New Lessons for Combinatorial Biosynthesis from Myxobacteria

The biosynthetic mta gene cluster responsible for myxothiazol formation from the fruiting body forming myxobacterium Stigmatella aurantiaca DW4/3-1 was sequenced and analyzed. Myxothiazol, an inhibitor of the electron transport via the bc 1-complex of the respiratory chain, is biosynthesized by a unique combination of several polyketide synthases (PKS) and nonribosomal peptide synthetases (NRPS), which are activated by the 4′-phosphopantetheinyl transferase MtaA. Genomic replacement of a fragment of mtaB and insertion of a kanamycin resistance gene into mtaA both impaired myxothiazol synthesis. Genes mtaC and mtaDencode the enzymes for bis-thiazol(ine) formation and chain extension on one pure NRPS (MtaC) and on a unique combination of PKS and NRPS (MtaD). The genes mtaE and mtaF encode PKSs including peptide fragments with homology to methyltransferases. These methyltransferase modules are assumed to be necessary for the formation of the proposed methoxy- and β-methoxy-acrylate intermediates of myxothiazol biosynthesis. The last gene of the cluster,mtaG, again resembles a NRPS and provides insight into the mechanism of the formation of the terminal amide of myxothiazol. The carbon backbone of an amino acid added to the myxothiazol-acid is assumed to be removed via an unprecedented module with homology to monooxygenases within MtaG.

Myxobacteria are Gram-negative soil bacteria that are assigned to the two suborders Cystobacterineae and Sorangineae. Both belong to the ␦-group of the Proteobacteria (1). They are distinguished from most other bacteria by their ability to glide in swarms, to feed cooperatively, and to form fruiting bodies upon starvation (2,3). In addition, they have been shown to produce a wide variety of secondary metabolites with unique structures and biological activities (for reviews, see Refs. 4 and 5). These include the electron transport inhibitors myxothiazol (6), stigmatellin (7), and myxalamids (5,8) produced by different strains of Stigmatella aurantiaca (Cystobacterineae) and the epothilones produced by Sorangium cellulosum (Sorang-ineae) (9) (structures are given in Fig. 1). Due to their antitumor activity, epothilones have attracted great attention (10 -12). Myxothiazol as well as epothilones contain a thiazole ring that is formed by the incorporation of cysteine into the polyketide backbone (13). Thiazoline and thiazolidine structures of bacitracin in Bacillus licheniformis (14) and the bacterial siderophores yersiniabactin and mycobactin have recently been shown to be biosynthesized by a NRPS 1 or a combined PKS/NRPS in Yersinia pestis and Mycobacterium tuberculosis (15)(16)(17). Combinations of PKS and NRPS are known from other microorganisms (18 -20) with two examples very recently reported from myxobacteria (Myxococcus xanthus (Ref. 21) and Sorangium cellulosum (Ref. 22)). No such combinations have been published so far for the formation of a thiazole coupled to a polyketide structure. In addition to the bis-thiazole moiety, myxothiazol has some unique features: the unusual leucine derived starter unit 3-methyl-butyryl-CoA (13) and the linear polyketide backbone, which includes a ␤-methoxy-acrylate and a terminal amide structure.
Little is known about the biochemistry of the formation of myxobacterial compounds and their corresponding gene clusters. A PKS gene (23) and a NRPS gene (24) from myxobacteria were detected by hybridization studies using heterologous gene probes or by Tn5 mutagenesis. Fragments of these genes have been cloned and sequenced. In contrast a great variety of PKS and NRPS genes from Actinomycetes and fungi have been shown to be involved in the biosynthesis of many secondary metabolites (for a review, see Ref. 25). We have established a method for cloning myxobacterial PKS and NRPS genes. Using this approach, both the stigmatellin-and the myxalamid biosynthetic gene clusters of S. aurantiaca Sg a15 were identified (22). The frequent occurrence of secondary metabolites in myxobacteria that seem to be synthesized by a combined PKS/ NRPS is remarkable (e.g. myxothiazol, epothilones, myxalamids; compare Fig. 1). With probes for PKS and NRPS genes, we detected DNA fragments hybridizing with both gene types in a cosmid library of S. cellulosum So ce90 (22) that produces epothilones. In addition one ORF showing homology to PKS and NRPS involved in the biosynthesis of myxovirescin has been reported recently (21). This report deals with the isolation and characterization of a gene cluster from S. aurantiaca DW4/3-1 that is involved in myxothiazol biosynthesis. It comprises genes encoding PKSs and NRPSs, inactivation of which impairs myxothiazol biosynthesis. In close proximity to this cluster is located mtaA, which encodes a putative Ppant transferase. The gene product is shown by insertional mutagenesis to be necessary for myxothiazol formation. MtaA presumably catalyzes the transformation of the apo-to the holo-form of the combined myxothiazol-PKS/NRPS. The presence of genes of the Ppant transferase family in the neighborhood of PKS or NRPS genes has been observed in other bacteria (26,27). The mta gene cluster is located adjacent to the developmental genes fbfA and fbfB that are involved in fruiting body formation of S. aurantiaca DW4/ 3-1 (28,29). The PKS/NRPS part of the gene cluster has several unique features. These can well be correlated to the 3-methylbutyryl-, bis-thiazole-, ␤-methoxy-acrylate-, and terminal amide moiety of the compound.

MATERIALS AND METHODS
Bacterial Strains and Culture Conditions-Bacterial strains and plasmids are described in Table I. Escherichia coli strains and S. aurantiaca DW4/3-1 (30) and its descendants were cultured as described previously (29).
Analysis of Secondary Metabolite Production in S. aurantiaca DW4/ 3-1-For the production of secondary metabolites, the strains were cultivated in Zein liquid medium (1% Zein, 0.1% peptone from casein tryptically digested, 0.1% MgSO 4 ⅐7H 2 O, 50 mM HEPES buffer, pH 7.0, and 1% of the adsorber resin XAD-16 (Rohm & Haas)). 100-ml batch cultures in 250-ml Erlenmeyer flasks were incubated at 30°C on a gyratory shaker at 160 rpm for about 3 days. As cell mass and adsorber resin contained secondary metabolites after the fermentation, second-ary metabolites were extracted from both together twice with acetone. The spectrum of secondary metabolites produced was determined in aliquots of concentrated acetone extracts by diode-array-detection HPLC analysis using a Hewlett Packard 1090 series II instrument. The conditions for the chromatography of the extract from wild type and mutant BS57 were: column ET 125 ϫ 2 mm and precolumn, Nucleosil 120-3 C 18 ; solvent: 0.2% acetic acid (A)/acetonitrile (B) gradient, 50% B at 5 min to 80% B at 15 min; flow rate 0.5 ml/min; detection range 200 -400 nm. For extracts from mutant BS47 in comparison to wild type growth, sample preparation and analysis was performed as described above except for the usage of Nucleosil 120-5 HPLC columns.
DNA Manipulations, Analyses, Sequencing, and PCR-Chromosomal DNA from S. aurantiaca was prepared as described (31). Southern analysis of genomic DNA was performed using the standard protocol for homologous probes of the DIG DNA labeling and detection kit (Roche Molecular Biochemicals). PCR was carried out using Taq polymerase (Life Technologies, Inc.) according to the manufacturer's protocol. 5% Me 2 SO was added to the mixture. Conditions for amplification with the Eppendorf Mastercycler gradient were as follows: denaturation for 30 s at 95°C, annealing for 30 s at 60°C and extension for 45 s at 72°C; 30 cycles and a final extension at 72°C for 10 min. Primers used for screening the cosmid library were: SEQBS24 -4, 5Ј-GTACGTCGCGT-CACCCTCCAG-3Ј; SEQBS24 -23, 5Ј-CGCCCGCCGCGCTTCTGCCAT-3Ј; SEQSKAF-4, 5Ј-GCTCCATGTTGGCTCGGTG-3Ј; SEQSKAF-5, 5Ј-CCACCTCGGCGCCAGACAAACCC-3Ј.
For double-stranded sequencing of the pBS23 and pBS24 inserts, a dye terminator kit (PE Biosystems) and an ABI Sequencer (Applied Biosystems) were used. pSK4 and pSK9 were sequenced using the T7 polymerase sequencing kit (Amersham Pharmacia Biotech). The inserts of pBS23, pSK4, and pSK9 were sequenced by primer walking. The insert of pBS24 was restricted by AluI, HpaII, and Sau3A, respectively. The AluI, HpaII, and Sau3A fragments were cloned into the SmaI, ClaI, and BamHI sites of pBCSK(Ϫ). The resulting subclones were sequenced using universal primers. Sequence gaps were closed by primer walking. The region between E25 and E201 was sequenced using primer walking from plasmid pESW8. Sequencing of cosmids E25 and E201 was performed by a shotgun approach as follows; sheared fragments of the two cosmids were subcloned separately into pTZ18R. At least 500 clones were selected from each cosmid library, plasmid DNA prepared (Qiagen) and sequenced using Big Dye RR terminator cycle sequencing kit (PE Biosystems) and UPO/RPO-primer (MWG-BioTech). The gels were run on ABI 377 Sequencers, and data were assembled and edited using the XGAP program (32). All other DNA manipulations were performed according to standard protocols (33). Amino acid and DNA alignments were done using the programs of the Lasergene software package (DNASTAR Inc.) and Clustal W (34).
Colony Hybridization-For colony hybridization of S. aurantiaca strains, 5 l of cell suspension (1 ϫ 10 8 cells/ml) was spotted onto a nylon membrane (Biodyne B, Pall). The membrane was incubated for 5 min on blotting paper prewetted with 0.5 M NaOH,1 M Tris/HCl, pH 8.0, and 1.5 M NaCl, 1 M Tris/HCl, pH 8.0, respectively. After drying for 15 min, the filters were UV-cross-linked using a Stratalinker (Stratagene). Bacterial debris was removed with tissue prewetted with 60°C prewarmed prehybridization solution. After a short incubation in 2ϫ SSC (1ϫ SSC is 0.15 M NaCl plus 0.015 M sodium citrate), prehybridization was carried out for 2 h at 60°C in 5ϫ SSC, 5ϫ Denhardt's solution, 0.5% SDS, and 50 g/ml denatured salmon sperm DNA. Hybridization was performed overnight at 60°C after addition of the DNA probes, which were 32 P-labeled with a nick translation kit (Roche Molecular Biochemicals). The filters were washed in 2ϫ SSC, 0.2% SDS twice for 30 min at 60°C.
Preparation and Screening of the Cosmid Library-The cosmid library was made as described for S. aurantiaca Sg a15 (22). Approximately 1800 cosmid-harboring single colonies were picked into 96-well microtiter plates, grown in LB medium overnight, and replicated. To one copy of the library, 25% glycerol was added and the plates were frozen at Ϫ80°C. Cell mass of 16 clones of each plate of the second copy was pooled, and a cosmid preparation of each pool performed. These "cosmid pools" were used as templates in PCR reactions with primer pair SEQBS24 -4 and SEQBS24 -23 specifically binding to the insert of pBS24. Pools giving a PCR product of the expected size (821 bp) were used as templates in a second round of PCR reactions employing primers SEQSKAF-4 and SEQSKAF-5, which bind to the insert of pBS23. Only pools not giving the PCR product of the expected size for the latter primer pair (768 bp) were further analyzed. The corresponding clones were picked from the first copy of the library and the cosmids were independently isolated. Analogous PCR reactions with single cosmids revealed that the first primer pair binds to the insert of cosmid E25, whereas the second pair does not.
Construction of the S. aurantiaca DW4/3-1 Mutant Strains BS47 and BS57-For characterization of mtaA, insertional mutant BS57 was constructed. Plasmid pSKAF that was obtained by inserting an EcoRI fragment which encoded fbfA, fbfB, mtaA, and part of mtaB into pB-SSK(Ϫ) was restricted by FseI. After removing the extending 3Јends by T4 DNA polymerase, the kanamycin resistance gene (neo) was inserted, which was obtained from pUC4KIXX (Amersham Pharmacia Biotech) after restriction with SmaI. The plasmid obtained, pBS31, contained the neo gene in the divergent orientation of the mtaA gene. pBS31 was linearized with ScaI and transferred into S. aurantiaca DW4/3-1 wild type by electroporation (22,34). In the kanamycin-resistant recombinant BS57, the wild type gene was replaced by the mutant mtaA gene, into which the neo gene was inserted (compare Figs. 2 and 3).
The 12-kbp EcoRI fragment of pSKAF was cloned into pBCSKSalI⌬ (constructed via restriction of pBCSK(Ϫ) with SalI, fill-in using the Klenow fragment of the DNA polymerase, and religation resulting in loss of the SalI restriction site in pBCSK(Ϫ)) precut with EcoRI, resulting in plasmid pBS26. After SalI restriction of pBS26, a 2.3-kbp SalI-SalI fragment was replaced by the neo gene from pUC4KIXX, resulting in plasmid pBS27. By electroporation pBS27 was introduced into S. aurantiaca strain DW4/3-1. The resulting electroporants were screened by colony hybridization using pBCSK(Ϫ) and the neo gene (derived from pUC4KIXX) as probe, respectively. BS47 (pBCSK(Ϫ)-negative and neopositive) was chosen for further analysis (compare Figs. 2 and 3).

Isolation and Identification of a Ppant Transferase Involved in Myxothiazol
Biosynthesis-Sequence analysis of the chromosomal regions adjacent to the S. aurantiaca DW4/3-1 developmental genes fbfA and fbfB (28,29) led to the detection of a 834bp open reading frame (ORF) designated mtaA (myxothiazol). The start codon of mtaA is located about 800 bp downstream of the fbfB stop codon (see Figs. 2 and 3). The codon bias is in accordance with other genes from myxobacteria (66.1/50.7/79% GϩC in the first, second, and third position, respectively; see Table II) (35). Sequence alignments reveal that mtaA encodes a predicted polypeptide with significant homology to Ppant transferases (26,27,36), such as AcpS and EntD from E. coli or Sfp from Bacillus subtilis (Fig. 4). AcpS is the first cloned and characterized Ppant transferase that catalyzes the conversion of the inactive apo form of the fatty acid synthase to the functional form of the enzyme. EntD and Sfp were originally reported to be specific for the activation of the enterobactinand the surfactin synthetase, respectively. Recently, it was reported that by coexpression Sfp is able to activate fragments of the erythromycin PKS expressed in E. coli (37). Inactivation of mtaA by insertional mutagenesis (compare Fig. 3) resulted in mutant BS57, which is impaired in myxothiazol formation. Mutant BS57 was analyzed for secondary metabolite production by diode-array coupled HPLC and HPLC-MS and did not form any detectable amount of myxothiazol (see Fig. 5; HPLC-MS data not shown). In addition at least one other yet unidentified substance could not be detected in the mutant.
Cloning, Identification, and Sequencing of the mta Gene Cluster-About 450 bp downstream of mtaA, a large open reading frame (ORF) was detected and designated mtaB. It encodes a gene product with significant similarity to PKSs. The gene begins with an ATG start codon, which is preceded by a putative ribosome binding site (AGGA), and the codon bias is in accordance with other myxobacterial genes like mtaA (72/52/ 77). Replacing an internal 2.3-kbp SalI-SalI fragment with the Tn5-derived kanamycin resistance cassette as illustrated in Fig. 3 inactivated the gene and resulted in strain BS47. Mutant BS47 was analyzed for secondary metabolite production by diode-array coupled HPLC and HPLC-MS. BS47 did not form a detectable amount of myxothiazol but still produced the unknown compound missing in BS57 (see Fig. 5; HPLC-MS data not shown).
Since only part of the PKS was encoded by plasmid pSKAF, a cosmid library of S. aurantiaca DW4/3-1 was constructed and screened as described under "Materials and Methods." Cosmid E25 was mapped and subcloned (see Fig. 2). A 7.3-kilobase pair HindIII/BglII fragment located at the end of E25 was cloned from genomic DNA of S. aurantiaca DW4/3-1 (pESW8) and used as a probe to isolate cosmid E201, which shares with E25 the sequence of the last SauIIIA restriction site. Cosmids E25 and E201 were shown to be colinear with genomic DNA of S. aurantiaca DW4/3-1 by Southern analysis (data not shown).
Partial sequences of the inserts of plasmids pE25-11, pE25-5.5, and pE25-2 (see Table I) were determined. They showed significant homology to PKSs and NRPSs. Subsequently, cosmid E25 and part of E201 were sequenced. Six large ORFs designated mtaB-mtaG were detected (see Fig. 2). The overall GϩC content of the sequenced mta region is 65.9%.
Structural Features of Myxothiazol Biosynthetic Genes-Sequence motifs typical for PKSs (20,38,39) and NRPSs (40 -42) were detected as shown in Figs. 2 and 4 and Table II. Enzy-  Table II). Two direct repeats of about 1.13 kbp were detected in MtaD and MtaF (AT domains). Directly downstream of this repeat in MtaF, a polypeptide fragment homologous to known SAM-dependent methyltransferases (MT) including the putative SAM binding site (41) is found. Following the AT domain of MtaE, a further polypeptide fragment homologous to MTs is located. In Fig. 4 both mta MT domains are aligned with core regions of C-MTs (internal domains) and O-MTs (single proteins) described for different PKS systems. MtaG harbors an unprecedented motif within NRPSs resembling flavin and F 420dependent monooxygenases/hydrogenases (MonoOx) (see Fig.  4). The MonoOx domain is homologous to bacterial luciferases (Prosite signature no. PDOC00397) like LuxA of Xenorhabdus luminescens (Protein Identification Database (PID) code g155414; 21.7% identity and 40% similarity), to the F 420 -dependent glucose-6-phosphate dehydrogenase Fgd of Mycobacterium leprae (PID code g3129987; 21.3% and 44%, respectively), to the mitomycin biosynthetic gene MitH from Streptomyces lavandulae (PID code g4731347; 24% and 40%, respectively), and to the methylenetetrahydromethanopterin reductase of Methanobacterium thermoautotrophicum Mer (PID code g2127699; 20.3% and 38%, respectively). Bacterial F 420 -dependent enzymes have been compared, and areas of FIG. 3. Genotypes of S. aurantiaca DW4/3-1 and mutants BS47 and BS57. E, S, and X indicate EcoRI, SalI, and XhoI restriction sites (further XhoI restriction sites are located in the region but not shown). (S/X) indicates the SalI restriction site, which was lost during construction of mutant BS47 (for the construction of the mutant strains, compare "Materials and Methods"). Southern analysis of chromosomal DNA restricted with EcoRI/SalI of S. aurantiaca DW4/3-1 (lanes 1 and 3), mutant BS57 (lane 2), and mutant BS47 (lane 4) simultaneously hybridized with inserts of pBS23 and pSK4 (see Table I). The insert of pSK4 covers mtaA and hybridizes with the 5.1-kbp EcoRI/SalI fragment. The insert of pBS23 represents the SalI/SalI fragment replaced by the neo gene in BS47. Fragment sizes are estimated using DNA Molecular Weight Marker III, DIG labeled (Roche Molecular Biochemicals). domains have been compared for their putative binding specificity for activated acids (compare Fig. 4 and "Discussion"). DISCUSSION Myxobacteria have been shown to be a valuable source of secondary metabolites with biological activity (4,5). Nevertheless, little is known about PKS and NRPS systems from this class of microorganisms. We here report the first completely sequenced PKS/NRPS biosynthetic gene cluster from myxobacteria the analysis of which reveals several unique features.

FIG. 4. Alignments of conserved active sites of Ppant transferases, ER-, AT-, MT-, and MonoOx-domains encoded in the mta gene cluster (Mta) with amino acid sequences of different Ppant transferases (HetI, Anabaena; Lpa14, Srf, Gsp, Bacillus spec. and EntD, E. coli) and ER-, AT-, MT-, and Monoox-domains and biosynthetic proteins of different gene clusters and organisms (erythromycin; (Ery) (68), niddamycin (Nid) (49), pikromycin (Pik) (69), rifamycin (Rif) (39), yersiniabactin (15) (YE HMWP, Y. enterolyticus; YP HMWP, Y. pestis), and lovastatin (LNKS) (59)). MtaGMOx,
The biosynthesis of the electron transport inhibitor myxothiazol follows a multi-step process. 3-Methylbutyrate, three acetates, two propionates, and two cysteines are condensed giving rise to the carbon framework of the molecule (13). This study demonstrates that the biosynthetic machinery for myxothiazol resembles a typical type I PKS (45) combined with three NRPS modules. Both PKSs and NRPSs need to be transformed from the inactive apo to the active holo form by 4Јphosphopantetheinylation (27). MtaA shows strong sequence similarity to Ppant transferases, and gene inactivation of mtaA leads to a mutant (BS57) not producing myxothiazol. Therefore, MtaA seems to be responsible for the posttranslational modification of MtaB and possibly of more proteins of the myxothiazol biosynthetic machinery in S. aurantiaca DW4/3-1. Moreover, the mutant also does not produce at least one additional, still unidentified substance made by the wild type strain that may be synthesized by a PKS and/or NRPS (compare Fig.  5). This indicates MtaA to be responsible for Ppant transfer to proteins catalyzing the biosynthesis of different (secondary) metabolites. A possible polar effect of the insertion of the neo gene into mtaA in mutant BS47 cannot be excluded. Nevertheless, the distance to mtaB suggests that mtaB is regulated differently from mtaA. In addition, a polar effect would not explain the missing occurrence of the unidentified substance(s) in the mutant.
The modular structure of type I PKSs usually starts with an AT or a CoA-ligase domain responsible for the recognition (and activation in case of the CoA-ligases) of the starter molecule followed by transfer of the activated substrate to the first ACP domain (compare the biosynthetic gene clusters of erythromycin (Ref. 38), rapamycin (Refs. 20 and 46), and rifamycin (Ref. 39)). In the case of mtaB, the modular organization looks different; the protein begins with the loading ACP(L), followed by keto synthase (KS) 1, two ATs (AT1 and AT2), dehydratase (DH) 1, ER1, keto reductase (KR) 1, and ACP1 (see Table II). To date, this kind of modular arrangement has only been found in case of the soraphen biosynthetic gene cluster from another myxobacterium (S. cellulosum), the sequence of which is available from U.S. patent 5693774. We suppose that MtaB resembles the first multienzyme component of the myxothiazol biosynthetic machinery. In accordance with this suggestion module 1 and the loading domain have presumably been intermixed. One AT would be responsible for 3-methylbutyryl-CoA recognition and for the transfer of the acyl residue to ACP(L), which would resemble the function of a loading domain. The other AT would load the first malonyl group onto ACP1, and KS1 would next condense both activated acyl groups, giving rise to the first diketide intermediate bound to ACP1. Subsequent action of KR and DH domains result in the first double bond of myxothiazol. Soraphen also contains an unusual starter moiety processed by the corresponding PKS (SorA) (benzoate derived from phenylalanine (Ref. 47)) and analogous functions for the first modules of SorA can be assumed. According to the structure of myxothiazol, one would predict the incorporation of a propionate unit by the next module. This would involve an AT specific for methylmalonyl-CoA in contrast to the malonyl-CoA-AT, which is needed for the first extension step. Both AT types have been compared and characterized for the erythromycin, niddamycin, and rifamycin PKS (39,48,49). In Fig. 4, typical malonyl-and methylmalonyl-CoA-ATs from Actinomycetes are compared with ATs of the mta gene cluster. We were unable to define specific regions for substrate binding of malonyl-CoA, methylmalonyl-CoA, or 3-methylbutyryl-CoA.
Derivatives of branched chain ␣-keto acids are known as the precursors of polyketides of the avermectin type (50), and a branched-chain ␣-keto acid dehydrogenase gene cluster (bkd-FGH) necessary for the formation of these compounds has been cloned from Streptomyces avermitilis (51). An analogous gene locus (esg) was identified in Myxococcus xanthus and could be shown to be necessary for cell-cell signaling during development (52,53). A homologous gene was detected in S. auran-tiaca DW4/3-1. 2 This suggests S. aurantiaca DW4/3-1 to have a similar activity involved in the synthesis of the myxothiazol starter molecule 3-methylbutyrate. In the first step of avermectin biosynthesis, the ␣-branched 2-methylbutyrate or isobutyrate is loaded onto the PKS (compare Ref. 54). These metabolites are derived from isoleucine or valine, respectively, whereas the putative starter molecule in myxothiazol biosynthesis 3-methylbutyrate is derived from leucine (13). AT1 of MtaB should be rather specific for 3-methylbutyrate and possibly for other ␤-branched acids because myxothiazol species that do not contain leucine-derived starters have not been isolated yet. 3 It is tempting to speculate that other activated  (55), indicating that the 3-methylbutyrate-AT of the mta gene cluster could be useful to generate new structures via combinatorial biosynthesis. Unfortunately, the sequence of aveA1 is not available from common data bases for a comparison to the putative 3-methylbutyrate-AT of the mta gene cluster. After the second chain extension step in myxothiazol biogenesis cysteine adenylation, condensation, cyclization, and bisthiazole formation takes place, giving rise to an intermediate that finally is extended by the right-hand part of the PKS. A NRPS responsible for thiazoline formation during biogenesis of the polypeptide bacitracin (14) and the combination of PKS and NRPS to form thiazoline and thiazolidine ring structures during the biosynthesis of yersiniabactin and mycobactin have been described recently (15)(16)(17). The occurrence of typical motifs of NRPSs responsible for adenylation, condensation, and cyclization in MtaC and MtaD (compare Ref. 41) is in accordance with the predicted biosynthetic machinery for myxothiazol. Interestingly, the bis-thiazole structure in myxothiazol seems to be formed by two independent adenylation and cyclization domains, whereas only one adenylation domain has been reported for three cysteine cyclizations in the yersiniabactin biosynthetic gene cluster (56). Additionally, the NRPS domains for the first ring are encoded on one pure NRPS gene, whereas the second thiazole ring seems to be derived from mtaD, which resembles an ORF encoding NRPS and PKS modules. With respect to thiazole ring formation from thiazoline by MtaC and MtaD, no differences to the known systems giving rise to the thiazoline structures bacitracin and yersiniabactin are obvious (e.g. an additional NAD binding domain). Possibly intermodular cross-talk with the ER domain present in module 1 of mtaB takes place. This domain could be inactive because the typical motif LXHXXXGGVGXXAXXXA important for functionality of this domain (49) is changed to LXLXXXGGLAXXLXXXL (compare Fig. 4) and myxothiazol analogs without the corresponding double bond have not been isolated. 3 Because the differences in the core region except for the H to L exchange are not very prominent, one could also speculate that this ER domain is involved in the formation of the thiazole ring(s) from the proposed thiazoline intermediate. Interestingly, it has been speculated that the reduction of one thiazoline to a thiazolidine ring in yersiniabactin may be performed by the ketoreductase (KR) domain of the PKS part of HMWP1 (16). In order to investigate whether this process may be catalyzed by an enzyme encoded downstream of the gene cluster, we have sequenced further downstream of mtaG. Within approximately 15 kbp, no gene with homology to oxidoreductases could be detected. Genes with putative function in primary metabolism were found (data not shown), indicating that the end of the gene cluster was reached. Alternatively, it cannot be excluded that this oxidative process may occur non-enzymatically (57).
The PKS genes mtaE and mtaF contain domains with homology to SAM-dependent methyltransferases. MT domains within PKSs have been reported recently in case of yersiniabactin (15) and lovastatin biosynthesis in Aspergillus terreus (58,59). Whereas these domains are responsible for C-methylation, the MTs of mtaE and mtaF seem to resemble O-methyltransferases. The latter contain a LDXGXG core sequence, whereas in C-MTs the D seems to be replaced by an E (compare Fig. 4). We suppose that the MT of MtaE catalyzes the O-methylation of the hydroxyl group during myxothiazol biogenesis. Interestingly, there is no KR or DH domain located in mtaF. This could well explain how the ␤-methoxy-acrylate is formed because the enol form of the ␤-keto intermediate can readily be methylated by the MT of MtaF (compare Fig. 6). The yersiniabactin MT1 in HMWP1 is located behind the AT domain of the protein like the MTs of MtaE and MtaF.
Unexpectedly, we identified another NRPS gene behind mtaF. Only two other PKS systems have thus far been characterized with terminal NRPS domains, but in both systems (biosynthesis of lovastatin (Ref. 58) and phthiocerol dimycoserosate, a Mycobacterium tuberculosis cell envelope lipid (Refs. 60 and 61)), the role of these domains remains unclear. None of the derived secondary metabolites has a terminal amide moiety such as that found in myxothiazol. For the polypeptide antibiotic nosiheptide, it has been shown using incorporation studies with 15 N-labeled amino acids that the nitrogen of its terminal amide is derived from serine (62). The genetic data presented in this study make it likely that during myxothiazol biosynthesis the myxothiazol acid coupled to the ACP of MtaF is extended with another amino acid (e.g. serine). The carbon skeleton of the terminal amino acid would then have to be removed, giving rise to myxothiazol. The myxothiazol precursor is presumably condensed at the PCP of MtaG, and the MonoOx domain within MtaG can act upon this precursor and oxidize the intermediate at the ␣-position of the amino acid (compare Fig. 6). Dealkylation of the alcohol amide could take place releasing myxothiazol, whereas the carbon skeleton of the terminal amino acid would have to be released from MtaG by subsequent action of the TE domain present at the end of the polypeptide. Terminal amides occur in many bioactive substances including most peptide hormones (63). Their formation in mammals and insects has been studied extensively (64), and it has been shown that they arise from the oxidative cleavage of C-terminal glycine-extended precursors. The enzyme responsible is a bifunctional, copper-and zinc-dependent monooxygenase that first hydroxylates the extended peptide and then dealkylates the alcohol amide. Similar biochemistry has been described recently for the formation of nicotinamide from nicotinuric acid by peptidylglycine ␣-amidating monooxygenase, which is involved in niacin biosynthesis from NADP (65). Despite these similarities, no homology between the described type of enzyme and the MonoOx domain of MtaG could be detected. Because of the sequence similarities with flavin and F 420 -dependent monooxygenases/hydrogenases, it seems likely that the ␣-position of the extended myxothiazol is hydroxylated by the MonoOx domain, which may also be responsible for the cleavage of the resulting alcohol amide. This could represent a general mechanism for the formation of terminal amides in other PKS and NRPS type structures, e.g. thiostrepton.
The presence of the MonoOx domain in MtaG in the middle of the adenylation domain may indicate that this is a good location to integrate further functionalities into NRPS modules because the MT domain in HMWP2 has been reported at almost the same position (16).
The occurrence of a single protein harboring PKS and NRPS modules (MtaD), of the O-MTs within PKSs (MtaE and MtaF), and of the monooxygenase within a NRPS (MtaG) provide evidence for combinatorial biosynthesis taking place in S. aurantiaca DW4/3-1. The direct repeat (ATs of MtaD and MtaF) in the gene cluster could indicate that a gene duplication has taken place to integrate further functionalities. This type of strategy in which modules rather than individual enzymatic domains are the building blocks for genetic manipulation has been suggested very recently (37). The data presented in this report show that myxobacteria have an incredible ability to mix and match biosynthetic genes, and the currently broadening knowledge about the PKS/NRPS systems in Gram-negative bacteria will hopefully unravel further genetic and biochemical features.