authentic THN [1] (Fig. 2A), indicating that under the improved conditions THN [1] was the only radioactive product from [ 14 (Fig. 3). The k cat value for THN formation was calculated to be 0.77 Ϯ 0.04 min Ϫ1 . A slight substrate inhibition, which was about a 10% decrease in velocity against the maxi- . Symbols connected by broken lines represent radioactivity measurements, and those connected by solid lines represent the UV absorbance detected at 280 nm. See "Experimental Procedures" for the assay and HPLC conditions. The inset shows the ratio of the amounts of products as determined by radioactivity measurements. mum rate, was observed at concentrations higher than 200 M (data not shown).
In the fungus Colletotrichum lagenarium, PKSI PKS, the type I PKS, is responsible for melanin production and synthesizes THN [1] solely from malonyl-CoA (25). In this case, the aldol reaction for the first ring closure might occur between the C-2 and C-7 positions of the pentaketide intermediate. This pathway was inferred from the chemical structure of the derailment product, ␣-acetylorsellinic acid, that was detected in the reaction in vitro. In the reaction mixture of RppA, however, no ion peak corresponding to the molecular mass of ␣-acetylorsellinic acid or its isomer was observed by LC/MS, and thus the ring-folding pattern of the pentaketide intermediate in the RppA reaction remains unclear. Another ion chromatogram for the pseudo-molecular ion [M Ϫ H] Ϫ :125 revealed that triacetic acid lactone was also synthesized by RppA (retention time 8.6 min on HPLC) in a negligible amount, although it was hardly detected by the autoradiogram experiment ( Fig. 2A). We also examined the possibility of incorporation of [1-14 C]acetyl-CoA into triacetic acid lactone or ␣-acetylorsellinic acid and its isomer formed as a result of different ring folding. However, the presence of [ 14 C]acetyl-CoA in the assay did not give any radioactive products (Fig. 2B).
Temperature and pH Dependence of RppA-We examined temperature and pH dependence of RppA by using [2-14 C]malonyl-CoA at a final concentration of 100 M. The optimal pH was about 7.5 in 100 mM Tris-HCl buffer (data not shown). Raising or lowering the pH by 1.0 caused an ϳ25% loss of activity. The highest enzyme activity was obtained at about 30°C at pH 7.5 in 50 -200 mM Tris-HCl buffer.
Starter Substrate Specificity of RppA-The starter specificity of RppA was examined by monitoring the incorporation of [ 14 C]malonyl-CoA into the condensation products. We used various acyl-CoA compounds with 4 -10 carbon lengths. At a low concentration (3.6 M) of the acyl-CoA compounds, only THN [1] was formed from malonyl-CoA, and none of the unnatural starter CoAs tested here were used as a starter substrate by RppA, which suggested a clear preference for malonyl-CoA to other acyl-CoAs as the starter bound to the active site pocket. Efficient incorporation of the unnatural starter CoAs was observed when the concentrations of the starter CoAs were increased to 200 M, which is 10-fold higher than that of malonyl-CoA as the extender substrate.
When butyryl-CoA [2e] was incubated with RppA along with [ 14 C]malonyl-CoA, two radioactive products, 3e and 4e, were obtained in addition to THN [1] (Fig. 4B). Comparison of the UV absorption and radiochromatograms in Fig. 4, A and B, showed that these two products represented the products formed from butyryl-CoA. The molecular weights (M r ) obtained by LC/APCIMS analysis of 3e and 4e were 154 (corresponding to the mass of a triketide) and 196 (a tetraketide), respectively. Similar product patterns were observed when isobutyryl-CoA [2d] and isovaleryl-CoA [2c] were incorporated as starter substrates (Fig. 4, C and D). The M r of 3d and 4d containing isobutyryl-CoA as the starter were 154 (a triketide) and 196 (a tetraketide), respectively. The M r of 3c and 4c containing isovaleryl-CoA as the starter were 168 (a triketide) and 210 (a tetraketide), respectively. The ratio of the triketide [3e] and the tetraketide [4e] was 24:14 in the butyryl-CoA-primed products, when the radioactivity levels were quantified, whereas that of the triketide [3c] and the tetraketide [4c] was 5:16 in the isovaleryl-CoA-primed products.
When hexanoyl-CoA [2b] was incorporated, an additional tetraketide (M r , 224; 5b), which migrated as a less polar compound than 4b, was observed in addition to a triketide (M r , 182; 3b) and a tetraketide (M r , 224; 4b) (Fig. 4E). Surprisingly, octanoyl-CoA [2a] as a starter yielded a small amount of a hexaketide [6a], a product that resulted from five cycles of condensation (Fig. 4F), which was confirmed by its ion peaks [M ϩ H] ϩ at 337 and [M Ϫ H] Ϫ at 335 by LC/APCIMS analysis. In addition to 6a, the octanoyl-CoA-primed reaction yielded a triketide (M r , 210; 3a) and two tetraketides (M r , 252; 4a and 5a). RppA also accepted acetoacetyl-CoA [2f], leading to formation of triacetic acid lactone [3f] (Fig. 4G). Thus, RppA exhibited broad starter substrate specificity toward aliphatic acyl-CoA compounds as summarized in Table I. On the other hand, under the assay conditions used in this study, no activity was detected with acetyl-CoA, decanoyl-CoA, benzoyl-CoA, phenylacetyl-CoA, crotonoyl-CoA, or tiglyl-CoA. Of the all acyl-CoAs examined, acetoacetyl-CoA gave the corresponding products in the greatest amount. The relative order of acyl-CoAs to be incorporated as a starter was determined to be acetoacetyl-CoA Ͼ hexanoyl-CoA Ͼ butyryl-CoA Ͼ isovaleryl-CoA Ͼ isobutyryl-CoA Ͼ octanoyl-CoA (Table I). A longer chain acyl-CoA, such as decanoyl-CoA, was not used as the starter, indicating that the upper limit of carbon chain length of the starter substrate is 8.
Incorporation of Methylmalonyl-CoA as an Extender Unit-When 200 M starter CoAs and 20 M [ 14 C]methylmalonyl-CoA were incubated with 21 g of RppA in a total of 200 l at 30°C for 2 h, a radioactive product was obtained only from the reaction with acetoacetyl-CoA [2f], although at a very low level (data not shown). Optimization of the reaction conditions, under which 200 M methylmalonyl-CoA, 200 M acetoacetyl-CoA, and 130 g RppA in a total of 600 l were incubated at 30°C for 2.5 h, led to production of 7 as a major peak (Fig. 4H). LC/APCIMS analysis of 7 gave a [M ϩ H] ϩ peak at 141 and a [M Ϫ H] Ϫ peak at 139, suggesting that 7 is a triketide expected from acetoacetyl-CoA as the starter and one methylmalonyl-CoA molecule as the extender.

DISCUSSION
The broad substrate specificity of RppA, belonging to type III PKS, implies that the supply of the starter substrate is a significant factor in determining the substrate in vivo. However, considering the remarkably low value of the apparent K m (0.93 Ϯ 0.1 M) for malonyl-CoA, together with the facts that RppA prefers malonyl-CoA and does not utilize other acyl-CoAs as a primer at a low substrate concentration, we suppose that malonyl-CoA is the natural substrate of RppA in the cell. This is consistent with the observation that S. griseus overexpressing RppA accumulates flaviolin, an auto-oxidized product of THN [1], as a single peak on HPLC. 3 The present study has demonstrated that RppA can accept C 4 to C 8 aliphatic acyl-CoAs as a starter substrate. RppA contained all of the key amino acid residues that are responsible for reaction priming and chain elongation (Cys-138, His-270, and Asn-303) (26 -28). Of these three residues, Cys-138 was shown to be essential for polyketide synthesis (8). It is widely accepted that the initial reaction step involves the loading of a starter acyl-CoA onto the thiol group of an active site cysteine, giving rise to an enzyme-bound acyl molecule via a thioester bond. This acyl transfer reaction is clearly a decisive step in determining the starter specificity of CHS-related enzymes. RppA is unusual among CHS-related enzymes in that it preferably uses malonyl-CoA as a starter unit. The difference of the residues surrounding the active site presumably accounts for this unique feature of RppA. Interestingly, RppA uses branched chain acyl-CoAs (isobutyryl-CoA [2d] and isovarelyl-CoA [2c]) but not aromatic acyl-CoAs (benzoyl-CoA and phenylacetyl-CoA) (Table I), although plant CHSs, in contrast, use both branched and aromatic acyl-CoAs (29 -34). The electronic and/or steric hindrance caused by a bulkier residue present in the substrate binding pocket in RppA could explain the differences in substrate specificity between RppA and CHS. Consistent with this idea, RppA accepted butyryl-CoA [2e] but not crotonoyl-CoA. Site-directed mutagenesis experiments to clarify the starter specificity are in progress.
RppA synthesizes several ␣-pyrones when some acyl-CoAs are used as starters. The mechanism of the final ring closure, 3  yielding pyrone, is the second feature that distinguishes RppA from CHS. CHSs produce bisnoryangonin and p-coumaroyltriacetic acid lactone as derailment products, as a result of hydrolysis of the polyketide chain from active site cysteine (or CoA) and subsequent non-enzymatic pyrone formation (35). However, the reaction mechanism by which RppA yields pyrone is supposed to be quite different from that of CHS. For generation of a carboxylic acid of the polyketide chain through the hydrolysis of the thioester bond from Cys-138 (or CoA) before the pyrone formation, an acidic condition to accomplish the dehydration of hemiacetal is necessary (Fig. 7A). All of the pyrones produced by RppA were detected by HPLC analysis without acidifying the reaction mixture (data not shown), al-though pyrone is poorly extracted with ethyl acetate under the assay condition used (pH 7.5). Furthermore, we could not detect any hemiacetal or carboxylic acid compounds by either HPLC/autoradiogram (Fig. 4) or LC/MS ion scanning analysis. We thus conclude that the release of the final product from Cys-138 (or CoA) occurs by intramolecular lactonization, as in the case of animal fatty acid synthase (36,37), or by Claisen condensation of the polyketide chain. Nucleophilic attacking of the carbonyl carbon of the thioester by the oxyanion at the C-5 position leads to the formation of ␣-pyrones, whereas that by the carbanion at the C-6 position yields phloroglucinols (Fig.  7B). This could proceed when the polyketide is attached to CoA, because there is no evidence of retransfer of the polyketide onto the active site cysteine after the final chain extension (22). It is possible that lactonization occurs after the release of the CoAattached polyketide from the enzyme. Type I PKSs possess thioesterase domains, which are responsible for release of the polyketide chain from the enzyme surface (38 -40). In CHSrelated enzymes, however, no such residues capable of hydration or intermolecular Claisen condensation of the polyketide chain have been identified (22,23), and therefore the mechanism of product release remains mysterious. Our results also show that the structure of the starter unit of the nascent polyketide chain contribute much toward the folding of the final product, because the reaction from isovarelyl-CoA [2d] yielded pyrone [4d] with no apparent formation of phloroglucinol, whereas the octanoyl-CoA [2a]-primed reaction produced phlorocaprylophenone [5a] (Fig. 6).
Interestingly, hexaketide 6a was produced from octanoyl-CoA [2a] and five molecules of malonyl-CoA. This is the first observation that CHS-related enzymes possess the ability to catalyze condensation of more than five units of malonyl-CoA. The size of the active site cavity physically limits the number of malonyl-CoA condensations, as implied from the x-ray crystal structures of 2-pyrone synthase and CHS (22,23). The carbon length (C 18 ) of 6a suggests that the cavity volume of RppA is quite large, although malonyl-CoA (natural substrate) cannot utilize the maximum potential of RppA. The stability of the active site pocket could be altered depending on the structure of the starter substrate incorporated into the growing chain, and therefore the variety in the chain length of products would arise. In the case of acetoacetyl-CoA [2f]-primed reaction, a short starter moiety, such as a methyl group, may be insufficient to be retained in the pocket and released before further condensation of malonyl-CoA.
In conclusion, RppA exhibits high activity toward 4-to 8-carbon straight and branched chain acyl-CoAs but little activity toward aromatic or unsaturated acyl-CoAs. The broad substrate specificity of RppA leads to production of tri-to hexaketide pyrones and tetraketide phloroglucinols, presenting a diverse product profile. The availability of the knowledge of substrate specificity of RppA is promising for the production of single-ring pyrones, phloroglucinols, or their derivatives by overexpressing mutant RppA enzymes in Streptomyces and E. coli cells supplemented with various starter compounds.