Introduction

Polyene macrolactams are an underexplored group of natural products that have only been found in actinomycetes. Examples isolated from Streptomyces sp., include viridenomycin,1 hitachimycin (stubomycin),2 BE-14106,3 aureoverticillalactam4 and ML-449,5 all of which display antibacterial and cytotoxic activities. Vicenistatin and incednine6, 7 display potent antitumor activity, and cyclomenol A, another streptomycete product, displays potent anti-inflammatory activity.8, 9 More recently, sceliphrolactam, an antifungal from a wasp-associated Streptomyces sp., has been reported.10 Micromonosporin A is the first polyene macrolactam to be isolated from a Micromonospora sp.;11 however, because of its instability and spontaneous degradation, the natural product itself was unable to be evaluated for biological activity. Interestingly, the fully reduced derivative exhibited antimalarial and antimycobacterial activitites.11 Salinilactam A was identified in the obligate marine actinomycete genera Salinispora, but due to overlapping olefinic resonances and spontaneous degradation, the structure was completed by analysis of the genome sequence in combination with analytical data.12 The wide range of potent activities exhibited by polyene macrolactams described to date make this class of compounds of interest for drug discovery efforts as well as synthetic endeavors.

Macrolactams are polyketide synthase/nonribosomal peptide synthase (PKS/NRPS) hybrids possessing varying degrees of unsaturation. Hitachimycin (stubomycin) was proposed to arise from a modified phenylalanine starter unit following incorporation of 13C-labeled phenylalanine,2 and a similar process can be proposed for virdenomycin (Figure 1a). An alternative starter unit is employed in the assembly of BE-14106,3 which is constructed from an unsaturated C9 β-aminoacyl starter unit of PKS/NRPS origin; established following gene inactivation studies, heterologous expression and incorporation studies.13 Again it is reasonable to propose that a similar pathway operates in the biosynthesis of aureoverticillalactam and ML-449 (Figure 1b). Vicenistatin and incednine are both biosynthesized from a modified glutamic acid starter as determined by labeling studies.14, 15 Cyclamenol A and micromonosporin A (Figure 1c) also appear to be constructed from a modified glutamic acid starter unit, while a lysine starter unit has been suggested for salinilactam A.12

Figure 1
figure 1

Structures of polyene macrolactams with (a) phenylalanine derived starter units, (b) aminoacyl starter units and (c) glutamate-derived starter units.

Full size image

During routine chemical screening (LC-DAD-ESIMS) of our chemical fraction library of marine actinobacteria, a UV signature characteristic of a polyene moiety (λmax 304, 342, 362 nm) was observed in chemical fractions of two microbes identified as Micromonospora strains CMS I1-30 and CMS I2-32. The peak of interest from both strains was attributed to a parent ion of identical MW and fragmentation, suggesting that both strains produced the same metabolite. Strain CMS I2-32 also exhibited potent biological activity in other fractions, making it a priority for genome sequencing. Here we report the identification of a polyene macrolactam, micromonolactam, in two Micromonospora strains, CMS I1-30 and CMS I2-32, as well as the identification and annotation of the biosynthetic genes for micromonolactam in strain CMS I2-32, which we propose is biosynthesized from a modified glutamate starter unit with 11 polyketide extensions.

Materials and methods

General experimental procedures

UV spectra were recorded on an HP1100 Diode Array Detector (Hewlett Packard (Agilent), Santa Clara, CA, USA) and ESI mass spectra were measured on a Waters Micromass ZQ with ESI (capillary voltage 3.5 kV; cone voltages 30 V/50 V; source temperature 100 °C; desolvation temperature 400 °C; cone gas flow 60 l h−1; desolvation gas flow 450 l h−1). All samples were run in positive ion mode with a mass range of 100–1400 Da and a scan time of 0.5 s. Accurate MW data were obtained using a quadrupole time-of-flight mass spectrometer (Applied Bioscience-MDS Sciex QStar XL, Applied Bioscience, Indianapolois, IN, USA). 1H, 13C and 2D homonuclear and heteronuclear NMR spectra were measured on an Avance 500 MHz NMR spectrometer (Bruker, Karlsruhe, Germany) using a 1.7-mm TXI probe head. Genomic DNA was sequenced using a Roche 454 Junior Sequencer (Roche).

Organisms

Strains CMS I1-30 and CMS I2-32 were isolated from ocean floor sediment samples collected off the coast of North Carolina. Sediment samples were collected in sterile Falcon tubes and treated with heat (55 °C for 1 h) and agitation (vortexed,1 min) in order to select for actinomycetes. Treated samples were then plated (100 μl) on duplicate agar plates, using various specialized saline media conditions to further select for marine actinomycetes. Plates were incubated for 4–6 weeks, after which potential actinomycetes were continually transferred to fresh media until pure colonies were obtained. Strains CMS I1-30 and CMS I2-32 were isolated using M6 medium (mannitol 500 mg, peptone 100 mg, agar 18 g, 29.25 p.p.t. seawater 1 l).

Fermentation

For the production of micromonolactam, CMS I1-30 and CMS I2-32 were cultured in P1 medium (starch 10 g, yeast 1 g, peptone1 g, 29.25 p.p.t. seawater 1 l).

DNA isolation

Isolation of genomic DNA from Micromonospora sp. CMS I2-32 was adapted from standard methods using the CTAB (cetrimonium bromide) procedure.16 A starter culture (5 ml) using M3 media (starch 10 g, yeast extract 4 g, peptone 2 g, 29.25 p.p.t. seawater 1 l) was incubated for 5 days at 30 °C. An aliquot (2.5 ml) of this culture was transferred to 50 ml of M3 media and incubated for a further 5 days at 30 °C with shaking at 200 r.p.m. The culture was centrifuged, and the collected cells were used for DNA isolation.

DNA sequencing and assembly

The genomic DNA sequence (10 × coverage, single-end reads) of Micromonospora strain CMS I2-32 was obtained from shotgun pyrosequencing by the Roche 454 GS Junior using titanium chemistry. The sequence reads were assembled using the Newbler software package17 into 868 contigs, with an average length of 8000 bp and total length of 6 944 130 bp.

Genome annotation

Annotation was performed using the RAST server.18 Prodigal and Artemis were also used to investigate open reading frames (ORFs) manually and in more detail.19, 20 Amino-acid sequences of potential gene products were compared with the SWISS-PROT and PIR databases using BLAST.21

Nucleotide sequence accession number

The DNA sequences determined in this study are deposited in the DDBJ, EMBL and GenBank databases under the accession numbers provided in Supplementary Information.

Results

Taxonomy of strains

Strain CMS I1-30 was identified as a sub species of Micromonospora carbonacea with a 98.2% sequence homology by phylogenetic analysis of a 500-bp segment of the 16S rDNA. Strain CMS I2-32 was identified as a sub species of Micromonospora peucetia with a 97.9% sequence homology by phylogenetic analysis of a 500-bp segment of the 16S rDNA. Phylogenetic analysis of the strains was performed by Accugenix (Newark, DE, USA) using partial 16S rDNA sequence comparison. Construction of a phylogenetic tree for comparison with sequence data for recently described marine actinomycete strains (e.g., Salinispora) in the literature suggests that CMS I1-30 and CMS I2-32 are closely related and both occur in a unique clade that falls between the obligate marine actinomycete genera Salinispora and Verrucosispora.

Large scale fermentation, extraction and isolation of micromonolactam

Strain CMS I1-30 was grown in seed medium for 7 days at 26 °C and an aliquot (2.0 ml) was used to inoculate 200 ml of fermentation medium in 1-L shake flasks, which were cultured at 200 r.p.m. and 30 °C for 5 days. XAD-16 resin (20 g l−1) was added to each flask at day 5, and the fermentation was continued for 24 h. The resin and cells were filtered using cheesecloth, washed with water to remove salts and extracted with a methanol gradient and a final elution of acetone to afford a crude extract (1.41 g), which was fractionated by reversed-phase C18 liquid chromatography eluting with increasing amounts of methanol in water. The fraction eluted with 80% methanol in water (105.9 mg) was purified by reversed-phase HPLC (Waters SunFire C18, Waters, Milford, MA, USA; 5 μm; 10 × 250 mm; flow rate of 2.5 ml min−1; detection at 214 and 304 nm; 20–35% CH3CN linear gradient over 5 min then 35–55% CH3CN linear gradient over 20 min). Final purification of the semi-pure polyene (tR=17.8 min) was performed using reversed-phase HPLC (Waters SunFire C18; 5 μm; 4.6 × 150 mm; flow rate of 1.0 ml min−1; detection at 214 and 304 nm; 20–40% CH3CN linear gradient over 5 min then 40–45% CH3CN linear gradient over 10 min to afford pure 1 (tR=10.8 min, 0.6 mg): light yellow solid; UV (MeOH) λmax 304, 342, 362 nm; ESIMS m/z 452 (M-H2O+H)+, 470 (M+H)+, 492 (M+Na)+, 939 (2M+H)+, 961 (2M+Na)+; HRESITOFMS m/z 492.2719 (calcd for C28H39NO5Na+, 492.2726).

The culture, production and initial extraction methods for strain CMS I2-32 were identical to the protocols used for CMS I1-30. A slight modification of the final steps led to the isolation of pure 1 as follows: The fraction eluted with 80% methanol in water (129.4 mg) was purified by reversed-phase HPLC (Waters SunFire C18; 5 μm; 10 × 250 mm; flow rate of 2.0 ml min−1; detection at 245 and 304 nm; 44% CH3CN isocratic conditions for 15 min then 44–100% CH3CN linear gradient over 5 min). Final purification of the semi-pure polyene (tR=9.4 min) was performed using reversed-phase HPLC (Phenomenex Gemini C18, Phenomenex, Torrance, CA, USA; 5 μm; 250 × 4.6 mm; flow rate of 1.0 ml min−1; detection at 245 and 304 nm; 35% CH3CN isocratic conditions for 20 min then 35–100% CH3CN linear gradient over 5 min to afford pure 1 (tR=17.0 min, 0.4 mg): light yellow solid; UV (MeOH) λmax 304, 342, 362 nm; ESIMS m/z 452 (M-H2O+H)+, 470 (M+H)+, 492 (M+Na)+, 939 (2M+H)+, 961 (2M+Na)+; HRESITOFMS m/z 492.2720 (calcd for C28H39NO5Na+, 492.2726). As compounds of this type can be quite labile, micromonolactam purification was performed in subdued light (laboratory lights were off and collection vials were covered in foil). Samples were also stored in a freezer in vials filled with N2 gas until needed for processing or analysis.

Structural elucidation of micromonolactam

The molecular formula of 1 was found to be C28H39NO5 by HRMS (10 double bond equivalents) and confirmed by the NMR data. The characteristic UV absorption (λmax=304, 342, 362 nm) was indicative of a polyene moiety and together these data suggested that this compound was a constitutional isomer of salinilactam A. The NMR data (Table 1) confirmed the presence of 28 carbons, including one carbonyl, 1 quaternary sp2, 21 methines (15 olefinic), 2 methylenes, and 3 methyl carbons. Three spin systems (bold lines in Figure 2a) present in 1 were determined by COSY, TOCSY and HMBC correlations (Figure 2a). A triene system, indicated by COSY and TOCSY correlations between six olefinic protons, was conjugated with a carbonyl, evidenced by the HMBC correlation from H-2 to a carbon at δC 168.1 (Table 1). In turn, H-2 was coupled to H-3, which appeared as a doublet of doublets at δH 6.93 (J2,3=14.9, 10.8 Hz), and established the E geometry of this double bond. Due to the complexity of the 1H NMR spectra in the olefinic region, coupling constants for protons H-4 through H-7 could not be ascertained, but NOESY correlations clearly illustrate the E geometry for all three double bonds in this triene system (Figure 2b). Additional 2D homonuclear correlations further extended this system from H-7 (δH 5.80) through to the H-9 oxymethine (δH 3.57), which was confirmed by HMBC correlations from the H-27 methyl doublet (δH 1.09 J8,27=6.8 Hz) to carbons at δC 40.9 (C-8 ), 76.7 (C-9 ) and 140.9 (C-7). COSY and TOCSY data indicated a second spin system containing three oxymethines, a methylene and two olefinic protons. The 1H resonance at δH 3.57 (assigned to H-9) integrated for two protons and the HSQC data indicated that these overlapping signals correlated with two identical carbon shifts at δC 76.7, making assignment more difficult. Within this spin system, an oxymethine resonance at δH 4.35 was assigned to H-13 based on coupling with an olefinic proton assigned as H-14 and an NOE correlation with a second olefinic proton assigned as H-15. An additional NOE correlation between the δH 4.35 resonance with another oxymethine at δH 3.69 helped identify this second signal as H-11, and it was deduced from the COSY data that H-11 and H-13 were separated by the single methylene group H2-12. An NOE correlation from H-11 was likely to H-9, rather than H-10, and this was supported by subsequent examination of the PKS domain organization in the biosynthetic cluster that indicated two chain extensions lacking dehydrogenase (DH) domains resulting in two sequential hydroxyl groups in the spin system. Thus another module also lacking a DH domain accounted for the third hydroxyl group at H-13 in this spin system. Notably, a P450 enzyme was identified within the micromonolactam biosynthetic cluster, and based on homology of the biosynthetic genes with those for salinilactam A in Salinispora tropica, most likely resulted in hydroxylation at C-10.12

Table 1 1H and 13C NMR shift assignments of micromonolactam (1) in CD3OD
Full size table
Figure 2
figure 2

(a) Key COSY, TOCSY and HMBC correlations establishing the planar structure of 1. (b) Key NOESY correlations for micromonolactam (1).

Full size image

COSY, TOCSY and HMBC data identified a third spin system that extended from C-17 through the C-26 methyl group. Five of the remaining olefinic protons were assigned to a triene moiety based on COSY and TOCSY data, which could be extended to include the H2-24 methylene at δH 2.28/2.59, a methine at δH 3.80 (H-25) and a doublet methyl at δH 1.29 (H-26). The chemical shift of the H-25 methine (δH 3.80, δC 47.9) indicated proximity to a heteroatom, most likely nitrogen. NOESY correlations indicated 18E and 20E, whereas a strong NOESY correlations between H2-24 and H-21, as well as H-22 and H-23, established the geometry as 23Z. This spin system could be extended beyond C-19 that showed an HMBC correlation with the H3-28 methyl protons, which appeared as a singlet (δH 1.85). Additional HMBC correlations from the singlet methyl resonance to a quaternary sp2 carbon (δC 136.0, C-18), the C-19 olefinic carbon (δC 133.5) and an as yet unassigned olefinic carbon at δC 139.0 indicated that this spin system continued with further conjugation. HSQC data showed that the latter olefinic carbon (δc 139.0, H-17) was bound to a proton at δH 6.36, which overlapped with two additional olefinic methines at 6.36 p.p.m. with carbon shifts of δC 128.8 and 133.5, the latter of which had been already assigned to H-15 (δC 133.5) in the second spin system. Further analysis of the biosynthetic genes for 1 supported the presence of a pentaene moiety consistent with the number of olefinic shifts observed. Thus the remaining unassigned olefinic resonance (δH 6.36, δC 128.8) was assigned to position 16 in the molecule, thereby completing the planar structure.

UV-LCMS analysis of surrounding chromatographic fractions indicated four additional minor compounds with nominal masses of 469 and similar retention times and UV profiles to that of micromonolactam. The two most minor components were still present in complex mixtures and could not be identified further. The remaining two were present in enough quantity to be further purified and examined by 1H NMR analysis, which indicated that they are likely isomers of micromonolactam. None of the 1H data collected corresponded to the reported values for salinilactam. Nonetheless, the possibility remains that salinilactam is present in CMS I2-32 but is expressed in minute amounts by the organism.

Bioactivity assessment of micromonolactam

Micromonolactam (1) was tested for antimicrobial activity using a disc diffusion assay. Test organisms included Gram-positive (Bacillus subtilis and Mycobacterium smegmatis) and Gram-negative (Escherichia coli) bacteria, yeast (Candida kefyr) and a filamentous fungus (Aspergillus niger). No activity was observed at the concentrations tested (0.5 μg μl−1 and 1.0 μg μl−1).

Identification of the micromonolactam gene cluster

Shotgun sequencing of Micromonospora sp. CMS I2-32 total DNA, resulted in 868 contigs after genome assembly with the Newbler program.17 The high %GC content and large genome size made sequencing and assembly difficult. The mean size of the contigs was 8000 bp, which separated natural product gene clusters over multiple contigs, and there was no indication of how the contigs related to the original genome sequence. In addition, sequencing errors introduced frame shifts within some of the larger genes.

Structural similarities between salinilactam A and micromonolactam suggested that the PKS genes, the amino-acid starter unit and tailoring enzymes involved in their biosynthesis should be similar. The genome of S. tropica has been sequenced and contains only a single modular PKS responsible for the biosynthesis of salinilactam A.12

Prodigal, a prokaryotic gene recognition and translation initiation site identification program, was used to identify ORFs within the contigs. The rationale for this was that modular biosynthetic enzymes are often encoded by genes that range from 4 kb up to 20 kb, the largest genes known in bacteria. By manually scanning the Prodigal output, it was simple to identify proteins over 1500 aa. Contigs 00164, 00306, 00396 and 00402 were identified as encoding such large proteins using Prodigal.19 The translated sequences were searched against BLAST and annotated using Artemis.21, 20 The sequences ranged from 79% to 88% identity with S. tropica CNB-440 proteins Strop_2768 and Strop_2778 –Strop_2781, the PKS proteins involved in salinilactam A biosynthesis (Table 2).

Table 2 Annotated contigs identified as containing genes homologous to the salinilactam gene cluster are listed in an order that highlights how the gene cluster organization observed in Micromonospora sp. CMS I2-32 aligns very closely with the order of the genes in S. tropica
Full size table

To identify the entire micromonolactam gene cluster, the draft genome sequence was uploaded to the RAST server, an automatic annotation tool for prokaryotic genomes.18 Contigs 00287, 00410, 00493, 00569, 00601, 00632, 00652, 00701 and 00858 were identified as encoding additional salinilactam-like PKS domains (Table 2). To discover the remaining genes, the salinilactam cluster in S. tropica was studied, spanning from Strop_2757 to Strop_2281 (Accession numbers YP_001159577.1 to YP_001159601.1). Analysis of the salinilactam gene cluster in S. tropica identified a discrete Type II thioesterase (Strop_2763). The sequence of this protein was used to BLAST against CMS I2-32 in RAST, identifying contig00114 which encodes 12 genes, including a discrete type II thioesterase, as well as 7 genes which are homologous to the salinilactam gene cluster range of Strop_2757 to Strop_2765 (Table 2). The remaining 7 kb of the cluster was found in a similar manner; the aminotransferase (Strop_2772) was used to BLAST Micromonospora sp. CMS I2-32, identifying contig00328, and the AMP-ligase (Strop_2774) was used in a BLAST search, identifying contig00641 (Table 2).

Discussion

The presence of micromonolactam, a polyene macrocyclic lactam in Micromonospora sp. CMS I2-32 was established using a combination of chemical and bioinformatics data. Over 80 kb of DNA homologous to the salinilactam gene cluster was observed, located over 15 contigs of the draft genome sequence of Micromonospora sp. CMS I2-32.

Analysis of the micromonolactam PKS domains in Micromonospora sp. CMS I2-32 identified five PKS genes encoding 48 domains, contained within 11 modules that were located over 12 contigs. Further analysis of the PKS fragments obtained from Micromonospora sp. CMS I2-32 indicated that at least 6 more domains are required for a complete PKS sequence, putting the total number of domains as 54 (Figure 3). The salinilactam PKS in S. tropica is reported to have 10 modules and 49 domains,12 suggesting a different amino-acid starter unit than micromonolactam, despite the noticeable similarities.

Figure 3
figure 3

Proposed PKS module and domain organization for micromonolactam biosynthesis. AT domains that activate malonate are labeled AT, while those that activate methylmalonate are labeled ATmm. The domains shaded in grey (The ACP and KS domains in module 1; the AT and DH domains in module 2; the AT domain in module 8; and the AT domain in module 11) are necessary for biosynthesis, based on the structure of micromonolactam, but were not observed in the genome sequence data. mmlA is proposed to consist of four modules that account for the introduction of four double bonds and a methylmalonate unit, similar to Strop_2768 in salinilactam A biosynthesis. mmlB is speculated to consist of a single module accounting for the fifth double bond in micromonolactam. mmlC is proposed to consist of at least two modules that introduce two hydroxyl groups in the sixth and seventh round of chain extension. mmlD is proposed to introduce the third hydroxyl group and second methylmalonate extender unit. mmlE is proposed to consist of at least two modules that introduce the sixth and seventh double bond. Module 11 introduces the final double bond and initiates polyketide cyclization and release. A full color version of this figure is available at The Journal of Antibiotics journal online.

Full size image

Aware that the differences in the number of PKS domains between the two macrolactams may be due to errors introduced by sequence assembly, the partial PKS sequences of micromonolactam were compared with the original sequence of Strop_2768 (Accession number YP_001159588.1). Examination of Strop_2768 also revealed a fourth PKS module, containing KS, AT, DH, KR, ACP domains, which had not been reported previously. This indicates that the biosynthesis of salinilactam A in S. tropica requires an additional round of polyketide chain elongation. Previous studies had proposed that salinilactam A was derived from a 5-aminohex-2-enoate starter unit and condensed with eight malonate and two methylmalonate units.12 However, our analysis of the highly similar PKS sequences from the two different organisms indicated that there are nine malonate and two methylmalonate units in both molecules, suggesting an alternate C5 starter unit, such as glutamate. Examination of other genes in the micromonolactam gene cluster identified a 2-oxoglutarate oxidoreductase (mmlX), which pointed towards a glutamate starter unit. Although a 2-oxoglutarate oxidoreductase gene is not present in the salinilactam gene cluster of S. tropica, a homolog is present elsewhere (Strop_0699).

Other macrolactams thought to utilize glutamate as the starter unit include vincenistatin, cyclamenol, micromonosporin and incednine (Figure 1c). Vicenistatin has been shown to be synthesized from glutamate via (2S,3S)-3-methylaspartate by labeling studies,23 while labeling studies have shown that incednine is biosynthesized from glutamate via β-glutamic acid,15 indicating two different mechanisms for the generation of the glutamate-derived starter unit (Figure 4).

Figure 4
figure 4

Two different pathways for the rearrangement of glutamate resulting in 3-aminobutyric acid ((a) salinilactam, incednine pathway) and 3-methylaspartate ((b) vincenistatin, cyclamenol pathway).

Full size image

Considering that micromonolactam and salinilactam A are more structurally similar to incednine than vincenistatin, based on the position of the C-26 methyl group relative to the amide group, it is likely that micromonolactam and salinilactam A process glutamate in the same manner as observed in incednine biosynthesis. Both gene clusters contain ORFs identified as encoding a class I and II aminotransferase (mmlI and Strop_2772) and an L-lysine 2,3 aminomutase (mmlH and Strop_2771) based on sequence homology.

If this group of related macrocyclic lactams are biosynthesized from a glutamate starter unit, then a glutamate 2,3-aminomutase enzyme is required. The first report of a glutamate 2,3-aminomutase was characterized in Clostridium difficile.24 The protein was homologous to lysine 2,3-aminomutase in many other species, but closer inspection revealed the lack of the lysyl-binding residues Asp293 and Asp330.24 Using this information, mmlH, the protein assigned as lysine 2,3-aminomutase in Micromonospora sp. CMS I2-32, was aligned with glutamate 2,3-aminomutase proteins identified from the NCBI protein database. The alignment includes lysine 2,3-aminomutase from Clostridium subterminale Sb4,25 in which the lysine-binding residues have been confirmed from the X-ray crystal structure (Figure 5).

Figure 5
figure 5

Amino-acid sequences for glutamate 2,3 aminomutases aligned using ClustalW.26 DMERI—Desulfosporosinus meridiei DSM 13257 (AFQ42647.1), DACID—Desulfosporosinus acidiphilus SJ4 (AFM39794.1), SGLYC—Syntrophobotulus glycolicus DSM 8271 (ADY56980.1), DDEHA—Desulfitobacterium dehalogenans ATCC 51507 (YP_006429495.1), DREDU—Desulfotomaculum reducens MI-1 (ABO50665.1), DDICH—Desulfitobacterium dichloroeliminans LMG P-21439 (AGA68574.1), MTHER—Moorella thermoacetica ATCC 39073 (ABC18788.1), IPOLY—Ilyobacter polytropus DSM 2926 (YP_003966579.1), CSUBT –lysine 2,3-aminomutase from C. subterminale Sb4 (2A5H_A), CMSI232 (mmlH)—Micromonospora sp. CMSI232, STROP (Strop_2771 )—Salinispora tropica (ABP55213.1). A full color version of this figure is available at The Journal of Antibiotics journal online.

Full size image

The alignment shows the conservation of the key active site residues characteristic of radical SAM enzymes (CxxxCxxC).27 The lysine-binding residues, characterized in the lysine 2,3-aminomutase, isolated from C. subterminale Sb4 (CSUBT in Figure 4), include Arg134, which binds the α-carboxylate group, Asp293 and Asp330, which bind the ɛ-ammonium group.25 The eight proteins identified as glutamate 2,3-aminomutase have Lys and Asn in place of Asp293 and Asp330, respectively, similar to the glutamate 2,3-aminomutase characterized in C. difficile, which binds glutamate, not lysine.24 Analysis of mmlH has identified the radical SAM motif CxxxCxxC and Arg134, indicative of an amino-acid substrate.24 However, mmlH lacks Asp293 and Asp330, indicating that it is unlikely to bind a lysine substrate.24

Therefore, we speculate that mmlH, identified as lysine 2,3-aminomutase in BLAST, is actually a glutamate 2,3-aminomutase and is involved in converting glutamic acid to β-glutamic acid. Then, decarboxylation by mmlI, a PLP (pyridoxal-phosphate)-dependent aminotransferase,28 would generate 3-aminobutyric acid (Figure 4a). This is consistent with the labeling studies of incednine by Eguchi and colleagues who speculated that salinilactam could be synthesized from glutamate and noted that an additional module for polyketide chain extension would be required.15 This module has now been identified in both the salinilactam A and micromonolactam gene clusters.

The proposed biosynthesis of micromonolactam is shown below (Figure 6). Glutamic acid is mutated to β-glutamate by mmlH. β-Glutamate is activated by mmlK, an AMP-dependent synthetase and ligase, and transferred to mmlM, an acyl-carrier protein. MmlI decarboxylates this intermediate to generate the 3-aminobutyric acid thioester. Eguchi and colleagues have discovered that the aminoacyl-ACP intermediate is protected by the addition of an acyl chain transferred by a second ATP-dependent ligase.14 The acyltransferase, mmlJ, transfers the primed starter unit to the PKS system, mmlA. Polyketide biosynthesis then proceeds with four chain elongation cycles, the third of which uses methylmalonate instead of malonate, and the intermediate is transferred to a second PKS system, beginning with mmlB (Figure 6).

Figure 6
figure 6

Proposed biosynthesis of micromonolactam in Micromonospora sp. CMS I2-32. A full color version of this figure is available at The Journal of Antibiotics journal online.

Full size image

In conclusion, we have isolated and identified a new macrocyclic lactam, named micromonolactam, obtained from two different Micromonospora species isolated from marine sediments. The genome sequencing data for one strain revealed the presence of a number of biosynthetic gene clusters, including one cluster for a polyunsaturated hybrid polyketide identified as micromonolactam. Together, the chemical, analytical and genetic data were used to establish the structure of the compound. The genetic data also revealed that the biosynthetic cluster contained 11 polyketide modules, indicating that the compound is biosynthesized from 9 malonate and 2 methylmalonate units. In addition, a series of genes associated with processing a glutamate starter unit were also observed, which is in line with a proposal for the related hybrid lactam incednine.15 Retro analysis of the genetic data for salinilactam A from S. tropica has uncovered an additional PKS module, establishing a common biosynthetic pathway for these metabolites.