Molecular Characterization of the Genes of Actinomycin Synthetase I and of a 4-Methyl-3-hydroxyanthranilic Acid Carrier Protein Involved in the Assembly of the Acylpeptide Chain of Actinomycin in Streptomyces *

Actinomycin synthetase I (ACMS I) activates 4-methyl-3-hydroxyanthranilic acid, the precursor of the chromophoric moiety of the actinomycin, as adenylate. The gene acmA of ACMS I was identified upstream of the genes acmB andacmC encoding the two peptide synthetases ACMS II and ACMS III, respectively, which assemble the pentapeptide lactone rings of the antibiotic. Sequence analysis and expression of acmA inStreptomyces lividans as enzymatically active hexa-His-fusion confirmed the acmA gene product to be ACMS I. An open reading frame of 234 base pairs (acmD), which encodes a 78-amino acid protein with similarity to various acyl carrier proteins, is located downstream of acmA. TheacmD gene was expressed in Escherichia coli as hexa-His-fusion protein (Acm acyl carrier protein (AcmACP)). ACMS I in the presence of ATP acylated the purified AcmACP with radioactivep-toluic acid, used as substrate in place of 4-MHA. Only 10% of the AcmACP from E. coli was acylated, suggesting insufficient modification with 4′-phosphopantetheine cofactor. Incubation of this AcmACP with a holo-ACP synthase and coenzyme A quantitatively established the holo-form of AcmACP. Enzyme assays in the presence of ACMS II showed that toluyl-AcmACP directly acylated the thioester-bound threonine on ACMS II. Thus, AcmACP is a 4-MHA carrier protein in the peptide chain initiation of actinomycin synthesis.

The assembly of the 4-MHA pentapeptide lactones proceeds in a nonribosomal mechanism (5,6). Two peptide synthetases of 280 and 480 kDa (ACMS II and ACMS III, respectively) were previously identified in actinomycin-producing Streptomyces chrysomallus carrying the five amino acid modules responsible for assembly of the pentapeptide lactone rings as indicated in Fig. 1 (7,8). ACMS II and III were shown to contain also the functions for epimerization of the valine in position 2 and for introduction of the methyl groups into sarcosine and N-methylvaline in positions 4 and 5, respectively (7)(8)(9).
4-MHA, the ultimate amino-terminal residue of the 4-MHA pentapeptide lactone sequence, is activated by a separate adenylating enzyme, ACMS I (10). It has a size of 45 kDa and can adenylate a variety of benzene carboxylic acids structurally related to 4-MHA, e.g. 4-methyl-3-hydroxybenzoic acid  or p-toluic acid (10,11). These compounds can replace 4-MHA in the 4-MHA pentapeptide lactone when fed to actinomycinproducing Streptomyces through competition with the endogenous 4-MHA, giving rise to new acyl pentapeptide lactones (12). ACMS I does not detectably bind its substrate as thioester, and therefore the corresponding binding domain would be located on an acceptor protein distinct from ACMS I. Incubation of peptide synthetase ACMS II purified from S. chrysomallus with threonine and p-toluyl-CoA-thioester or p-toluyl-pantetheine-thioester, both nonnatural substrates mimicking a 4Јphosphopantetheine thioester, resulted in the formation of covalently bound p-toluyl-threonine (8). ACMS II also catalyzed, albeit with lower efficiency than in the latter cases, formation of p-toluyl-threonine from threonine and chemically synthesized p-toluyl-adenylate. This suggested that ACMS II, which cannot activate 4-MHA, would possess a binding domain with a 4Ј-phosphopantetheine cofactor (ACP domain) as an acceptor site for 4-MHA (8).
Cloning of the genes of ACMS II and III (acmB and acmC) showed that they lie closely linked in tandem on the chromosome of S. chrysomallus (13). Their analysis revealed an organization of the ACMSs into two and three modules, respectively, necessary for pentapeptide lactone assembly. Inspection of the protein sequence of ACMS II, however, showed that the conserved motif of the 4Ј-phosphopantetheine cofactor attachment site occurred only twice, namely in the two amino acid modules of this protein. Thus, the third 4Ј-phosphopantetheine cofactor, which was initially postulated to lie in front of the first (threonine) module (8), is not located on ACMS II. From the previous work on acmB and acmC in the acm gene cluster we got evidence that the 5Ј-end of the gene acmA of ACMS I is located upstream of the gene acmB of ACMS II and points in the opposite direction (13). Here we show the cloning and expression of acmA and the analysis of the gene product. More importantly, we found directly downstream of the gene of ACMS I a small open reading frame (ORF) encoding a small ACP (acmD), which is the protein harboring the missing third 4Ј-phosphopantetheine cofactor required for condensation of 4-MHA with threonine on ACMS II.

EXPERIMENTAL PROCEDURES
Chemicals and Radiochemicals-4-MHA was synthesized as described (10 Strains and Cultures-Streptomyces lividans 1326 was from the John Innes strain collection (Norwich, United Kingdom). It was grown and kept on R2YE (14). Submerged growth was for up to 4 days at 30°C with shaking (250 rpm) in 300-ml Erlenmeyer flasks containing 100 ml of YEME medium (14). Flasks were equipped with steel springs as baffles. Escherichia coli strains for cloning were DH5␣ (15) and JM109 (16), and the strain for expression was M15, from Qiagen. They were grown according to standard protocols.
Plasmids, DNA Manipulations, and Cloning and Sequencing Procedures-Techniques for DNA isolation, manipulation, and transformation were as described by Sambrook et al. (17) and Hopwood et al. (14). The cosmid cosA1 carrying the ACMS gene cluster from S. chrysomallus ATCC 11523 was described previously (13). The plasmids used for subcloning fragments were pTZ18U (Amersham Pharmacia Biotech), pSP72 (Promega), pSL1180 (Amersham Pharmacia Biotech). Sequence determinations were performed with a Taq cycle sequencing kit (United States Biochemicals-Amersham Pharmacia Biotech sequencing kit US71001 or US78500) on plasmid DNA. Fragments were subcloned into pTZ18U. Sequence comparisons, multiple sequence alignments, and identity scores were computed with CLUSTAL W (18) or with the FASTA data base search results (19). The plasmid for expression of acmA and acmB in S. lividans was pIJ702 (20). For expression of acmA in E. coli, expression vector pQE30 (Qiagen) was used, and for expression of acmD in E. coli, pQE32 (Qiagen) was used. For PCRs, Vent DNA polymerase (BioLabs) was used according to the manufacturer's instructions.
Heterologous Expression of acmA-Expression of acmA in E. coli was as amino-terminal hexa-His-tagged fusion protein. The acmA was amplified with suitable restriction ends by PCR using cosmid cosA1 (13) as template. Forward and reverse primers were FACMA1 (5Ј-TAAGAG-GAAGCTGGATCCGCCGATAAATGGTG-3Ј) and RACMA2 (5Ј-TAG-GCGTGGATCCCGTCGACCGAGGTGAA-3Ј), respectively. The resulting 1.6 kilobase pair PCR fragment was digested with BamHI and SalI and ligated into pQE30. In this construct the amino-terminal end of ACMS I would change from MADK-to MRGSHHHHHHGSADK-. Transformation into E. coli JM109 yielded plasmid pACMA1. After restriction analysis and control sequencing on both ends of the fragment, pACMA1 was transformed into E. coli strain M15. Cultures of M15/pACMA1 (1.6 liters of 2ϫ YT medium, 100 g/ml ampicillin, 25 g/ml kanamycin) were grown at 30°C to an A 600 of 0.9 and then induced with 2 mM isopropylthiogalactoside. Cells were harvested after a further 14 h of incubation at 30°C.
For the expression of acmA as carboxyl-terminal hexa-His-tagged fusion protein in S. lividans, acmA was placed as a translational fusion in frame into the ATG start codon of the melC1 gene of Streptomyces plasmid pIJ702 (20), which is under control of the mel promoter. The ATG start codon of melC1 is contained in the unique SphI restriction site of the plasmid. The acmA was amplified with synthetic primers to generate a matching 5Ј-SphI site in its start codon and to create a hexa-His encoding sequence between the last codon and the stop codon. PCR was performed with primers FACEX2 (5Ј-GAGGGCATGCATAT-GGCCGATAAATGGTGGGGGGAA-3Ј) and RACEX2 (5Ј-GAAGATCTT-CAGTGGTGGTGGTGATGGTGCGAGGCCCCCTTGAGCTCAGCGGG-3Ј) using subclone pA1sub27 containing a 2.8-kilobase SalI fragment from cosA1 cloned into pTZ18U (Fig. 2). A 1.45-kilobase fragment was obtained, which, after cleavage with SphI and BglII, was ligated in pSP72 cleaved with SphI and BglII. After transformation in DH5␣, plasmid pACMA10 was obtained. Sequencing on both ends of the insert confirmed the correctness of the modified ends of acmA. pACMA10 was digested with SphI and BglII, and the excised 1.45-kilobase SphI/BglII fragment was isolated and ligated into SphI/BglII-cleaved pIJ702. Transformation into S. lividans resulted in plasmid pACMA11. The recombinant protein encoded by pACMA11 is changed at the amino terminus from MADK-to MHMADK-and at the carboxyl terminus from -AS to -ASHHHHHH. Microsequencing of the purified protein later confirmed the new amino-terminal sequence (not shown). For expression of acmA in S. lividans encoded by pACMA11 as hexa-Histagged fusion protein, the transformed strain was grown at 30°C (6 liters of YEME medium, 10 g/ml thiostrepton) for 4 days and then harvested by suction filtration.
Heterologous Expression of acmB-The acmB was expressed in S. lividans from plasmid pACM5 as described previously (13).
Heterologous expression of acmD (AcmACP)-The gene acmD encoding the 4-MHA carrier protein was expressed in E. coli as aminoterminal hexa-His-tagged fusion protein. The acmD gene was engineered by PCR with synthetic primers FACPI1 (5Ј-CCGCATGCTCTC-GAAGGACGACATCAGGGCGAT-3Ј) and RACPI1 (5Ј-CGAGATCTGT-CGTCGGGGCGGTCGCGGCGC-3Ј) using pA1sub27 as template. This generated a 276-base pair fragment with a SphI site (5Ј) and the BglII site (3Ј) necessary for ligation into plasmid pTZ18U cleaved with SphI and BamHI. After transformation into E. coli, the resultant plasmid was cleaved with SphI and KpnI, and the excised approximately 250base pair fragment was ligated into expression vector pQE32 cleaved with SphI/KpnI. Successive transformation into JM109 and in M15 yielded plasmid pACPI-Q1. Sequence analysis of the cloned fragment in pACPI-Q1 confirmed that in the encoded protein, the amino terminus is changed from MISK-to MRGSHHHHHHGIMLSK-. Cultures of M15/pACPI-Q1 (1.8 liters of 2ϫ YT medium, 100 g/ml ampicillin, 25 g/ml kanamycin) were grown at 30°C to an A 600 of 0.7 and then induced with 0.1 mM isopropylthiogalactoside. Cells were harvested after further growth for 14 h at 30°C.
Heterologous Expression of a Putative 4Ј-Phosphopantetheine Transferase Gene (ORF C, 761 base pairs) flanking the nosiheptide resistance gene of Streptomyces actuosus (21)-The gene was amplified py PCR with oligonucleotides A (5Ј-AGCGCGGAATCGACTGAGGATCCAT-GACGGCCCGACA-3Ј) and B (5Ј-TGTTGCTCTCAAGCTTGGTCAGAT-CACGA-3Ј). Template for PCR was a genomic 8.5-kilobase BamHI fragment in clone 202, kindly provided by H. G. Floss and colleagues. The PCR-generated fragment was digested with BamHI and HindIII at the sites introduced by the oligonucleotides A and B (underlined) and inserted in the E. coli expression plasmid pQE30 (Qiagen), resulting in plasmid pPAN-5. Expression of ORF C from pPAN-5 was performed as described for acmD.
Protein Purification-All operations were carried out at 0 -4°C.
Purification of Recombinant ACMS I Encoded by pACMA11 from S. lividans-About 35 g of freshly harvested mycelium of S. lividans carrying pACMA11 suspended in 70 ml of Buffer A (10% glycerol (w/v), 50 mM KPO 4 buffer (pH 7.0), 1 mM EDTA, 4 mM DTE, 1 mM PMSF, 1 mM benzamidine) was passed twice through a French press at 10,000 psi. After DNase I treatment (30 g/ml) and with gentle stirring for 90 min, the suspension was centrifuged (30 min at 25,000 ϫ g). The supernatant (60 ml) was passed through a DEAE-cellulose column (8 ϫ 3.5 cm) to which ACMS I does not bind at neutral pH. Fractions with enzymatic activity were pooled (240 ml). Addition of 28 ml of 1 M Tris-HCl, pH 8.0, shifted pH from 7.0 to 8.0, and the solution was applied to a nickel-chelate resin column (2.8 ϫ 1.6 cm). After washing the column with Buffer NW (see above), bound proteins were eluted with a 150-ml linear gradient of 0 -250 mM imidazole in Buffer NW. Fractions of 3.5 ml were collected, and the fractions with enzymatic activity was pooled. This pool was diluted with 5 volumes of 15% glycerol (w/v), 4 mM DTE, 1 mM EDTA, 1 mM PMSF, 1 mM benzamidine and applied onto an -aminohexyl-Sepharose 4B (Sigma) column (6.5 ϫ 1 cm) for enzyme concentration. After washing with Buffer C (15% glycerol (w/v), 50 mM Tris-HCl (pH 8.0), 1 mM EDTA, 4 mM DTE, 1 mM PMSF, 1 mM benzamidine), the bound protein was eluted with a step of 0.3 M NaCl in Buffer C. The protein (total volume, 6 ml) was purified in 1.5-ml portions to homogeneity by gel filtration on an Superdex TM 75 column (Amersham Pharmacia Biotech) previously equilibrated with Buffer B (15% glycerol (w/v), 100 mM Tris-HCl (pH 8.0), 1 mM EDTA, 4 mM DTE, 1 mM PMSF, 1 mM benzamidine). Protein with apparent homogeneity was in the rear half of the activity peak. The purified recombinant protein could be stored at Ϫ80°C for at least 6 months without detectable loss of activity.
Purification of 4-MHA Carrier Protein (AcmACP) Expressed in E. coli-Approximately 12 g of cell paste was suspended in 34 ml of Buffer NI (see above). After passage of the suspension through a French press at 10,000 psi, DNase I (25 g/ml) was added, and the suspension was left on ice for 15 min with gentle stirring. After centrifugation (15 min at 20,000 ϫ g), the supernatant was applied to a nickel-chelate column (3 ϫ 1.6 cm) equilibrated with Buffer NI. The column was washed with 15 ml of the same buffer and 80 ml of Buffer NW until no more protein eluted from the column. A 100-ml linear gradient of 0 -500 mM imidazole in Buffer NW afforded elution of hexa-His-tagged protein. The bulk of protein appeared at 100 -200 mM imidazole, and the corresponding fractions were pooled. 1 M DTE was added to give 4 mM final concentration, and 4.5-ml portions (3 mg protein) were gel filtrated on Ultrogel AcA54 (40 ϫ 25 cm column dimensions) in Buffer B (see above). After this step, enzyme appeared to be pure as judged from SDS-PAGE. Typically, about 4 mg of total protein were obtained by this procedure. Protein could be stored at Ϫ80°C for 4 weeks without loss of p-toluic binding activity in Buffer B.
Purification of Recombinant ACMS II Encoded by pACM5 from S. lividans-Partially purified recombinant ACMS II was prepared as described previously (13). Enzyme from the AcA34 gel filtration step was used.
Purification of Recombinant 4Ј-Phosphopantetheine Transferase (Holo-ACP Synthase) from S. actuosus Encoded by pPAN5-The resultant hexa-His-tagged fusion protein was purified analogous to AcmACP as described above.
Analytical Methods-SDS-PAGE was performed according to Laemmli (22). Protein determinations were done according to Bradford Sequencing of the region upstream of the gene of ACMS II (acmB) led to identification of acmA as indicated by an arrow. Downstream of acmA, partially overlapping its 3Ј-end, follows acmD, which encodes the 4-MHA carrier protein AcmACP. The SalI fragment in plasmid pA1sub27 was the template for PCR amplifications of acmA and acmD. (23). Antibodies against ACMS I from S. chrysomallus were raised in a rabbit by two administrations of 100-g portions of the purified enzyme (at a 1-month interval) (courtesy of Prof. F. J. Fehrenbach, Robert-Koch-Institut, Berlin, Germany). The serum was used without further purification in dilutions of 1:1000 to 1:3000 for Western blot analysis by semidry blot standard techniques (provided by Biometra and Schleicher & Schuell). Anti-rabbit-alkaline phosphatase conjugate (Sigma) served as second antibody for the purpose of nitroblue tetrazolium/5-bromo-4chloro-3-indolyl phosphate detection (both from Roche Molecular Biochemicals). Thin-layer chromatography of enzymatically formed acyl adenylates was performed as described (10).
The recombinant 4-MHA carrier protein (AcmACP) was tested for thioester formation with p-[1-14 C]toluic acid (used as 4-MHA substrate analogue) by measuring the formation of trichloroacetic acid-precipitable radioactivity. One unit of AcmACP was defined as the amount of enzyme that covalently binds 1 nmol of p-toluic acid in 30 min at 29°C. Standard assay contained 14 mM p-[1-14 C]toluic acid, 10 mM ATP, 45 mM MgCl 2 , 280 nM purified recombinant ACMS I and 100 l of AcmACP-containing protein fraction. Buffer was Buffer B (see above) and incubation was at 29°C for 30 min. After addition of 6% trichloroacetic acid and leaving on ice for 15 min, precipitated proteins were collected by suction filtration on membrane filters ME 25 (Schleicher & Schuell). After washing and drying, filter bound radioactivity was counted in a Packard 1600CA scintillation counter. To visualize charged AcmACP as labeled band in SDS-polyacrylamide gels, precipitated protein was resuspended in 70 l of of 15% (w/v) glycerol, 100 mM Tris-HCl (pH 8.0), 1 mM EDTA, 1 mM PMSF, and 1% SDS. 2 l of 40% sucrose-0.25% bromphenol blue was added. 5-l aliquots were usually counted by liquid scintillation counting, and 10-l portions were separated by SDS-PAGE (17.5% polyacrylamide). Gels were stained with Coomassie Brilliant Blue R250 (Serva) and, after drying, exposed to x-ray film (NIF100, Konica).
Measurement of activity of recombinant ACMS II was as described previously (13).
The assay to demonstrate the transfer of p-toluic acid from p-toluic acid-AcmACP-thioester to ACMS II-threonine thioester was the same as described above to measure charging of AcmACP but with the additional presence of ACMS II and its substrate threonine. After addition of 5% trichloroacetic acid, protein was allowed to precipitate overnight on ice. Control reactions involved omission of one component from the reaction in each case. Precipitated proteins were collected, washed twice with 7% trichloroacetic acid and once with 6 ml of 3% trichloroacetic acid, and finally resuspended in a 350-l solution of 15% (w/v) glycerol, 100 mM Tris-HCl (pH 8.0), 1 mM EDTA, 1 mM PMSF, and 1% SDS. 9 l of 40% sucrose-0.25% bromphenol blue was added. A 10-l aliquot was analyzed by liquid scintillation counting, and 8 l was applied onto a 5% SDS-polyacrylamide gel. Staining, drying, and autoradiography was as above.
Incubation of AcmACP with the holo-ACP synthase from S. actuosus was performed in a total volume of 600 l (Buffer B) at 30°C for 45 min in the presence of 28 M AcmACP, 1 M holo-ACP synthase, 290 nM ACMS I, 10 mM p-[1-14 C]toluic acid, 10 mM ATP, 20 mM MgCl 2 and 0.2 mM CoA. 80-l aliquots were precipitated with 2 ml of 7% trichloroacetic acid and 50 l of 3% bovine serum albumin at several time points.
The trichloroacetic acid-precipitable and filter bound radioactivity was measured as described above (filter GF92, Schleicher & Schuell).
Nucleotide Sequence Accession Number-The nucleotide sequences from S. chrysomallus obtained in this study has been assigned Gen-Bank TM accession number AF134587 for acmA and AF134588 for acmD.

RESULTS
Cloning and Analysis of the ACMS I Gene-The genes of the actinomycin synthetases II and III, acmB and acmC, are tandemly arranged on the S. chrysomallus chromosome (13). Analysis of the region upstream of acmB revealed, at a distance of 430 base pairs of noncoding DNA, the start of an ORF in opposite direction to acmB (Fig. 2). The 5Ј-end of this ORF encodes an amino acid sequence that is identical to the aminoterminal protein sequence of ACMS I, previously determined by microsequencing of the protein (13). From this it became evident that this ORF is the gene coding for ACMS I (acmA). Sequencing further into the 3Ј-direction revealed the complete ORF of a total length of 1419 base pairs, which would encode a protein of 472 amino acids, with a calculated molecular mass of 51.5 kDa and a pI of 7.36. This fits with the previous determined molecular mass of native ACMS I (52-53 kDa (10)) but is 6.5 kDa larger than the size of denaturated ACMS I estimated from SDS-PAGE (45 kDa) (11). The codon usage of acmA is typical for streptomycete genes showing a strong bias for GϩC-rich codons with a GϩC-contents of 90.4% at the third base in codons. The overall GϩC content of the gene is 71.4%.
Analysis of the deduced protein sequence revealed similarity (between 20 and 30% identity) with a number of aromatic and heteroaromatic carboxylic acid adenylating enzymes, acyl-CoA  ligases and the activation domains of various peptide synthetases (Table I). Among these are the aromatic carboxylic acidactivating enzymes EntE and YbtE from E. coli and Yersinia pestis, respectively (24 -26). Furthermore, the enzyme has sequence similarity (20% identity) to the hydroxypicolinic acidadenylating enzyme SnbA from Streptomyces pristinaespiralis and to coumarate CoA ligase (24% identity) from Petroselinum crispum, a member of the acyl-CoA ligases (27,28). The conservation in these sequences was always highest in the five so-called core regions characteristic for adenylating domains (29) (Table I), which leaves no doubt about the nature of the cloned gene. As predicted from the biochemical analyses of ACMS I obtained from S. chrysomallus (10,11), no 4Ј-phosphopantetheine attachment site was found in the enzyme sequence that distinguishes this enzyme from typical peptide synthetases, which bind their substrates as thioesters after activation as adenylate.
Expression of acmA-To further characterize the acmA gene product, the gene was expressed in E. coli as an amino-terminal hexa-His-tagged fusion protein. Western blotting of protein extracts of E. coli M15 carrying the expression construct pACMA1 (see under "Experimental Procedures") with antibodies directed against ACMS I from S. chrysomallus showed the presence of the expressed protein as a 45-kDa band in SDS-PAGE (not shown). Nickel-chelate chromatography of total ex-tract from 6 M guanidine hydrochloride-treated cells and in parallel to the soluble fraction of nondenatured cells revealed that by far, most of the protein was in the insoluble fraction. Testing the soluble fraction revealed enzymatic activity as judged from the ATP-pyrophosphate exchange dependent on 4-MHA, 4-MHB, or p-toluic acid at an appreciable activity. The enzyme was purified with reference to the purification of the original protein from S. chrysomallus (see under "Experimental Procedures"). However, this resulted in very low enzyme yields (final total yield, 0.2% at a specific activity of 1.6 nanokatal/mg of protein), which made production of the enzyme in E. coli not amenable for our purposes.
To address the production of active enzyme in higher yields, acmA was engineered as a translational fusion into the ATG start codon of the melC1 gene on plasmid pIJ702, and the resultant plasmid, pACMA11, was transformed into S. lividans. In contrast to the construct used in E. coli, here the enzyme carried the hexa-His tail at its carboxyl-terminal end. Analyzing total and cytosolic protein extracts from the transformed strain and the control strain harboring pIJ702 by Western blotting revealed the presence of ACMS I after expression of acmA in S. lividans (Fig. 3). Moreover, the protein was exclusively in the soluble fraction (not shown). Total activity based on the protein present in the starting material was measured by the 4-MHB dependent ATP-pyrophosphate ex- change and was found to be 3-fold higher than total enzyme activity in extracts of S. chrysomallus X2-18, an actinomycinhigh producer (11). The purification is shown in Table II. Remarkably, about 50% of total activity did not bind to nickelchelate matrix due to either masking of the hexa-His tail or its removal through proteolytic activities. Nevertheless, steps following such as adsorption to aminohexyl-Sepharose with subsequent elution with a salt step and gel filtration on Superdex TM 75 afforded purification of the enzyme to homogeneity with 12% yield and a specific activity of approximately 20 nanokatal/mg of protein. This specific activity is about 3-fold higher than estimated for the wild type enzyme, which apparently is due to the much shorter purification procedure of the recombinant protein (four versus eight steps). The recombinant enzyme was indistinguishable from the wild type enzyme with respect to acyl adenylate formation both from various benzene carboxylic acids and in the ATP-pyrophosphate exchange reaction dependent on these substrates (11). Likewise, the wild type enzyme, the recombinant ACMS I, did not catalyze the formation of a CoA thioester from any of the benzene carboxylic acids tested including 4-MHA.
Cloning and Sequencing of acmD, an ORF Transcriptionally Coupled to acmA-In the course of sequencing the ACMS I gene acmA, immediately downstream of acmA, an ORF was identified that overlaps with its ATG start the stop codon of acmA indicating transcriptional and translational coupling between the two genes (Fig. 2). This suggested a close functional link between the gene products of acmA and that ORF, which was named acmD. The acmD gene has a length of 236 base pairs, encoding a protein of 78 amino acids in length with a calculated molecular mass of 8691 Da and a pI of 3.83. Comparison of the deduced amino acid sequence of acmD with protein sequences in the data bank revealed similarity with various ACPs from bacteria involved in fatty acid and polyketide synthesis (Table III). These carry 4Ј-phosphopantetheine as prosthetic group attached to a conserved serine, which is also present in the acmD gene product (Table III). There was also similarity with DltC, the D-alanine carrier in lipoteichoic biosynthesis in Lactobacillus casei (30), the ACP domain of the entB gene product of the enterobactin synthesis system in E. coli (31,32), and the amino-terminal ACP extradomain of the yersiniabactin synthetase HWMP2, a peptide synthetase involved in the biosynthesis of the acyltripeptide yersiniabactin in Yersinia strains (33,34). The gene product of acmD appeared to be a suitable candidate for carrying the missing 4Ј-phosphopantetheine cofactor accepting 4-MHA in actinomycin biosynthesis and was therefore named AcmACP.
Expression of acmD as a Hexa-His-tagged Fusion Protein-For expression of acmD as hexa-His-tagged fusion protein the expression plasmid pACPI-Q1 (using vector pQE 32) was constructed. After transformation of E. coli strain M15 cultivation was performed at 30°C and induction was with low concentrations of isopropylthiogalactoside to allow gradual expression of the gene. SDS-PAGE analysis of protein extracts of the E. coli strain transformed with plasmid pACPI-Q1 revealed an abundant protein of approximately 10 kDa, which was missing in the control strain carrying pQE32 (Fig. 4, lanes 1 and 2, respectively). The purification of the protein was by nickel-chelate chromatography and Ultrogel AcA 54 gel filtration, which yielded an appreciable amount of pure protein (Fig. 4, lane 3). The total yield was 67%.
The ability of the protein to bind p-[ 14 C]toluic acid in the presence of ACMS I and ATP was measured by the formation of trichloroacetic acid-stable radioactivity. Covalent substrate binding was also used to demonstrate the charged protein as a labeled band in SDS-PAGE as described under "Experimental Procedures." Fig. 4, lanes 4 and 4Ј, shows that the protein was charged with p-[ 14 C]toluic acid when ACMS I and ATP were present. No radiolabeling of the protein was seen when ATP (Fig. 4, lanes 5 and 5Ј) or ACMS I (Fig. 4, lanes 3 and 3Ј) were absent in the reaction mixture. These findings indicate that the AcmACP does not activate p-toluic acid per se and was charged only in case of prior activation of p-toluic acid by ACMS I. An estimate for the extent of 4Ј-phosphopantetheinylation was obtained by comparing the amount of radioactive substrate covalently bound to the protein and the total amount of Ac-mACP protein present in the assays. In two AcmACP preparations after gene expression in E. coli, only 10% of the AcmACP was found to bind ACMS I activated p-toluic acid. This apparently points to a low extent of phosphopantetheinylation of AcmACP in the foreign host E. coli. To increase the amount of phosphopantetheinylated AcmACP (holo-AcmACP) and also to demonstrate that the p-toluic acid became bound to AcmACP via 4Ј-phosphopantetheine as prosthetic group, the acylation of AcmACP by ACMS I was performed in the additional presence of a holo-ACP synthase and CoA (35). We chose for these experiments the holo-ACP synthase from S. actuosus (21), which was expressed from its gene as a hexa-His-tagged fusion protein in E. coli as described under "Experimental Procedures." The holo-ACP synthase was added to the AcmACP/ ACMS I incubations in 1 M concentrations and incubated for different lengths of time with and without CoA. The data in Fig. 5 clearly show that with increasing length of incubation, the amount of p-toluic acid covalently bound to AcmACP increased until the final level of 95-100% mol of p-toluic acid/mol of AcmACP was reached. By contrast, in the absence of CoA the basal level of 10% remained constant during the whole 40 min of incubation, which was also the case when no holo-ACP synthase was present (the latter not shown). These data illustrate the 1:1:1 stoichiometry between the enzyme, its cofactor, and the substrate p-toluic acid. Moreover, varying the ACMS I concentration (from 20 to 450 nM) in short (2 and 4 min) incubations with AcmACP revealed direct proportionality between the amount of ACMS I added and p-toluyl-AcmACP formed (not shown).
p-Toluyl-AcmACP Is Substrate of ACMS II-Once we had identified the acmD gene product as 4-MHA binding protein, we set out to elucidate the possible interaction between the AcmACP and ACMS II in the transfer of the aromatic carboxylic acid to the threonine moiety (i.e. the first amino acid in the pentapeptide lactone ring) activated by ACMS II. In order to exclude possible contamination of ACMS II with AcmACP when isolated from actinomycin-producing S. chrysomallus, we used the recombinant ACMS II expressed in S. lividans (13). This ensured that all components used for the enzyme reaction were derived from an actinomycin-free background. Data in Fig. 6 (lanes 4 and 4Ј) show that upon incubation of ACMS II with AcmACP, ACMS I and their substrates threonine, p-[ 14 C]toluic acid, and ATP, the bands of both ACMS II and AcmACP were significantly labeled. When the same experiment was performed without ATP (Fig. 6, lanes 3 and 3Ј) or without the AcmACP (lanes 5 and 5Ј), labeling of neither protein band was observed. This indicates that binding of p-toluic acid is ATP-dependent and that labeling of ACMS II only occurs from the charged AcmACP. In the absence of threonine, ACMS II became only faintly labeled, possibly due to the presence of residual trace amounts of threonine that had been loaded to the enzyme previously in in vivo conditions. (Fig. 6, lanes 2 and 2Ј) This findings clearly show that the p-toluic acid bound to AcmACP directly acylates the threonine covalently bound to ACMS II. Lanes 6 and 6Ј in Fig. 6 show that when ACMS I is absent, neither band of AcmACP nor ACMS II are labeled, which clearly reveals that charging of both AcmACP and ACMS II is dependent on the prior activation of p-toluic acid as adenylate. DISCUSSION We have shown previously that the 4-MHA pentapeptide lactones, the penultimate precursors of the bicyclic actinomycins, are assembled by the two peptide synthetases, ACMS II and III. These enzymes from actinomycin-producing S. chrysomallus activate the five amino acids of the peptide lactone ring as adenylates and thioesters, whereas 4-MHA is activated by a separate 4-MHA adenylating enzyme, ACMS I. ACMS I does not bind its substrate as thioester as revealed by enzymatic testing and, as shown here, because of the absence of a 4Јphosphopantetheine attachment site in its sequence. Furthermore, ACMS I has no activity as an acyl-CoA ligase (11). It was shown previously that purified ACMS II binds threonine, the first amino acid of the pentapeptide chain, as thioester, which can be acylated with nonnatural substrates such as p-toluyl-CoA or p-toluyl-phosphopantetheine thioesters, yielding p-toluylthreonine (8). This suggested that 4-MHA (represented in these experiments by the structural analogue p-toluic acid) in natural conditions would be thioesterified to an unknown carrier, which most probably contained a 4Ј-phosphopantetheine cofac- tor (8). Moreover, purified ACMS II from S. chrysomallus reacted, when charged with threonine, with chemically synthesized p-toluic acid-adenylate under formation of p-toluylthreonine, suggesting that the 4Ј-phophopantetheine cofactor would be located on ACMS II. This view was supported by the fact that covalent binding of p-toluic acid to enzyme could be inhibited by sulfhydryl blockers, such as dibromopropanone, n-ethylmaleinimide, or iodoacetamide (8). Sequencing the ACMS II gene, however, revealed that ACMS II had only two 4Ј-phosphopantetheine attachment sites, each located in the ACP domain of the threonine and valine modules (13).
The data presented here clearly show that the postulated additional 4Ј-phosphopantetheine cofactor is located on a small 4-MHA carrier protein (AcmACP) encoded by acmD located downstream of the ACMS I gene acmA in the ACMS gene cluster (Fig. 2). The functional studies done with the purified protein expressed in the foreign host E. coli clearly reveal that AcmACP is specifically acylated by ACMS I with p-toluic acid (or 4-MHA). The acylated AcmACP then is used as substrate by ACMS II in the acylation of the covalently bound threonine. This is most probably with the assistance of the putative acyltransferase domain containing the HHIVMDAFG motif in the amino-terminal region of ACMS II found in front of the threonine module (Fig. 7). These results clearly assign the role as carrier in the transfer and thioesterification of 4-MHA to the acmD gene product and not to an ACP extradomain located on ACMS II.
The involvement of small ACPs and of ACP extradomains in peptide biosynthesis systems or related systems in bacteria has been realized only recently. A first example is the 8.9-kDa Dalanine carrier protein from Lactobacillus casei, which carries D-alanine through the cell membrane to membrane-associated lipoteichoic acid and, in cooperation with an acceptor, forms a D-alanine ester with lipoteichoic acid (30,36). In the biosynthesis of the trilactone, enterobactin, which is a trimer of 2,3-dihydroxybenzoyl-serine (2,3-DHB-serine), is assembled by EntE, a dihydroxybezoic acid adenylating enzyme, and EntF, a serine-activating peptide synthetase (25,37). Their roles are analogous to those of ACMS I and ACMS II in the assembly of 4-MHA-threonine (reviewed in Ref. 6). Recent work has revealed that EntB, the isochorismate lyase that catalyzes the last step of 2,3-DHB synthesis carries an ACP extradomain of 100 amino acids in length (32). This ACP domain was identified as binding domain for 2,3-DHB, which could be modified with 4Ј-phosphopantetheine cofactor in vitro by the holo-ACP synthase EntD (32). EntB was subsequently shown to be an essential compound in the synthesis of the whole enterobactin in a reconstituted cell-free system (33). The role of EntB thus is analogous to the role of AcmACP described here. Interestingly, the iron-chelating acyl tripeptide yersiniabactin contains salicylic acid at its amino-terminal end. Sequence analyses of the yersiniabactin gene cluster in Yersinia pestis (26) showed a gene encoding YbtE, the salicylic acid-activating enzyme, and two genes encoding the peptide synthetases, HMPW1 and HMPW2, responsible for yersiniabactin peptide assembly. HMPW2 activates cysteine, the first amino acid in the yersiniabactin biosynthesis sequence (38), and was shown to be the partner of YbtE in a fashion similar to EntE and EntF. Remarkably, the salicylyl-ACP domain in the yersiniabactin system is fused to the amino terminus of HMPW2, which transfers salicylic acid to the covalently bound cysteine with formation of salicylyl-cysteine. The same domain organization as in HMPW2 with respect to the ACP domain had initially been postulated by our group for ACMS II based on the biochemical analysis of this enzyme prepared from S. chrysomallus (8). However, 4-MHA peptide synthesis differs from these examples in that AcmACP is a separate small acyl carrier protein and not fused to any protein of the same pathway. Fusion of ACPs to other proteins of the same pathway may ensure their coexpression. In the case of actinomycin biosynthesis, fusion of the AcmACP to another enzyme of the pathway or an ACMS did not happen in S. chrysomallus. Apparently, cotranscription and cotranslation of acmA and acmD suffices to ensure the coordinate expression with each other and the other actinomycin biosynthesis genes. The separate acyl carrier proteins or extradomains in the biosynthesis systems of aromatic acylpeptides and peptide lactones represent the functional link between the adenylating enzymes and the large modular peptide synthetases, which consist of similar repeating units (modules) activating amino acids only. ACPs or ACP domains carrying the aromatic carboxylic acids can specifically acylate the activated substrates residing on a peptide synthetase. Remarkably, CoA is not used as carrier. This appears also to be the case in the formation of several macrocyclic polyketides, which contain aromatic or pseudoaromatic carbox-ylic starter units, such as shikimic acid or 3-amino-5-hydroxybenzoic acid in the biosynthesis of rapamycin and FK506 (39,40) or rifamycin (41), respectively. The corresponding polyketide synthases contain starter modules with a domain with similarity to acyl-CoA ligases and, in addition, an ACP domain. It may be concluded that the ACP domain is directly acylated by the CoA ligase domains without intermediation of the aryl CoA derivative. Formation of CoA thioesters of aromatic carboxylic acids in bacteria thus may be a route only in the catabolism of aromatic compounds.