3-Amino-5-hydroxybenzoic Acid Synthase, the Terminal Enzyme in the Formation of the Precursor of mC7N Units in Rifamycin and Related Antibiotics*

The biosynthesis of ansamycin antibiotics, like rifamycin B, involves formation of 3-amino-5-hydroxybenzoic acid (AHBA) by a novel variant of the shikimate pathway. AHBA then serves as the starter unit for the assembly of a polyketide which eventually links back to the amino group of AHBA to form the macrolactam ring. The terminal enzyme of AHBA formation, which catalyzes the aromatization of 5-deoxy-5-amino-3-dehydroshikimic acid, has been purified to homogeneity from Amycolatopsis mediterranei, the encoding gene has been cloned, sequenced, and overexpressed in Escherichia coli. The recombinant enzyme, a (His)6 fusion protein, as well as the native one, are dimers containing one molecule of pyridoxal phosphate per subunit. Mechanistic studies showed that the enzyme-bound pyridoxal phosphate forms a Schiff’s base with the amino group of 5-deoxy-5-amino-3-dehydroshikimic acid and catalyzes both an α,β-dehydration and a stereospecific 1,4-enolization of the substrate. Inactivation of the gene encoding AHBA synthase in theA. mediterranei genome results in loss of rifamycin formation; production of the antibiotic is restored when the mutant is supplemented with AHBA.

AHBA Synthase Assay-AminoDHS solution (500 l, 0.6 mM) and buffer B were mixed with a portion of the protein solution (50 -300 l) to give a total volume of 1 ml. After 1 h of incubation at 28°C, the reaction was stopped by the addition of 200 l of 15% trichloroacetic acid solution. After centrifugation, the production of AHBA was assessed by measuring the increase in A 296 relative to a blank. Protein concentration was determined with Bio-Rad protein assay solution, using bovine serum albumin as a standard.
To determine the kinetic parameters of the native or recombinant enzyme, initial velocities were measured by incubating 50 or 100 l of enzyme solution at 33°C with 0.1-1.1 mM aminoDHS in buffer C in a total volume of 1 ml and following the absorbance at 310 nm. The reaction was linear for at least 10 min.
Ammonium Sulfate Fractionation-Wet cells of A. mediterranei (25 g) were suspended in 100 ml of buffer B and disrupted by two passages through a French pressure cell. The cellular homogenate was centrifuged at 39,000 ϫ g for 20 min at 4°C. The supernatant was adjusted to 50% ammonium sulfate saturation by the slow addition, with stirring on ice, of powdered (NH 4 ) 2 SO 4 , followed by stirring for an additional 30 min. The precipitate was discarded. The (NH 4 ) 2 SO 4 concentration was then raised to 70%, and the protein pellet was dissolved in a minimal amount of buffer C and dialyzed for 4, 10, and 10 h against three changes of 40 volumes of the same buffer.
Anion Exchange Chromatography-Preswollen DE 52 resin was equilibrated in buffer C and packed at a bed height of 8 cm (bed volume of 40 ml) following the manufacturer's protocol. The dissolved 50 -70% ammonium sulfate precipitate (15-20 ml, 300 -350 mg of protein) was applied to the column at a flow rate of 48 ml/h. After the sample had been loaded, the column was washed until the A 280 was constant. A linear gradient of 50 -350 mM KCl in buffer C (total volume of 240 ml) was applied, and fractions of 4 ml were collected. The fractions containing the AHBA synthase activity (fractions [31][32][33][34][35][36][37][38] were pooled (28 -32 ml) and dialyzed overnight against buffer C.
Hydrophobic Interaction Chromatography-A Phenyl-Sepharose CL-4B suspension in buffer C containing 30% ammonium sulfate was packed in a 2.5 cm ϫ 10-cm column to a bed height of 8 cm (bed volume 40 ml). The sample (28 -32 ml, 90 -100 mg of protein) from DE 52 chromatography in buffer C 30% saturated with ammonium sulfate was applied to the column at a flow rate of 48 ml/h. The column was then washed with 40 ml of buffer C 30% saturated with ammonium sulfate, followed by elution of AHBA synthase with a linear gradient of 30 -0% ammonium sulfate in buffer C (total volume 160 ml). Fractions of 2 ml were collected, and the fractions containing AHBA synthase (fractions [47][48][49][50][51][52][53][54][55] were pooled (14 -16 ml) and dialyzed for 4 and 10 h against two changes of 40 volumes of buffer C.
Gel Filtration-Sephadex G-200 resin in buffer C containing 150 mM KCl was packed in a column (2.5 cm ϫ 50 cm) at a flow rate of 10 ml/h (bed volume 220 ml). The enzyme solution from the Phenyl-Sepharose column, concentrated to a volume of 2 ml in buffer C containing 150 mM KCl, was applied to the column, which was then eluted with the same buffer at a flow rate of 10 ml/h. Fractions of 3 ml each were collected, and the fractions of high specific activity (fractions [43][44][45][46][47][48] were combined, concentrated to a volume of 2 ml, and dialyzed for 4 and 10 h against two changes of 40 volumes of buffer C. Anion Exchange FPLC on Mono Q-The AHBA synthase solution (8 mg of protein) from the gel filtration was loaded onto a Mono Q column (bed volume 1 ml). The column was washed with buffer C for 5 min at 0.5 ml/min and then eluted with a linear gradient of 50 -250 mM KCl in buffer C for 5 min at 1 ml/min. Fractions were collected every 2 min.
Step gradients were continued for 10 min with 250 mM KCl and for 15 min with 1 M KCl at a flow rate of 0.5 ml/min. The fractions containing high AHBA synthase activity, eluting just before 250 mM KCl, were pooled.
Hydrophobic Interaction FPLC-Ammonium sulfate was added to the AHBA synthase pool from the Mono Q FPLC to give 30% saturation. The solution (1.6 mg of protein) was applied to a Phenyl-Superose FPLC column (bed volume of 1 ml) at 0.5 ml/min. The column was flushed with buffer C containing 30% saturated ammonium sulfate for 3 min at 0.5 ml/min, followed by elution with a linear gradient of 30 -0% saturated ammonium sulfate in buffer C over 30 min. Fractions were collected every 2 min, and the fractions containing high AHBA synthase activity, eluting at 20 -22 min, were pooled. Protein from this step was subjected to SDS-PAGE, and the band corresponding to AHBA synthase was cut out and transferred electrophoretically to Immobilon PVDF transfer membrane (Millipore Corp.). The membrane was used for gas phase microsequencing at the protein analysis facility of the Department of Biochemistry, University of Washington.
DNA Sequencing and Sequence Analysis-DNA sequencing was performed using the Sequenase kit (U. S. Biochemical Corp.) and [␣-35 S]dCTP (NEN Life Science Products) according to the manufacturer's protocol. Sequencing reactions were analyzed on polyacrylamide gels (8% (v/v) acrylamide, 5% bisacrylamide, 8 M urea, 45 mM Tris borate, pH 8.0, and 1 mM EDTA). SK, KS, T 3 , or T 7 primers (Stratagene) for dsDNA sequencing reactions and M13 primer (Stratagene) for ssDNA sequencing reactions were used. ssDNA was prepared using M13K07 (Promega) following the manufacturer's protocol. DNA and protein sequences were analyzed using the University of Wisconsin Genetics Computer Group (UWGCG) program version 7.3 (10).
Inactivation of AHBA Synthase Gene-A primary disruption of the AHBA synthase gene was engineered using a marker-replacement suicide vector, pSKϪ/AHBA2. In this plasmid, the hygromycin resistance gene (hyg) of pIJ963 (11), which resides on a 1.75-kb BglII fragment, had been inserted into the 2.3-kb XhoI fragment of pSKϪ/AHBA1 ( Fig.  2A) at the only BglII site, which lies at the N-terminal part of the AHBA synthase gene, in an orientation such that transcription of the hyg gene would occur in the same direction as that of the AHBA synthase gene, leaving 1.2 and 1.1 kb of homologous DNA flanking this insertion to the left and right, respectively. Through electroporation (12), approximately 20 transformants of A. mediterranei S699 that were resistant to hygromycin were obtained per microgram of heat-treated denatured pSKϪ/AHBA2. Southern hybridization of the transformants (data not shown) demonstrated that they arose by the expected single crossover, either upstream, named HGF001, or downstream, named HGF002, of the AHBA synthase gene, in approximately equal numbers. HGF001 produced normal rifamycin B yields; however, all HGF002 recombinants showed delayed production and reduced yield of rifamycin B (approximately half the yield of wild-type A. mediterranei S699). One of the HGF001 strains was chosen for maintenance on synthetic agar medium (13) lacking hygromycin. After propagation through three subsequent generations, about 0.5-1% of the colonies showed sporulation 1-2 days earlier and lost the ability to produce rifamycin B. Southern hybridization with six random choices of these rifamycin B nonproducing colonies (HGF003-1 to -6) (data not shown) confirmed that pSKϪ/ AHBA2 had integrated into the chromosome in HGF001, and that all SCHEME I. Proposed pathway of formation of the mC 7 N unit in rifamycin.
six HGF003 strains had undergone the second crossover event to replace the endogenous AHBA synthase gene with one that was truncated with the insertion of the hygromycin resistance gene marker (Fig. 2B).
Expression of the AHBA Synthase Gene-pSKϪ/AHBA1, which contains the 2.3-kb XhoI fragment of A. mediterranei DNA carrying the AHBA synthase gene, was digested with EcoRI. The resulting 1.6-kb EcoRI fragment was resolved on 0.8% agarose gel and ligated into pRSET digested with EcoRI using T 4 DNA ligase. The ligation products were transformed into BL21(DE3)/pLysS. Several colonies of E. coli BL21(DE3)/pLysS/pRSET(AHBA) grown on LB agar plates containing carbenicillin (50 g/ml) and chloramphenicol (34 g/ml) were inoculated into 10 ml of LB medium containing the same antibiotics. After 5-6 h of growth at 37°C with shaking (280 rpm), 1 ml of these cultures was used to inoculate 100 ml of LB medium containing 1 M sorbitol, 2.5 mM betaine, 50 g/ml carbenicillin, and 34 g/ml chloramphenicol. These cultures were incubated at 30°C until the A 600 reached 1.0. Isopropyl-1-thio-␤-D-galactopyranoside was then added to a final concentration of 0.1 mM, and the expression level during further incubation was checked by SDS-PAGE and assay of AHBA synthase activity.
Purification of Recombinant AHBA Synthase-A 500-ml overnight culture of E. coli BL21(DE3)/pLysS/pRSET(AHBA) in LB medium containing 1 M sorbitol, 2.5 mM betaine, carbenicillin, and chloramphenicol (A 600 ϭ 1.7) was transferred to 5 liters of the same medium in a 10-liter fermentor. Fermentation was carried out at 30°C and 230 rpm with an aeration rate of 3 liters/min. At an A 600 of 1.2, isopropyl-1-thio-␤-Dgalactopyranoside was added to a final concentration of 0.1 mM. Aeration rate and stirrer speed were increased to 6 liters/min and 280 rpm, respectively. The cells were harvested after 24 h of growth (A 600 ϭ 1.7) and washed with buffer D.
The washed cells were suspended in 100 ml of buffer A. The cell suspension was passed through a French pressure cell twice and centrifuged. (NH 4 ) 2 SO 4 was added to the supernatant, and the 35-55% precipitate was collected, dissolved in buffer C, and dialyzed against the same buffer overnight. The dialyzed protein solution was loaded onto a column of DE52 anion exchange resin (bed volume 45 ml), pre-equilibrated with buffer C, at a flow rate of 48 ml/h. After washing with one bed volume of buffer C, the column was eluted with a gradient of 50 -350 mM KCl in buffer C (total 350 ml).
The fractions containing AHBA synthase were collected and resolved on nickel resin (Novagen) following the manufacturer's protocol. The nickel resin was packed into an FPLC column (bed volume of 7 ml) and washed first with three bed volumes of distilled water, second with five volumes of charge buffer (50 mM NiSO 4 ), and finally with three volumes of binding buffer (20 mM Tris-HCl, pH 8.0, containing 5 mM imidazole and 500 mM KCl). The protein solution was loaded onto the column at a flow rate of 1 ml/min, which was then washed with 10 volumes of binding buffer and 6 volumes of washing buffer (20 mM Tris-HCl, pH 8.0, containing 60 mM imidazole and 500 mM KCl). Bound protein was eluted with a linear gradient of 60 -100 mM imidazole over 30 min, with AHBA synthase eluting at about 80 mM imidazole.
ESI-MS Analysis of AHBA Synthase-To a 60-l sample of purified recombinant AHBA synthase (1 mg of protein/ml, Tris buffer, pH 7.5) were added at room temperature, in 15-min intervals, three portions of a few micrograms each of NaBH 4 . After 1 h total reaction time, the pH was adjusted to 7.0 by addition of a few microliters of trifluoroacetic acid. The sample as well as a 60-l sample of unreduced enzyme were desalted by HPLC on a Vydak C-4 column (Rainin) using a linear gradient of water/0.06% trifluoroacetic acid to acetonitrile/0.06% trifluoroacetic acid (1 ml/min over 60 min). The fractions containing the protein (30 -35 min) were pooled, lyophilized, and dissolved in 50 -60 l of 60% acetonitrile/40% water/0.06% trifluoroacetic acid. Samples of 20 l were then introduced into the mass spectrometer at a flow rate of 16 l/min and data were acquired, averaging over 15-20 scans/sample. Data processing involved the maximum entropy deconvolution (Max-Ent) procedure (14,15).
Tritiated NaBH 4 Reduction of AHBA Synthase-To 2 mg of purified recombinant AHBA synthase in 0.5 ml of Tris buffer, pH 7.5, was added aminoDHS (1.2 mM) and an excess of tritiated NaBH 4 (300 mCi/mmol). After 20 min, the reaction was quenched by addition of 6 N HCl to pH 2. The protein was removed by centrifugation, and the pH was adjusted to 8 with 1 N NaOH. The supernatant, which at this point contained about 23 Ci of tritium, was incubated with alkaline phosphatase and then subjected to preparative layer chromatography (silica gel, n-butanol/ water/acetic acid 2:1:1). The majority of the radioactivity coincided with authentic aminoSA (R F ϭ 0.43), the reduction product of the free substrate, but a broad band of radioactivity (ϳ5%) at a higher R F (0.55-0.7) was seen, whereas no radioactivity was evident at the position of authentic 5-deoxy-5-(N-pyridoxylamino)shikimic acid (N-pyridoxyl-aminoSA) (R F ϭ 0.30). The broad high R F band upon rechromatography separated into two bands, one cochromatographing with authentic Npyridoxyl-AHBA (R F ϭ 0.58) and the other (R F ϭ 0.67) unidentified, but not identical with authentic pyridoxine (R F ϭ 0.44).
A reference sample of N-pyridoxyl-AHBA was synthesized by mixing 7.6 mg of AHBA and 13.3 mg of PLP⅐H 2 O, each in 0.5 ml of 0.05 M phosphate buffer, pH 7. After 15 min, when according to TLC Schiff's base formation was complete, excess NaBH 4 was added and the mixture kept at room temperature for 1 h with occasional shaking. The product was isolated by ion exchange chromatography (Bio-Rad AG 1X8, formate, successive elution with 3 N HCOOH, 0.1 N HCl, and 1 N HCl) and preparative layer chromatography (silica gel, n-butanol/water/acetic acid 2:1:1, R F ϭ 0.26). The resulting N-phosphopyridoxyl-AHBA was then dephosphorylated by incubation with alkaline phosphatase and the N-pyridoxyl-AHBA purified by preparative layer chromatography in the same system. N-Pyridoxyl-aminoSA was prepared analogously from aminoSA and PLP.

RESULTS
Enzyme Purification-AHBA synthase was purified from 54h-old mycelia of A. mediterranei strain S699 as summarized in Table I. The enzyme activity was assayed by measuring the amount of AHBA formed after 1 h incubation of the protein with 0.3 mM aminoDHS at 28°C, followed by addition of trichloroacetic acid. In the early stages of purification, AHBA was quantitated by an inverse isotope dilution GC-MS assay (9), in the later stages by measuring the increase in A 296 , the absorption maximum of AHBA at an acidic pH. The six-step 180-fold purification (Table I) gave AHBA synthase of a specific activity of 72 units/mg protein in 5% overall yield. The protein at this stage was judged by SDS-PAGE to be homogeneous ( Fig. 1).
Properties-The native molecular mass of AHBA synthase was estimated by gel filtration as 74 kDa and by nondenaturing PAGE as slightly higher than that of bovine serum albumin (66,700 Da). Elution of the enzymatically active band from nondenaturing PAGE and reanalysis by SDS-PAGE gave a molecular mass of 39 kDa, suggesting that the native enzyme is a dimer. Kinetic parameters were determined using partially purified enzyme from the DE 52 column step, assaying activity by following the change in A 310 during the initial 10-min linear phase of the reaction. The enzyme is most active at 33°C and has a pH optimum of 7.5. AHBA synthase retained its activity over a broad range of temperature and pH. Over 84% of the maximum activity of AHBA synthase was maintained over a temperature range from 28 to 50°C and a pH range from 7.0 to 9.0. The K m value for aminoDHS was determined from Lineweaver-Burk plots as 0.164 mM.
c Cumulative yield is given in parentheses.

3-Amino-5-hydroxybenzoic Acid Synthase
3,4-dideoxy-4-amino-D-arabino-heptulosonic acid 7-phosphate (aminoDAHP, 1 mM), aminoSA (1 mM) or 3-deoxy-D-arabinoheptulosonic acid 7-phosphate (DAHP, 1 mM), none of the compounds except DAHP affected the enzyme activity, with DAHP showing 40 to 50% activation of the enzyme. It had been shown by others (16) that DAHP synthase from A. mediterranei is inhibited by rifamycin. Activation of AHBA synthase by DAHP provides another piece of evidence that the AHBA biosynthetic pathway is related to the shikimate pathway, and suggests some type of cross-regulation of the two pathways. Partial Amino Acid Sequence Analysis-To obtain partial amino acid sequence information for the construction of oligodeoxynucleotide primers for the cloning of the AHBA synthase gene, the protein from the Phenyl-Superose FPLC step was further purified by SDS-PAGE. The specific region containing the AHBA synthase band was cut out from the gel after visualization with Coomassie Blue and electrophoretically transferred to PVDF membrane (17,18). Gas-phase microsequencing of the intact protein revealed a 10-amino acid Nterminal sequence H 2 N-N-A-R-K-A-P-E-F-P-A (sequence 1). Internal amino acid sequences were determined after in situ CNBr cleavage (19,20) of the enzyme on the PVDF membrane. The cleavage products were separated by reverse-phase HPLC (21) on a C 4 column, and two peptides were chosen for amino acid sequence analysis. They yielded sequences of 26 and 16 amino acids, Cloning of AHBA Synthase Gene-Based on the above partial amino acid sequences of AHBA synthase, three degenerate oligodeoxynucleotides were designed (Table II), taking into account the preferred codon usage (22) of genes from organisms with GC-rich DNA, like A. mediterranei or Streptomyces (Ͼ90% G or C in the third base). These oligonucleotides were used as PCR primers to amplify a 500-bp (combination primers 1 and 2) and a 250-bp (combination primers 1 and 3) region of genomic DNA from A. mediterranei S699. Both PCR products hybridized strongly to the same bands of restriction-digested genomic DNA, establishing them to originate from the same region. Sequencing of the two PCR products revealed that the deduced amino acid sequence of the 257-bp PCR product contained the last two amino acids of the N-terminal peptide sequence and the first six amino acids of the internal peptide sequence 3, neither of which had been used in the construction of the primers. On the other hand, the deduced amino acid sequence of the 505-bp PCR product contained the first 13 amino acids of the internal peptide sequence 2, which had not been encoded by the primer, but did not have the N-terminal amino acid sequence corresponding to oligonucleotide 1 at the 5Ј end of the PCR product. Instead, the 505-bp PCR product also had the same sequence as the oligonucleotide 2 at the 5Ј end, suggesting that oligonucleotide 1 might bind nonspecifically within a region where the AHBA synthase gene is located. Furthermore, none of the deduced amino acid sequence corresponding to the peptide 3 was found in the 505-bp PCR product.
Two nondegenerate primers (primers 4 and 5, Table II) were then synthesized, representing base sequences near the 3Ј and 5Ј ends, respectively, of the 505-and 257-bp PCR products. PCR with these primers gave a ϳ700-bp DNA fragment; sequencing revealed a 717-bp base sequence, which included the coding region of internal peptide 3. This 717-bp PCR product was then used to isolate the complete AHBA synthase gene from a cosmid library of A. mediterranei S699 genomic DNA. The library was constructed from partially Sau3A-digested A. mediterranei S699 genomic DNA, which was cloned into cosmid vector pOJ446 (23) restriction digested with BamHI and HpaI. The titer of the library was 2 ϫ 10 5 colonies/g of DNA and the average insert size was 30 -40 kb. Screening with the 717-bp PCR probe led to the isolation of 6 colonies from 2000 colonies screened, which contained DNA hybridizing strongly to the probe. Southern hybridization analysis of restriction digests of the inserts from the isolated six colonies utilizing the 717-bp PCR product identified a 2.3-kb XhoI fragment carrying the AHBA synthase gene.
Sequencing and Sequence Analysis of the AHBA Synthase Gene-The 2.3-kb XhoI fragment was subcloned into XhoIdigested pSKϪ. Two white colonies were selected, in which the 2.3-kb XhoI fragment was present in two different orientations. A detailed restriction map was obtained ( Fig. 2A) using several restriction enzymes, including EcoRI, PstI, SmaI, SacI, and SalI. XhoI/EcoRI (ϳ800 bp), EcoRI/SacI (ϳ1 kb), and SmaI/ XhoI (ϳ700 bp) digests of the 2.3-kb fragment were subcloned into pSKϪ or M13mp18 and sequenced using pSKϪ, M13mp18, or sequence-specific primers. The nucleotide sequence of the AHBA synthase gene was determined in both directions, and the restriction sites used for subcloning were verified by determination as part of an overlapping sequence.
The nucleotide and deduced amino acid sequences of the 2.3-kb XhoI fragment, shown in Fig. 3 stream of the coding sequence of the AHBA synthase gene. The sequences 69 bp upstream of the initiation codon were found to average 58 mol% GϩC. The deduced amino acid sequence included the previously determined N-terminal and internal amino acid sequences of the enzyme.
Screening of protein sequence data bases with the deduced amino acid sequence of AHBA synthase using the FASTA and BLAST (28) programs showed homology to the products of a series of genes implicated (29) primarily in transamination (Fig. 4) or dehydration/deoxygenation (Fig. 5) reactions involved in deoxyhexose biosynthesis (cf. Refs. [42][43][44][45]. Of the products of these genes, at the time of our work, only the AscC protein had been studied in detail. AscC encodes an enzyme, E1, containing a pyridoxamine phosphate (PMP) cofactor and an iron-sulfur cluster, which catalyzes the 3-deoxygenation of CDP-4-keto-6-deoxy-D-glucose in the biosynthesis of ascarylose in Yersinia pseudotuberculosis (46 -48). Ascarylose is one of the antigenic determinants in the cell wall polysaccharides of Y. pseudotuberculosis (43). However, the sequence homology between AscC and the AHBA synthase gene does not include the iron-sulfur cluster motif in AscC. In view of this partial homology to a known PMP-containing enzyme, the amino acid sequence of AHBA synthase was compared with those of representative other pyridoxal-phosphate (PLP) and PMP enzymes (49 -51). The alignments revealed that AHBA synthase contains a typical PLP binding motif with a conserved aspartate (Asp-159) and the active site lysine (Lys-188), which presumably binds the cofactor as a Schiff's base (Fig. 4). In the PMPcontaining AscC and the deduced sequences of related gene products, this lysine is replaced by a histidine (Fig. 5). This led to the expectation that AHBA synthase is a PLP enzyme, although catalysis via a PMP cofactor would be mechanistically equally plausible.
Involvement of AHBA Synthase Gene in Rifamycin Biosynthesis-The isolation and characterization of the gene encoding AHBA synthase reported above sets the stage for the cloning of the rifamycin biosynthetic gene cluster based on the paradigm (52) that in Actinomycetes the genes encoding the biosynthesis of a given antibiotic are clustered in the genome of the producing organism. Thus, analysis of the DNA surrounding the AHBA synthase gene should reveal other rifamycin biosynthesis genes, provided that the AHBA synthase gene is indeed essential for rifamycin formation. To verify this point, we carried out a gene disruption experiment. An inactivated version of the AHBA synthase gene was constructed by insertion of a 1.7-kb BglII hyg-containing DNA fragment from pIJ963 (11) into the unique BglII site in pSKϪ/AHBA1 ( Fig. 2A). To reduce the restriction limitation and increase the integration efficiency, this suicide vector was denatured and then introduced into A. mediterranei S699 by electroporation. Successive selection first for single-crossover mutants (Hyg r ) and then for a second crossover gave a double-crossover mutant (Hyg r ) (Fig.  2B) in which the functional AHBA synthase gene had been replaced by the inactivated version. This mutant was unable to produce rifamycin B, the normal metabolite of the wild-type, but production was restored to wild-type levels by supplementation of the culture with AHBA. When the mutant culture was supplemented with [7-13 C]AHBA (89% 13 C), the resulting rifamycin B contained at least 81% 13 C, confirming the absence of endogenous AHBA synthesis in the mutant. Thus, the cloned AHBA synthase gene is indeed essential for rifamycin formation, justifying the expectation that it is part of the rifamycin biosynthesis gene cluster of A. mediterranei.
Expression of the AHBA Synthase Gene in E. coli-To establish whether AHBA synthase is indeed a PLP enzyme and to generate larger amounts of enzyme for mechanistic studies, the AHBA synthase gene was overexpressed in E. coli. An EcoRI restriction digest of pSKϪ/AHBA1 containing the 2.3-kb XhoI restriction fragment produced a 1.6-kb EcoRI fragment in which 23 nucleotides from the start of the coding region of the AHBA synthase gene had been deleted. The 1.6-kb EcoRI fragment was cloned into the expression vector pRSET for expres-FIG. 2. Genetic maps of the AHBA synthase gene. A, restriction map of the region of the A. mediterranei S699 chromosome and of pSKϪ/AHBA1, encompassing the AHBA synthase gene. The orientation and extent of the AHBA synthase gene were deduced from the nucleotide sequence (see Fig. 3). B, A. mediterranei HGF003 carries a mutated AHBA synthase gene into which has been inserted a 1.7-kb BglII fragment carrying the hygromycin resistance gene (hyg) through in vitro and in vivo gene replacement (see text). C, the expression of (His) 6 -AHBA synthase under the control of the T7 promoter in pRSE-T(AHBA). An EcoRI DNA fragment, from pSKϪ/AHBA1, which carried the N-terminally truncated AHBA synthase gene was ligated into EcoRI-digested pRSET B (Invitrogen). The T7 promoter (p T7 ), ribosome binding sites (RBS; underlined) and indicated N-terminal amino residues (in one-letter code) of the recombinant AHBA synthase were inherited from the vector pRSET B.

3-Amino-5-hydroxybenzoic Acid Synthase
PMP to the fermentation broth, however, did not lead to an increase in enzyme activity. The final AHBA synthase activity was 13 times higher than in cell-free extracts of A. mediterranei S699. The recombinant enzyme was purified 38-fold by ammonium sulfate precipitation, DE 52 chromatography and adsorption on a nickel column as summarized in Table III to give nearly homogeneous protein (Fig. 6) in 46% yield.
Properties of Recombinant AHBA Synthase-Although the recombinant enzyme was obtained as a fusion protein and lacked the first six N-terminal amino acids, it showed very similar behavior to the enzyme isolated from A. mediterranei. The native molecular mass of the recombinant AHBA synthase was determined as 80 kDa and that of the denatured protein as 44 kDa by gel-filtration and SDS-PAGE. These values are somewhat low, compared with the molecular mass of 46,101 calculated from the DNA sequence. Nevertheless, the recombinant enzyme, as the native one, is most likely a dimer with subunits of identical molecular mass. The K m value of the recombinant enzyme for aminoDHS was 0.133 mM, compared with 0.164 mM for the native enzyme. The temperature and pH optima, 37°C and 8.5, respectively, also differed only slightly or predictably from those of the native enzyme, 33°C and 7.5.
The recombinant enzyme displayed an absorbance maximum at 418 nm, typical of a PLP-lysine Schiff's base (53). Reduction with sodium borohydride abolished the enzymatic activity and led to a change in the absorbance maximum to 330 nm, characteristic of PMP derivatives. PLP forms an adduct with cysteine that absorbs at 330 nm, and this property was used to estimate the amount of PLP bound per mole of AHBA synthase holoenzyme (54). Solutions of AHBA synthase were mixed with 200 mM cysteine in 10 mM HCl, and the A 330 was measured. By comparison to standard solutions of PLP, the AHBA synthase was found to contain 0.6 Ϯ 0.05 mol of PLP/mol of enzyme subunit. The protein concentration was measured by both the Bradford method and by UV, using the absorption coefficient calculated ( The two conserved PLP-binding residues, an aspartate and a lysine, which presumably hydrogen-bond to N1 and form a Schiff's base with the PLP cofactor, respectively, are indicated by vertical arrows. Aligned proteins are from A. mediterranei (AHBA) (the present work), S. griseus (StrC: a L-glutamine-dependent scyllo-inosose aminotransferase; 32.0% identity; 38.1% similarity) ( (35), Streptomyces alboniger (Pur4: a putative aminotransferase; 28.7.0% identity; 35.8% similarity) (36), S. griseus (StsA: a putative L-alanine dependent N-amidino-3-keto-scyllo-inosamine aminotransferase; 31.7% identity; 39.8% similarity) (30). The values of identity and similarity refer to comparisons of the different proteins with AHBA.
trum of the mixture the two species were clearly resolved, confirming that the holoenzyme must be a symmetrical dimer in which each subunit carries one molecule of PLP.
Mechanistic Role of Pyridoxal Phosphate-To demonstrate a role of PLP in the catalytic mechanism of AHBA synthase, we examined the effect of an inhibitor. Gabaculine (5-amino-1,3cyclohexadienelcarboxylic acid), a naturally occurring amino acid isolated from Streptomyces toyocaenis (56), is an irreversible inhibitor of many PLP-requiring aminotransferases (57). Rinehart et al. (58) and Ganem (59) had, in fact, suggested that this compound might be an intermediate on the biosynthetic pathway to mC 7 N units. In view of the structural similarity of gabaculine and aminoDHS, we checked the effect of gabaculine on AHBA synthase. Incubation of the recombinant enzyme with gabaculine at various concentrations led to irreversible inactivation of the enzyme. The inactivated enzyme did not recover activity upon dialysis against buffer, buffer containing PLP, or buffer containing substrate. The irreversible inhibition of AHBA synthase by gabaculine was time-dependent and showed a biphasic pattern in which the initial, more rapid phase was followed by a second phase of slower decline in activity (data not shown). At an inhibitor concentration of 15 M, the t1 ⁄2 (the preincubation time required for loss of 50% of enzyme activity) was about 20 s at 33°C. The gabaculineinactivated enzyme showed a UV absorption maximum at 330 nm. These data clearly implicate the enzyme-bound PLP in the catalytic mechanism of the enzyme (Scheme II).
The involvement of a PLP Schiff's base in the catalytic process was further demonstrated by reduction of the enzyme-   substrate complex with tritiated sodium borohydride followed by denaturation of the enzyme, acidic hydrolysis of the products and treatment with alkaline phosphatase. The major tritiated product from the reduction of AHBA synthase preincubated with aminoDHS was identified by comparison with an independently synthesized unlabeled sample as 3-(N-pyridoxylamino)-5-hydroxybenzoic acid (N-pyridoxyl-AHBA) (Scheme III). A second, unidentified product was noted; this was not identical with an authentic reference sample of 5-deoxy-5-(Npyridoxylamino)shikimic acid. The isolation of tritiated N-pyridoxyl-AHBA provides strong support for the intermediacy of Schiff's bases between the bound PLP cofactor and the nitrogen of the substrate, intermediates, and product in the catalytic mechanism. Furthermore, the fact that reduction of the enzyme-substrate complex produced N-pyridoxyl-AHBA suggests that the intermediate Schiff's base(s) predominating at equilibrium must be relatively product-like, i.e. give reduction products at least capable of aromatizing to the AHBA derivative upon acid hydrolysis.
Stereospecificity of 6-Deprotonation-The mechanism of AHBA formation emerging from the above experiments involves dehydration of aminoDHS by a PLP-catalyzed ␣,␤-elimination (Scheme IV). A second necessary step is the loss of a proton from C-6 of aminoDHS in a 1,4-conjugate enolization. This deprotonation may be enzyme-catalyzed, and hence stereospecific, or it may occur nonenzymatically, e.g. by spontaneous aromatization of a dienone released from the enzyme after the PLP-catalyzed dehydration of aminoDHS. To probe this issue, we synthesized (6R)-and (6S)-[6-2 H 1 ]aminoDHS from (6R)-and (6S)-[6-2 H 1 ]shikimic acid (60) by the same reaction sequence used to prepare the unlabeled compound (9), i.e. conversion to aminoSA via the epoxide followed by enzymatic oxidation at C-3 with E. coli shikimate dehydrogenase. Incubation of these two substrates with AHBA synthase in a cellfree extract of A. mediterranei followed by derivatization and GC-MS analysis of the resulting AHBA (9) showed Ͼ80% enrichment of deuterium from the 6R and Ͻ5% enrichment from the 6S isomer. Hence, the 6-deprotonation of aminoDHS by AHBA synthase is stereospecific, and thus enzyme-catalyzed, and involves elimination of the pro-6S hydrogen. DISCUSSION Following the establishment of a possible pathway for the formation of AHBA, the precursor of the mC 7 N starter unit of ansamycin antibiotics, by cell-free experiments (9), we decided to focus on AHBA synthase as the first enzyme of this pathway to investigate in detail. Two considerations prompted this choice. 1) AHBA synthase was the only enzyme that seemed unique to this pathway; unlike the other presumed enzymes it has no equivalent in the normal shikimate pathway. Hence, it was expected that the gene encoding it would represent the most unique probe to isolate ansamycin and mitomycin biosynthetic gene clusters. 2) The chemical aromatization of amino-DHS by acid, base, or buffer treatment proceeds in the same way as that of DHS (61,62), i.e. it produces exclusively protocatechuic acid (63). Since AHBA synthase generates quantitatively AHBA from the same substrate, the enzyme must completely redirect the aromatization chemistry of aminoDHS; the mechanism by which it achieves this was not obvious.
The studies reported here have revealed this mechanism (Scheme IV) by showing that the enzyme employs PLP catalysis. The genetic evidence that the enzyme is a PMP or, more likely, PLP enzyme was corroborated by analysis of the recombinant protein which showed the presence of a PLP-Schiff's base. NaBH 4 reduction experiments confirmed the involvement of a Schiff's base between the cofactor and the amino group of the substrate and ruled out a possible alternative mechanism proceeding through a Schiff's base between a PMP form of the enzyme and the C-3 carbonyl group of the substrate. The reaction mechanism of this enzyme thus can be suggested to resemble those of PLP enzymes catalyzing ␣,␤-elimination reactions in amino acid metabolism, such as serine dehydratase or tryptophanase, i.e. the initial enzyme-substrate Schiff's base first undergoes deprotonation at the ␣-carbon to give a quinoid intermediate, which then ejects the electronegative ␤-substituent, possibly after protonation to make it a better leaving group (64). The other required part of the reaction, abstraction of a proton from C-6 in a vinylogous enolization, is also catalyzed by the enzyme, as evidenced by the fact that this deprotonation is stereospecific. However, it is not evident from the data whether this step occurs before or after the 4,5-dehydration of the substrate.
Interestingly, the enzyme fits the general paradigm of PLP enzymes that, with few exceptions, all the group (proton) transfer reactions take place on only one face of the enzyme-substrate complex (65,66). The proton at C-5, the OH-group at C-4, and the pro-S proton abstracted from C-6 all are located on the same face of the substrate molecule. Although this may well be fortuitous, it is tempting to suggest that it could reflect the operation of a proton recycling mechanism in which the same enzyme base situated, as in other PLP enzymes, on the Si face of the planar PLP-substrate complex successively mediates the deprotonation at C-5, the protonation of the leaving group 4-OH and the deprotonation at C-6 (65, 66). On the other hand, the enzyme may well only catalyze the deprotonations at C-5 and C-6; if the vinylogous enolization precedes abstraction of the C-5 proton, then the resonance energy gained by aromatization should provide enough driving force for the spontaneous ejection of the 4-OH, i.e. the PLP catalysis may be limited to the generation of the resonance-stabilized carbanion at C-5.
In this context, it is interesting to note that the sequence homology between AHBA synthase and the PLP enzymes catalyzing ␣,␤-elimination and ␤-replacement reactions is not particularly strong. Rather, AHBA synthase is most closely related to the deduced amino acid sequences of the products of a family of genes (they have been called the secondary metabolic aminotransferase or "SMAT" genes; Ref. 44), many of which are found SCHEME II. Proposed mechanism of inhibition of AHBA synthase by gabaculine. SCHEME III. Reduction of the AHBA synthase-substrate complex with tritiated sodium borohydride.
in antibiotic biosynthesis gene clusters, which contain a typical PLP or PMP binding motif and seem to be involved primarily in the formation of antibiotic deoxysugar moieties, but some possibly also in regulatory functions (Refs. 29 and 42; cf. Refs. [43][44][45]. As the alignments in Figs. 4 and 5 show, these genes neatly fall into two subgroups. One carries a modified PLP binding motif with two adjacent histidines in place of the conserved active site lysine, and most of them also encode a series of cysteines which represent an iron-sulfur cluster motif (Fig.  5). This group mostly seems to be involved in 3-deoxygenation reactions of deoxysugars; it includes ascC, the gene encoding the well characterized (46 -48) enzyme E1 from Y. pseudotuberculosis, which catalyzes the 3-deoxygenation reaction in the biosynthesis of ascarylose. Another member of this subgroup, graG, has recently been shown by inactivation experiments to be involved in the biosynthesis of the 2,3,6-trideoxysugar, Lrhodinose. 2 The second subgroup (Fig. 4) encodes a typical PLP binding site and lacks the iron-sulfur cluster motif. Its members have been suggested to be involved in transamination reactions leading to amino deoxysugars and aminocyclitols. Only recently has transaminase activity of one gene product from this family been demonstrated. The protein expressed from stsC from Streptomyces griseus was shown to catalyze the 2 D. Tornus and H. G. Floss, unpublished results.  (30); AHBA from A. mediterranei (the present work); BplF from B. pertussis (31); EryC1 from Saccharopolyspora erythraea (33); TylB from S. fradiae (34); DnrJ from S. peucetius (35); DegT from B. stearothermophilus (32). SCHEME IV. Proposed mechanism of AHBA formation from aminoDHS.

3-Amino-5-hydroxybenzoic Acid Synthase
transamination between glutamine and scyllo-inosose to give scyllo-inosamine, the precursor of the aminocyclitol moiety of streptomycin (30). As the phylogenetic tree in Fig. 7 illustrates, these two subgroups clearly represent two separate branches of this superfamily of genes which diverged some time ago. AHBA synthase not only shares the PLP, as opposed to the PMP binding motif with the proposed transaminases, but is in general much more closely related to the PLP than to the PMP subgroup, although the dehydration chemistry it catalyzes formally resembles part of the chemistry of E1 more than the chemistry of transamination. The membership of an enzyme like AHBA synthase in this subgroup of PLP genes/enzymes suggests that some other members of this subgroup may also encode enzymes that catalyze reactions other than transaminations.
The cloning of the AHBA synthase gene and the demonstration that this gene is essential for rifamycin biosynthesis provides the means for the isolation of gene clusters encoding the biosynthesis of rifamycin as well as other antibiotics biosynthesized from AHBA. Sequence analysis of cosmid clones of A. mediterranei DNA isolated with the AHBA synthase gene as a probe and extended by chromosome walking has revealed numerous other genes involved in rifamycin biosynthesis. Their sequence and functional analysis will be reported in forthcoming publications. 3 The AHBA synthase gene from A. mediterranei has also been used, in a collaboration with the group of E. Leistner, Bonn, to clone homologous genes from Streptomyces collinus Tü 1892 presumably involved in the biosynthesis of ansatrienin (mycotrienin) and naphthomycin, 4 and from Actinosynnema pretiosum, presumably involved in ansamitocin biosynthesis. 5 Similarly, D. Sherman and co-workers have used the AHBA synthase gene from A. mediterranei to clone the mitomycin biosynthetic gene cluster from S. lavendulae. 6 The AHBA synthase gene thus represents another useful addition to the growing inventory of strategic gene probes, which can serve to identify biosynthetic pathways for key natural product classes and to isolate the gene clusters encoding them.