Cell Subset Targeting of Glycosylated Polyketides Revealed by Multiplexed Phenotypic Screening of Natural Product Fraction Libraries and Bioengineered Analogs

Abstract

Bioactive natural products are small molecules produced by organisms of all domains to influence and respond to environmental conditions. These molecules have evolved to recognize both intra and inter cellular targets and the conserved homology of protein folds and motifs shared across all organisms allows for serendipitous activity in non-native contexts. The importance of the intersection of biological and chemical space within this privileged subset of compounds is evidenced by the fact that a majority of clinically approved small molecule therapeutics are natural products or derivatives. However, the workflow for natural product chemistry remains laborious and resource intensive. One barrier is the structural complexity of natural products, which limits the utility of medicinal chemistry in improving pharmacological properties and developing analogs as chemical probes. A second issue is that discovery of an active compound depends on the nature of the assay. A natural product may be inactive in one screen but active if a different readout or other assay condition is chosen. Finally, given an active extract, purifying and identifying the responsible constituent efficiently is challenging. This work presents research into the above areas. First, a combination of targeted biosynthetic gene inactivation, precursor directed biosynthesis, and semi-synthesis was used to generate analogs of the apoptolidins, a family of cytotoxic glycosylated polyketides. Analogs were characterized in terms of potency and cellular uptake and localization. Second, a general method for multiplexing natural product screening assays across multiple readouts and cell types is demonstrated. Data from multiplexed assays is linked to cheminformatic characterization of extracts to rank order chemical species for isolation. Applications of this approach identified heretofore unknown differential targeting of lymphoid, myeloid, and blast populations in primary AML patient samples by apoptolidin A, ciromicin B, and ciromicin A, respectively.