Biosynthesis of Dolastatin 10 in Marine Cyanobacteria, a Prototype for Multiple Approved Cancer Drugs

Abstract

Dolastatin 10, a potent tubulin-targeting marine anticancer natural product, provided the basis for the development of six FDA-approved antibody-drug conjugates. Through screening of cyanobacterial Caldora penicillata environmental DNA libraries and metagenome sequencing, we identified its biosynthetic gene cluster. Functional prediction of ten enzymes encoded in the 39-kb cluster supports the dolastatin 10 biosynthesis. The non-heme diiron monooxygenase DolJ was biochemically characterized to mediate the terminal thiazole formation in dolastatin 10.

Graphical Abstract

Cyanobacteria are among the most ancient organisms on Earth¹ and are ubiquitous across natural habitats, perhaps partly due to their capacity to synthesize a spectrum of secondary metabolites.² We and others have identified hundreds of structurally and functionally diverse natural products from marine benthic filamentous cyanobacteria over the past decades.³ The continuous exploration of these microorganisms can unearth a wealth of novel natural products primed for pharmaceutical innovations. Among those compounds discovered from marine cyanobacteria, dolastatin 10 (Dol-10) has the greatest clinical impact (Fig. 1). Its synthetic analogs (Fig. 1), monomethyl auristatins E (MMAE) and F (MMAF), have succeeded as potent cytotoxic payloads in antibody-drug conjugates, leading to six FDA-approved medications, of which five are currently on the market: brentuximab vedotin, polatuzumab vedotin, enfortumab vedotin, disitamab vedotin, tisotumab vedotin (all MMAE based and marketed) and belantamab mafodotin (MMAF based and discontinued). Natural Dol-10 analogs have also emerged, showcasing structural diversity in both the types and number of components.⁴ These compounds exhibit significant cytotoxicity,^5–8 with Dol-10 targeting the vinca site of the tubulin complex to specifically disrupt tubulin-microtubule equilibrium in multiple types of cancer cells in the nM to pM range.⁹ Tubulin interacting compounds are an important class of anticancer agents, such as paclitaxel, vincristine, maytansine, eribulin, colchicine, and gatorbulin-1 that we recently isolated from the cyanobacterium Lyngbya cf. confervoides.^{10, 11}

Figure 1.

Structures of dolastatin 10 and selected natural and synthetic analogs.

Dol-10 is a linear pentapeptide first isolated from an Indian Ocean collection of the sea hare Dolabella auricularia in 1987 (Fig. 1).⁸ Subsequently, Dol-10⁷ and the homolog symplostatin 1⁶ were isolated from Symploca spp. in Guam and Palau (Fig. 1). The taxonomy of the producer of these compounds was later reclassified as Caldora penicillata.¹² Following the discovery from cyanobacteria from the Pacific Ocean, we found these compounds also in marine cyanobacteria from Florida and the Caribbean,^12–13 suggesting their ubiquitous presence and providing a reliable compound source. The isolation of Dol-10 in a much higher yield than from the sea hare and observations that D. auricularia feeds on cyanobacteria indicated that the biosynthesis should be encoded in the cyanobacterial genome. Indeed, these modified peptides possess typical cyanobacterial structural motifs.⁷ The structural constituents of Dol-10 include an N-terminal (S)-dolavaline (Dov), succeeded by (S)-valine, (3R,4S,5S)-dolaisoleuine (Dil), (2R,3R,4S)-dolaproine (Dap) and culminating with C-terminal (S)-dolaphenine (Doe) (Fig. 1).

Caldora penicillata is a pantropical marine cyanobacterium. It was originally identified as Symploca hydnoides or Symploca sp. but molecular phylogenetic studies revealed that C. penicillata is molecularly distinct from Symploca and Lyngbya species.¹² Importantly, C. penicillata produces several unique molecules with medically important bioactivities, such as the class I histone deacetylase inhibitor largazole,¹⁴ anticancer agents Dol-10 and symplostatin 1,^{6, 7} and the neuromodulators kimbeamides.¹⁵ However, the biosynthetic bases of these molecules have not been discovered from this prolific producer. Here, we report the biosynthetic gene cluster (BGC) of Dol-10 identified through a combination of the screening of fosmid libraries of environmental C. penicillata samples and metagenome sequencing. We biochemically characterized one oxidative decarboxylase that catalyzes the formation of Dol-10’s C-terminal Doe. Further bioinformatics analysis reveals the distribution and variations of the dol BGC in publicly available cyanobacterial genomes.

The structure of Dol-10 suggests its biosynthesis emerged from a cluster involving both polyketide synthases (PKSs) and non-ribosomal peptide synthetases (NRPSs), further enriched by N- and O-methylations. To identify this cluster, we constructed two metagenomic fosmid DNA libraries from two C. penicillata subsamples collected at Big Pine Ledges, Florida (July 10, 2014). LC-MS and GNPS analysis confirmed the production of Dol-10 in all subsamples (FK14-30A, B, C and E) and symplostatin 1 as well as four additional analogs, including the putative N-monomethyl Dol-10 (also named monomethyl auristatin D, MMAD), in two subsamples (FK14-30C and E) (Fig. S1A–B). The libraries were screened with degenerate primers designed on the conserved motifs of ketosynthase (KS) and adenylation (A) domains of PKSs and NRPSs (Fig. S2).^{16, 17} From each library, only one single clone was identified with two pairs of primers. Both clones (named S16) carried the same insert after sequencing both insert ends. The complete sequence of the 36.5-kb insert as determined subsequently by Illumina sequencing consisted of 13 open reading frames (ORFs) including two PKS genes, two NRPS genes and one partial NRPS gene (Fig. 2A), and suggested the absence of more biosynthetic genes toward the 5’ end. An intensive re-examination of the fosmid libraries using an array of sequence-specific primers failed to unearth clones with sequences overlapping with the S16. Pursuing an alternative approach, we sequenced metagenomic DNA isolated from one subsample FK14-30C via the Illumina MiSeq platform. A total of 2.9 Gb of sequences with a GC content of 49% were obtained from 50 million paired-end reads (2 x 250).¹⁸ However, the putative dol BGC was not predicted from the assembly using AntiSMASH.¹⁹ Employing the Paired-read Iterative Contig Extension (PRICE) tool,²⁰ we successfully augmented the 5’ end of the previously identified insert sequence by roughly 9 kb (Fig. 2A). The resulting fragment (GenBank accession number OR852429) spans 45,877 bp and carries 15 predicted ORFs (Table S1), with the first 10 ORFs encoding deduced functions for Dol-10 biosynthesis, therefore forming the dol BGC (Fig. 2A).

Figure 2.

The biosynthesis of Dol-10. A: The gene organization of an identified 45,877-bp fragment including 10 predicted dol genes whose predicted functions are indicated. B: Scheme of the proposed Dol-10 biosynthetic pathway. Two O-methylations are indicated in red while the potential C-methylation is in blue.

The dol cluster comprises genes for three methyltransferases (MTs: DolA, DolD, DolE), four NRPSs (DolB, DolC, DolG, DolI), two PKSs (DolF, DolH), and one putative nonheme diiron monooxygenase (DolJ). DolB is a monomodular NRPS with a domain organization of A-MT-thiolation (T). Its A domain is predicted to activate l-Val (Table S2), while its MT domain presumably catalyzes the double N-methylations, thereby forming Dov (Fig. 1), or monomethylation as observed in MMAD (Fig. S1B). The biosynthesis of Dol-10 proceeds further with the sequential incorporation of l-Val and l-Ile by the bimodular NRPS DolC with expected A domain specificity (Table S2). The MT domain of the DolC’s second module likely catalyzes an N-methylation on the loaded l-Ile (Fig. 2B). Subsequently, the monomodular PKS DolF with a domain organization of KS-AT (acyltransferase)-KR (ketoreductase)-T extends the growing chain by one acetate unit. The presence of the motif sequence of RAFH in DolF’s AT domain suggests malonyl (M)-CoA as substrate (Figs. 2B and S3A), while its KR domain presumably introduces an S-OH group (Fig. S4),²¹ followed by an O-methylation to align with the determined Dol-10 structure (Figs. 1 and 2B). Next, DolG elongates the intermediate with l-Pro (Table S2), and DolH incorporates a reduced acetate unit (Fig. 2). Unlike DolF, the AT domain of DolH carrying the motif sequence of VASH presumably uses methylmalonyl (MM)-CoA as substrate (Figs. 2B and S3A), while its KR domain likely generates an S-OH group for subsequent methylation (Fig. S4). To our knowledge, MM-CoA specific AT domain is unique among known cyanobacterial PKSs. Phylogenetic analysis of DolH’s AT domain reveals a small clade comprising its homologs from cyanobacteria Symploca, Caldora and Aetokthonos species (Fig. S3B), none of which have been biochemically characterized yet. Of the three MTs for the Dol-10 biosynthesis, DolD and DolE were grouped with other O-MTs in the phylogenetic analysis, while DolA was in the C-MT clade (Fig. S5). Therefore, DolD and DolE might catalyze two O-methylations, while it is unclear if DolA catalyzes C-methylation to produce Dap (Fig. 2B). The order of methylations also remains elusive. Lastly, DolI presumably incorporates l-Phe and l-Cys in order (Table S2) and converts the loaded l-Cys into a thiazoline ring by its cyclization (Cy) domain, similar to BarG in the biosynthesis of barbamide²². A terminal thioesterase (TE) domain hydrolyzes the tethered intermediate. The product might then be oxidatively decarboxylated, potentially by the DolJ, resulting in Dol-10’s final thiazole (Fig. 2B). LynB7, sharing 78% identity to DolJ, was recently shown to mediate the same transformation in lyngbyapeptin B biosynthesis,²³ which may also occur in barbamide biosynthesis by BarH.²² DolJ, BarH and LynB7 are members of the nonheme diiron monooxygenase family (Fig. S6),²³ which catalyzes a range of oxidation reactions (e.g., hydroxylation).²⁴ A subsequent gene, in the opposite direction to dolJ, encodes for a cyclase (Orf11), succeeded by a presumed arsenite transporter (Orf12). Their roles in Dol-10 biosynthesis are unclear.

To validate the identified dol cluster, we prepared recombinant DolJ as well as BarH in E. coli (Figs. S7–S8)and synthesized compounds 1-2 as putative enzyme substrates (Figs. 3A, and S9–S18). When Fe²⁺ and ascorbic acid were present, DolJ effectively converted 1 into N-methylated Doe, which is confirmed by comparing with the synthetic standard 3 (Figs. S19–26) in LC and LC-MS analysis (Figs. 3B and S27–28). This reaction can proceed via β-hydroxylation or a radical intermediate (Fig. S29A). Indeed, a hydroxylated intermediate was observed in the enzyme reaction terminated after 10 min by LC-MS analysis (Fig. S29B–C), which almost completely converted to compound 3 after 20 min, supporting the β-hydroxylation path. The kinetics for 1 under determined optimal conditions gave a K_m = 70.2 ± 4.0 µM and k_cat = 0.58 ± 0.01 s^–1 (Figs. S30–31), suggesting comparable overall catalytic efficiencies of DolJ and LynB7.²³ The diastereomeric analog 2 was not active in the DolJ reaction (Fig. 3A–B). We also prepared recombinant BarH (Fig. S7) and it showed the same specificity toward compounds 1 and 2 as DolJ (Fig. 3B). The predicted protein structure of DolJ by Alphafold 2.0²⁵ exhibited a high predicted local distance difference test score that ranged from 85 to 98 excluding the initial three residues at the N-terminus and a single residue at the C-terminus, suggesting a high quality of structural prediction. DolJ structurally resembles the hydroxylase PtmU3 in the biosynthesis of platensimycin and platencin (PDB ID: 6OMR; RMSD = 4.1 Å across 352 residues, Figs. S6 and S32–33).²⁶ Compound 1 was docked into DolJ’s predicted active site using Autodock Vina,²⁷ and the C5 of its thiazoline moiety was at a 4.0 Å distance from the second iron (Fig. 3C). This proximity is congruent with the distances observed (4.7-5.1 Å) in the characterized co-complexes of PtmU3.²⁶ The relatively shallow substrate channel supported the use of compound 1 to mimic the putative native substrate of DolJ. Compared with the modeled structures of DolJ and BarH, the displacement in the α3-helix of LynB7 implies a preferential accommodation of a more compact substrate (Fig. S34), agreeing with a recent report.²³ Our biochemical analysis of DolJ and BarH validated the identified dol BGC.

Figure 3.

Biochemical characterization of DolJ and BarH. A: Structures of compounds 1-3. B: LC-traces of DolJ and BarH reactions with compounds 1 and 2 as substrates. The control reaction was the same as the full reaction except for containing heat-inactivated DolJ. C: Compound 1 (green) was docked into the active site of predicted DolJ structure (tv-blue), which was superimposed with Chain B of the structure of PtmU3 (PDB ID: 6OMR) with its substrate (PTN compound 11, cyan)²⁶ and two iron atoms (red spheres). The site for hydroxylation in both compounds is shown in pink, close to the second iron.

In exploring publicly available genomes and metagenomes, we identified four dol-like BGCs within marine and freshwater cyanobacterial genomes (Fig. 4). Notably, one BGC from Symploca sp. SIO1B1 mirrors our archetype and is predicted to synthesize symplostatin 1 (Table S2), therefore named the sym BGC. Unlike the dol BGC, the sym BGC incorporates one mbtH at the 5’-end and one additional MT gene at the 3’-end. The putative cyclase gene (orf11) was found in both BGCs. The recently reported aetokthonostatin BGC (aes) was also identified from two Aetokthonos hydrillicola strains in our analysis (Fig. 1).²⁸ The aes BGC resembles that in Symploca sp. SIO1C2, with one key exception of the split of one hybrid pks-NRPS gene in SIO1C2 into one PKS and one NRPS gene in the aes BGC. Another notable difference between these two BGCs are their gene orders. Except for the dol BGC, all four dol-like BGCs encode one additional MT (Fig. 4). During the aetokthonostatin biosynthesis, this MT AesK converts the terminal N-MeIle of monomethylaetokthonostatin into the N,N-diMeIle moiety.²⁸ The variations of these BGCs suggest a nuanced evolutionary adaptation in secondary metabolite biosynthesis.

Figure 4.

Identification of the dol-like BGCs. The homologous dol BGCs were identified in the marine and freshwater cyanobacterial genomes. The cluster alignment was performed by Clinker²⁹ and the biosynthetic genes encoding enzymes with >50% sequence identity were connected by the black-to-white shades according to the identities.

In conclusion, we have discerned the BGC of Dol-10 from the metagenomes of environmental C. penicillata samples. The role of an oxidative decarboxylase in forming the terminal thiazole moiety of Dol-10 and barbamide has been established biochemically. This research paves the way for identifying new strains producing Dol-10, related BGCs expressing Dol-10 analogs, and for fostering advances in synthetic biology and biocatalysis to generate Dol-10 and its analogs.

Data Availability

The data underlying this study are available in the published article and its Supporting Information.