Isolation and characterization of the versicolorin B synthase gene from Aspergillus parasiticus. Expansion of the aflatoxin b1 biosynthetic gene cluster.

Versicolorin B synthase catalyzes the side chain cyclization of racemic versiconal hemiacetal (7) to the bisfuran ring system of (−)-versicolorin B (8), an essential transformation in the aflatoxin biosynthetic pathway of Aspergillus parasiticus. The dihydrobisfuran is key to the mutagenic nature of aflatoxin B1 (1). The protein, which shows 58% similarity and 38% identity with glucose oxidase from Aspergillus niger, possesses an amino-terminal sequence homologous to the ADP-binding region of other flavoenzymes. However, this enzyme does not require flavin or nicotinamide cofactors for its cyclase activity. The 643-amino acid native enzyme contains three potential sites for N-linked glycosylation, Asn-Xaa-Thr or Asn-Xaa-Ser. The cDNA and genomic clones of versicolorin B synthase were isolated by screening the respective libraries with random-primed DNA probes generated from an exact copy of an internal vbs sequence. This probe was created through polymerase chain reaction by using nondegenerate polymerase chain reaction primers derived from the amino acid sequences of peptide fragments of the enzyme. The 1985-base genomic vbs DNA sequence is interrupted by one intron of 53 nucleotides. Southern blotting, nucleotide sequencing, and detailed restriction mapping of the vbs-containing genomic clones revealed the presence of omtA, a methyltransferase active in the biosynthesis, 3.3 kilobases upstream of vbs and oriented in the opposite direction from vbs. The presence of omtA in close proximity to vbs supports the theory that the genes encoding the aflatoxin biosynthetic enzymes in A. parasiticus are clustered.

itself does not pose a major health threat; however, renal and hepatic oxidative detoxification of AFB 1 -contaminated foods by P450 enzymes yields aflatoxin-15,16-exo-epoxide (see Scheme I, 2), a highly toxic mutagen (4,5). It has been shown that the epoxide targets guanine residues and selectively alkylates the N-7 position of this purine in double-stranded DNA (6,7). Depurination of the alkylated base has been correlated to bladder cancer in laboratory mice (8 -10), teratogenic effects in chicken embryos (11), and liver cancer in humans (12)(13)(14). A direct connection between DNA damage and the incidence of human cancer has been established to originate at mutational hot spots of the p53 gene, a tumor suppressor gene whose altered sequence has been associated with approximately 50% of all human cancers (15,16). Aflatoxin B 1 has been found to be responsible in particular for G 3 T transversions at codon 249 of the p53 tumor suppressor gene in hepatocarcinogenesis (17,18) (Scheme I).
The aflatoxin biosynthetic pathway is notably long and complex (Scheme II). Although the formation of polyketide natural products is initiated normally by acetate, a specialized fatty acid synthase apparently acts in the case of aflatoxin to generate a six-carbon hexanoyl starter unit. This primer is homologated by successive malonyl condensations to give, after intramolecular aldol condensation, cyclization, and oxidation, norsolorinic acid (3) (19 -21). Simple redox changes in the hexanoyl side chain yield the internal ketal averufin (4) (22)(23)(24). Oxidation at C-2Ј of 4 induces migration of the anthraquinone nucleus from C-1Ј to C-2Ј to afford hydroxyversicolorone (5) containing the first furan ring (25,26). Preparatory to formation of the second furan ring, oxygen is inserted into the carbon chain of 5 by a proposed Baeyer Villiger-like reaction to give versiconal acetate (6) (27,28). Support for this mechanism has come from a fermentation conducted in an 18 O 2 -containing atmosphere in which the isotopic label (*) was specifically incorporated at the ester oxygen (Scheme II) as shown in 6 (28). A cell-free system of A. parasiticus has been described in which an esterase catalyzed the hydrolysis of this terminal acetate to give versiconal (7), which was cyclized to (Ϫ)-versicolorin B (8) (29). Tracing the fate of 18 O label (*) from 6, it was shown that heavy isotope was retained in 7 without loss in the critical cyclization to versicolorin B (8) (27,28). In hemiacetals 5, 6, and 7, the chiral C-2Ј center is benzylic and adjacent to a masked aldehyde. This is an intrinsically labile stereocenter, and each of these three compounds is isolated as a racemate (25,30,31) (Scheme II).
The cyclization of versiconal hemiacetal (7) can be carried out nonenzymatically in the presence of acid to yield versicolorin B (8) as its racemate, which is designated historically as versicolorin C (32,33). At neutral pH this chemical process is slow and cyclization is catalyzed in vivo by versicolorin B synthase (VBS) to give optically active (Ϫ)-versicolorin B (8) (29 -33). 2 The absolute configuration installed in this cyclase-catalyzed step is preserved in the bisfuran throughout the remainder of the biosynthetic pathway to AFB 1 . The stereochemical match of this structure when metabolically activated as the exo-epoxide 2 and intercalated into right-handed helical DNA is essential to successful covalent adduct formation (35). These are key events in the tumorgenesis of this natural product. Preliminary purifications of VBS have been reported (36 -38), but an improved protocol yielding homogeneous protein has been achieved. 2 Detailed kinetic analyses of the reaction catalyzed by this enzyme reveal that from the stereochemically equilibrating mixture of enantiomers of 7, the 2ЈS-configured hemiacetal is specifically cyclized by VBS to (Ϫ)-versicolorin B (8). 2 Formation of the dihydrobisfuran is completed in the oxidative desaturation of versicolorin B (8) to versicolorin A (9) (29,39). The subsequent steps of the biosynthetic pathway are significantly less well understood. Cleavage of the anthraquinone nucleus and cyclization, decarboxylation, and dehydration afford the xanthone 10 (40). Successive O-methylations are known to occur at C-5 and C-7 to give O-methylsterigmatocystin (11) (41)(42)(43)(44). This intermediate is further cleaved oxidatively, demethylated, cyclized, and decarboxylated to ultimately afford aflatoxin B 1 (1) (45)(46)(47).
Although the mechanisms of these deep-seated molecular rearrangements in the post-versicolorin A segment of the pathway are not known, important progress has been made recently to identify the first genes in A. parasiticus that encode proteins involved in the biosynthesis of aflatoxin (48 -51). Preliminary evidence has been gathered to suggest that these genes are substantially clustered (48 -51), contrary to earlier reports (52)(53)(54)(55). A probable polyketide synthase (pksA) and two fatty acid synthase (fas-1A and fas-2A) genes have been identified by sequence homology and gene disruption experiments (51). The localization of two genes, a ketoreductase (nor-1) acting immediately after the formation of norsolorinic acid (3) and ver-1, whose gene product participates in the oxidative cleavage of versicolorin A (9), has been determined by gene disruption and complementation (48,49,56). Combined with the cloning of one of the purified O-methyltransferases (omtA), the direct linkage of these genes has been determined to be within 45 kb of one another (see Fig. 5) (51). In this paper we describe the isolation of the gene encoding versicolorin B synthase (vbs) from both cDNA and gDNA libraries derived from A. parasiticus. Comparison of the sequences reveals the presence of a single intron in the latter. Translation of the mature mRNA gives a protein of 70,226 Da, in modest agreement with the 78-kDa apparent molecular mass of wild-type VBS as judged by its relative electrophoretic mobility. 2 Alignment of the translated amino acid sequence of VBS with protein sequences compiled in protein data bases revealed a marked homology to several flavindependent oxidases and dehydrogenases. This relationship was unexpected because VBS does not catalyze a redox reaction. Finally, mapping of vbs gDNA clones has allowed the locus of this gene to be established about 3.3 kb upstream of omtA and separated from it by an apparent cytochrome P450 monooxygenase 3 approximately 1400 bp in length of unknown function. These findings expand the experimentally determined dimensions of the apparent aflatoxin gene cluster and unambiguously define the function and location of the gene encoding versicolorin B synthase. Purification and Sequence Determination of Lys-C-generated VBS Proteolytic Fragments-VBS was purified as described previously. 2 VBS was further purified by reverse-phase HPLC on a Vydac C 4 column equilibrated in 0.2% trifluoroacetic acid. The protein solution was dialyzed overnight against 5 mM NH 4 HCO 3 to remove salts. Purified VBS (250 g) was then subjected to automated protein sequence analysis for amino-terminal sequence determination.

Materials-Restriction
VBS was denatured, reduced, and S-alkylated with iodoacetamide for Lys-C proteolysis as described by the supplier (Boehringer Mannheim O, 0.1% trifluoroacetic acid) for 10 min. The HPLC trace of the Lys-C-generated proteolytic fragments of VBS was compared with two control HPLC traces: Lys-C endopeptidase autodigestion and undigested VBS. Two major VBS peptide fragments were collected and subjected to automated sequence analysis on an Applied Biosystems 470A gas-phase sequencer.
Isolation and Analysis of mRNA from Fungal Cells-Conidia of A. parasiticus, SU-1 (ATCC 56775, NRRL 5862), were inoculated into Adye and Mateles medium (58) and grown at 28°C for 48 or 60 h on a rotary shaker (200 rpm). The resulting mycelia were filtered through cheese cloth, rinsed with 250 ml of 0.85% NaCl, and quickly frozen in a liquid nitrogen-cooled mortar. The frozen mycelia were pulverized with a pestal to a fine powder under liquid nitrogen. Total RNA was extracted from the mycelia with guanidine hydrochloride and sodium lauryl sarcosinate (59). mRNA was isolated by poly(dT)-cellulose (1 ml) chromatography twice (59).
gDNA Preparation from A. parasiticus SU-1-High molecular mass (Ն50 kb) A. parasiticus gDNA was prepared by a modified procedure of Cihlar and Sypherd (60) described by Horng et al. (61). Residual RNA was removed by a second incubation with DNase-free RNase A (final concentration 0.1 mg/ml) for 5 h at 37°C.
Construction of gDNA and cDNA Libraries-The genomic DNA isolated from A. parasiticus SU-1 was partially digested with Sau3AI and size fractionated with a 10 -40% sucrose gradient ultracentrifugation. Fractions containing fragments of 9 -15 kb were pooled together and precipitated with sodium acetate and ethanol. The gDNA fragments were partially filled in with Klenow fragment to generate a two-base overhang. Lambda FixII DNA that had been previously digested with XhoI was partially filled in with Klenow fragment to leave a compatible two-base overhang to accommodate the genomic DNA fragments. The partially filled in gDNA was ligated to the treated Lambda FixII DNA and packaged using Gigapack® II Gold packaging extract (Stratagene). The packaged phage were then propagated in the restrictive P2 host E. coli (P2PLK-17) to an original titer of 5.4 ϫ 10 4 pfu/ml containing 95% recombinant phage. The primary gDNA library was then amplified in E. coli cells (LE392) to 1 ϫ 10 8 pfu/ml.
A. parasiticus SU-1 48-h mRNA was used to prepare a cDNA library using the Uni-ZAP XR vector and packaged using the Gigapack® Gold II packaging extract (Stratagene). The packaged phage were then propagated in E. coli cells (PLK-FЈ) to an original titer of 4.4 ϫ 10 6 pfu/ml containing 98% recombinant phage, and the primary cDNA library was subsequently amplified in E. coli cells (PLK-FЈ) to 1 ϫ 10 9 pfu/ml.
Hybridization Experiments with Degenerate Probes-The A. parasiticus cDNA and gDNA libraries were screened by plaque hybridization with seven radiolabeled degenerate probes (7NC, 8NC, 9NC, 10C, 13C, 14C, and 15C; Fig. 1). Degenerate probes for vbs were designed from sequenced fragments of LysC-endopeptidase-treated VBS, taking into account Aspergillus nidulans and Aspergillus niger codon preferences. 4 The radiolabeled probes were generated by end-labeling using T4 polynucleotide kinase (New England Biolabs) with [␥-32 P]ATP (Amersham Corp., 6000 or 10 mCi/ml) (62). The probes were separately purified from unincorporated [␥-32 P]ATP on a Nuctrap column (Stratagene). Each library was plated onto LB agar plates and transferred to Duralon membranes according to the manufacturer's instructions (Stratagene). The membranes were screened according to normal hybridization techniques (59) using hybridization temperatures ranging from 37 to 45°C for up to 40 h. Low stringency washes were typically carried out with 1 ϫ SSC at temperatures ranging from 25 to 37°C for 2-8 h (59). The membranes were autoradiographed with Kodak X-OMAT film at Ϫ80°C using intensifying screens.
mRNA-mediated PCR of a vbs Fragment from First-Strand Synthesis-Amplification of a vbs gene fragment with PCR using an mRNA template was first carried out using MMLV reverse transcriptase from a First-Strand Synthesis kit to generate an mRNA-DNA heteroduplex template (Stratagene). The heteroduplex was subjected to typical PCR reaction conditions as follows. A 100-l PCR reaction mixture contained 10 mM Tris⅐HCl (pH 8.8), 50 mM KCl, 1.5 mM MgCl 2 , 0.001% (w/v) gelatin, 200 M of each dNTP, 1 M each of a noncoding and coding primer listed in Fig. 1, heteroduplex template (100 ng), and Taq DNA polymerase (2.5 units). The reactions were performed in 1.5-ml microfuge tubes by a "hot-start" procedure whereby the tubes were placed in the thermocycler at 95°C before adding the Taq DNA polymerase. Temperature cycling program A outlined in Table I was used. The reaction products were examined on a 1.5% agarose gel. The PCR amplification of the vbs gene fragment was carried out using both 48and 60-h mRNA. The clean PCR product obtained using primers 8NC and 10C was ligated to SmaI-digested pBluescriptII SK(Ϫ). Positive clones were sequenced to verify the orientation and sequence of the ligated genes (RPF1-13NC and RPF1-17C).
PCR Amplification of VBS gDNA and cDNA Library Probes-PCR amplification of the genomic vbs gene fragment was conducted using two primers based on the nucleotide sequences obtained from the RPF1 clones (primer 21C and primer 22NC; see were carried out using purified recombinant DNA from the genomic library, temperature cycling program A (cycles 2-4), and normal PCR conditions as described by Lundberg et al. (82). The reaction product (615 bp) was examined on a 1.5% agarose gel. The same protocol was used to generate the cDNA vbs fragment; however, the cDNA library was used as the source of template DNA.
Screening of the cDNA Library Using PCR-generated vbs Probes-The A. parasiticus SU-1 cDNA library was screened by plaque hybridization with radiolabeled vbs cDNA probes. Radiolabeled vbs probes were generated using [␣-32 P]dCTP with the Random Primed DNA Labeling kit (Life Technologies, Inc.) using the 21C/22NC cDNA PCR fragment as template DNA. The library was transfected into E. coli (P2PLK⅐FЈ) and transferred to Duralon membranes (Stratagene). The Duralon membranes were preincubated at 51°C in Quik Hyb solution (Stratagene) for 30 min before hybridizing with radiolabeled probe for 1 h at 51°C. The membranes were washed at 25°C for 15 min in 2ϫ SSC followed by a 30-min stringent wash at 56°C in 0.1ϫ SSC. The membranes were then placed on Kodak X-OMAT film for overnight exposure and identification of positive clones. Clones were further purified by conducting secondary and tertiary screens.
Plaques of positive clones were cored from the stock plates and placed in SM buffer (59). The positive clones were further verified by PCR using the phage stock solution as the source of template DNA and primers 21C and 22NC to amplify a 615-bp vbs fragment. A second gene fragment of approximately 750 bp was obtained from a PCR reaction using oligonucleotides 8NC and 10C. Generation of the double-stranded cDNA plasmid clones by in vivo excision of phagemid particles with helper phage R408 was carried out as described by the manufacturer (Stratagene). Rescued cDNA plasmids were amplified in E. coli (XL1-Blue). Plasmid pCVBS241e (Fig. 2) was used as template DNA for versicolorin B synthase cDNA sequencing.
Screening of gDNA Library Using PCR-generated vbs Probes-The A. parasiticus SU-1 gDNA library was screened by plaque hybridization with radiolabeled 21C/22NC vbs gDNA probes by a method analogous to the one described for the cDNA library. Plaques of positive clones were cored from the stock plates and placed in SM buffer. The positive clones were further verified to contain vbs by PCR using the phage stock solution as the source of template DNA and primers 21C and 22NC by amplifying an appropriate length vbs fragment (615 bp). A second gene fragment of approximately 800 bp was obtained from a PCR reaction using oligonucleotides 8NC and 10C. The gDNA plasmid clone, pGVBS4.5, was constructed by ligating the 4.5-kb NotI/KpnI fragment from a gDNA clone, g62b, into the NotI/KpnI site of pBluescriptII SK(Ϫ) following established procedures (Fig. 3) (59). pGVBS4.5 was amplified in E. coli (XL1-Blue) and served as template DNA for doublestranded sequencing.
Restriction Mapping of gDNA Clones-Representative procedures for restriction mapping by single and multiple digests are described by Ausubel et al. (63). Lambda clones (52a, 55c, 56a, 57a, 62b, and 63c) were individually digested with combinations of the following restriction enzymes: BamHI, EcoRI, HinDIII, KpnI, NdeI, NcoI, NotI, SacI, SalI, SmaI, XbaI, and XhoI and separated on both 1.5 and 0.75% agarose gels. Single and double restriction digestions containing 0.5 l of each enzyme and 500 ng of DNA were incubated at the lowest optimal temperature for 2 h. A second series of digestions using gel-purified XbaI fragments resolved further complexities.
The DNA from SalI and XbaI digests was resolved by agarose gel electrophoresis and transferred onto a nitrocellulose membrane by capillary (Southern) transfer and cross-linked to the surface using a UV cross-linker (Stratagene, 1200 J) (62,63). Multiple filters were probed with 5Ј-␥-32 P-radiolabeled oligonucleotide probes (1.25 ϫ 10 5 cpm/ml) for vbs (25C , Table II) and omtA (Omt1-2NC, Table II). Prehybridizations and hybridizations were conducted with Quik Hyb solution in sealed bags at 42°C for 1 and 7 h, respectively (Stratagene). The membranes were washed with 2 ϫ SSC/0.1% SDS at 25°C for 15 min followed by a single wash in 0.1 ϫ SSC/0.1% SDS at 42°C for 1-1.5 h. The membrane was autoradiographed at Ϫ80°C for 5-18 h using intensifying screens and Kodak X-OMAT film.
PCR Analysis of gDNA Clones-To test for the presence of vbs, ver-1, and omtA, polymerase chain reactions were assembled with appropriate primers using master mixes and a hot start protocol. The required oligonucleotide primers (2.5 l, 50 pmol each), DNA template (1.0 l, 100 ng), and mineral oil (ϳ60 l) were loaded into microfuge tubes followed by the "lower" master mix (44 l) and briefly centrifuged. The microfuge tubes were placed in the thermocycler, which had been previously heated to 95°C. After the initial denaturization step (5 min), the "upper" master mix (50 l) was rapidly added, and the PCR cycling was initiated.
The lower master solution contained for each reaction deionized distilled water (32.  Tables I and II. Samples were resolved on agarose gels (0.4 -1.5% agarose).
Nucleotide Sequencing and Analysis-Specific restriction fragments were subcloned into the plasmid vector pBluescriptII SK(Ϫ) by standard methods. Single-stranded DNA for sequencing was obtained from these subclones by rescue from plasmid-bearing cells with helper phage VCSM13 according to the supplier's protocol (Stratagene). Plasmid DNA for double-stranded sequencing was purified by the Qiagen plasmid purification procedure. DNA sequencing was accomplished using Sequenase-2.0 DNA polymerase as described by U. S. Biochemical Corp. and either commercially available or custom-synthesized oligonucleotide primers. Sequence data were compiled manually and analyzed using the DNA Strider program. RESULTS VBS, a homodimeric protein of 78-kDa subunits, catalyzes the dehydrative cyclization of racemic versiconal hemiacetal (7) to optically active versicolorin B (8), the step penultimate to desaturation of the tetrahydrobisfuran to the dihydrobisfuran present in (Ϫ)-versicolorin A (9) (29 -33). 2 This unique structural feature is conserved through the subsequent intermediates of aflatoxin B 1 (1) biosynthesis (Scheme II) and is the seat of the progressively severe carcinogenic properties of these metabolites. VBS was purified to homogeneity from A. parasiticus SU-1 (ATCC 56775) by methods established in this laboratory, 2 but failed to give amino-terminal sequence data by automated methods. Although homogenous enzyme was submitted for amino acid sequence analysis (250 pmol), amino acid intensities corresponding to Յ10 pmol of enzyme were observed, suggesting that the amino terminus of the native protein was post-translationally modified. To circumvent this problem, the protein was digested with Lys-C endopeptidase and two of approximately 20 VBS peptide fragments were isolated by reverse-phase HPLC. These two peptide fragments both gave reproducible amino acid sequence data and credible stoichiometry (Fig. 1).
The two peptide sequences were used to design seven oligonucleotide probes for hybridization and PCR experiments. The degeneracy of these probes was minimized by comparing codon usage in A. nidulans and A. niger structural genes to compile a table of codon biases. 4 Plaque hybridization experiments did not provide reproducible results using the partially degenerate probes 7NC, 8NC, 9NC, 10C, 13C, 14C, and 15C synthesized by automated methods (Fig. 1). However, PCR-generated nondegenerate probes were later substituted for these degenerate probes in the hybridization experiments, as described under "Experimental Procedures," to lead to the successful cloning of VBS.
A vbs cDNA gene fragment, obtained from the reverse transcriptase-mediated PCR, was amplified using PCR primers 8NC and 10C and estimated to be approximately 750 bp in length. No PCR product was obtained using any other primer combination from the set of primers shown in Fig. 1. PCR amplification of the approximately 750-bp fragment was observed using both 48-and 60-h mRNA. This gene fragment was subcloned into pBluescriptII SK(Ϫ) generating clones with the insert oriented in both directions (pRPF1-17C and pRPF1-13NC). Single-stranded DNA was prepared by infecting the plasmid-borne XL1-Blue cells with R408 helper phage. Direct sequencing was performed using dideoxy sequencing methods in both directions using clones containing inserts in opposite orientations. Partial nucleotide sequences from the coding (pRPF1-17C) and the noncoding strands (pRPF1-13NC) can be seen in Fig. 4. Translation of the two nucleotide sequences of both pRPF1-17C and pRPF1-13NC concurred with the amino acid sequence data obtained from each of the two peptide fragments isolated from the Lys-C endopeptidase treatment of VBS (Fig. 4, pRPF1-13NC, and pRPF1-17C nucleotide sequences).
From these nucleotide sequences, nondegenerate primers 21C and 22NC were prepared to serve as oligonucleotide primers for PCR experiments with both the gDNA and cDNA libraries (Fig. 4). A 615-bp internal fragment was successfully amplified by PCR from both the cDNA and gDNA libraries. This PCR fragment was used to generate oligonucleotide probes for plaque hybridizations as described under "Experimental Procedures." Further PCR analysis of the gDNA and cDNA clones with primers 8NC and 10C afforded two discrete gene products from each set of clones approximately 800 and 750 bp in length, respectively. The approximately 50-bp difference between the PCR fragments derived from cDNA and gDNA templates was attributed to the presence of an intron within the gene fragment, which was later verified by DNA sequence comparison. From the lambda Uni-ZAP XR cDNA library prepared from 48-h A. parasiticus mRNA, approximately 150,000 plaques were screened. Sixteen positive cDNA clones were isolated and verified to be identical through restriction mapping. Approximately 150,000 plaques were screened from the lambda FixII gDNA library of A. parasiticus yielding six positive gDNA clones. Each was verified to contain vbs by PCR analysis, restriction mapping, Southern analysis, and/or nucleotide sequencing.
Further investigations were undertaken with the genomic lambda clones, which successfully assigned the orientation of vbs and a probable cytochrome P450 monooxygenase (cyp) with respect to earlier portions of the putative aflatoxin gene cluster. Southern and PCR analysis positively identified the presence of omtA in two clones (56a and 62b) and the absence of ver1 in all of the isolated gDNA clones. Although the vbs gene was verified by PCR to be present in each clone following library screening, subsequent examinations attested to a significant truncation of the 5Ј terminus of vbs in two clones (52a and 55c). Together, the genomic clones 52a and 62b contained approximately 30 kb of overlapping genomic sequence, as measured by restriction mapping (Fig. 5). The distance between vbs and omtA was measured by PCR and verified by DNA sequence analysis. Primer combinations Omt1-2NC ϩ 56NC and Omt1-4NC ϩ 56NC (Table II) gave 2.78-and 4.35-kb PCR products, respectively, which are in agreement with the known 1.49-kb separation between the Omt1 primers (66). Employing the four omtA primers (Tables I and II) pRPF1-17 coding strand (b). Experimentally determined protein sequences (underscored) correlated with the superimposed, translated nucleotide sequences. Degenerate primers 8NC and 10C are indicated above the nucleotides pRPF1-13 and pRPF1-17. Nondegenerate primers 21C and 22NC were constructed after sequencing both pRPF1-13 and pRPF1-17. proximately 3.3 kb of each other, in opposite orientations (Fig.  5). Nucleotide sequence data from a 1.3-kb XbaI/KpnI genomic DNA fragment overlapping with the reported 5Ј upstream region of omtA (J. Yu, 1993, GenBank accession number L25834) and an apparent cytochrome P450 monoxygenase, 3 approximately 1400 bp in length, established the clustered nature of the three genes.
The genomic nucleotide sequence of versicolorin B synthase has been determined and is contained within 2610 bp of phage clone 62b. The transcribed cDNA clone possessed a continuous open reading frame of 1932 bp, as well as 20 bp of 5Ј-nontranslated and 161 bp of 3Ј-nontranslated regions. Comparison of the combined cDNA and genomic DNA sequences revealed that the coding region is interrupted by a single 53-bp intron (Fig.  6). The intervening sequence, which has been observed in other eukaryotic genes, shared the consensus regions 5Ј-(exon)/ GTARGY . . . NRCTRAN . . . YAG/(exon)-3Ј (68 -70). The Hogness box, TTTAAA, was seen Ϫ92 nucleotides from the vbs start codon. In addition, two putative CAAT promoter sequences (70) were detected at Ϫ162 and Ϫ224 nucleotides. A pyrimidine-rich motif, commonly associated with fungal promoters, was located between Ϫ72 and Ϫ60 nucleotides upstream of the start codon (70,71). A common trend found in this sequence and many other filamentous fungi genes was an adenine at the third nucleotide upstream of the start codon (70 -72). At the 3Ј-terminus, a polyadenylation tail was appended at position ϩ161 from the end of the stop codon. This site does not correspond to the canonical poly(A) site, although this is not unusual in fungal genes (73). The polyadenylation consensus sequence was represented by 5Ј-AATTAATA-3Ј, 126 nucleotides after the stop codon.
Translation of the coding sequence provides a protein of 643 amino acids with a molecular mass of 70,271 Da and a calculated isoelectric point of 5.06. These values differ from those observed for the native protein (36 -38, 83). 2 The monomeric molecular mass of native VBS as estimated by SDS-polyacrylamide gel electrophoresis and size exclusion chromatography is approximately 78 kDa, with an experimentally determined isoelectric point of 4.7 Ϯ 0.1. Recent work in our laboratory has demonstrated that the native protein is N-glycosylated (data not shown). From the translated amino acid sequence, there are three potential N-glycosylation sites with the motif Asn-Xaa-Thr or Asn-Xaa-Ser. The discrepancy in molecular mass and pI can be attributed to the post-translational modification of the native protein.
The amino acid sequence of the VBS protein was found to have significant homology to many flavin-dependent oxidases and dehydrogenases through BEAUTY (75) and BLAST (76) searches of the Brookhaven protein, SWISS-PROT, PIR, and GenBank data bases (Table III). Choline dehydrogenase and glucose oxidase provided the highest correlations among the homologous proteins identified. Specifically, the BEAUTY search identified the greatest homology with proteins in the GMC oxidoreductase family (77) (cluster ID 3015) and other flavin-dependent oxidases and dehydrogenases (Table III). Interestingly, glucose oxidase from A. niger, like VBS, is homodimeric and has a similar molecular mass. The former has eight potential N-glycosylation sites and is so modified in at least two of these (78). Glucose oxidase has a pI of 4.1 Ϯ 0.1 (74). The results from the GAP alignment of VBS (644 amino acids) and choline dehydrogenase (557 amino acids) showed 34% identity and 56% similarity over the entire VBS amino acid sequence. The GAP alignment to glucose oxidase (583 amino acids) showed 38% identity and 58% similarity over the entire VBS protein sequence (79). Strong regions of homology were observed in the nucleotide phosphate binding sites and the active sites of the GMC family of oxidoreductases (77). Alignments of each of these regions are shown in Fig. 7. An x-ray crystal structure of glucose oxidase from A. niger has been reported at 2.3 Å resolution (78). One FAD molecule is bound in each identical subunit, and these reside near the dimer interface in a ␤␣␤-motif showing high structural conservation. Significant hydrogen bonding interactions are evident to the FAD, particularly to the ribose and phosphate groups. The principal interactions between the protein and the former are seen in Glu 72 , Gly 49 , and Gly 123 in the amino-terminal region. The first two of these correspond to exact amino acid matches in the aligned VBS sequence, whereas the third does not. His 102 , thought to be hydrogen bonded to the ribose 2Јoxygen in glucose oxidase, has been replaced by a tyrosine in VBS. The diphosphate group is involved in several hydrogen bonds, in part to water molecules and to Thr 52 , which has been replaced by alanine in the aligned VBS sequence, although the threonine can be found at the amino-terminal adjacent site. Although displaced by one residue, this threonine aligns with threonine or serine in all other members of the GMC oxidoreductases summarized in Fig. 7. Surrounding these Thr/Ser residues is the GXGXXG motif characteristic of this protein family (79). This sequence motif is associated with phosphate binding and is fully conserved in VBS.
So, although important interactions between glucose oxidase and FAD show strong correlations in the structure of VBS, a striking 23-amino acid gap exists between Gly 138 and Phe 139 . This is a significant deletion in the middle of the potential FAD binding domain and a gap not present in any of the currently known GMC family of flavoproteins. Moreover, two amino acid contacts to FAD in glucose oxidase lying carboxyl-terminal to this gap do not map to identical residues in VBS. In this connection it is noteworthy that homogeneous VBS does not contain a bound flavin chromophore and preliminary kinetic evidence suggests that FAD, FMN, and glucose have little or no inhibitory effect on the cyclization of versiconal (7) to versicolorin B (8) (35,80). 2

DISCUSSION
VBS catalyzes the dehydrative cyclization of (Ϯ)-versiconal hemiacetal (7) to set the absolute configuration of (Ϫ)-versicolorin B (8) and, hence, aflatoxin B 1 (1) (28,29,38). 2 This key cyclization reaction in aflatoxin biosynthesis has been demonstrated by isolation and purification of the native protein (36 -38) and by expression of vbs in S. cerevisiae to afford fully active enzyme. 5 Design of PCR probes from amino acid sequence data derived from the pure protein allowed the VBS gene to be isolated from both cDNA and gDNA libraries of A. parasiticus. The unexpectedly high homology of the translated protein to flavin-dependent enzymes as glucose oxidase (74) and choline dehydrogenase (80) leaves open the question as to whether this relatively large 78-kDa protein may harbor a second activity, presumably oxidative, in the aflatoxin pathway. However, truncation in the region of presumed FAD binding and the absence of bound flavin chromophore in the native protein suggests that VBS may not have such a role.
Little is known at present about the complexity of aflatoxin biosynthesis at the level of individual proteins and the possibility of dual catalytic roles, but an insight into how difficult this understanding may become has already been encountered. The ver-1 mutant, Wh-1 (ATCC 36537), is blocked in the conversion of versicolorin A (9) to demethylsterigmatocystin (10) (81). The genetic defect has been localized in complementation and gene disruption experiments (48) to reside in ver-1. However, whereas the involvement of ver-1 in aflatoxin biosynthesis has been securely established in vivo, the function of its gene product has defied demonstration in vitro (48,56). Vidal-Cros et al. (34) observed a 56% protein sequence identity be-tween VER-1 and scytalone reductase, an enzyme that catalyzes an aryl dehydroxylation in the biosynthesis of melanin. Furthermore, a 52% similarity exists between the ver-1 gene product and the presumed ketoreductase from the Streptomyces actIII gene (48). The transformation of the anthraquinone 9 to the xanthone 10 involves oxidative ring cleavage, rearrangement, deoxygenation, and decarboxylation and may well require several enzymes (57). The degree to which these proteins act individually or in a tightly ordered or even physically associated manner remains to be established. No intermediates in this process have been isolated or, indeed, in the equally cryptic xanthone 11 3 coumarin 1 transformation (45). Finally, the clear demonstration that 6-deoxyversicolorin A is not an intermediate in aflatoxin biosynthesis is at loggerheads with the function assumed for VER-1 based on protein sequence information (29).
Recent work by Linz, Bhatnagar, Payne and co-workers (51) has established partial organization of the aflatoxin B 1 biosynthetic genes in A. parasiticus. In contrast to the earlier findings of Papa (53)(54)(55), whose work identified several linkage groups in A. parasiticus for AFB 1 -related genes, it appears that like a growing number of other secondary metabolites, aflatoxin is a further example of a natural product whose biosynthetic genes are clustered. These workers have identified the organization of several demonstrated and pre-5 J. C. Silva, and C. A. Townsend, manuscript in preparation.  Alcohol oxidase (methanol oxidase) 120 2.0e-30 (6)  spԽQ00922ԽALOX_CANBO Alcohol oxidase (methanol oxidase) 108 1.9e-29 (6) spԽP13006ԽGOX_ASPNG Glucose oxidase precursor 97 2.2e-29 (6) gbԽI11354Խ 97 2.2e-29 (6) pirԽS32156ԽS32156 Mandelonitrile lyase 113 1.5e-23 (6) sumed AFB 1 genes: pksA, nor-1, fas-1A, fas-2A, aflR, aad, ord1, ord2, and omtA, although only a few of these have been well characterized. The data presented in this paper further define the extent of clustering of the aflatoxin B 1 biosynthetic genes. Southern analysis and restriction mapping of the vbs gDNA clones resulted in the discovery that omtA, a later gene in the aflatoxin B 1 biosynthetic pathway involved in the S-adenosylmethionine-dependent formation of 11, was located within one of the vbs gDNA clones (Fig. 5) (J. Yu, 1993, GenBank accession number L25834). We have demonstrated that vbs and omtA are within 3.3 kb of each other by PCR and Southern analysis (Fig. 5). We have also identified a probable cytochrome P450 gene between vbs and omtA. These results link the earlier genes of the biosynthetic pathway to the later genes to form an enlarged and apparently contiguous gene cluster responsible for the biosynthesis aflatoxin B 1 and unambiguously defines the locus of vbs.