2-Hydroxysorangiadenosine: Structure and Biosynthesis of a Myxobacterial Sesquiterpene–Nucleoside

Myxobacteria represent an under-investigated source for biologically active natural products featuring intriguing structural moieties with potential applications, e.g., in the pharmaceutical industry. Sorangiadenosine and the here-discovered 2-hydroxysorangiadenosine are myxobacterial sesquiterpene–nucleosides with an unusual structural moiety, a bicyclic eudesmane-type sesquiterpene. As the biosynthesis of these rare terpene–nucleoside hybrid natural products remains elusive, we investigated secondary metabolomes and genomes of several 2-hydroxysorangiadenosine-producing myxobacteria. We report the isolation and full structure elucidation of 2-hydroxysorangiadenosine and its cytotoxic and antibiotic activities and propose a biosynthetic pathway in the myxobacterium Vitiosangium cumulatum MCy10943T.

The genes in the myxobacterium M. xanthus DK1622, which encode the proteins involved in leucine degradation and the alternative biosynthesis of isovaleryl-CoA [2] have been described previously. For that reason, the genome sequence of Vitiosangium cumulatum MCy10943 T was investigated accordingly to find homologues genes. All genes encoding proteins which are required for the leucine degradation pathway and the alternative biosynthesis of isovaleryl CoA from M. xanthus DK1622 and their corresponding homologues in Vitiosangium cumulatum MCy10943 T are listed in Table S1.  Table S1.

Formation of sesquiterpene scaffolds and eudesmane-type sesquiterpenes
The formation of the sesquiterpene scaffold starts from farnesyl pyrophosphate (FPP) with subsequent cyclization via a sesquiterpene cyclase, which is known to catalyze the formation of diverse (poly)cyclic sesquiterpene skeletons [5]. Six different initial cyclization reactions are possible starting from farnesyl diphosphate [5] (Fig. S15A). These include the direct conversion of farnesyl diphosphate to (E,E)-germacrenyl cation via 1-10-cyclization or 1-11 cyclization to the (E,E)-humulyl cation. The 2-E double bond of FFP can be isomerized via ionization leading to the farnesyl cation, followed by reattachment of diphosphate at C-3 to yield nerolidyl diphosphate (NPP) (Fig. S15B). This allows NPP a 1,6-cyclization to the bisabolyl cation, a 1,7cyclization to the cycloheptenyl cation, a 1,10-cyclization to the (E,Z)-germacrenyl cation or a 1,11-cyclization to the (E,Z)-humulyl cation (Fig. S15C).
According to the featured eudesmane-type sesquiterpene structure, a logical proposal would start from FFP to yield via a 1,10-cyclization the (E,E) germacrenyl cation (Fig. S16A). The germacrenyl cation could also be cyclized to intermedeol, such as described in the literature for intermedeol cyclases (Fig. S16B) [6]. However, the secondary metabolome of MCy10943 T shows no signal corresponding to the sum formula or mass of intermedeol or a hydroxylated eudesmadiene intermediate. On the contrary, three compounds with the mass of 205 m/z and the sum formula C15H25 are present in the crude extract of MCy10943 T , which correspond to the non-hydroxylated eudesmadiene intermediates (Fig. S16C, Fig. S17). Cyclization of eudesmadiene begins with a 1-10 bond S17 formation to yield germacrenyl A which is then protonated at C6 to facilitate C2-C7 bond formation resulting in the eudesmane carbocations.

Metabolome-genome correlation
Initial in silico analysis already excluded two of the antiSMASH-identified biosynthetic gene clusters (BGCs) to produce 1 and 2, since both terpene cyclases share high sequence similarity to the geosmin terpene cyclase. In addition, a study investigating different terpene cyclases from S. cellulosum So ce56 suggested, that the geosmin cyclase gene (sce1440) is specific in terms of its production profile [7]. In order to narrow down the number of suitable BGCs for the production of 1 and 2 we performed metabolomic profiling of the myxobacterial strains Cystobacter ferrugineus Cb fe23, Cystobacter fuscus Cb fe15 (DSM 52655) and Cystobacter fuscus DSM 2262 T . These three strains harbor according antiSMASH and manual evaluation similar terpene gene clusters to No. 3, 5 and 6 (Tab. 1). In the respective crude extracts, no metabolites corresponding to the retention time, exact molecular mass and sum formula connected to 1 and 2 was found in the LC-MS chromatogram of these three myxobacterial strains (Fig. S11 and S12). In addition, the gene clusters No. 3, 5, 6 were further excluded, since as proposed above, each gene cluster is missing at least either a gene encoding an oligoprenyl transferase or a cytochrome P450 gene. Taken together, genomic investigation combined with metabolomic profiling seemingly reduced the likelihood that the terpene gene clusters No. 3, 5 and 6 are responsible for the production of 1 and 2.

S21
However, since genes responsible for the production of the terpenes might be poorly expressed in some organisms and genes pivotal for the formation of the terpene scaffold can be located outside of the biosynthetic core region (terpene or diterpene cyclase genes), it cannot be entirely excluded that the terpene gene clusters No. 3, 5 and 6 are in some way involved in the biosynthesis of 1 and 2. (SBCb004) is reduced to sora7-9, sora11-12, sora18-19 and sora21-22. Figure S18 shows this truncated 2hydroxysorangiadenosine BGC in Cystobacter sp. strain MCy9101 (SBCb004) and Figure S9 and S10 demonstrate that this truncated BGC is presumably sufficient to produce 1 and 2 in low concentration.