Table ​Table11 contains a list of the strains and plasmids used in this work. Escherichia coli XL1-Blue, DH5α, and ET12567 were grown at 37°C in Luria-Bertani medium supplemented with either ampicillin (Amp; 100 μg/ml) or apramycin (50 μg/ml) when necessary (18). Cultures of S. cinnamonensis were grown in YEME medium (19) at 30°C for the isolation of genomic and plasmid DNAs, preparation of protoplasts, and fatty acid analysis. R₂YE medium (19) was used for the preparation of Streptomyces spore suspensions and for the regeneration of protoplasts after transformation. For the dl-perdeuterated valine feeding experiments, the following liquid medium was used: yeast extract (0.3%)–malt extract (0.5%)–peptone (0.3%)–glucose (1.0%), pH 7.2.

Isolation of S. cinnamonensis genomic DNA was performed according to standard methods (19). Genomic DNA was partially digested by Sau3A1 to yield a majority of fragments of around 30 to 40 kb. Sucrose gradient centrifugation was run at 36,000 rpm for 16 h at 4°C in a Sorvall TH-641 Rotor (Sorvall Ultracentrifuge; Du Pont, Wilmington, Del.). Fractions containing 30-to 40-kb fragments were collected and precipitated. Cosmid SuperCos I DNA was prepared according to the instruction manual (Stratagene, La Jolla, Calif.). Twenty microliters of a ligation mixture of genomic DNA (1 μg) and SuperCos I (0.2 μg) was packaged with the Gigapack III XL lambda extract kit (Stratagene) and then transfected into the E. coli XL1- Blue strain. The E. coli XL1-Blue cells competent for transfection were prepared in Luria-Bertani medium with 0.2% (wt/vol) maltose–10 mM MgSO₄. About 3,600 recombinant colonies were chosen as a working cosmid library of S. cinnamonensis and subsequently screened using a radiolabeled pHL1 probe that included the ccr gene and 0.5 kb of meaA gene (24). Radioactive DNA labeling was performed with a RTG DNA labeling beads (-dCTP) kit from Pharmacia (Piscataway, N.I.), with [α-³²P]dCTP (3,000 Ci/mmol) from Amersham Life Science (Arlington Heights, Ill.). Nonradioactive probing of cosmid subclones with pHL1 was accomplished using digoxigenin DNA labeling and detection kits (Boehringer, Mannheim, Germany). Southern and colony hybridizations were performed according to established methods (36) with a Porablot NY amp nylon membrane (Macherey-Nagel, Duren, Germany). Conditions for prehybridization and hybridization were slightly modified from standard protocols, with final washes for Southern hybridization and colony hybridization being carried out at 68°C in 0.5× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate). Small-scale isolation of high-copy-number plasmid DNA from E. coli was accomplished by the TENS method (51). Large-scale E. coli plasmid preparation was carried out with the Wizard Plus SV Minipreps DNA Purification System (Promega, Madison, Wis.). Low-copy-number cosmid DNA was isolated with Nucleobond AX100 cartridges (Macherey-Nagel). Isolation of DNA fragments from agarose gels was carried out with QIAquick Gel Extraction Kit (Qiagen, Santa Clarita, Calif.). Restriction digestion, phenol extraction, ethanol and isopropanol precipitation, treatment of DNA with the Klenow fragment of DNA polymerase I and alkaline phosphatase, and T4 DNA ligation were accomplished by following standard protocols (36). Preparation and transformation of competent E. coli cells were performed by standard methods (36).

Positive cosmids identified with pHL1 were selected, and subclones were prepared with the high-copy-number vector pBluescript SK. DNA sequencing was performed with an ABI PRISM 377 DNA Sequencer (Perkin-Elmer, Foster City, Calif.) according to the conditions recommended by manufacturer. The sequence was analyzed for open reading frames (ORFs) by using CODONPREFERENCE in Genetics Computer Group or FramePlot 2.3 software available on the Internet (http://www.nih.go.jp/~jun/cgi-bin/frameplot.pl). Protein database searches were performed with BLAST and FASTA, protein comparisons were made with COMPARE, and protein sequence alignments were made with LINEUP and PILEUP (all from the University of Wisconsin Genetics Computer Group Package).

Streptomyces protoplasts were transformed in the presence of 25% polyethylene glycol 1000 (19). For overexpression of the meaA gene in S. cinnamonensis, plasmid pZR8 was constructed by PCR with two synthetic oligonucleotides as primers, using cosmid DNA pME291 as a template. Primer 1 (5′-ATACTCTAGAGGAACATGACAGAGC-3′) and primer 2 (5′-ATCTAAGCTTACGAAGGCGGTTGATGG-3′) contained XbaI and HindIII sites (underlined), respectively. The PCR product was purified directly from the gels (Qiagen), digested, and subcloned into pSE34 to produce pZR8. S. cinnamonensis protoplasts were then transformed with pZR8. The mutAB genes from A. mediterranei were released from pYL28 (49) and cloned into the XhaI/HindIII sites of pSE34, resulting in pZC32.

A shuttle vector pKC1139 containing the temperature-sensitive Streptomyces origin of replication from pSG5 was used to construct a meaA gene disruption plasmid (5, 25). A BamHI site suitable for the insertion of the erythromycin resistance gene (ermE) was created within the meaA ORF by PCR. Two fragments corresponding to the first and second halves of the meaA gene were PCR amplified from the positive cosmid clone pME291 by using the following primers: primer 1 (described above), primer 2 (described above), primer 3, 5′-ATCTGGATCCAAGATGTCCTCGTAC-3′, and primer 4, 5′-ATCTGAATTCCGACGGCTCGCACGTCGT-3′. Primer 3 contained a BamHI site, and primer 4 contained an EcoRI site (underlined). Primers 1 and 3 produced a 1.0-kb PCR fragment with XbaI and BamHI cohesive ends, and primers 2 and 4 produced a 1.0-kb PCR fragment with EcoRI and HindIII cohesive ends. The two fragments were cloned into the corresponding sites of pBluescript, resulting in the formation of pZR6. A 1.6-kb BglII ermE gene fragment was excised from pIJ4026 and cloned into the BamHI site of pZR6. The resulting pZR7 plasmid, with ermE transcribed in the same direction as meaA, was then isolated and confirmed by restriction analysis. A 3.6-kb XbaI/HindIII fragment containing the disrupted meaA gene and ermE insert was released from pZR7 and cloned into pKC1139 to yield pZR22, a meaA gene disruption plasmid. In order to increase transformation and homologous recombination efficiency, this plasmid was denatured by 1.0 N NaOH according to the method described by Oh et al. (29) before it was used to transform S. cinnamonensis protoplasts. The meaA mutant of S. cinnamonensis was then obtained following standard protocols (24) and confirmed by PCR.

S. cinnamonensis cultures were grown in YEME medium at 30°C for 72 h, and the fatty acids were extracted and analyzed as described previously (45). When required, perdeuterated valine (200 mM) was added to the media at the time of inoculation.

Fermentations of S. cinnamonensis C730.1 cells and mutant derivatives were carried out in a two-stage fermentation process as described previously. The second-stage production medium was either a glucose-soybean meal (24) or an oil-based medium composed of soybean meal (1.5%), glucose (2.5%), soybean oil (1.5%), methyl oleate (2.0%), lard oil (1.0%), CaCO₃ (0.3%), FeSO₄ · 7H₂O (0.03%), KCl (0.01%), and MnCl₂ 4H₂O (0.003%). When required, valine (0.1 or 0.8 M) was added in three equal portions at 24, 60 and 72 h during fermentation. Monensins were isolated and quantitated by high-performance liquid chromatography (HPLC) analysis as described previously (34).

The conditions for the production of monensins A and B used in the labeling experiments were identical to those described above. A 30 mM 1:3 mixture of ethyl [3,4-¹³C₂]acetoacetate and unlabeled ethyl acetoacetate was added batchwise to the liquid cultures in three equal portions after 24, 28, and 72 h of growth. Monensin A was purified from organic extracts of fermentations as described previously (34) and analyzed by ¹³C spectroscopy with proton decoupling.

Earlier work demonstrated that the C730.1 strain produces monensins A and B in a ratio of about 1:1 under standard laboratory growth conditions (24,34). Neither icm inactivation nor mutAB inactivation has any effect on the monensin A-to-monensin B ratio or the total monensin titer (24, 44). A ccr mutant, however, causes a substantial decrease in this monensins A-to-B ratio from 1:1 to approximately to 1:4 (24). This change was attributed to a decrease in the pools of ethylmalonyl-CoA (formed by the action of CCR) relative to those of methylmalonyl-CoA (24). The S. cinnamonensis meaA gene was PCR amplified and used to generate pZR8, in which the meaA gene was expressed from the strong constitutive ermE promoter (46). This plasmid was transformed into S. cinnamonensis C730.1 (wild type), as well as into icm (44), ccr (L1) (24), and mut (44) mutants. The monensin titers and monensin A-to-monensin B ratio for each of the resulting transformants was determined (Fig. ​(Fig.2).2). In all experiments, the overall monensin titers were not significantly altered. Plasmid-based overexpression of meaA, however, did significantly decrease the monensin A-to-monensin B ratio in the C730.1 strain and the mutAB mutant. These results are consistent with the expression of meaA leading to an increase in the pool of methylmalonyl-CoA relative to that of ethylmalonyl-CoA and suggestive of a role of MeaA in methylmalonyl-CoA formation. In the ccr mutant (L1), this ratio is already significantly reduced from that of the wild-type strain and only marginally increased by plasmid-based meaA overexpression. Surprisingly, meaA overexpression in the icm mutant did not change the monensin A-to-monensin B ratio.

MCM is a well-studied enzyme and has long been thought to play an important role in providing methylmalony-CoA for polyketide formation in many organisms, including the erythromycin-producing S. erythraea (21), rifamycin-producing A. mediterranei (49), and monensin-producing S. cinnamonensis (6). Indeed, overexpression of MCM-encoding genes mutAB in S. cinnamonensis wild-type strains has recently been found to increase both the overall monensin titers and the amount of monensin B relative to that of monensin A (49). The same phenomenon was observed in this study with plasmid-based overexpression of the A. mediterranei mutAB genes in the meaA mutant (Fig. ​(Fig.4).4). In this case, the monensin titers were increased almost twofold (relative to those of either the C730.1 or meaA mutant) and the monensin A-to-monensin B ratio switched from 4:1 to 1:5. The overexpression of mutAB genes in the mutant WD2 (which produced only 10% of the monensin titers of the C730.1 or WM2 strain) led to a remarkable 20-fold increase in monensin titers. As in the case of the meaA mutant, plasmid-based expression of mutAB in the WD2 strain led to a significant decrease in the monensin A-to-monensin B ratio. Thus, in either the WM2 or WD2 strain, the shift toward the monensin A analog caused by insertional inactivation of meaA can be reversed by plasmid-based expression of either meaA or mutAB. Furthermore, expression of either of these genes can be used to restore monensin production in a meaA icm mutant (WD2) to at least wild-type levels. These observations are consistent with MeaA playing a major role in methylmalonyl-CoA formation.

Isotope labeling experiments with both polyether- and macrolide-producing streptomycetes have shown that valine can be catabolized to isobutyryl-CoA, and subsequently to methylmalonyl-CoA, for antibiotic biosynthesis (23, 30) (Fig. ​(Fig.1).1). Media supplemented with valine and isoleucine can sometimes stimulate macrolide antibiotic production, while increasing NH₄⁺ concentrations can have negative effects on both macrolide production and valine dehydrogenase activity (31, 32), all suggesting that branched-chain amino acid catabolism may be an important source of building blocks for macrolide biosynthesis. Valine at 0.1 M was fed to fermentations of the meaA icm mutant in three separate aliquots (at 24, 48, and 72 h of fermentation in the production media). This experiment was also carried out using valine at a final concentration of 0.8 M. Feeding with 0.1 M valine had no significant effect on the monensin production by the WD2 mutant, while feeding with 0.8 M valine led to only moderate increases in the monensin titers (25% increase) (Fig. ​(Fig.4).4). In the latter case, the monensin titers were 40% of the level achieved by the wild-type C730.1 strain and the monensin A-to-monensin B ratios were essentially equivalent (1:1). Thus, the valine catabolic pathway in the meaA icm S. cinnamonensis mutant does not appear to be efficient process for producing methylmalonyl-CoA.

Biosynthetic studies in S. cinnamonensis have previously demonstrated that dual-labeled [1,3-¹³C₂]acetoacetyl-CoA (generated by carrying out fermentations in the presence of the corresponding labeled ethyl acetoacetate) can be converted into [1,2-¹³C₂]methylmalonyl-CoA in the absence of either ICM or MCM (44). The pathway by which this intact conversion, which requires a carbon skeleton rearrangement, occurs was undetermined. It was, however, suggested that MeaA might play a role in this pathway (44). To investigate this possibility, ethyl [3,4-¹³C₂]acetoacetate was added to fermentations of C730.1 and the meaA mutant (WM2). Strong ¹³C doublets surrounding the natural abundance signal for both C-32 and C-33 of monensin A were observed (Fig. ​(Fig.5)5) for the WM2 mutant, which indicated the simultaneous presence of a ¹³C label at both carbons. The size of these doublets was substantially greater than the natural abundance signal and indicated an approximate 10-fold enrichment of ¹³C at these carbons due to the intact utilization of the ¹³C-labeled acetoacetate. The natural abundance ¹³C signals for all of the carbons of monensin A derived from C-2 and C-3 of methylmalonyl-CoA were similarly surrounded by enriched doublets. In this case, however, there was only about a twofold enrichment of ¹³C at these positions due to the intact utilization of the ¹³C-labeled acetoacetate. Indistinguishable labeling patterns were observed for monensin A isolated from the C730.1 strain grown under identical conditions. Thus, insertional inactivation of meaA affects the pool of methylmalonyl-CoA but does not seem to significantly affect the amount generated from acetoacetyl-CoA or related metabolites. These data combined with data from the previous study (44) indicate that acetoacetyl-CoA can be converted intact to methylmalonyl-CoA in S. cinnamonensis in the absence of either ICM, MCM, or MeaA. A similar study was also carried out with the icm meaA double mutant (WD2). In order to generate sufficient monensin A for nuclear magnetic resonance analysis, this experiment was carried out using the WD2/pZC32 strain (plasmid-based expression of mutAB). Similar low levels of intact labeling of the methymalonyl-CoA-derived positions of monensin from labeled ethyl acetoacetate were also seen in this experiment (data not shown), indicating that conversion of acetoacetyl-CoA to methylmalonyl-CoA can proceed in the absence of both ICM and MeaA.

Sequence analysis predicts that S. cinnamonensis meaA encodes a 74-kDa protein with a perfectly conserved coenzyme B₁₂ binding motif at the C terminus and 36 to 40% amino acid sequence identity to the two known coenzyme B₁₂-dependent mutases (MCM and ICM) (6, 27, 47). Isobutyryl-CoA and methylmalonyl-CoA (substrates for ICM and MCM, respectively) differ only in the oxidation state of one carbon. The two enzymes show significant amino acid sequence homology, particularly in the region shown from the crystal structure of the P. freundenreichii subsp. shermanii MCM shown to be involved in substrate binding (26). Most of these amino acids are also conserved in MeaA (data not shown). In MCM, the Tyr^A89 conserved in all MCM sequences and Arg^A208 are located near the bottom of the substrate binding hole and appear to form hydrogen bonds with the carboxyl group of the methylmalonyl-CoA substrate. These residues are not conserved in ICM (the substrate in this case contains the reduced methyl carbon substituent) but are conserved in the predicted amino acid sequences of all known MeaA proteins. Thus, the MeaA substrate might also contain two oxidized and possibly carboxylated substituents, with one activated as a CoA thioester.

The role of meaA in providing methylmalonyl-CoA for monensin biosynthesis was studied by two approaches. Firstly, meaA was overexpressed from a multicopy plasmid using the strong constitutive ermE promoter. This system has been used successfully in the past for overexpression of the ccr gene in S. cinnamonensis (24). Overexpression of the meaA gene resulted in a significant increase of methylmalonyl-CoA-derived monensin B compared to the ethylmalonyl-CoA-derived monensin A in the C730.1 type strain and mutAB-blocked mutants. These observations are consistent with an increased pool of methylmalonyl-CoA relative to that of ethylmalonyl-CoA and suggest that MeaA is able to generate methylmalonyl-CoA. The same conclusion was reached by generating a meaA mutant of S. cinnamonensis and observing that it made significantly higher ratios of monensin A compared to monensin B. No significant change in total monensin titers was observed in either of these experiments, indicating that under these conditions, another process provides the methylmalonyl-CoA for monensin biosynthesis. This observation was not made for the meaA icm mutant that produced less than 10% of the monensin titers than the corresponding control strains. Evidence that this decrease was due to a limitation in methylmalonyl-CoA was provided by the observation that plasmid-based expression of either mutAB or meaA restored monensin production at least to the same levels as those seen in the control. The reason that inactivation of insertional inactivation of meaA has a more pronounced effect on the icm mutant than on the C730.1 strain is unclear at present. It is interesting to note that only in the icm mutant did plasmid-based overexpression of meaA not lead to an increase in the monensin A-to-monensin B ratio. Regardless, these experiments have clearly indicated that alterations in the levels of MeaA in S. cinnamonensis can lead to significant changes in the amounts of methylmalonyl-CoA available for monensin biosynthesis and that its role can be efficiently replaced by MCM. Previous studies of MeaA in M. extorquens AM1 (9, 37) and S. collinus (15) have indicated that MeaA is neither an MCM nor an ICM. Thus, it seems likely that under the tested conditions, this methylmalonyl-CoA is being produced by a pathway involving a different type of coenzyme B₁₂-dependent mutase.

The intact conversion of [1,3-¹³C₂]acetoacetyl-CoA into the methylmalonyl-CoA-derived positions of monensin has long been considered to be indicative of a pathway (Fig. ​(Fig.1)1) involving reduction to butyryl-CoA, isomerization to isobutyryl-CoA, and subsequent oxidation (38, 44). The retention of all three deuteriums of [¹³CD₃] acetate into the methyl group of methylmalonyl-CoA in S. cinnamonensis is also consistent with such a process and inconsistent with other pathways involving MCM or PCC and MMT (Fig. ​(Fig.1)1) The recent observation that intact conversion of labeled acetoacetate into methylmalonyl-CoA was not affected by insertional in activation of the icm gene was surprising and suggested that another enzyme-catalyzed process must account for this conversion. In the present study, the same low levels of intact conversion of acetoacetyl-CoA into methylmalonyl-CoA were observed for meaA and meaA icm mutants and a wild-type strain, suggesting that meaA is not involved in converting acetoacetyl-CoA or related metabolites (butyryl-CoA) to methylmalonyl-CoA. Furthermore, it is evident from these data and a previous analysis (44) that a process not involving the cloned icm, mcm, or meaA genes is involved in this conversion (Fig. ​(Fig.1).1). One possible explanation might be the presence of a second copy of icm expressed under certain conditions. Recently, a putative second icm gene was discovered in the genome of S. coelicolor (accession no. CAB71920 and CAB71846 for the large and small subunits in the NCBI GenBank database, respectively). This gene contains two ORFs (designated the mutB and icmA genes, respectively) encoding 531 and 159 residues, respectively, which are similar in size to IcmA and IcmB of both S. cinnamonensis and S. coelicolor (33, 48). mutB encodes a protein with 51% identity to IcmB, while icmA encodes a protein with 68% identity to IcmA of S. coelicolor. Unlike the first icmAB genes characterized (33, 48), these two ORFs are separated by only 40 bp, suggesting that they may be translationally coupled. An alignment of this putative ICM with the MCM of P. freundenreichii subsp. shermanii (data not shown) revealed that many of the same amino acids lining the active site of MCM are conserved. The Tyr^A89 and Arg²⁰⁷ present in MCM and in MeaA are absent in both this putative S. coelicolor ICM as well as the S. cinnamonensis ICM (data not shown). Thus, sequence data suggest, but do not prove, that there may be two copies of the genes encoding ICM in S. coelicolor. Such a possibility cannot be completely ruled out for S. cinnamonensis. However, it has been demonstrated that there is no detectable ICM activity in the S. cinnamonensis icm mutant (48). Furthermore, the fatty acid analyses carried out in the presence of valine in the present study showed that there is no detectable in vivo conversion of isobutyryl-CoA to butyryl-CoA in either the icm or icm meaA mutants using a minimal medium. We were unable to make this determination in a complex fermentation medium because only trace amounts of the straight-chain fatty acids required for this analysis were produced. The manner in which butyryl-CoA and acetoacetyl-CoA are converted into methylmalonyl-CoA in fermentations of these mutants thus remains uncertain.