Genome mining based on transcriptional regulatory networks uncovers a novel locus involved in desferrioxamine biosynthesis

Abstract

Bacteria produce a plethora of natural products that are in clinical, agricultural and biotechnological use. Genome mining has uncovered millions of biosynthetic gene clusters (BGCs) that encode their biosynthesis, the vast majority of them lacking a clear product or function. Thus, a major challenge is to predict the bioactivities of the molecules these BGCs specify, and how to elicit their expression. Here, we present an innovative strategy whereby we harness the power of regulatory networks combined with global gene expression patterns to predict BGC functions. Bioinformatic analysis of all genes predicted to be controlled by the iron master regulator DmdR1 combined with co-expression data, led to identification of the novel operon desJGH that plays a key role in the biosynthesis of the iron overload drug desferrioxamine (DFO) B in Streptomyces coelicolor. Deletion of either desG or desH strongly reduces the biosynthesis of DFO B, while that of DFO E is enhanced. DesJGH most likely act by changing the balance between the DFO precursors. Our work shows the power of harnessing regulation-based genome mining to functionally prioritize BGCs, accelerating the discovery of novel bioactive molecules.

Fig 1. a, Predicted gene regulatory network of Streptomyces coelicolor based on 17 well-known regulators. Each node in the network represents a (regulatory) gene, and every edge represents a PWM predicted regulatory interaction between nodes. The edges colored in dark gray indicate strong PWM prediction scores, while the lighter gray shades represent weaker interactions. Matches within BGC regions are depicted as triangles. In six regions (black circled), the matches fall within a co-expressed region, highlighting their functional relation to these compounds. b, Representation of the four co-expressed regions, including the locations of their detected TFBSs as colored dots. All predicted TFBSs have been experimentally validated in pre-existing work. The data underlying this figure can be found at https://zenodo.org/records/15106944.

Fig 2. a, Anti-correlation of gene expression between dmdR1 and its predicted regulon. Left: Pearson correlation coefficients (PCCs) between dmdR1 and all genes with a DmdR1 position weight matrix (PWM) score greater than 15 in their regulatory region. The vertical dashed line marks the refined PWM score threshold of 22.875. The horizontal dotted lines mark PCC = ±0.43, corresponding to an adjusted p-value of 0.05. Right: Target genes immediately downstream of a predicted DmdR1 binding site, ordered by decreasing PWM score. Plus and minus indicate the strand of the target gene. Genes marked with an × did not have significant co-expression with dmdR1. Binding site details are given in S2 Table, and the raw data underlying this figure can be found at https://zenodo.org/records/15106944. b, The putative regulon of DmdR1 in Streptomyces coelicolor M145. White dots indicate predicted DmdR1 binding sites. Genes are labeled by SCO number and colored by putative function. Clusters are drawn to scale, and arrows represent the direction of transcription.

Fig 3. New model for biosynthesis of desferrioxamines B and E.

a, Extracted ion chromatograms for m/z values corresponding to DFO-related metabolites in culture extracts of the knock-out mutants of SCO4048 (desJ), SCO4049 (desG), and SCO40450 (desH) compared to the parent Streptomyces coelicolor M145 strain. The desG mutant fails to produce DFOB, while a 16-fold decrease in DFOB biosynthesis was seen in desH mutants (cf. S3 Fig). Details on the mass spectrometry data can be found at https://zenodo.org/records/15106944. b, Proposed biosynthetic pathway for assembly of desferrioxamines E and B. Main biosynthetic enzymes presented in bold face. DesG and DesH balance intracellular N-hydroxy-N-succinylcadaverine (HSC) and N-hydroxy-N-acetylcadaverine (HAC) concentrations by converting HSC to HAC. In the absence of DesG and/or DesH, the cells likely fail to produce sufficient levels of HAC, thereby strongly attenuating the production of DFOB. Although DesC has been shown to be able to catalyze the acetylation of N-hydroxycadaverine in vitro, the enzyme can only modestly compensate for the loss of DesH in vivo, underlining the important role played by DesG and DesH in DFOB production (S4 Fig).