Inhibitors of the Hsp90 molecular chaperone are showing considerable promise as potential molecular therapeutic agents for the treatment of cancer. Here we describe novel 2-aminothieno[2,3-d]pyrimidine ATP competitive Hsp90 inhibitors, which were designed by combining structural elements of distinct low affinity hits generated from fragment-based and in silico screening exercises in concert with structural information from X-ray protein crystallography. Examples from this series have high affinity (IC50 = 50-100 nM) for Hsp90 as measured in a fluorescence polarization (FP) competitive binding assay and are active in human cancer cell lines where they inhibit cell proliferation and exhibit a characteristic profile of depletion of oncogenic proteins and concomitant elevation of Hsp72. Several examples (34a, 34d and 34i) caused tumor growth regression at well tolerated doses when administered orally in a human BT474 human breast cancer xenograft model.
In the past 15 years, fragment-based lead discovery (FBLD) has been adopted widely throughout academia and industry. The approach entails discovering very small molecular fragments and growing, merging, or linking them to produce drug leads. Because the affinities of the initial fragments are often low, detection methods are pushed to their limits, leading to a variety of artifacts, false positives, and false negatives that too often go unrecognized. This Digest discusses some of these problems and offers suggestions to avoid them. Although the primary focus is on FBLD, many of the lessons also apply to more established approaches such as high-throughput screening.
We have designed four generations of a low molecular weight fragment library for use in NMR-based screening against protein targets. The library initially contained 723 fragments which were selected manually from the Available Chemicals Directory. A series of in silico filters and property calculations were developed to automate the selection process, allowing a larger database of 1.79 M available compounds to be searched for a further 357 compounds that were added to the library. A kinase binding pharmacophore was then derived to select 174 kinase-focused fragments. Finally, an additional 61 fragments were selected to increase the number of different pharmacophores represented within the library. All of the fragments added to the library passed quality checks to ensure they were suitable for the screening protocol, with appropriate solubility, purity, chemical stability, and unambiguous NMR spectrum. The successive generations of libraries have been characterized through analysis of structural properties (molecular weight, lipophilicity, polar surface area, number of rotatable bonds, and hydrogen-bonding potential) and by analyzing their pharmacophoric complexity. These calculations have been used to compare the fragment libraries with a drug-like reference set of compounds and a set of molecules that bind to protein active sites. In addition, an analysis of the overall results of screening the library against the ATP binding site of two protein targets (HSP90 and CDK2) reveals different patterns of fragment binding, demonstrating that the approach can find selective compounds that discriminate between related binding sites.
Fragment-based drug discovery typically requires an interplay between screening methods, structural methods, and medicinal chemistry. X-ray crystallography is generally the method of choice to obtain three-dimensional structures of the bound ligand/protein complex, but this can sometimes be difficult, particularly for early, low-affinity fragment hits. In this Perspective, we discuss strategies to advance and evolve fragments in the absence of crystal structures of protein-fragment complexes, although the structure of the unliganded protein may be available. The strategies can involve other structural techniques, such as NMR spectroscopy, molecular modeling, or a variety of chemical approaches. Often, these strategies are aimed at guiding evolution of initial fragment hits to a stage where crystal structures can be obtained for further structure-based optimization. Essentially, all models are wrong, but some are useful.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.