Heterologous expression, isolation, and characterization of versicolorin B synthase from Aspergillus parasiticus. A key enzyme in the aflatoxin B1 biosynthetic pathway.

Aflatoxin B1 is a potent environmental carcinogen produced by certain strains of Aspergillus. Central to the biosynthesis of this mycotoxin is the reaction catalyzed by versicolorin B synthase (VBS) in which a racemic substrate, versiconal hemiacetal, is cyclized to an optically active product whose absolute configuration is crucial to the interaction of aflatoxin B1 with DNA. Attempted over-production of VBS in Escherichia coli led principally to protein aggregated into inclusion bodies but also small amounts of soluble but catalytically inactive enzyme. Comparisons to wild-type VBS by SDS-polyacrylamide gel electrophoresis and after N-glycosidase F treatment revealed that extensive glycosylation accounted for the mass discrepancy (7,000 ± 1,500 Da) between the native and bacterially expressed proteins. Several over-expression systems in Saccharomyces cerevisiae were surveyed in which one that incorporated a secretion signal was found most successful. VBS of indistinguishable mass on SDS-polyacrylamide gel electrophoresis and kinetic properties from the wild-type enzyme could be obtained in 50-100-fold greater amounts and whose catalytic behavior has been examined. The translated protein sequence of VBS showed three potential N-glycosylation sites (Asn-Xaa-Ser/Thr) consistent with the modifications observed above and unexpectedly revealed extensive homology to the ADP-binding region prominently conserved in the glucose-methanol-choline (GMC) family of flavoenzymes. Over-production of VBS in yeast marks the first aflatoxin biosynthetic enzyme to be so obtained and opens the way to direct study of the enzymology of this complex biosynthetic pathway.

The study of secondary metabolism has advanced rapidly in the last decade from whole-cell experiments to the purification and characterization of individual biosynthetic enzymes. In fewer instances the genes that encode not only these proteins but also all or part of the biosynthetic machinery required to synthesize a fully elaborated natural product have been detected and sequenced. Among these is aflatoxin B 1 (AFB 1 , 1 ; Scheme 1, 9), a potent environmental carcinogen produced by the common molds Aspergillus flavus, Aspergillus parasiticus, and Aspergillus nominus. Its oxidative activation and induction of mutations, notably in the p53 gene, has brought into sharp focus the relation of exposure to this mycotoxin and the incidence of human disease (1)(2)(3). AFB 1 is synthesized by a complex pathway involving many chemical steps requiring up to 20 different enzymes (4,5).
Contrary to earlier reports (6 -8), the genes in A. parasiticus responsible for AFB 1 formation appear to be clustered (9,10). Two of these have been localized by complementation of classically derived blocked mutants and confirmed by gene disruption (11,12). Similarly, an O-methyltransferase has been purified and its corresponding gene cloned (13). All three of these genes have been linked to within 45 kilobase pairs of each other (9). Homology data from the genomic DNA regions flanking these genes has revealed among other things a pair of yeastlike fatty acid synthase subunits (12) and a putative polyketide synthase (14). An important tool has been developed to help determine the function of these encoded proteins by gene disruption, allowing the insertion of DNA at specific genomic sites to generate rationally blocked mutants (15). Recent application of this method identified the pair of fatty acid synthase subunits identified above as composing a specialized synthase whose function is to prepare an unusual hexanoyl starter unit (Scheme 1, bold in 1) for further elaboration by the proposed polyketide synthase (10, 14 -16). Biochemical experiments have both foreshadowed (17) and confirmed these implications (18).
In this paper we report a yeast expression system for the over-production of proteins involved in AFB 1 biosynthesis. While inactivation of biosynthetic genes affords important information about the function of the encoded proteins in the study of natural product biosynthesis, these observations alone may not be sufficient to completely understand their biosynthetic role or the extent of their association, if any, required for catalysis. A case in point is VER-1, a classical A. parasiticus mutant that accumulates versicolorin A (7) (19). Complementation has identified the gene ver-1 as the site of lesion in this mutant (11). Insertional inactivation of this gene leads to the accumulation of versicolorin A (7), confirming this identification. The apparent function of VER-1 in the synthesis of demethylsterigmatocystin (8, R ϭ RЈ ϭ H) may be assumed from a 56% amino acid sequence identity to skytalone reductase, an enzyme well-documented to catalyze an aryl dehydroxylation in the biosynthesis of melanin (20). The conversion of anthra-quinone 7 to xanthone 8 involves ring cleavage, rearrangement, deoxygenation, and decarboxylation and very well may require several enzymes as has been presumed to be the case in A. nidulans (21). The degree to which these proteins interact to achieve catalysis remains to be established. In the event that 6-deoxyversicolorin A (7, 6-OH 3 6-H) is not incorporated into AFB 1 (9) strongly suggests that aryl dehydroxylation is not the first step in xanthone formation (contrary to the implications drawn from the previously described homology data) (20). Therefore, study of these individual proteins and, if necessary, in association with one another may be required to arrive at a complete understanding of their true catalytic roles in the pathway (9,18). The development and application of an overproduction system suitable for these studies is described for the first time for VBS, an enzyme central to the biosynthesis of AFB 1 (22).
VBS has recently been isolated from A. parasiticus as a dimer with an apparent monomeric molecular mass of 78 kDa as estimated by SDS-PAGE (22). In the absence of metals and cofactors, VBS catalyzes the dehydrative cyclization of racemic versiconal hemiacetal (6) to optically active (Ϫ)-versicolorin B (6) (22)(23)(24). In this reaction the intrinsic stereochemical lability of the substrate (the chiral center is benzylic and adjacent to a masked aldehyde) provides a mechanism of enantiomeric interconversion with a t1 ⁄2 of ϳ45 min (22). VBS efficiently binds the correct antipode of 5 (K m ϭ 1 M) and catalyzes tetrahydrobisfuran ring formation with concomitant loss of water. The absolute configuration of the tetrahydrobisfuran set in this reaction is critical to the DNA interactive properties of metabolically activated AFB 1 (9) ultimately produced (25)(26)(27).
The VBS gene has been cloned and sequenced from both cDNA and gDNA libraries of A. parasiticus and has been found to bear significant similarity at the transcribed amino acid level to several flavin-dependent oxidases and dehydrogenases, notably glucose oxidase from Aspergillus niger (38% identity and 58% similarity) (9). This was a surprising finding since VBS does not catalyze a redox reaction and does not bind FAD or FMN despite an apparent, albeit truncated, FAD binding domain (9). These observations have led us to over-produce this protein for the purpose of further mechanistic study.
Isolable amounts of the native enzyme are low, between 15 and 30 g of VBS per liter of A. parasiticus culture (22,28). Attempts to over-produce catalytically active VBS in Esche-richia coli were unsuccessful due to the formation of inclusion bodies and low amounts of soluble but inactive protein. Attempts to re-fold and reactivate the solubilized VBS failed. Reports of expression of glucose oxidase cloned from A. niger into Saccharomyces cerevisiae encouraged our efforts to use yeast as a host for the heterologous expression of versicolorin B synthase (29). S. cerevisiae has been used with increasing frequency for the secretion of heterologous proteins (30 -35) in expression vectors having various secretion signals to facilitate the active transport of the over-produced proteins (29, 36 -38). Secretion of an over-produced protein simplifies isolation and eliminates exposure to endogenous proteolytic enzymes allowing a greater yield and is thought to yield more uniform co-and post-translational modifications (39). This strategy has also been exploited with the development of "super-secreting" S. cerevisiae strains that provide the added advantage of a 5-250fold enhancement in protein accessibility compared with that of wild types (38, 40 -43). In this paper we report the purification and characterization of recombinant VBS from A. parasiticus over-produced in S. cerevisiae in a protocol of general utility to enable understanding of aflatoxin biosynthesis at the enzymic level.
CSPD ® development reagents for alkaline phosphatase, and biotinylated SDS-PAGE molecular weight markers were all purchased from Tropix (Bedford, MA). Western blots were performed using a Bio-Rad Trans-Blot Electroblotter and Tropix, polyvinylidene fluoride (Tropifluor ) membranes. Mouse monoclonal antibody, 12CA5, for the influenza hemagglutinin epitope tag, 3xf, was purchased from Boehringer Mannheim.
Construction of pT7-7VBS E. coli Expression Vector-The vbs gene was obtained from the vbs cDNA clone, c241e (9). A restriction site was introduced by PCR using a 33-bp oligonucleotide primer (5Ј-AAGAATT-TAATTTTTCATATGGGACGAAACTGG-3Ј) for the 5Ј-end of the gene and the plasmid-based T7 oligonucleotide primer (5Ј-GTAATACGACT-CACTATAGGGC-3Ј), which lies downstream of the poly(A) region at the 3Ј-end of the cDNA clone, c241e contained within the pBSIISK(Ϫ) vector. The N-terminal primer contained an NdeI restriction site (underlined nucleotides) which included the translation initiation codon of the vbs gene (bold and underlined nucleotides). The PCR reactions were carried out under normal conditions as described by Lundberg et al. (44) using 30 cycles of 1 min, 95°C denaturation step, a 30-s, 53°C annealing step, and a 3-min and 30-s, 74°C polymerization step. The vbs PCR product was tailed with dCTP using terminal deoxynucleotidyltransferase following the procedure outlined by Smith et al. (45) . The modified vbs PCR product was subcloned into dGTP-tailed PstI cut pBSIISK(Ϫ) to serve as the carrier vector for the recombinant vbs (45,46). The recombinant pBSIISK(Ϫ) vector, pBSIISK(GC)vbs, was digested with NdeI and XhoI to excise the vbs fragment, purified by electroelution as described by the manufacturer (Bio-Rad), and subcloned into the NdeI and SalI sites of pT7-7 (47,48) to give the E. coli vbs expression vector pT7-7vbs. The sequence of vbs in the pT7-7vbs expression vector was verified using either commercially available or custom-synthesized oligonucleotide primers.
Expression of E. coli-derived VBS-Cultures of E. coli K-38 cells (100 ml) containing both pGP1-2 and pT7-7vbs (47,48) were prepared by inoculating a single colony into 500 ml of LB medium (90 g/ml ampicillin and 50 g/ml kanamycin) and grown on a rotary shaker (200 rpm) at 30°C to ϳ3.2 ϫ 10 7 cells/ml (A 600 ϭ 0.40). Expression of VBS was induced by placing the flask of cells in a 42°C water bath for 20 min. Expression of VBS was continued after cooling the cells from the 42°C induction temperature to 25°C and allowing the cells to grow for 3-4 h on a rotary shaker at 200 rpm, at 25°C. The cells were harvested by centrifugation at 5,000 ϫ g for 10 min, frozen in liquid nitrogen, and stored at Ϫ80°C (48).
Partial Purification of E. coli-derived VBS Inclusion Bodies-Cells (1 g wet weight) were suspended in 10 ml of breaking buffer (50 mM potassium phosphate (pH 7.4), 20% glycerol, 1 mM EDTA, and 0.25 mM PMSF) and incubated for 30 min on ice. Acid-washed glass beads (ϳ2 g, 0.05-0.10 mm diameter) were added, and the cell suspension was vortexed five times for 1-min intervals and set on ice for 2-min intervals. The cell debris and inclusion bodies were collected by centrifugation (10,000 ϫ g) for 45 min and washed twice with breaking buffer supplemented with 0.25% Triton X-100 and 10 mM EDTA. The precipitate containing the inclusion bodies was solubilized for 12 h at room temperature in 10 ml of breaking buffer supplemented with 6 M urea. The solution was centrifuged (10,000 ϫ g) for 45 min, and then the supernatant was dialyzed stepwise (2 h) against low salt buffer (5 mM potassium phosphate (pH 7.4), 20% glycerol, 1 mM EDTA, and 0.25 mM PMSF) containing 3, 1.5, 0.75, and 0 M urea (49,50).
Construction of the S. cerevisiae Expression Vector, pCGSVBS-The secretion vector pCGSVBS ( Fig. 1) was constructed from the plasmid pCGS681 and carried the cDNA of the bgh gene immediately downstream of the SUC2 secretion signal (39). Using the source vector pCGS681 required removing the insert BGH cDNA fragment. This particular expression vector allows fusion of the gene of interest directly to the SUC2 secretion signal sequence using the available NcoI site in the vector immediately downstream of the signal sequence. However, the existence of a second NcoI site in the expression vector complicated its utility. The VBS secretion vector pCGSVBS was thus constructed by breaking down the source plasmid pCGS681 into two fragments because of the second NcoI site within the URA3 selectable marker. The plasmid pCGS681 was treated with EcoRI and SalI, and the resulting two fragments, named the TPI/BGH fragment and the pCGS fragment, were purified from low melting agarose (51). The TPI promoter-containing fragment TPI/BGH was subcloned into the EcoRI and SalI sites of pBSIISK(Ϫ), which contains no NcoI sites, to make pBSTPI/BGH. The linear YEp (yeast episomal plasmid) portion of the treated pCGS681, the pCGS fragment (with EcoRI and SalI overhangs), was purified with low melting agarose and used later to make the final construct. The carrier vector, pBSTPI/BGH, was treated with both NcoI and SalI to remove the BGH cDNA insert from the TPI promoter region. The pBSTPI fragment, containing the SUC2 signal sequence immediately downstream of the TPI promoter, was purified from the digestion mixture through a 1% low melting agarose gel and used for further cloning.
The vbs insert for the yeast expression vector was prepared from two fragments owing to the presence of two NcoI sites within the coding DNA. The first vbs fragment, the 5Ј-end, was produced by PCR, and the second vbs fragment, the 3Ј-end, was obtained as a restriction fragment from the cDNA clone, c241e (9).
A restriction site was introduced by PCR into the first vbs fragment, the 5Ј-end of the vbs gene, using a 42-bp oligonucleotide primer (5Ј-CCATGGGACGAAACTGGTTCCAGGTAACAGCTATGGCGGTGG-3Ј) as the upstream primer and an 18-bp internal vbs oligonucleotide primer, 31NC (5Ј-ACGGCAGCCATTTGTCCC-3Ј), as the downstream primer. The downstream primer was homologous to a region 606 bases downstream of the start ATG (underlined). The upstream primer was designed to engineer a single NcoI site at the translation initiation codon (ATG). This change was effected through incorporation of a cytosine just upstream of the start codon (bold) and through alteration of the nucleotide C29T (bold/underlined nucleotide) within the coding sequence to remove an NcoI site immediately downstream of the start codon but not change the amino acid encoded at this site. The PCR reaction was carried out under normal conditions as described by Lundberg et al. (44) using 30 cycles of 1 min, 30 s, 95°C denaturation step, a 1-min, 43°C annealing step, and a 2-min 74°C polymerization step. The PCR product, vbsC29T31NC (ϳ600 bp), was cut with EspI and purified using low melting agarose (51).
The 3Ј-end of the vbs gene was obtained from the vbs cDNA clone, c241e. The cDNA clone was cut with XhoI and was then treated with calf intestinal alkaline phosphatase. The linearized XhoI cut cDNA clone was further cut with EspI, and the 3Ј-end of the gene (ϳ1900 bp) was purified using low melting agarose (51).
The two purified vbs fragments were ligated using T4 DNA ligase and subsequently phosphorylated using T4 polynucleotide kinase. The kinase reaction mixture was ligated into the cut pBluescript vector carrying the TPI promoter region for the S. cerevisiae expression vector pBSTPI (NcoI/SalI) with T4 DNA ligase. The ligated product pBSTPI/ VBS was then cut with EcoRI and SalI allowing the TPI/VBS fragment (ϳ2900 bp) to be purified using low melting agarose (51). Finally, the secretion vector pCGSVBS ( Fig. 1) was completed by ligating the EcoRI/ SalI cut TPI/VBS fragment from pBSTPI/VBS to the EcoRI/SalI cut pCGS fragment isolated from the pCGS681 plasmid. The sequence of vbs in the pCGSVBS expression vector was verified using either commercially available or custom-synthesized oligonucleotide primers.
Construction of S. cerevisiae Expression Vector, pCGSVBS3xf-A NotI site was engineered into the cDNA clone c241e just upstream of the stop codon of vbs. Two PCR primers, NotITag1R (5Ј-GCGGCCGC-CCTGCCCAGCCATCATTTCAGC-3Ј) and NotITag2RE (5Ј-GCGGC-CGCTAGACCACGGTTGATCAGGTCGC-3Ј), were designed to incorporate a NotI site just upstream of the stop codon and an additional nucleotide (cytosine) upstream of the NotI site. These primers were used in separate PCR reactions with T3 and T7 plasmid primers to amplify two fragments of vbs from c241e. The two PCR products were ligated after treatment with NotI and then purified from low melting agarose. The ligated PCR product (VBS(C/NotI)) was then cut with EcoRI and XhoI for ligation into EcoRI/XhoI cut pUC19, to produce pUC19VBS(C/NotI). The plasmid pUC19VBS(C/NotI) was linearized with NotI and the 3xf NotI cassette (epitope tag) was ligated into the vector to create pUC19VBS3xf. The plasmid pUC19VBS3xf was linearized with KpnI and filled-in using Klenow fragment. The blunt-ended, linearized pUC19VBS3xf was then digested with EspI, and the 3Ј-end of the modified vbs gene fragment containing the epitope tag was purified from low melting agarose. The secretion vector pCGSVBS was linearized with KpnI and filled-in using Klenow fragment. The blunt-ended linearized pCGSVBS was then digested with EspI, and the YEp portion of pCGSVBS (EspI/blunt) vector was purified using low melting agarose. The vbs3xf fragment (EspI/blunt) was then ligated into the YEp portion of pCGSVBS (EspI/blunt) to create the epitope-tagged secretion expression vector, pCGSVBS3xf (Fig. 2).
Yeast Transformation and Growth-Yeast transformations were carried out according to Hinnen et al. (52) All transformants were obtained on (ϪURA) selective plates containing 3% glucose. Inoculae were grown in uracil-selective medium containing 3% glucose for 48 h and diluted 1:100 into uracil-selective medium containing 3% glucose and potassium phosphate/citric acid buffer or ammonium phosphate buffer for expression (53). Expression cultures were grown for 90 h on a rotary shaker (175 rpm) at various temperatures (22.5, 25.0, 27.5, and 30.0°C) and were harvested by centrifugation (5,000 ϫ g) at 4°C for 10 min. Secreted VBS activity was determined after diluting media 10-fold in 50 mM potassium phosphate (pH 7.4) with 20% glycerol and 1 mM EDTA.
Western Analysis of VBS3xf-For Western blot analysis of the epitope-tagged fusion protein VBS3xf, an aliquot of the expression medium (1.0 ml) was denatured in SDS Gel Loading Buffer (50 mM Tris (pH 6.8), 2% ␤-mercaptoethanol, 2% SDS, 10% glycerol, and 0.1% bromphenol blue) and subsequently analyzed by SDS-PAGE. Prestained and biotinylated molecular mass standards were simultaneously electrophoresed. The proteins were then transferred onto a polyvinylidene fluoride membrane and were detected using mouse monoclonal antibody 12CA5 for the influenza hemagglutinin epitope tag, 3xf (Figs. 6 and 7).
Purification of S. cerevisiae-derived VBS-All steps were carried out at 0 -4°C. The yeast expression vector encoding the suc2 signal sequence fused to the coding sequence of VBS from A. parasiticus under the constitutive control of the TPI promoter was used to express and secrete VBS in S. cerevisiae. Functional VBS was easily detected in the culture medium after 72-96 h of growth. The cells were removed by centrifugation, and the clarified medium was concentrated ϳ20-fold by ultrafiltration using the CH2-System from Amicon equipped with a 30,000 molecular mass cut-off spiral wound membrane. The sample was then ultradialyzed against 10 volumes of 50 mM potassium phosphate (pH 7.4) with 20% glycerol, 1 mM EDTA, and 0.25 mM PMSF. The resulting protein extract was further concentrated (to ϳ20 ml) using an Amicon stirred cell apparatus with a 30,000 molecular mass cut-off disc membrane. The extract was allowed to dialyze into 5 mM potassium phosphate (pH 7.4), 200 mM glycine, 20% glycerol, 1 mM EDTA, and 0.25 mM PMSF. Following extensive dialysis, the extract was clarified by centrifugation and resolved in a Rotofor ® IEF chamber in the presence of 2.5% Ampholine (Pharmacia Biotech Inc.), pH 4 -6 at Ϫ10°C. After 5 h at constant power (12 watts), 20 IEF fractions were collected and analyzed for VBS activity. The active fractions (between pH 4.4 Ϯ 0.1 and 4.9 Ϯ 0.1) were individually treated with 5.0 M NaCl to adjust the final salt concentration to 500 mM NaCl and then dialyzed into 50 mM potassium phosphate (pH 7.4) with 20% glycerol to remove the ampholites and stored at Ϫ20°C.
Mono Q FPLC-A Mono Q HR5/5 FPLC anion exchange column (5 ϫ 50 mm; Pharmacia) was equilibrated at room temperature with 50 mM potassium phosphate (pH 7.5) and 20% glycerol (0.5 ml/min). One ml of active Rotofor ® sample was loaded, and the column was rinsed with ϳ50 ml of buffer. A 2-h linear gradient (60 ml) was then initiated from 0 to 300 mM sodium chloride, and the eluent was monitored at ϭ 280 nm. VBS eluted between 135 and 145 mM NaCl (Table I and Fig. 4).
Determination of K m and V max for Yeast-derived VBS-From a solution of [ 3 H]versiconal hemiacetal dissolved in assay buffer (ϳ50 mM, 1% HMPA), serial dilutions were carried out to obtain six different substrate concentrations, each of which was determined radiochemically and found to be 71.6, 6.7, 2.9, 1.5, 1.2 and 1.0 M, respectively. These were incubated in quadruplicate for 12.5, 12, 11.5, 11, 10.5, and 10 min, respectively, with and without purified expressed VBS (6.0 g/ml). The results from these assays were averaged and calculated as M versicolorin B produced per min from incubations with enzyme over that without enzyme. The apparent K m and V max for expressed VBS were calculated using a Lineweaver-Burk analysis of the data (55) and found to be 6.5 Ϯ 0.6 M and 0.096 Ϯ 0.006 M/min, respectively; the k cat for expressed VBS was ϳ2.1 Ϯ 0.2 s Ϫ1 .
Evaluation of the Effects of Metals on Enzyme Activity-VBS (5 g/ml) was preincubated at 37°C for 20 min with metal ions at 1.0 mM in 0.05 ml of 10 mM potassium phosphate buffer (pH 7.0) containing 20% glycerol. Tubes containing only VBS and phosphate buffer were used as controls. After preincubating for 20 min, an equal volume of 30 M [ 3 H]versiconal hemiacetal (specific activity 9.3 ϫ 10 12 dpm/mol) in 20 mM potassium phosphate buffer with 20% glycerol was added. VBS activity was determined as described above. The effect of metal ions on VBS activity was reported as the amount versicolorin B produced relative to the control (Table II).
Chemical Modification of S. cerevisiae-derived VBS-Initial screening for cysteine active-site amino acid residues with heterologously expressed VBS was performed in the presence of 10 mM chemical modifying reagent (Table II) at 37°C in 50 mM Tris hydrochloride (pH 7.4), 20% glycerol, and 1 mM EDTA (56). Screening for the presence of acidic amino acid residues was monitored in the presence of 10 mM EDC in 50 mM MES (pH 5.0), 50 mM glycine ethyl ester, 20% glycerol, and 1 mM EDTA (56,57). Examination of the participation of histidine activesite amino acid residues was conducted in the presence of 10 mM DEPC in 50 mM Tris hydrochloride (pH 7.0), 20% glycerol, and 1 mM EDTA (56,58). Both EDC and DEPC chemical modifying reagents were added from 100 mM stock solutions. The chemical modifying reagents were preincubated with VBS (5 g/ml) for 30 min at 37°C before assaying for VBS activity. In protection experiments, the preincubation mixture contained radioinactive versiconal hemiacetal (10 M) in addition to the chemical modifying reagent, before 10-fold dilution in Assay Buffer ( VBS was inactivated with DEPC diluted with ethanol. The ethanol concentration in the reaction mixture never exceeded 1% (v/v). Concentration-dependent inactivation of VBS (3 g/ml) with DEPC was carried out in 50 mM potassium phosphate (pH 6.0) with 20% glycerol by preincubating at 25°C for 30 min with increasing concentrations of DEPC (1.0, 3.00, and 10.0 mM). Time-dependent inactivation of VBS with DEPC was carried out for 30 min at 25°C in the presence of 10 mM DEPC, whereas aliquots were removed at 3, 12.5, 25, and 30 min to test remaining VBS activity (Fig. 6). Concentration-dependent inactivation of VBS (3 g/ml) with EDC was carried out in 150 mM MOPS (pH 6.0) containing 50 mM glycine ethyl ester, 20% glycerol, and 1 mM EDTA by preincubating at 25°C for 30 min with increasing concentrations of EDC (1.0, 3.00, and 10.0 mM) (56,57). Time-dependent inactivation of VBS by EDC was carried out using 1 mM EDC hydrochloride, whereas aliquots were withdrawn at 3, 12.5, 25, and 30 min to test remaining VBS activity (Fig. 7).

RESULTS
Utilization of the pT7-7/pGP1-2 dual plasmid expression system (47,48) allowed the tightly regulated expression of inactive VBS in E. coli at levels between 30 and 40% of total E. coli protein. Although the levels of expressed protein were impressive, we were unable to obtain cell-free extracts with catalytically active VBS. Further analysis of the protein extracts revealed the substantial production of inclusion bodies and small amounts of soluble protein upon heat-shock induction. Attempts were made to denature and refold the aggregated protein by dialysis and elution through a Sephadex G-25 desalting column (49,62). While catalytically active protein was never obtained, despite several attempts, it was observed that the E. coli-derived VBS has an apparent molecular mass of ϳ70 kDa, similar to that predicted by translation of the gene sequence, but significantly less than the 78-kDa mass found for the wildtype enzyme (22). Three potential N-glycosylation sites had been noted earlier from the deduced protein sequence of VBS (Asn-Xaa-Ser/Thr) (9) suggesting that the discrepancy in molecular masses between bacterially expressed protein and native VBS owed to post-translational carbohydrate attachment. Wild-type VBS was treated with (PNGase F) to test this hypothesis and found to give a protein whose behavior on SDS-PAGE was indistinguishable from the E. coli-produced protein (see below and Fig. 2). An alternative host for heterologous expression of VBS which was capable of the apparently necessary post-translational modification was sought.
S. cerevisiae has been used as a host for the heterologous expression of eukaryotic genes including those from Aspergillus. Tatsumi and co-workers (33) have expressed functional alkaline protease in S. cerevisiae identical to the native protein using cDNA encoding prepro-alkaline protease from Aspergillus oryzae. Further evidence of its utility has been demonstrated by Hata and co-workers (31) who have successfully expressed glucoamylase from A. oryzae in an active form. Most promising among these, however, was a report by Frederick and co-workers (29) who were able to express glucose oxidase from A. niger in S. cerevisiae. This success was particularly relevant because we had previously determined that VBS was most homologous to glucose oxidase of the GMC oxidoreductases whose gene sequences are known (9). In addition, glucose oxidase shares many physical characteristics to those of VBS. 1) Both proteins are dimers of similar molecular weight; 2) both proteins have similar degrees of N-glycosylation; 3) both proteins have prominent amino acid identities in the highly conserved FAD binding motif characteristic of FAD-dependent oxidoreductases; 4) both proteins have a similar pI; and last, 5) both proteins come from the closely related Aspergillus species.
The plasmid pCGS681 is a 2-m-based S. cerevisiae plasmid carrying the URA3 auxotrophic selectable marker (37,39). It carries the ColE1 origin of replication from pBR322 and the ampicillin resistance marker for E. coli propagation of the plasmid. Expression is controlled by the constitutive S. cerevisiae TPI promoter, which has the suc2 (sucrase) secretion signal downstream of the promoter. Heterologous expression and secretion of VBS was accomplished by subcloning the coding cDNA of vbs into the YEp portion of the pCGS681 yeast expression vector to make pCGSVBS, thus fusing the SUC2 secretion signal sequence to the ATG-start of the versicolorin B synthase gene (Fig. 1).
In order to facilitate monitoring small amounts of heterologously expressed protein in S. cerevisiae, an epitope tag was engineered into the 3Ј-end of the VBS coding sequence just before the stop codon to make a second expression plasmid, pCGSVBS3xf (Fig. 2). The epitope tag 3xf was derived from the influenza hemagglutinin protein and possessed the repeating amino acid sequence of YPYDVPDYA. The monoclonal antibody 12CA5 recognizes this primary amino acid sequence and was used for Western analysis of the protein derived from the yeast expression plasmid pCGSVBS3xf. The epitope tag was subcloned into the VBS sequence of the expression plasmid as an inframe 119-bp NotI cassette (Fig. 2). The fragment of DNA (3xf) provided the coding DNA sequence for the epitope tag, repeated three times. Western analysis of the epitope-tagged expression system revealed the constitutive expression of VBS (Fig. 3). However, VBS activity was not obtained from the epitope-tagged expression system, presumably due to C-terminal modification of the protein. However, catalytic activity was realized using the nonepitope-tagged expression vector pCGS-VBS (Fig. 1).
Heterologous VBS secretion was measured from a variety of yeast strains to select one that optimally produced VBS. Together with three wild-type yeast strains, four super-secreting mutants were tested (39,40). It was found that the supersecreting mutant CGY2998 (39) produced approximately 25fold more VBS than any of the wild-type strains of S. cerevisiae. This significant increase in VBS production made the detection of active VBS in the crude expression shake flasks noticeably easier and was the first step toward successfully isolating expressed VBS from S. cerevisiae.
In unbuffered shake flask cultures no active VBS was isolable due to the natural pH decrease in growing S. cerevisiae cultures. To minimize this effect, two separate buffered (ϪURA) dropout media were developed which also allowed determination of the optimal pH for heterologous secretion of VBS. Over 90 h of expression the pH of both buffered media was stable to within 0.4 pH units. It was observed that the optimal pH for expression was 6.5 and that there was no significant difference in active protein expression using either of the two buffering systems. There was approximately a 2-fold increase in expression levels of active VBS when the expression of VBS was carried out at pH 6.5 versus pH 5.75-6.25.
The effect of growth temperature on VBS secretion was also investigated in shake flask experiments. When normalized for activity (units/g (lyophilized dry cell weight)), expression conditions were optimized at 25°C. The increase in yield was approximately 30% compared with either 22.5 or 27.5°C. Utilizing the super-secreting S. cerevisiae strains under the optimal conditions for VBS production (25°C and pH 6.5), we have been able to produce 2-8 mg of active VBS per liter of culture.
Purification of the S. cerevisiae-derived VBS was carried out employing ultrafiltration followed by preparative isoelectric focusing (IEF) and Mono-Q anion exchange FPLC (Fig. 4 and Table I). The generated protein extract of the medium was concentrated and filtered through a 30,000 molecular mass cut-off spiral wound membrane before being ultradialyzed. The resulting protein extract was prepared for isoelectric focusing by further concentration before subsequent dialysis. The yeastderived VBS was efficiently separated from the majority of the contaminating secreted yeast proteins. In an electric potential the majority of these proteins precipitated between pH 5.0 and 7.0. IEF experiments indicated that the yeast-derived VBS possessed a pI of 4.6 Ϯ 0.1 (native pI ϭ 4.7 Ϯ 0.1) (22). Homogeneous yeast-derived VBS was afforded by Mono-Q FPLC anion exchange chromatography. After the column was equilibrated in 50 mM potassium phosphate with 20% glycerol (pH 7.5), a linear gradient from 0 to 300 mM NaCl gave the active protein at a chloride concentration between 135 and 145 mM. An average of 1.2 mg of heterologously expressed VBS was obtained per liter of yeast expression culture (50 g of cells/liter, wet mass). The purification of VBS was carried out in two steps with an overall 185-fold purification ( Fig. 4 and Table I). Homogeneity was determined by visualization of a single 78,000 Ϯ 1,500-Da protein band in a Coomassie Blue-stained SDS-PAGE gel, experimentally indistinguishable from the wild-type enzyme (Fig. 5).
Treatment of both native and yeast-derived VBS with PN-Gase F resulted in deglycosylated proteins of indistinguishable molecular mass as determined by SDS-PAGE (70,000 Ϯ 1,500 Da). The molecular mass shift of 7,000 Ϯ 1,500 Da is believed to reflect similar patterns of glycosylation. This shift in apparent molecular mass is not seen when E. coli-derived VBS is treated with PNGase F. In fact, the molecular mass observed with both PGNase-treated and nontreated E. coli-derived VBS appears to be that expected from the translated cDNA of vbs (ϳ70,000 Da) (9).
A radiochemical assay was developed to accurately monitor the production of versicolorin B (6) by VBS. The ability to incorporate acetate into AFB 1 (9) was exploited as an efficient means to obtain radiolabeled substrate. A simple method of separating the substrate and product from the VBS assay mixture was developed using a 1:1 mixture of diethyl ether and hexanes with 0.05% acetic acid (v:v) to selectively elute versicolorin B (6), whereas the considerably more polar unreacted versiconal hemiacetal (5) remained bound to the silica gel. As long as proper controls were simultaneously performed, the amount of enzymically derived product could be accurately distinguished from that resulting from trivially acid-catalyzed cyclization (22).  Analysis of the effect of changing substrate concentrations on initial velocities was used to determine the kinetic parameters of V max and K m for the yeast-derived VBS. The selection of substrate concentrations was made after preliminary experiments gave a rough estimate of the K m , and these were assayed in triplicate with dilute active solutions of purified yeast-derived VBS (6.0 g/ml). The initial velocities expressed as M/ min of product, versicolorin B, formation were analyzed using the Lineweaver-Burk equation (55). The V max obtained under these conditions was 0.096 Ϯ 0.006 M/min, and the K m , indicated by the racemic substrate concentration at half-maximal velocity, was found to be 6.5 Ϯ 0.6 M. The turnover number was estimated to be ϳ2.1 Ϯ 0.2 s Ϫ1 for each subunit of VBS. For native versicolorin B synthase these kinetic parameters were determined to be 0.14 Ϯ 0.01 M/min and 2.4 Ϯ 0.3 M for the V max and K m , respectively, and a k cat of 2.5 s Ϫ1 (22). Therefore, as was similarly observed for glucose oxidase, whatever differences in glycosylation pattern may exist between yeast-and A. parasiticus-derived proteins (29), they have little measurable effect on the kinetic behavior of VBS.

FIG. 3. Western blot of the yeast Invsc2 cell-free extracts containing the heterologously expressed, epitope-tagged VBS grown in Buffered Medium A (lanes A-F) and Buffered Medium B (lanes G-L) at various time points (see "Experimental
Based on the pH-rate profile of VBS (22), we designed chemical modification studies to provide further evidence for the identities of the VBS active-site amino acid residues. Initial studies utilizing N-ethylmaleimide, iodoacetamide, and O-methylisourea were not supportive of the involvement of cysteine as an active-site amino acid residue (Table II). While the effect of Hg 2ϩ on VBS activity suggested the involvement of cysteine in the active site, these results were not paralleled with the effects of Cu 2ϩ or any of the cysteine-specific chemical modifying reagents. Other metals were tested to determine whether the dehydrative cyclization required metal ion. The presence of metals did not seem to enhance the activity of VBS, ruling out this possibility.
While the concentrations required were high, further chemical modification studies with yeast-derived VBS gave tentative evidence of the involvement of both histidine and aspartate and/or glutamate (Figs. 6 and 7). Both in the presence of DEPC and EDC expressed VBS activity was diminished relative to control reactions. Protection studies employing unlabeled versiconal hemiacetal with DEPC and EDC support the view that the chemical modification was in fact occurring in or near the active site. Observations from the pH-rate analysis of the native enzyme (22) indicate the presence of two ionizable groups within the active site of VBS whose pK a values are 3.7 Ϯ 0.1 and 7.3 Ϯ 0.1 in agreement with the active-site amino acid residues suggested by the chemical modification data. Careful analysis of the pH/activity curve suggests one or two acidic residues play a role in catalysis consistent with the mechanistically similar enzymic glycosidation reactions in which these enzymes typically have one to two acidic amino acid residues at the active site to carry out the general acid-base chemistry of glycosidic bond formation (63). While these data must be interpreted with caution, one could visualize up to two acidic amino acid residues conducting the same type of catalysis and possibly the presence of a histidine residue (Fig. 8).
With further evidence for the identities of the active-site residues of versicolorin B synthase in hand, efforts were made to test simple substrate analogues to aid the design of future mechanism-based inactivators. Among those analogues examined ( Fig. 9) were 2-hydroxyphenylacetic acid lactone (10), 2,5dihydroxyphenylacetic acid lactone (11), emodin (12), versicol-  orin B (6), and the benzofurandiol 13, a close structural analogue of the right-hand half of the normal substrate. The compounds were rapidly surveyed for their relative extents of inhibition as a function of their concentration. Most of these had no effect on the cyclization reaction rate. Notably, emodin (12), containing the trihydroxylated anthraquinone analogous to the substrate, showed no inhibition (Table III), whereas the product of the reaction, versicolorin B (6), did. However, most revealing among these experiments 13, which mimics the Cring and side chain of the substrate 5, was found to decrease the efficiency of the normal reaction. More detailed investigation of the inhibitory behavior of 13 clearly indicated a competitive pattern with a K i of 96.0 Ϯ 18.0 M (Fig. 10).
Further insight into the nature of the active site was sought in the striking amino acid sequence similarity between VBS and glucose oxidase from A. niger, as well as other members of the GMC family (64) of flavin-dependent oxidoreductases (9). While the dehydrative cyclization catalyzed by VBS requires no redox cofactor, the structural similarity between the tricyclic anthraquinone of versiconal (5) and the flavin nuclei of FAD, FMN, and riboflavin was nonetheless intriguing. That glucose oxidase also binds the substrate, glucose, in the vicinity of the flavin compounded our interest as its size and composition was not unlike the side chain of versiconal. Thus, despite the more than 20-amino acid deletion evident from alignment of the apparent flavin-binding domains of VBS and glucose oxidase and other GMC flavoproteins, the thought was entertained that these protein structural modifications could direct this region to versiconal binding and the comparatively simple cat-alytic task of acid-promoted cyclization chemistry.
Inhibition experiments were conducted with FAD, FMN, riboflavin, and glucose, and, to our disappointment, none of these, even at quite high concentrations (for example, glucose at 100 mM), gave any detectable inhibition of the cyclization reaction. This outcome is consistent with the observations above made with emodin (13) and suggest that the anthraquinone nucleus is not of primary importance to substrate binding (a finding perhaps not unexpected as progress through the biosynthetic pathway involves six other anthraquinone-containing intermediates). DISCUSSION Versicolorin B synthase, VBS, plays a pivotal role in the biosynthesis of the mycotoxin aflatoxin B 1 (9) by catalyzing the cyclodehydration of (Ϯ)-versiconal hemiacetal (5) to (Ϫ)-versicolorin B (6) to establish the absolute configuration of the dihydrobisfuran ring system (22)(23)(24) essential to DNA interaction (25,65,66). Initial attempts to over-produce this enzyme in E. coli gave mainly inclusion bodies and small amounts of catalytically inactive protein. Recognizing by SDS-PAGE and treatment with PNGase F that ϳ10% of the apparent mass of the wild-type enzyme owed to post-translational glycosylation, over-production of VBS in S. cerevisiae was attempted. Of several systems surveyed, one incorporating a secretion signal gave the best yield of protein and whose molecular mass and kinetic properties were virtually identical to the wild-type en-  zyme (22). The gene encoding VBS has been cloned and sequenced from gDNA and cDNA libraries of A. parasiticus (9). The translated sequence of VBS gave three potential sites of N-glycosylation (Asn-Xaa-Ser/Thr) in keeping with the post-translational modifications observed above and shows striking similarities to the GMC family of flavin-dependent oxidoreductases (64). The extent of identity was greatest with glucose oxidase (35, 67) (GCG BESTFIT analysis: 38% identical, 58% similar) among the available protein sequences. A partially deglycosylated derivative of this enzyme has been crystallized (68), and its x-ray structure shows the presence of carbohydrate residues at the dimer interface. Like glucose oxidase, VBS forms a homodimer, and it may be suggested, but not proved, that glycosidation of the protein could be important to catalytic activity by promoting dimer formation or facilitating correct folding of the protein. In the event, it is curious that VBS, which does not catalyze a redox reaction, should show such high homology to the flavin binding domain of the GMC family of flavoproteins. In this connection, it is interesting to note that computerassisted (BESTFIT) sequence alignments also reveal 33% identity and 61% similarity to mandelonitrile lyase, which takes part in the enzymic degradation of cyanogenic glycosides (69). This enzyme catalyzes the decomposition of the cyanohydrin to benzaldehyde and hydrogen cyanide, a reaction, like that carried out by VBS, that does not involve oxidation/reduction. A number of mandelonitrile lyase enzymes have been isolated and studied (70). Interestingly, some require oxidized flavin (71) (FAD) for activity, whereas others do not (72). Unfortunately, only one primary protein sequence is currently available from this class, one of the FAD-requiring lyases (73). As more gene/protein sequences become available, it will be of interest to see whether those that do not bind flavin retain, like VBS, vestigial structural elements consistent with evolution from a flavin-binding ancestor. Of those mandelonitrile lyase enzymes that do bind flavin, it does not participate in catalysis (74), and its role is unclear. It has been proposed to be important to proper folding of the protein to achieve a catalytically active form (75).
The low pI and broad pH stability of both the wild-type and yeast-derived VBS have allowed its substantial purification by preparative isoelectric focusing in tandem with conventional ion exchange chromatography steps. Chemical modification experiments are consistent with the participation of carboxylic acid residues and a possible histidine in catalysis, an observation in accord with pH-rate profile data and isotopic labeling studies (76). The S. cerevisiae secreting system described has made possible 50 -100-fold improvement in the production of VBS compared with the wild-type A. parasiticus and confers advantages of ease of isolation and uniformity of post-translational glycosylation. This first example of eukaryotic over-expression of an individual biosynthetic gene opens the way to direct study of other enzymes of aflatoxin biosynthesis and will complement the information available from insertional inactivation (11,15) to generate blocked mutants of the pathway.