Triple Hydroxylation of Tetracenomycin A2 to Tetracenomycin C in Streptomyces glaucescens OVEREXPRESSION OF THE tcmG GENE IN STREPTOMYCES LIVZDANS AND CHARACTERIZATION OF THE TETRACENOMYCIN A2 OXYGENASE*

Nucleotide sequence analysis of the tcmG gene has suggested that the TcmG protein is responsible for the triple-hydroxylation of tetracenomycin (Tcrn) A2 to Tcm C in Streptomyces glaucescens (Decker, H., Motamedi, H., and Hutchinson, C. R. (1993) J. Bacterial. 175, 3876-3886). The heterologous expression of the tcmG gene in Streptomyces lividans and the purification and charac- terization of TcmG protein, which we have named Tcm A2 oxygenase, are described here. NH,-terminal amino acid analysis of the purified enzyme led to the revision of the translational start site of tcmG to a TTG, 33 base pairs downstream of the GTG site assigned initially on the basis of nucleotide sequence analysis. Tcm A2 oxy- genase is a monomeric protein in solution and contains 1 mol of non-covalently bound FAD; the apoenzyme can be partially reconstituted in vitro by addition of FAD. Tcm A2 oxygenase exhibits an optimal pH of 9.0-9.5 and prefers NADPH over NADH as an electron donor. The apparent K‘, of the enzyme for Tcm A 2 , NADH, and NADPH are 1.81 f 0.38,260 19, and m, respectively, and the apparent

have elucidated all of its biosynthetic intermediates ( Fig. 1 B ) (9)(10)(11). We also characterized several key enzymes of the pathway, including the Tcm F2 polyketide synthase , the Tcm F2 cyclase (151, and the Tcm F1 monooxygenase (16) and cloned (17) and analyzed (18-23) the nucleotide sequences of the complete gene cluster for the biosynthesis of 1 (Fig. 1B). These studies have provided detailed insights into the biochemistry and genetics of the biosynthesis of 1 in S. glaucescens, which can serve as a model for the formation of aromatic polyketides in general. During this work, we have suggested that the three cis-hydroxy groups at the 4, 4a, and 12a positions of 1 are introduced by hydroxylation of Tcrn A 2 , 6, an unprecedented process catalyzed by the TcmG protein (20). The latter idea was further supported by complementation experiments with the Tcm C non-producing mutant S. glaucescens WMH1089 that contains a 180 base pair deletion mutation that has been mapped to the tcmG gene (20); the production of 1 was restored upon transformation of the WMH1089 strain with pWHM126 that carries most of tcmG (20). Moreover, the results of in vivo "0, feeding experiments have indicated that only the 4-and 12a-OH groups of 1 are derived from molecular oxygen, leaving the 4a-OH to arise presumably from water (6).
To extend our investigation of the biosynthesis of 1, in particular to understand the underlying enzymatic reaction mechanisms of the pathway, we studied the hydroxylation of 6 to 1 in vitro and report here the overexpression of tcmG in Streptomyces lividans and the purification and characterization of the Tcm A2 oxygenase. NH,-terminal amino acid analysis of the purified enzyme has led to a revision of the tcmG translational start site to a TTG codon, 33 base pairs downstream of the GTG codon assigned initially on the basis of nucleotide sequence analysis (20). Our results establish the stoichiometry of the conversion of 6 to 1 and prove that this reaction is catalyzed by the enzyme encoded by tcmG. Tcm A2 oxygenase was found to be a monomeric flavoprotein containing 1 mol of non-covalently bound FAD and to require molecular oxygen and reduced nicotinamide cofactors.
EXPERIMENTAL PROCEDURES General-UV-VIS spectra were recorded on a Hitachi U-3000 spectrophotometer (San Jose, CA). Refrigerated centrifugation was done in a Sorvall RC-5B superspeed centrifuge (Newtown, CT). A Pharmacia FPLC system was used for enzyme purification and all FPLC columns were purchased from Pharmacia Biotech Inc. HPLC was done with a Waters model 501 pump system (Marlborough, MA) and a Waters 484 variable wavelength absorbance detector. Enzyme incubations were performed in a GCA Precision shaking water bath (+ 0.1 "C) (Precision Scientific Inc., Chicago, IL). Fermentations were carried out in a rotary shaker-incubator (Series 25, New Brunswick Scientific CO

FIG. 1. Tcm C and related naphthacenequinones (A) and biosynthetic pathway of Tcm C in S. glaucescens ( E ) .
are described elsewhere; pSP72 and pGem7zf were obtained from Promega Corporation (Madison, WI). The ermE+ promoter containing plasmids pWHM63, pWHM64, and pWHM65 were gifts from G. Meurer.' Thiostrepton was obtained from Sal Lucania at the Squibb Institute for Medical Research (Princeton, NJ). Unless specified, common chemicals, restriction enzymes, DNA ligase, and other materials for recombinant DNA procedures were purchased from standard commercial sources and used as provided.
DNA Isolation a n d Manipulation-Plasmid DNA from Escherichia coli was prepared according to Lee and Rasheed (26). For plasmid DNA isolations from Streptomyces spp., the cells were lysed according to Hopwood et al. (271, then treated as described by Lee and Rasheed (26). Agarose gel electrophoresis, restriction enzyme digestion, DNAligation, and preparation of competent E. coli DH5a cells and their transformation were performed by established methods (28). DNA was purified from agarose gels with the QIAEX kit as directed by the manufacturer (QIAGEN Inc., Chatsworth, CAI. Protoplasts of Streptomyces spp. were prepared and transformed by the methods of Hopwood et al. (27).
Construction of pWHM68, pWHM72, and pWHM73-To prepare pWHM68, a 2.1-kb EcoRI-Hind111 fragment from pWHM1018 (20) that contains tcmG was cloned into the same sites of pSP72 to give pWHM66. The later was digested with HindIII and EcoRV and the resulting 2.1-kb fragment was cloned into the HindIII-SmaI sites of pWHM63 to give pWHM67, from which a NsiI-XbaI fragment was moved into the XbaI-PstI sites of pWHM3 to yield pWHM68 (Table I).
To construct pWHM72 and pWHM73, polymerase chain reactions  CCGCCAGCGGGCTCCCTC-3', 1.0 ng), 5 pl of formamide, 1 p1 of BSA (1 mg/ml), 36 p1 of H,O. The reaction mixture was covered with three drops of mineral oil, boiled for 5 min, and placed at 70 "C. To this mixture was added 4 pl of the dNTP mixture (final concentrations for dCTP and dGTP were 60 p and for dATP and dTTP, 40 p~, respectively) and 1 pl of Taq polymerase (4.5 units). The PCR temperature program was as follows: 24 cycles of 40 s at 96.5 "C and then 2.5 min at 71 "C; after the last amplification cycle, 40 s at 97 "C and then 7 min at 71 "C. After cycle 12, a n additional 0.5 pl of Taq polymerase (2.25 units) was added. The final amplified 306-base pair fragment was purified by agarose gel electrophoresis, digested with HindIII and SacI, and ligated with the 4.2-kb HindIII-SstI fragment obtained from partial digestion of pWHM1018 to yield pWHM69. The tcmG gene was transferred from pWHM69 as a SphI-XbaI fragment into similar sites of pWHM64 and pWHM65 to give pWHM70 and pWHM71, from which the 2.1-kb EcoRI fragments were cloned into the same site of pWHM3 to yield pWHM72 and pWHM73, respectively (Table I).
Protein Analysis-Protein concentrations were determined by the Bradford (29) method with BSA as the calibration standard. Pure Tcm A2 oxygenase also was quantified by UV absorption a t 280 nm where the molar absorbance index (ez8,, ", , , I is 75.7 mM-l.cm-l (this value was calculated from the amino acid sequence deduced for the apoenzyme from tcmG). The molecular weight of the enzyme subunit was determined by SDS-PAGE using the Life Technologies, Inc. protein molecular weight standards of myosin H-chain 200,000, phosphorylase b 97,400, BSA 68,000, ovalbumin 43,000, carbonic anhydrase 29,000, p-lactoglobulin 18,400, and lysozyme 14,300. SDS-PAGE was performed according to the method of Laemmli (30) or on the PhastSystem (Pharmacia) as described by the manufacturer and the gels were Coomassie Blue-stained (31). The abundance of each band was then quantified on a Molecular Dynamics model 300A Computing Densitometer (Sunnyvale, CA). The molecular weight of the native TcmG was determined by gel filtration chromatography on a Superose 6 HR 10/30 column in 20 m~ sodium phosphate, pH 7.2, 1 mM DTT, 150 mM NaCl with a flow rate of 0.4 ml/min and the column was calibrated with blue dextrin 2 x lo6, alcohol dehydrogenase, 150,000, BSA 66,000, carbonic anhydrase 29,000, and cytochrome c 12,400 purchased from Sigma.
Enzyme Assays-The substrate 6 and authentic product 1 were isolated from S. glaucescens WHM1089 and S. glaucescens GLA.0, respectively, and characterized as described elsewhere (1,9,10). ethanolamine-HC1 buffer, pH 9.5, in the presence of enzyme (10-50 pl), was incubated at 25 "C. The assay was initiated by addition of 6 and terminated by addition of solid NaH,PO, to saturation and extraction with EtOAc (2 x 400 pl). The EtOAc extracts were collected and concentrated in oacuo to dryness, then the residue was dissolved in 50-120 p1 of methanol and analyzed by TLC or HPLC. SiO, plates were used and developed in CHClfleOH (955, v/v); under these conditions 6 and 1 have a n Rf of 0.77 and 0.28 and, under UV light, display a characteristic yellow and blue fluorescence, respectively. This TLC method was used throughout the purification to monitor the enzyme activity qualitatively. Alternatively, a HPLC method was developed that provided a quantitative analysis of the enzymatic synthesis of 1 from 6. Assay samples were analyzed by HPLC on a Nova-Pak C,, column (Waters) developed with a linear gradient from CH,CN/H,O/AcOH (20:80:0.1%, v/v) to CH,CN in 10 min followed by additional 5 min at 100% CH,CN at a flow rate of 2 mumin with UV detection at 280 nm. The column was calibrated with authentic 1 and 6 that, under these conditions, have retention times of 6.5 and 12.0 min, respectively. The HPLC assay method was used in all studies with the following modifications. For the pH dependence study, the assays were performed in 50 mM Tris-HC1 buffer, pH 6.5-9.0, and 50 mM ethanolamine-HCl buffer, pH 8.0-10.5, respectively, in the presence of 11.2 pg of TcmG. For determination of the kinetic parameters, the assays were done with the concentration of 6 varied from 0.5 to 30 PM, 1.0 mM NADPH, and 2.93 pg of TcmG, or with concentrations of NADH or NADPH varied from 50 to 1.5 mM, 100 p~ 6, and 5.85 pg of TcmG, respectively, in 50 m~ ethanolamine-HC1 buffer, pH 9.0, for durations that yielded a linear relationship between product formation and time. The apparent kinetic constants of K ' , and V m U were determined by a nonlinear regression analyses (32) based on the Marquardt-Levenberg algorithms.
Enzyme Purification-All steps were carried out a t 4 "C except for the brief time when the enzyme was on the FPLC columns that were at room temperature.
Step 1. Preparation of cell-free extract: cultures of S. liuidans transformed with the tcmG expression plasmids were grown in R2YENG media (9, 17) with thiostrepton (10 pg/ml) in a 2-liter baffled Erlenmeyer flask. After incubation at 30 "C and 300 revolutions/min for 3 days, cells were harvested by centrifugation (13,600 x g , 20 min, 4 "C) and washed sequentially with 0.5 M NaCl and 0.1 M sodium phosphate buffer, pH 7.2, with centrifugation as necessary to yield approximately 15 g cellsfliter (wet weight). The washed cells were suspended in 100 mM sodium phosphate buffer, pH 7.2, 2 mM DTT, 0.1 mM phenylmethylsulfonyl fluoride, 1 mM EDTA, 10% glycerol (10 mug cells). Lysozyme (1 mg/ml) was added, and the mixture was left to incubate at room temperature for 2 h. To this viscous slurry, solid MgCl, (5 mg/ml) and DNase (1 pg/ml) were added. The resulting slurry was incubated on ice for 1 h, and cell debris were removed by centrifugation (27,500 x g , 20 min, 4 "C) to yield a cell-free extract. The enzyme activity in preparations from this and succeeding purification steps was followed by the TLC assay method.
Step 2. Ammonium sulfate fractionation: the cell-free extract was brought to 41% saturation (234 gfliter) by addition of solid ammonium sulfate. The suspension was stirred for 1 h and centrifuged as above to remove the precipitate. The resulting supernatant was brought to 62% saturation (375 ghiter) with solid ammonium sulfate and was stirred for 1 additional h. Centrifugation as above afforded a pellet that had the enzyme activity.
Step 3. Sephacryl S-200 HR column: the ammonium sulfate pellet was dissolved in a minimum volume of 20 mM sodium phosphate buffer, pH 7.2, 1 mM DTT, 150 mM NaCl and applied to a Sephacryl S-200 HR column (2.6 x 60 cm). The column was eluted at the flow rate of 2 mumin with the same buffer, and 5-ml fractions were collected.
Step 4. Mono Q HR 10/10 column: fractions containing the enzyme m~ Tris-HC1, pH 8.0, 1 m~ DTT and applied to a Mono Q HR 10/10 activity after gel filtration chromatography were dialyzed against 25 column. After washing with 25 mM Tris-HC1 buffer, pH 8.0, 1 mM D m , the column was developed a t a flow rate of 2 mVmin with a linear 60-ml gradient from 0 to 0.6 M NaCl in the same buffer, and 2-ml fractions were collected.
Step 5. Alkyl Superose HR 5/5 column: the active fractions after anion exchange chromatography were brought to 1.2 M ammonium sulfate by addition of solid ammonium sulfate and were applied to a n Alkyl Superose HR 5/ 53 column. The column was washed with 50 m~ sodium phosphate, pH 7.2,l mM DTT, 1.2 M (NH,),SO,, then developed at a flow rate of 0.5 mumin with a linear 15-ml gradient from 1.2 to 0 M (NH,),SO, in the same buffer, and 0.5-ml fractions were collected. The final preparation of active enzyme was stored at -20 "C, and no significant loss of enzyme activity was observed over a 4-week period.
NH,-terminal Sequence Determination-A portion of the purified TcmG protein (16 pl, 20 pg) was loaded onto a ProSpinm sample preparation cartridge (Applied Biosystems) and washed according to the instructions provided by the manufacturer. A fraction of the protein bound-polyvinylidene difluoride membrane of the ProSpinTM device was used directly for the NH,-terminal amino acid sequence determination by automated Edman degradation chemistry at the University of Wisconsin Biotechnology Center (Madison, WI).
Prosthetic Group Determination-Pure TcmG protein (200 pl, 224 pg) in 25 mM Tris-HC1, pH 8.0, was boiled for 5 min, cooled on ice immediately, and centrifuged in an Eppendorff centrifuge for 20 min to pellet the denatured protein. The supernatant was subjected to HPLC analysis on a C,, column to quantify any flavin prosthetic group released by this process (33). The column was calibrated with FAD and FMN, respectively, and was developed with the following program: flow rate of 2 mumin with UV-VIS detection at 450 nm in 5 mM NH,OAc buffer, pH 6.5, 10% MeOH (v/v) in the first 5 min followed by a 20-min linear gradient from 10  assayed to determine the residual enzyme activity. For in oifro reconstitution ofthe apoenzyme. a 500-pl solution containing 1.65 nmol ofthe apoenzyme, 16.5 nmol of either FAD or FMN, respectively, in 25 mM Tris-HCI, pH 8.0. 1 m y DTT was incuhated on ice for 1.5 h (35,361; 1.65 nmol of the holoenzyme was similarly treated with FAD or FMN as controls. Both FAD and FMN were purified by HPLC on C,, column under the conditions described above to ensure no contamination of other flavin derivatives in the commercial materials (34). The resulting reconstituted enzyme was assayed directly hy the HPLC method withnut attempting to remove the excess FAD or FMN.

RESULTS
Heterologous Exprrssion of thr tcmG Gene in S. lividans-Since Tcm A2 oxygenase activity was not detected in cellular extracts from either the S. glaucmcrns GLA.0 wild-type strain or the WMH1094 (20) Tcm C non-producing mutant that hiotransformed 6 to 1 effectively in vivo, we expressed the tcmG gene in S. 1ividan.s to facilitate the isolation and characterization of Tcm A2 oxygenase. pWHM1019, pWHM72, and pWHM73 were made in the high copy number vector pWHM.7 (25) so as to place the expression of tcmG under the control of the tcmG promoter (21) with the tcmG RBS from S. glaucescens (20) or of the ermE* promoter (12) in combination with a RRS from either Strrptom.yces antihioticus (37) or S. lioidans (.78), since it is known that the ermE" promoter displays the strongest activity among several common Streptomyces promoters studied (39) and that the sequence of the RBS can also have a distinct effect on the level of gene expression." pWHM68 was constructed to examine the effect of placing the rrmE'" and tcmG promoters in tandem on the level of expression of tcmG.
All four plasmids were introduced by transformation into S.
Iioidans 1326 (24), and the levels of tcmG expression were assayed by SDS-PAGE of cell-free extracts prepared and analyzed as described under "Experimental Procedures." A distinctive hand migrating with a size of 60,000 Da was observed in all four cases (Fig.  2 A , lanes 1-4); this band was absent in a sample from the control culture of S. lividans (pWHM3) (Fig.  2A, lane 5 ) . The apparent size of the expressed protein is consistent with the size of 61,694 Da predicted from the nucleotide sequence of tcmG (20). Whereas the abundance of TcmG was approximately the same from the plasmids in which tcmG expression was under the control of either ermE* or tcmG promoter alone ( Fig. 2A,  Purifirntion of thr TrmC Prntcin from S . 1ir.irIlnns IpWHM6iR)"Since S. lirlidans rpWHM681 rxprrssrd trmG most effectively among the constructs trsted, a crll-frrr extract \vas prepared from this recombinant strain for the purificntitrn of TcmG. The enzyme activity was fractionatrd hy addition of (NH,,),SO., to the cell-frre extract a t 41-62'; sat.urntion. and more than RFir> of the enz.yme activitv was recovrrrtl in thr (NH,,),SO., precipitated pellet. Furthrr purification was prrformed as descrihed undrr "Exprrimental Procrdurrs" hv sizr exclusion tSephacryl S-200 HRI. anion-exchangr (Mono Q H R 10/10), and hydrophohic interaction (Alkyl Suprrosr HR 5 / 3 1 chromatography. Thew procrdurrs gave purr TcmG \vith an overall 20-fold purification in more than 70r; yirld fTahlr 111 (approximately 40 mgllitrr of isolated yirld,. Thr purifird protein was homogeneous when examined hy SDS-PAGE wherr it migrated as n single hand of 60.000 Da I Fig. 213 I. NH,-trrminnl Srqurncr Drtrrminntion-To confirm that thr isolated protein whose purification was guitlrd hy Tcm A2 oxygenase activity was indeed the trmG product. a portion of t h r purified protein was suhjected to amino-trrminal srqurncing and, to our surprise, the NH, terminus was found to hc, STEEVPVLIV-, which is 12 amino acids shortrr than that prrdicted from the tcmC sequence ( Fig. 3 A ) (201. Although it is known that post-translational processing of thv nnscrnt protrin occasionally can result in loss of a short prptidr frapnrnt 1.10. 41), re-examination of the tcmC sequencr rrvralrd a possihlr new translational start site, thr 7°K; that is 88 hasr pairs downstream of the original GTG translational start sitr (Fig.   3 A ) (20). This 'ITG is preceded by a putativr RRS of GAAA(;-GCTGC, which compares flavorahly with GAAA(;GA(XT that is complementary to the 3'-end of 16 S rRNA of S. 1it.idnns ( 3 8 I and predicts an NH, terminus idrntical to thr onr tlrtrrminrd from the purified TcmG protein (Fig. 3A I rxcluding thr nlrt residue. Therefore, we revised the translational start sitr of tcmC to this I T G codon, and Fig. 8B shows thr drducerl amino acid sequence ofTcmG. The new NH, terminus ofTcmG. in fact. aligns well with several othrr hactrrial oxygrnasrs 1201. Sincr the amino terminus of the isolated protrin was found to hr Srr. it can he calculated from the rrvisrd nucleotidr srqurncr that the Tcm A2 oxygenasr has a molrcular mnss of 60,429 D:I with a n PI of 5.57. Molecular Wpight Drtrrminathn-Thr nativr form ofTcm A2 oxygenase has a M , of 61.000 as drterminrd by grl filtration chromatography on a Superose 6 HR 10/80 column.
Stoichiomctr?, of thr Trm A2 O q E r n n s r Cntnl.vrrrl l!,vdro.r.v-Iation of 6 to I-No discrete intermrdiatr was drtrctrd in t h r conversion of 6 to 1 catalyzed hy Tern A2 oxygmasr undrr all A. Originally assigned TcmG>fMet-P-V-S-D-R-P-K-G-C-I-   conditions studied. As summarized in Table 111, the enzyme utilizes either NADH or NADPH as electron donors and requires molecular oxygen; removing 0, by exchange with nitrogen inhibited the hydroxylation completely, as did heat denaturation.

Revised TcmG>jMet-S-T-E-E-V-P-V-L-I-V-
pHDependence-Tcm A2 oxygenase displayed an optimal pH of 9.0-9.5 in 50 m M Tris-HC1 or 50 nm ethanolamine-HC1 buffer as shown in Fig. 4. While a decrease of 1 unit below the optimal pH caused an approximately 50% loss of the specific activity, an increase of 1 unit above the optimal pH resulted in complete loss of the enzyme activity. It is not known, however, if the loss of enzyme activity at pH > 9.5 resulted from deprotonation of specific amino acid residues at the active site, from denaturation of the protein, or from decomposition of NADPH.
Prosthetic Group Investigation-Most of the known oxygenases have either a flavin or heme as their prosthetic group or require a metal ion for the activation of molecular oxygen. Since analysis of the nucleotide sequence of tcmG has revealed a conserved domain for flavin binding (20) and a flavin has characteristic absorption maxima at 375 and 450 nm (361, we determined the UV-VIS absorption spectrum of the purified Tcrn A2 oxygenase (Fig. 5). These data show that Tcm A2 oxygenase is a flavoenzyme. To further identify the nature of this flavin prosthetic group, a solution of TcmA2 oxygenase (224 pg) was heat denatured to release the non-covalently bound flavin (33), which then was analyzed by HPLC on a C,, column. Under the given conditions, most of the known flavin derivatives such as FAD and FMN are well separated (341, as shown in Fig. 6A.   Fig. 6B shows that the prosthetic group released from Tcm A2 oxygenase is FAD, which was further confirmed by co-chromatography with a mixture of authentic FAD and FMN (Fig. 6 0 .
From the results of HPLC analysis, calibrated with authentic FAD, it was established that the molar ratio of apo-TcmG/FAD is 1:l. This value agrees reasonably well with a 1:0.73 ratio of apo-TcmG/F'AD, determined spectroscopically based on the molar absorbance indexes of E; , "" and €!Em, respectively.
Preparation of Apo-Tcm A2 Oxygenase and Its in Vitro Reconstitution with FAD-After establishing that Tcm A2 oxygenase contains 1 mol of non-covalently bound FAD as a prosthetic group, we explored ways to prepare the apo-Tcm A2 oxygenase. The best results were obtained by dialysis of Tcm A2 oxygenase in 100 nm potassium phosphate buffer, pH 4.0, in the presence of 2 M KC1 (35, 36). The protein was completely denatured by this treatment as indicated by its precipitation and was resolubilized and presumably refolded by subsequent dialysis in 25 nm Tris-HC1, pH 8.0, 1 m M DTT. The apoenzyme was colorless, in contrast to the characteristic yellow color of the holoenzyme, suggesting the removal of the flavin prosthetic group, and possessed very little of the initial enzyme activity ( Table IV). The apo-Tcm A2 oxygenase could be partially reconstituted in uitro (35, 36), but only by FAD; addition of either FAD or FMN to the holo-Tcm A2 oxygenase under parallel conditions resulted in a very small change in activity. Kinetics-Assuming that the 0, concentration was constant in the assay solution, kinetic analyses were carried out on the basis of a pseudo-first-order treatment with a steady-state approach. Thus, the effect of the initial concentration of 6 on the formation of 1 was determined at the concentration of NADPH 2 10 K2mpH, and the effect of NADH or NADPH was determined at the concentration of 6 2 10 KEmA2. Velocities were then fitted to the Michaelis-Menten equation (32) and the apparent K ' , for 6, NADH, and NADPH were found to be 1.81 '' 0.38, 260 -c 19, and 82.1 2 17 p~, respectively, with an apparent V m m of 14.7 f 1.1 nmol Tcm C/min-mg. DISCUSSION The enzymatic mechanism for the introduction of angular hydroxy groups like 4a-OH and 12a-OH of 1 into many other naphthacenequinone, angucycline, and anthracycline antibiotics (4)(5)(6)(7)(8)42) is unknown, and the origins of such groups have been studied previously only by in vivo feeding experiments with or "0-containing precursors (6,8,(43)(44)(45)  bition of the oxygenase with P-450 inhibitors (46,47). At least three pathways can be proposed for their introduction. They could simply be retained from the carbonyl groups of polyketide precursors, such as acetate, malonate, etc., without going through an aromatic intermediate like 6, as in the urdamycins (42) or 5, whose 4a-OH was found to be derived from acetate (8). Alternatively, they could be introduced late in the biosynthetic pathway by an oxygenase, as proposed for 1 in Fig. 7, acting as either a monooxygenase (route a ) or a dioxygenase (route b).

In vitro reconstitution of apo-Tcm A2 oxygenase with FAD and FMN
The purification and characterization of Tcm A2 oxygenase supports the hypothesis that the triple hydroxylation of 6 to 1 is catalyzed by a single enzyme. This enzyme requires molecular oxygen and is able to use either NADH or NADPH. Since the apparent V',JK', for NADPH (0.179) is more than 3-fold larger than that for NADH (0.0565), we conclude that Tcm A2 oxygenase prefers NADPH under physiological conditions. NH,-terminal amino acid analysis of the purified Tcm A2 oxygenase confirmed that it was encoded by tcmG but showed that translation begins at a rare TTG codon instead of the much more frequently used GTG or ATG translational start site (48) assigned initially on the basis of nucleotide sequence analysis (Fig. 3A) (20). SDS-PAGE analysis showed that the purified enzyme displayed a single band with M, 60,000, and gel filtration chromatography suggested that the native enzyme has an M, 61,000, indicating that Tcm A2 oxygenase is a monomeric protein in solution.
The exact mechanism of the Tcm A2 oxygenase catalyzed hydroxylation of 6 to 1 is not clear yet since the data reported here do not discriminate between a monooxygenase and a dioxygenase mechanism. As proposed in Fig. 7 (route a ) , two of the three oxygens could be introduced stepwise from molecular oxygen if the enzyme acts like a monooxygenase. The first monooxygenase activity could hydroxylate 6 to hydroquinone 7 that could be further oxidized by the second monooxygenase activity to yield epoxyquinone 8; cis opening of oxirane ring by a H,O molecule could introduce the third oxygen to yield dihydroxyquinone 9 that could be finally reduced to 1. In contrast, two of the three oxygens could also be introduced in a concerted fashion from molecular oxygen if the enzyme acts as a dioxygenase (Fig. 7, route b ) where a likely stable intermediate would be the epoxysemiquinone 10; cis opening of its oxirane ring by a H,O molecule could introduce the third oxygen to yield 1. Both mechanisms are consistent with the results of an in uiuo 1 8 0 2 feeding experiment that has demonstrated that the oxygens of the 4-OH and 12a-OH groups come from molecular oxygen and the 4a-OH group presumably comes from H,O (6). The monooxygenase pathway (Fig. 7, route a ) is supported by the amino acid sequence similarity between TcmG (20) and other bacterial hydroxylases, such as those found in the oxytetracycline producer Streptomyces rimosus (49)  daunorubicin producer Streptomyces peucetius (50) that act on tetracyclic aromatic substrates similar to 6. Monooxygenases that oxidize hydroquinone to form epoxyquinones also have been isolated recently from Streptomyces LL-C10037 (51) and Streptomyces MPP 3051 (51), and a monooxygenase activity for the direct oxidation of an anthraquinone to an epoxyanthraquinone has been reported in Streptomyces rosa var. notoensis . Moreover, 7 has been isolated as a stable metabolite by refluxing an acidic solution of 1 (521, and a similar structure like 9, in fact, has been isolated from Streptomyces olivaceus TU 2353 (4,5), the producer of 2, as the minor metabolite elloramycin E. (It was not established if elloramycin E is the direct precursor of 2 or results from a facile oxidation of 2.) The dioxygenase pathway (Fig. 7, route b ) is consistent with our inability to detect any discrete intermediate in the in vitro hydroxylation of 6 to 1 (10 might be spontaneously hydrolyzed to 1 upon release from the enzyme). We favor the dioxygenase mechanism since in this case the monomeric Tcm A2 oxygenase would only need to recognize one substrate, 6, instead of recognizing at least three substrates, 6, 7, and 9, as in the monooxygenase mechanism, assuming that the conversion of 8 to 9 is also a spontaneous process. Furthermore, a dioxygenase mechanism similar to route b has recently been established for the vitamin K-dependent carboxylase (53, 541, although the latter enzyme does not possess any prosthetic group or require NAD(P)H cofactors.
Many bacterial oxygenases are flavoenzymes, and FAD and FMN are the most common forms of the flavin prosthetic group identified (36). It is known, however, that unusual flavin derivatives are utilized as cofactors in Streptomyces spp. For instance, a 5-deazaflavin is required for the anhydrotetracycline oxygenase from S. rimosus (491, and an unknown flavin is implicated in a tylosin reductase from Streptomyces frudiae (55). Tcm A2 oxygenase contains 1 mol of FAD that binds to the apoenzyme non-covalently, as judged by its release from the protein upon heat denaturation (33). These facts led to the preparation of apo-Tcm A2 oxygenase and its attempted reconstitution in vitro. Apparently the apoenzyme is much less stable than the holoenzyme, yet the holoenzyme can be reconstituted by FAD only, albeit to a low degree ( Table IV). The latter observation re-enforces the conclusion that Tcm A2 oxygenase uses FAD exclusively as its prosthetic group. It is known that upon removal of the flavin prosthetic group from some flavoenzymes the apoenzyme becomes less stable (35,36, 561, as observed here with the apo-Tcm A2 oxygenase, which failed to be reconstituted completely after being kept at -20 "C for 3 days. Although it is not clear why exogeneous FAD or FMN slightly activates or inhibits the enzyme (Table IV), the inhibitory effect of exogeneous flavins has been reported for other flavoenzymes and interpreted as resulting from competition between the free flavin and enzyme-bound flavin for NAD(P)H as electron donor (55). On the other hand, it is also known that a non-covalently bound flavin prosthetic group can be lost in the course of protein purification (57,58), leading to a fortuitous activation of the enzyme preparation when it is supplemented by externally added flavin.
It is interesting to point out the difference between the single and double promoter systems in the efficiency of expression of tcmG. The revised tcmG translational start site, unfortunately discovered after the expression vectors were made, abolished the influence of the two different RBSs in these constructs since the GAAAGGCTGC RBS that precedes the TTG translational start site was presumably utilized in all constructs. Therefore, since the expression of tcmG under the control of either tcmG or ermE* promoter alone gave an approximately equal abundance of TcmG ( Fig. 2A, lanes 2 4 ) whereas the tandem ermE*::tcmG promoters resulted in approximately 5-fold higher level of tcmG expression (Fig. 2A, lane 11, the two promoters placed in tandem have an additive effect on expression. Similar effects of dual promoters on gene expression have been seen in other cases (59).