Biosynthesis of Lyngbyastatins 1 and 3, Cytotoxic Depsipeptides from an Okeania sp. Marine Cyanobacterium

Abstract

Lyngbyastatins (Lbns) 1 (1) and 3 (2) belong to a group of cyclic depsipeptides that inhibit cancer cell proliferation. These compounds have been isolated from different marine cyanobacterial collections, while further development of these compounds relies on their lengthy total synthesis. Biosynthetic studies of these compounds can provide viable strategies to access these compounds and develop new analogs. In this study, we report the identification and characterization of one Lbn biosynthetic gene cluster (BGC) from the marine cyanobacterium Okeania sp. VPG18-21. We initially identified 1 and 2 in the organic extract by mass spectrometry and performed the targeted isolation of these compounds, which feature a (2S,3R)-3-amino-2-methylpentanoic acid (MAP) and a (2S,3R)-3-amino-2-methylhexanoic acid (Amha) moiety, respectively. Parallel metagenomic sequencing of VPG18-21 led to the identification of a putative Lbn BGC that encodes six megaenzymes (LbnA-F), including one polyketide synthase (PKS, LbnE), four non-ribosomal peptide synthetases (NRPSs, LbnB-D and F) and one PKS-NRPS hybrid (LbnA). Bioinformatic analysis of these enzymes suggested that the BGC produces 1 and 2. Furthermore, our biochemical studies of three recombinant adenylation domains uncovered their substrate specificities, supporting the identity of the BGC. Finally, we identified near-complete Lbn-like BGCs in the genomes of two other marine cyanobacteria.

Graphical Abstract

Marine cyanobacteria are an important source of structurally diverse bioactive natural products (NPs).^1–3 The discovery of these compounds has recently been supported by the applications of genome mining and rapid mass spectrometry (MS)-based approaches.⁴ The genome-based approaches provide comprehensive insights into potential NPs whose biosynthetic gene clusters (BGCs) are encoded in microbial genomes. Commonly encountered cyanobacterial BGCs belong to polyketide synthases (PKSs), nonribosomal peptide synthetases (NRPSs), their hybrids and ribosomally synthesized post-translationally modified peptides (RiPPs)^1,3 PKSs catalyze the sequential decarboxylative condensation of acyl-CoA building blocks,^5,6 whereas NRPSs assemble amino acids into the peptide backbone.⁷ Based on known enzyme reactions, new NPs can be predicted directly from the increasingly available (meta)genomes of cyanobacteria. At the same time, the MS-based metabolomics approach is powerful in detecting compounds and/or compound classes in an extract of microorganisms.⁸ The combination of HRMS with molecular ion isotopic patterns and MS fragmentation analyses has made it possible to develop tentative structural information about even unknown compounds.⁹ The integration of genomics and metabolomics studies can link compounds to their corresponding BGCs with relatively high accuracy.^8–11

Lyngbyastatins (Lbns) are cyanobacterial, modified cyclodepsipeptides belonging to three different compound classes with structural characteristics of certain dolastatins previously isolated from the sea hare Dolabella auricularia, although it is now known that the actual biosynthetic producers of these dolastatins are cyanobacteria.¹² Specifically, Lbns 1 (1) and 3 (2) closely resemble dolastatins 11 and 12 (Figure 1),^13–15 cytotoxic cyclic depsipeptides that target the actin cytoskeleton.^15,16 Indeed, at 0.2 μg/mL, 1 led to the complete loss of filamentous (F)-actin in smooth muscle A-10 cells,¹⁵ while its MIC toward KB and LoVo cells was 0.1 and 0.5 μg/mL, respectively. Compound 1 also demonstrated moderate solid tumor selectivity in vitro against L1210, H116, colon 38 and mammary 17/Adr cell types.¹⁵ In addition, Lbn2 is an analog of the cytotoxin dolastatin G,¹⁷ while Lbns 4–10 are serine protease inhibitors related to dolastatin 13.^18–20 Compound 1 is the founding member of this compound family and was first isolated as an inseparable mixture of two C-15 epimers from the extract of a Lyngbya majuscula/Schizothrix calcicola assemblage (Figure 1).¹⁵ Later, 2 was isolated from Lyngbya majuscula Harvey ex Gomont (Oscillatoriaceae) strains.¹⁴ The structures of 1 and 2 were elucidated by NMR, MS, and Marfey’s analyses, and both carry a 4-amino-2,2-dimethyl-3-oxopentanoic acid unit (Ibu, Figure 1). However, both R- and S-Ibu epimers seem to be present in isolated 1 and 2, whose NMR spectra show the extensive broadening and doubling of signals.^14,15 To elucidate the absolute configuration of the Ibu unit in Lbns, total synthesis of 1 has been conducted²¹ but challenges were encountered in the preparation of the sterically pure Ibu-N-Me-l-Ala unit. Interestingly, even though the desired S-Ibu containing unit was eventually generated,²² the final product was an inseparable mixture of 1 (1.8 mg) and 15-epi-1 (1.8 mg),²¹ suggesting Ibu epimerization during the Lbn synthesis. NMR and molecular modeling studies found that the rotation energy barrier around the amide bond of the Ibu-N-Me-l-Ala unit is lower than most amides, which may cause the broadness in the NMR spectra of 1 and 2, but suggested that natural 1 likely carries only the R-Ibu unit.²¹ Another structural feature of 1 and 2 is the presence of an unnatural amino acid component, (2S,3R)-3-amino-2-methylpentanoic acid (MAP) and (2S,3R)-3-amino-2-methylhexanoic acid (Amha), respectively (Figure 1). Other analogs of 1 and 2 have also been reported (Figure 1). For example, Ibu-epidemethoxylyngbyastatin 3 was isolated from Leptolyngbya sp. collected from the SS Thistlegorm shipwreck in the Red Sea.²³ Dolastatin 12 and its C-15 epimer Ibu-epidolastatin were isolated as promising anticancer molecules from the extract of the cyanobacterial assemblage of Lyngbya majuscula/Schizothrix calcicola.¹⁵ Majusculamide C and 57-normajusculamide were isolated from Lyngbya majuscula growing in the lagoon of Enewetak Atoll in the Marshall Islands,^24,25 while cytotoxicity-guided fractionation of the organic extract from a Fijian Lyngbya majuscula led to the discovery of desmethoxymajusculamide C.²⁶ The high structural similarities of these compounds suggest that they share a similar biosynthetic basis (Figure 1).

Figure 1.

Chemical structures of selected lyngbyastatin (Lbn) analogs belonging to the dolastatin 11/12 class. MAP: (2S,3R)-3-amino-2-methylpentanoic acid; Amha: (2S,3R)-3-amino-2-methylhexanoic acid; Ibu: 4-amino-2,2-dimethyl-3-oxopentanoic acid.

In the present study, we characterized the Lbn biosynthesis in the marine cyanobacterium Okeania sp. VPG18-21 (hereafter VPG18-21). Our LC-HRMS analysis confirmed the presence of 1 and 2 in the sample extract. Further, we performed the metagenome sequencing of VPG18-21, which led to the identification of a putative 52-kb Lbn BGC encoding six megaenzymes, LbnA-F. Our biochemical studies characterized the substrate specificities of three recombinant adenylation (A) domains from LbnA and the second and third module of LbnD, which agree with their substrates predicted from the chemical structures of 1 and 2. Furthermore, we identified two putative BGCs from the genomes of cyanobacteria Okeania sp. SIO1I7²⁷ and Moorea sp. SIO1G6²⁷ (now Moorena that is used hereafter), both of which likely express depsipeptides carrying a 3-amino-2-methyl fatty acid moiety.

Results and Discussion

Isolation and structure elucidation of 1 and 2

VPG18-21 was lyophilized and extracted with EtOAc-MeOH (1:1). The dried organic solvent extract was further fractionated by normal and reversed-phase (RP) chromatography. Subsequent RP-HPLC purification yielded 0.8 mg of 1 (0.38 % of the dried weight of the lipophilic extract) and 2.5 mg of 2 (1.19 %), respectively. The structures of these compounds were confirmed by HR-MS and ¹H NMR analysis (Figures S1–S2), which match the reported values.^14,15 The measured optical rotations of the two isolated compounds were also supportive of their same absolute configuration as 1 and 2.

The identification of 1 and 2 was further assisted by MS/MS fragmentation analysis (Figures 2 and S3). The ring-opening under experimental conditions occurred between N-methyl-alanine and Ibu, which provided the sequence of five of the nine amino acids (Ibu/N, O-diMe-Tyr/N-Me-Val/Gly-N-Me-Leu) observed as fragments b₄ through b₈. The sequence of the fragments was [b₈+H]⁺ = 858.5330 (calc. 858.5336) for 1 and 872.5500 (calc. 872.5492) for 2 indicating the loss of the Ibu unit, followed by the fragment [b₇+H]⁺ = 667.4385 (calc. 667.4389) for 1 and 681.4543 (calc. 681.4546) for 2, which corresponds to the loss of Ibu and N,O-dimethyl-tyrosine (N,O-diMe-Tyr) (Figures 2 and S3). The next fragment was identified as [b₆+H]⁺ = 554.3554 (calc. 554.3549) for 1 and 568.3697 (calc. 568.3705) for 2, which extended the cleaved sequence by an N-methyl-valine (N-Me-Val). The next fragment was determined to result from the loss of glycine (Gly) shown as the fragment [b₅+H]⁺ = 497.3331 (calc. 497.3334) for 1 and 511.3486 (calc. 511.3491) for 2. The fifth amino acid was determined by the fragment [b₄+H]⁺ = 370.2336 (calc. 370.2337) for 1 and 384.2495 (calc. 384.2493) for 2 as N-methyl-leucine (N-Me-Leu). The N-methyl-alanine (N-Me-Ala) unit was placed next to the Ibu unit based on the fragment with m/z 418.2330 for 1 and 418.2334 for 2 (calc. 418.2337), which was determined as the N-Me-Ala-Ibu-N,O-diMe-Tyr sequence.

Figure 2.

MS² fragmentation of 1 and 2 isolated from the extract of VPG18-21.

The fragments observed by a second ring-opening between the 2-hydroxy-3-methylvaleric acid (Hmva) and the MAP/Amha units provided the order of the remaining amino acids, allowing the verification of both structures. The observed fragment at m/z 281.1857 for 1 and 281.1856 for 2 (calc. 281.1860) was determined as N-Me-Leu-Gly-Hmva and subsequent dehydration, indicating the sequence of the three amino acids (Figures 2 and S3). The only unit not accounted for was the MAP/Amha which was placed between N-Me-Ala and the Hmva units since the remaining amino acid sequence was established as N-Me-Ala-Ibu-N,O-diMe-Tyr-N-Me-Val-Gly-N-Me-Leu-Gly-Hmva.

Identification of the putative Lbn BGC from the metagenome of VPG18-21

To obtain insights into the Lbn biosynthesis, we performed the metagenome sequencing analysis of VPG18-21 using the Illumina NovaSeq 6000 system. A total of 12.5 Gb of sequences were obtained from 83.6 million clean paired-end reads. Our assembly included a metagenomic binning step to enrich low GC cyanobacterial sequences²⁸ and generated 9,511 contigs, of which 151 were over 25 kb and 32 were over 50 kb. The longest contig was 135 kb in length. We next identified 25 putative BGCs from the metagenome assembly,²⁹ spanning all major secondary metabolite classes (Table S1), particularly six polyketides, five NRPs, and two hybrids. The Lbn structures suggest a PKS/NRPS hybrid biosynthetic pathway (Figure 3). From the two hybrid BGCs, we identified a putative 52-kb Lbn BGC from a single 87-kb contig (Table S2), which encodes one PKS/NRPS hybrid (LbnA), four NRPSs (LbnB, LbnC, LbnD, and LbnF) and one PKS (LbnE). This BGC is flanked with transposase genes at both ends (Table S2).

Figure 3.

Proposed biosynthetic pathway of 1 and 2. A schematic representation of the Lbn BGC was shown at the top. PKS genes and domains are colored in orange, while those of NRPSs are in light blue. M-CoA: malonyl-CoA; KS: ketosynthase; AT: acyltransferase; T: thiolation; AmT: aminotransferase; MT: N-, C- or O-methyltransferase whereas *indicates an inactive domain; A: adenylation; KR: ketoreductase; C: condensation; TE: thioesterase.

LbnA is predicted to synthesize and incorporate the MAP/Amha and Hmva moieties during the biosynthesis of 1 and 2 (Figure 3). LbnA contains nine PKS and NRPS domains in the order of ketosynthase (KS), acyltransferase (AT), thiolation (T), methyltransferase (MT), aminotransferase (AmT), condensation (C), adenylation (A), ketoreductase (KR), and thiolation (T) domain. Butyric acid and propionic acid are predicted to be activated and loaded to synthesize MAP and Amha, respectively. The incorporation of the aliphatic acid component has been observed in the biosynthesis of many cyanobacterial NPs, e.g., scytocyclamide,³⁰ malyngamide,³¹ hapalosin,³² columbamides,⁴ jamaicamides,³³ and puwainaphycin.³⁴ To our knowledge, the use of propionate as the starting unit in cyanobacterial natural product biosynthesis has not been reported yet, but it is known for the biosynthesis of erythromycin.³⁵ Furthermore, lysine propionylation is a common post-translational modification in cyanobacteria and this reaction requires propionyl-CoA as a substrate,³⁶ indicating the availability of propionate in cyanobacterial cells. Although future studies are needed to characterize the origin of short-chain fatty acid precursors, they can be recruited from the primary metabolism in multiple ways. For example, both fatty acyl-AMP ligases (FAALs)³⁷ and acyl-acyl carrier protein synthases (aACPS)^32,33 can activate fatty acids using ATP as a co-substrate. Furthermore, aliphatic acids can be activated by type III PKSs³⁸ or lipoyltransferases.³¹ However, genes encoding these known enzymes are not found in or around the putative Lbn BGC (Figure 3), suggesting that butyric acid and propionic acid may be activated by a non-pathway specific machinery. Indeed, we identified one putative FAAL and one aACPS homolog in the metagenome assembly of VPG18-21 (Table S3), though the exact enzyme involved in this biosynthetic step is still unclear. Activated butyric acid and propionic acid can be transferred to one standalone T domain for the reaction of the LbnA KS domain. The LbnA AT domain is predicted to specifically select malonyl-CoA for the synthesis of a C5 or C6 polyketide intermediate (Figure S4), which is then methylated at C-2 by the LbnA MT domain (Figure 3).³⁹ The PLP-dependent AmT domain of LbnA then likely converts the 3-oxo into the 3-amino group, generating the MAP and Amha moiety in 1 and 2, respectively. The cis-acting AmT has been proposed to introduce the amino group for the biosynthesis of multiple cyanobacterial lipopeptides, e.g., nostophycin,³⁹ microcystin,⁴⁰ and scytocyclamide,³⁰ supporting its functional assignment in the Lbn biosynthesis. The reactions of both LbnA MT and AmT are expected to be stereospecific as observed in the structures of Lbn analogs (Figure 1), but it is not predictable from their amino acid sequences. The C-terminus of LbnA possesses one KR-containing NRPS module (C-A-KR-T). The KR domain presumably reduces a 2-keto acid substrate tethered to the T domain to generate a 2-OH group for the following ester bond formation. A similar NRPS domain organization has been found to incorporate a 2-hydroxy acid moiety into several cyanobacterial NPs, e.g., hectochlorin,³⁹ microcystin,⁴¹ cryptophycins,⁴² and looekeyolides.⁴³ Of note, one key aspartate residue (Asp235) in the LbnA A domain, which interacts with the α-amino group of amino acid substrates, is mutated to favor the binding of 2-keto or 2-hydroxy acid (Table S4). Based on the chemical structures of 1 and 2 (Figure 1), we predict the A domain of LbnA uses (3S)-Me-2-oxovaleric acid ((S)-Mova) as its substrate, which is then converted into Hmva (Figures 1 and 3). In addition to LbnA, LbnE is the other Lbn PKS with a domain organization of KS-AT-MT-T. The LbnE AT domain is predicted to specifically activate malonyl-CoA, while its MT domain is expected to introduce an α-methyl group (Figure S4).

LbnB, LbnC, LbnD, and LbnF are the four predicted NRPSs. LbnB is a monomodular enzyme whose A domain is predicted to activate Gly (Table S3). It carries a predicted MT domain but likely loses the catalytic activity because the conserved, cofactor-binding motif GxGxG is mutated to GxKxG (Figure S5).⁴⁴ Indeed, the determined structures of Lbns do not carry N-methylation on this Gly moiety (Figure 1). LbnC has a domain organization of C1-A1-MT1-T1-C2-A2-MT2-T2 and l-Leu and Gly are predicted to be activated by its A1 and A2 domain, respectively (Table S3). LbnC contains two MT domains. In line with the Lbn structures, its MT2 is predicted to be inactive due to the mutation in the conserved GxGxG motif (Figure S5), similar to the LbnB MT domain. LbnD carries three NPRS modules and is expected to sequentially activate and incorporate l-Val, l-Tyr and l-Ala (Table S3). LbnD also has one and two MT domains in the first and second NRPS module, respectively, resulting in the N-methylation on l-Val and both N- and O-methylation on l-Tyr (Figure 3). A single NRPS module embedded with two MT domains has previously been observed in VatN that incorporates N,O-diMe-l-Tyr in the biosynthesis of vatiamide A and B.⁴⁵ The second module of LbnD LbnD2 shows 63% identity and 76% similarity to VatN. LbnF is a monomodular enzyme with a domain organization of C-A-MT-T-TE. Its A domain is predicted to activate and incorporate l-Ala (Table S3), while its TE domain is likely responsible for macrocyclization to produce the final products (Figure 3). Collectively, our bioinformatic analysis suggested that the identified BGC is responsible for the Lbn biosynthesis following the collinearity rule of PKSs and NRPSs.

Biochemical characterization of the substrate scopes of three recombinant A domains

To further validate the predicted Lbn BGC, recombinant A domains from LbnA and the second and third NRPS module of LbnD, LbnD2 and LbnD3, were prepared in E. coli (Figure S6). The substrate specificity of these three A domains was then assessed with a panel of selected natural amino acids and unnatural substrates in the hydroxylamine assay (Figures 4 and S7).⁴⁶ The recombinant A domain of LbnA showed the highest activity toward (2S,3S)-Hmva (Figures 4A and S7), agreeing with the bioinformatics prediction (Table S3). Its next best substrate was the other predicted substrate (S)-Mova (89% relative activity), followed by (2R,3S)-Hmva (75% relative activity), l-alloisoleucine (l-alle, 70% relative activity), (R)-Mova (65% relative activity), and l-Ile (54% relative activity). This result agrees with the presence of Hmva in the majority of isolated Lbn analogs (Figure 1), and the 2-keto group of activated substrates can be converted to 2-OH with the S configuration by the LbnA KR domain (Figure 3). Notably, the A domain of LbnA demonstrated considerable flexibility in activating substrates with both R and S configurations at C-2 and C-3 positions (Figures 4A and S7), suggesting its potential synthetic use to expand the chemical diversity of peptidic compounds. On the other hand, the successful incorporation of the activated building block in the NRP synthesis also requires the catalysis of other NRPS domains, particularly the C domain.^47,48

Figure 4.

Relative activity of different substrates activated by recombinant A domains of LbnA (A), LbnD2 (B), and LbnD3 (C). The activity of the best substrate is set as 100% for normalizing other substrates. The data represent means ± s.d. of three independent experiments. 2-Hydroxy-3-methylvaleric acid: Hmva; Me-2-oxovaleric acid: Mova; l-alloisoleucine: l-alle.

As expected, the A domain of LbnD2 favored l-Tyr (87% relative activity) as its substrate but its best substrate was O-Me-l-Tyr (Figure 4B). In contrast, neither N-Me-l-Tyr nor N,O-diMe-l-Tyr showed more than 10 % relative activity in comparison with O-Me-l-Tyr. These results suggested that the N-methylation modification likely occurs after the loading of l-Tyr to the T domain and subsequent O-methylation. Other tested natural amino acid substrates were also not activated significantly by the A domain of LbnD2, except for l-Leu which showed about 33% relative activity of O-Me-l-Tyr (Figure 4B). The A domain of LbnD3 activated both l-Ala and d-Ala (95% relative activity) at a similar level, followed by l-Cys (24% relative activity) (Figure 4C). Previous studies isolated 1 and 2 that carry both C-15-R and -S diastereomers (Figure 1).¹⁵ These diastereomers would be produced if both d- and l-Ala can be activated and incorporated by LbnD3. d-Ala can be produced from l-Ala by alanine racemase and is an essential component of cyanobacterial peptidoglycan.⁴⁹ Although we observed the flexibility of LbnD3 A domain in activating both d- and l-Ala in vitro, it is the C domain that catalyzes amino acid incorporation into NRP intermediates using either d- or l-amino acyl/peptidyl T donors. Variations in the conserved HHxxxD motif indicate the selectivity of the C domain toward the d- or l-substrate.⁵⁰ The presence of the HHIASDG motif in the C domain of LbnD3 suggested that it likely prefers an l-amino acyl/peptidyl T donor (Figure S5). Furthermore, the T domain of LbnD3 contains the conserved GGHSL motif (Figure S5), instead of GGDSI that is found in T domains associated with an epimerase and C domain favoring a d-amino acyl/peptidyl T donor.⁵⁰ In this regard, it remains ambiguous if LbnD3 truly incorporates both d- and l-Ala in the biosynthesis of Lbns. Nonetheless, the results of our biochemical studies of the three Lbn A domains validated the identified Lbn BGC.

Identification of BGCs of cyanobacterial NPs carrying a 3-amino-2-methyl fatty acid moiety

The Lbn group has multiple analogs isolated from different cyanobacterial collections but their BGCs have not been reported yet. Our attempt to mine Lbn-like BGCs from publicly available genomes and metagenomes, however, yielded only several Lbn biosynthetic enzyme homologs from different contigs of multiple Okeania species, but not complete BGCs. Because LbnA is likely responsible to generate and incorporate the featured 3-amino-2-methyl fatty acid moiety of 1 and 2, we turned to identify LbnA-containing BGCs by performing the sequence similarity network (SSN) analysis of LbnA.⁵¹ This analysis identified 10 LbnA homologs (Figure S8). We then manually inspected the upstream and downstream regions of these homologs in the genomes of corresponding strains. The LbnA homologs were found to be associated with partial BGCs in the draft genomes of Okeania sp. SIO2B3, Symploca sp. SIO2E6, Symploca sp. SIO2C1 and Okeania sp. SIO3H1. Supportively, a recent metabolomics study reported that Symploca species produce Lbn3 (2).²⁷ On the other hand, two near-complete LbnA-containing BGCs were identified from the genomes of Okeania sp. SIO1I7 and Moorena sp. SIO1G6 (Figure 5). The BGC of Okeania sp. SIO1I7 encodes one PKS, five NRPSs and one PKS/NRPS hybrid (Figure 5A, Table S5). A total of 10 PKS/NRPS modules suggest that this BGC produces a cyclic depsipeptide of the same size as Lbns (Figure 3). In addition, the predicted substrates of its NRPS A domains (including the A domain of ORF2) and PKS AT domains are identical to those from the Lbn BGC, except for a change from l-Tyr to l-Phe for the second NRPS module of ORF5 (Table S6). Of note, the l-Phe moiety appears in desmethoxymajusculamide C (Figure 1). On the other hand, the first PKS module of this BGC carries a GCN5-related acetyltransferase (GNAT) domain. In the biosynthesis of curacin A, the GNAT domain decarboxylates malonyl-CoA and then loads the resultant acetyl-CoA to an adjacent T domain.⁵² A similar acetyl transfer reaction can happen here, while the preceding CMT domain may further modify the loaded acetyl group (Figure 5A). Based on the predicted substrates of AT and A domains and the conservation of key motifs in other domains (Figure S9, Table S6), we deduced a putative chemical structure of NP encoded by this BGC (Figure 5A). The compound is the same as desmethoxymajusculamide C except that it contains 3-amino-2-methyl fatty acid instead of MAP.

Figure 5.

Organization of two mined Lbn-type BGCs and the predicted structures of their encoded compounds. (A). The Lbn-type BGC mined from the genome of Okeania sp. SIO1I7. The GenBank accession numbers of ORF1-8 are NET25238 to NET25245. The moiety incorporated by its LbnA homolog is colored in blue, and l-Phe is in red. (B). The Lbn-type BGC mined from Moorena sp. SIO1G6. The GenBank accession numbers of FAAL, DS, and ORF1-7 are NET65453 to NET65461. The moiety incorporated by its LbnA homolog is colored in blue, while the terminal alkyne is shadowed in green. In addition, d-Phe generated after epimerization is colored in red. GNAT: GCN5-related N-acetyltransferases; FAAL: fatty acid-AMP ligase; DS: fatty acid desaturase.

The BGC from the genome of Moorena sp. SIO1G6 encodes one PKS and six NRPSs, which together have seven modules (Figure 5B, Table S7). This BGC also encodes a pair of FAAL and fatty acid desaturase (DS), both of which show ~93% identity to VatA and VatB, respectively.⁴⁵ VatA and VatB activate hexenoic acid and generate a terminal alkyne for the biosynthesis of vatiamides. Despite the lack of a standalone T domain, this pair may catalyze the same reactions here. Based on the predicted substrates of AT and A domains and the conservation of key motifs in other domains (Figure S9, Table S8), we deduced a putative structure of the compound encoded by this BGC (Figure 5B). Of note, given the presence of FAAL and DS, this compound is predicted to carry the terminal alkyne, but the length of its acyl tail is unknown. Lbns and predicted NPs encoded by two newly mined BGCs all contain the 3-amino-2-methyl fatty acid moieties but with different acyl lengths. The Lbn BGC and two newly mined LbnA-containing BGCs employ different ways to activate and incorporate the 3-amino-2-methyl fatty acid moieties, highlighting the presence of diverse strategies for similar biotransformation in nature.

Conclusions

Lbn analogs have been detected in diverse cyanobacterial samples, but their biosynthesis is still unclear. In this study, we identified 1 and 2 from the marine cyanobacterium Okeania sp. VPG18-21 as confirmed by MS and NMR analysis. Metagenomic studies led to the identification of a putative Lbn BGC from VPG18-21. Our bioinformatics and subsequent biochemical studies suggested that the identified BGC is responsible for the biosynthesis of 1 and 2. Notably, as the A domain of LbnD3 activates both d- and l-Ala in vitro (Figures 1 and 4C), our study indicated that Ibu epimers observed after acid-hydrolysis of the isolated Lbns could be biosynthetically formed, although a simple epimerization on C-15 might occur during isolation and purification processes.^14,15 In addition, we identified two putative Lbn-type BGCs from the genomes of Okeania sp. SIO1I7 and Moorena sp. SIO1G6. These BGCs likely produce new cyanobacterial depsipeptides carrying a 3-amino-2-methyl fatty acid moiety similar to Lbns. Collectively, these results for the first time offer biosynthetic information on Lbns and lay the basis for the production of new Lbns derivatives using wild-type or engineered biosynthetic enzymes or heterologous expression of native or engineered BGCs.

Experimental Section:

Strains and General Materials

Escherichia coli DH5α was routinely used as a host for DNA cloning and sequencing. E. coli BL21 Rosetta (DE3) pLysS was used for protein expression. Transformed E. coli cells were cultured in Lysogeny Broth medium (LB) or Terrific Broth (TB) with a final concentration of kanamycin at 50 μg/mL. Okeania sp. VPG18-21 was collected at 20 m depth from the American Tanker shipwreck in Apra Harbor, Guam on April 3, 2018 and maintained frozen (−20 °C) in RNAlater. The sample is stored at the Smithsonian Marine Station. Its 16S RNA shows 99% of identity and 99.5% of similarity to that of Okeania sp. OdA4 (GenBank ID: LC534998.1).

NMR data were collected on a Bruker Avance II 600 MHz, high resolution 5 mm cryoprobe spectrometer operating at 600 MHz for ¹H, using residual solvent signals (δ_H 7.26; δ_C 77.0 ppm CDCl₃) as internal standards. HRMS data were obtained using Q Exactive Focus with electrospray ionization (ESI). Low-resolution electrospray ionization mass spectrometry (LRESIMS) data were obtained on an Applied Biosystems MDS SCIEX 3200 Q TRAP LC/MS/MS system using the Turbo Spray ion source in positive ionization mode. LRMS analysis was used for optimizing the conditions for obtaining the specific mass fragments and data was recorded for the validation of compounds. The optical rotations were recorded on a Rudolph Research Analytical Autopol III automatic polarimeter. The isolation was performed on a Shimadzu HPLC instrument with two LC-20AD pumps coupled with an SPD-M20A detector.

Extraction and Isolation

The cyanobacterial sample (VPG18-21) was lyophilized and extracted using 1:1 EtOAc/MeOH to provide 210 mg of extract. The extract was partitioned between hexanes and aq. MeOH, followed by partitioning the polar layer between n-BuOH and H₂O. The butanol fraction was subjected to a silica column using CH₂Cl₂ and increasing amounts of MeOH, giving ten fractions. The SiO₂ fraction eluting with 4% MeOH in CH₂Cl₂ was further purified using analytical reversed-phase HPLC (synergi hydro C18 column 250 × 4.6 mm, 1.0 mL/min; UV detection at 220 and 240 nm) with a H₂O/MeCN gradient (20% for 5 min, followed by 20–100% over 20 min, then 100% for 10 min) to afford 1 t_R 23.1 min and 2 t_R 23.6 min. The extract afforded 0.8 mg (0.38%) and 2.5 mg (1.19%) of 1 and 2, respectively, in the form of an amorphous white solid.

Lyngbyastatin 1 (Lbn1, 1): White, amorphous solid; [α]²¹_D = −11 (c 0.07, MeOH); [reported [α]²⁷_D = −17 (c 0.3, MeOH)]¹⁴; ¹H NMR data, see Figure S2; HRESIMS m/z [M + H]⁺ 999.6119 (calcd. for C₅₁H₈₃N₈O_12, 999.6125).

Lyngbyastatin 3 (Lbn3, 2): White, amorphous solid; [α]²¹_D = −32 (c 0.2, CHCl₃); [reported [α]²⁷_D = −62 (c 0.12, CHCl₃)]¹⁴; ¹H NMR data, see Figure S3; HRESIMS m/z [M + H]⁺ 1013.6282 (calcd. for C₅₂H₈₅N₈O_12, 1013.6281).

ESIMS Analysis

Compounds 1 and 2 were injected into the mass spectrometer using a 1:1 H₂O/MeCN with 0.1% formic acid. The spectra were collected in positive mode. The [M + H]⁺ was selected as the precursor (m/z 999.6 for 1 and 1013.6 for 2) for MS2 analysis. Source parameters were CUR 15, CAD Medium, IS 5000, TEM 050, GS1 55, and GS2 40. For both compounds, the dependent parameters were DP 56, EP 6, CE 30 and CXP 5.

Metagenome of VP18-21

The metagenomic DNA of VP18-21 for PCR amplification was obtained using MP Biomedicals FastDNA SPIN Kit for Soil (Santa Ana). High molecular weight DNA for genome sequencing was extracted using phenol followed by purification with the DNeasy PowerClean Pro Cleanup kit (Qiagen). The metagenome sequencing was performed using the Illumina NovaSeq 6000 sequencing platform with SP flow cell, 2×150 bp at the UF Interdisciplinary Center for Biotechnology Research. Reads were cleaned up with the Cutadapt program⁵³ to trim off sequencing adaptors and low-quality bases with a quality phred-like score < 20. Reads < 40 bases were excluded from the assembly. The metagenomic data set was obtained as a modest complex due to the association of other heterotrophic bacteria with the filament sheath of cyanobacteria in the source. Therefore, the sequence data were binned to obtain low GC cyanobacterial sequences (44% GC) separated from higher GC contaminants. The cleaned reads were assembled using SPAdes 3.14.1 with the parameters -only-assembler -k 55, 77, 99, 127 –meta⁵⁴ and metaWRAP.⁵⁵ Assembly statistics were generated using metaQuast from the Quast (Quality Assessment Tool) 5.0.0 package.⁵⁶

Bioinformatic Analysis and Prediction of the Biosynthetic Pathway

BGCs in the assembled metagenome of VP18-21 were identified with antiSMASH 6.0.²⁹ Motifs of domains in NRPSs and PKSs were determined after protein sequence alignment by ClustalW,⁵⁷ and the corresponding figures were prepared with Jalview 2.27.⁵⁸ The specificity codes of A domains were predicted with NRPSpredictor.⁵⁹ EFT-SSN analysis of LbnA was performed with cut off, e-value 5, length range (1000–9000) and a filter value of 780. The corresponding genomes of identified homologs of LbnA were analyzed with antiSMASH 6.0²⁹ to search for relevant BGCs.

Cloning, Expression, and Purification of A Domains

The A domains of lbnA, lbnD2 and lbnD3 were amplified with primers shown in Table S9 using Q5 DNA polymerase (New England Biolabs). PCR products were cloned into pET28b(+) (Novagen) and the resultant constructs were verified by DNA sequencing. To express proteins, overnight cultures were diluted 1:100 in 1 L of LB broth containing 50 μg/mL kanamycin and grown at 37 °C for 2 h before IPTG induction. The culture was then grown at 18 °C with shaking (180 rpm) for 18 h. Cells were harvested by centrifugation, resuspended in the lysis buffer (20 mM Tris-HCl, pH 8, 300 mM NaCl, and 20 mM imidazole), and lysed by sonication for six 10 s bursts at 20 mAmps using a UPC 2000U sonicator (Ultrasonic Power Corporation). Recombinant proteins were purified using nickel chelate chromatography (Qiagen). Proteins were eluted with the elution buffer (20 mM Tri-HCl, pH 8, 300 mM NaCl, and 200 mM imidazole). Purified proteins were desalted with a PD-10 column in the storage buffer (50 mM Tris-HCl, pH 8, 1 mM EDTA, 100 mM NaCl, 10 mM MgCl₂, 1 mM DTT, and 10% glycerol). Protein concentrations were determined by Bradford protein assay, and protein purity was assessed in SDS-PAGE analysis.

Characterization of the Substrate Specificity of Recombinant A Domains

The reaction mixture (200 μl) contained 25 mM Tris-HCl, pH 8, 15 mM MgCl₂, 2.25 mM ATP, 150 mM hydroxylamine, 3 mM substrate, and 5 μM purified A domain. The reaction with boiled enzyme was used as the negative control for each reaction. The reactions were performed in 96-well plates at 30 °C for 1 h and stopped by adding 250 μL of the stopping solution containing 10% (w/v) FeCl₃, 3.3 % (v/v) trichloroacetic acid, and 0.7 M HCl. The absorbance at 540 nm of each reaction was recorded on a BioTek Multi-Mode Microplate Reader (Agilent). The absorbance for the negative control was subtracted to calculate the absolute absorbance for each reaction. Each reaction was repeated at least three times and data were reported as mean ± standard deviations (s.d.).