Fosmidomycin biosynthesis diverges from related phosphonate natural products

Abstract

Fosmidomycin and related molecules comprise a family of phosphonate natural products with potent antibacterial, antimalarial and herbicidal activities. To understand the biosynthesis of these compounds, we characterized the fosmidomycin producer, Streptomyces lavendulae, using biochemical and genetic approaches. We were unable to elicit production of fosmidomycin, instead observing the unsaturated derivative dehydrofosmidomycin, which we showed potently inhibits 1-deoxy-D-xylulose-5-phosphate reductoisomerase and has bioactivity against a number of bacteria. The genes required for dehydrofosmidomycin biosynthesis were established by heterologous expression experiments. Bioinformatics analyses, characterization of intermediates and in vitro biochemistry show that the biosynthetic pathway involves conversion of a two-carbon phosphonate precursor into the unsaturated three-carbon product via a highly unusual rearrangement reaction, catalyzed by the 2-oxoglutarate dependent dioxygenase DfmD. The required genes and biosynthetic pathway for dehydrofosmidomycin differ substantially from that of the related natural product FR-900098, suggesting that the ability to produce these bioactive molecules arose via convergent evolution.

Figure 1.. The fosmidomycin family of phosphonate natural products.

(A) Structures of fosmidomycin and related antibiotics. Key chemical differences are noted in red. (B) The biosynthetic pathway for FR-900098.

Figure 2.. S. lavendulae produces dehydrofosmidomycin.

The bottom (red) spectrum shows the ³¹P NMR spectrum of spent media from S. lavendulae. The middle (green) spectrum shows the same spent media from S. lavendulae after addition of 1 mM fosmidomycin. The top (blue) spectrum shows the spent media from S. lavendulae after addition both 1 mM fosmidomycin and 1 mM dehydrofosmidomycin. Representative data from three independent replicates.

Figure 3.. Characterization of the dfm gene cluster.

(A) The dehydrofosmidomycin (dfm) gene cluster from S. lavendulae. The putative function of each gene is shown. A Genbank-formatted file of an integrative plasmid with the full dfm cluster, as well ca. 15 kb of flanking sequences upstream and downstream of the cluster is provided in a Supplementary Note. The extent of the various deletion derivatives used for heterologous expression studies is indicated above the diagram. Putative promoter regions are indicated by the thin red arrows. (B) ³¹P NMR spectra for extracts from the parent strain (S. albus J1074, red spectrum), the heterologous expression strain (S. albus J1704/dfm, mustard spectrum), and the heterologous expression strain spiked with 1 mM dehydrofosmidomycin (blue spectrum). The asterisk indicates dehydrofosmidomycin. Other compounds that were verified by ³¹P NMR and HRMS are indicated (see Supplementary Fig. 7). The predicted deoxdehydrofosmidomycin NMR peak was not confirmed due to lack of a synthetic standard. Representative data from three independent experiments. (C) ³¹P NMR spectra for extracts from heterologous expression strains carrying the indicated plasmids. The position of peaks corresponding to specific metabolites is indicated. Representative data from three independent experiments. HRMS analysis of these extracts confirming the presence of the presence of these metabolites is presented in Supplementary Fig. 7. (D) Structures of additional phosphonates produced by the heterologous expression strain.

Figure 4.. Feeding studies with 2AEP and L-methionine-(methyl-¹³C).

(A) ³¹P NMR spectra for extracts from the parent strain (S. albus J1074, red spectrum), the heterologous expression strain carrying the full dfm gene cluster (S. albus J1074/dfm, green spectrum), and an expression strain incapable of 2AEP biosynthesis due to deletion of phosphonopyruvate decarboxylase (S. albus J1074/ΔLM) grown without (teal spectrum) and with 1 mM 2AEP (purple spectrum). Representative data from three independent experiments. (B) HRMS analysis of the same three strains showing the EIC for dehydrofosmidomycin (180.00–180.01 Da). Peaks corresponding to dehydrofosmidomycin are indicated with an asterisk. Representative data from three independent experiments. (C) HRMS analysis for S. lavendulae grown with no additives (red) or with added unlabeled L-methionine (green) or ¹³C-labeled methionine (blue). Peaks corresponding to unlabeled and doubly labeled dehydrofosmidomycin are indicated. Representative data from three independent experiments.

Figure 5.. Conversion of trimethyl-2AEP to methyldehydrofosmidomycin by DfmD.

(A) The rearrangement of trimethylhydrazine-propionate (THP) to 3-amino-4-(methylamino)butanoic acid (AMBA) catalyzed by γ-butyrobetaine dioxygenase (BBOX) is shown at the top. The analogous reaction catalyzed by DfmD is shown below. (B) ³¹P NMR analysis of the DfmD reaction. The blue spectrum is from a no-enzyme control. The red spectrum is from a reaction containing purified DfmD. The teal spectrum is from the DfmD-containing reaction spiked with synthetic methyldehydrofosmidomycin. The asterisk marks the putative methyldehydrofosmidomycin peak. The integration values (∫) show the abundance of each compound relative to the phosphate buffer. Representative data from three independent experiments. (C) The EIC for methyldehydrofosmidomycin from the DfmD-containing reaction (red) and no-enzyme control (blue). Representative data from three independent experiments.

Figure 6.. Proposed biosynthetic pathway for dehydrofosmidomycin.

Putative intermediates and enzymes are shown. See text for detailed discussion.