Isolation of Verrucosins A-E from a Marine Verrucosispora sp. Reveals a Unifying Biosynthetic Hypothesis for Linear and Macrocyclic Polyketides

Abstract

As part of our long-standing program evaluating the biosynthetic complexity and biomedical potential of natural products from marine microbes, our attention was drawn to culture extracts from a Verrucosispora sp. (strain TAA-831), which produced multiple compounds with unique UV absorbance signatures and HRMS data. Large-scale fermentation and targeted isolation afforded verrucosins A-E (1-5), a mixture of linear and macrocyclic polyketides whose structures were determined through a synergistic combination of experimental, computational, and genomic approaches. The conserved sequence of methyl malonate and malonate motifs across the verrucosins implied a shared biosynthetic origin despite structural divergence as linear and cyclic congeners. Targeted genome mining revealed a lone type I/type III hybrid polyketide synthase biosynthetic gene cluster, vrs, that is likely responsible for verrucosin production. This revelation demonstrates for the first time that linear 3,5-dihydroxybenzenic (1 and 2) and cyclic ansamycin (3-5) polyketides can be naturally produced by a single biosynthetic gene cluster. The identification of the vrs cluster and bioinformatic prediction of the stereoselectivity of the embedded reductive domains within the modular type I polyketide synthase reinforced the NMR and computational stereochemical assignments for the co-isolates, particularly the stereochemically complex linear verrucosins (1 and 2).

1

A. New verrucosins A–E ( 1–5 ) produced by Verrucosispora sp. strain TAA-831. B. Previously reported polyketide natural products derived from 3,5-DHBA, including the baulamycins A and B (6 and 7, revised stereochemistry), kendomycin (8), kendomycins B–D (9–11), and cinnamomycin A (12).

2

Planar structure assignments of verrucosins A (1) and C (3) by interpretation of one- and two-dimensional NMR data. Bold blue lines indicate assignment by COSY NMR analysis, while red arrows indicate assignment by analysis of HMBC NMR data.

3

Fragment A NMR chemical shift analysis for verrucosins A and B (1 and 2). A. Benzylic coupling constant and principal component analysis of Fragment A NMR data (blue box) for known 2-alkyl-1-phenyl-1,3-propanediols. , B. DP4+ probabilistic analysis of Fragment A triol diastereomers, anti-anti-syn (isomer 1, blue box) and anti-anti-anti (isomer 2), compared with NMR experimental data for verrucosin A (1).

4

Fragment B NMR chemical shift analysis for verrucosins A and B (1 and 2). A. Direct assignment using absolute chemical shift difference of geminal methylene proton pairs (H-4a and 4b in 1 and 2, H-6a and 6b in 1 only). B. Principal component analysis of key Fragment B anti-anti and syn-syn NMR data (blue box) shows verrucosin A (1) in good agreement with syn-syn baulamycin B (7).

5

A. Key through-space ROESY NMR correlations used to assign relative configuration and three-dimensional lowest energy conformer for verrucosin E (5). B. Comparative TDDFT-calculated (verrucosin E, 5a and 5b) and experimental ECD spectra for verrucosins C–E (3–5).

6

A. Clinker similarity analysis of characterized kendomycin BGCs sourced from MIBiG, , kmy (accession: BGC0002671) and ken (accession: BGC0001066) and vrs clusters (GenBank accession: CP192240). B. The vrs BGC and biosynthetic hypotheses for coproduction of verrucosin A (1) and cyclic verrucosins 3–5. Overall, the vrs BGC exhibits a high degree of synteny with the known kendomycin BGCs, ken and kmy, with only minor variations in gene order and composition (Figure A). The major structural variation between the cyclic PKs isolated from TAA-831 and those reported from the ken and kmy BGCs is the size of the macrocycle, with verrucosins 3–5 possessing 16-membered macrocycles compared with 18-membered ones found in kendomycins 8–11 (notably there is a single report of one other 16-membered kendomycin; however, there is no publicly available genomic data associated with this compound nor the producing strain). The vrs BGC is perfectly congruent with this two-carbon difference, with the type I PKS assembly line (vrs19, 20, and 22) comprised of a loading module and seven elongating modules, one module less than the ken and kmy assembly lines responsible for the 18 carbon macrocycle kendomycins.