Summary Chemical libraries paired with phenotypic screens can now readily identify compounds with therapeutic potential. A central limitation to exploiting these compounds, however, has been in identifying their relevant cellular targets. Here, we present a two-tiered CRISPR-mediated chemical-genetic strategy for target identification: combined genome-wide knockdown and overexpression screening as well as focused, comparative chemical-genetic profiling. Application of these strategies to rigosertib, a drug in phase III clinical trials for high-risk myelodysplastic syndrome whose molecular target had remained controversial, pointed singularly to microtubules as rigosertib’s target. We showed that rigosertib indeed directly binds to and destabilizes microtubules using cell biological, in vitro, and structural approaches. Finally, expression of tubulin with a structure-guided mutation in the rigosertib-binding pocket conferred resistance to rigosertib, establishing that rigosertib kills cancer cells by destabilizing microtubules. These results demonstrate the power of our chemical-genetic screening strategies for pinpointing the physiologically relevant targets of chemical agents.

Quinazolinone-based anticancer agents were designed, decorated with functional groups from a 2-methoxyestradiol-based microtubule disruptor series, incorporating the aryl sulfamate motif of steroid sulfatase (STS) inhibitors. The steroidal AB-ring system was mimicked, favoring conformations with an N-2 substituent occupying D-ring space. Evaluation against breast and prostate tumor cell lines identified 7b with DU-145 antiproliferative activity (GI 300 nM). A preliminary structure-activity relationship afforded compounds (e.g., 7j GI 50 nM) with activity exceeding that of the parent. Both 7b and 7j inhibit tubulin assembly in vitro and colchicine binding, and 7j was successfully co-crystallized with the αβ-tubulin heterodimer as the first of its class, its sulfamate group interacting positively at the colchicine binding site. Microtubule destabilization by 7j is likely achieved by preventing the curved-to-straight conformational transition in αβ-tubulin. Quinazolinone sulfamates surprisingly showed weak STS inhibition. Preliminary in vivo studies in a multiple myeloma xenograft model for 7b showed oral activity, confirming the promise of this template.

Microtubule-targeting agents (MTAs) like taxol and vinblastine are among the most successful chemotherapeutic drugs against cancer. Here, we describe a fluorescence anisotropy-based assay that specifically probes for ligands targeting the recently discovered maytansine site of tubulin. Using this assay, we have determined the dissociation constants of known maytansine site ligands, including the pharmacologically active degradation product of the clinical antibody-drug conjugate trastuzumab emtansine. In addition, we discovered that the two natural products spongistatin-1 and disorazole Z with established cellular potency bind to the maytansine site on β-tubulin. The high-resolution crystal structures of spongistatin-1 and disorazole Z in complex with tubulin allowed the definition of an additional sub-site adjacent to the pocket shared by all maytansine-site ligands, which could be exploitable as a distinct, separate target site for small molecules. Our study provides a basis for the discovery and development of next-generation MTAs for the treatment of cancer.

Highlights d Pharmaceutical-grade rigosertib kills cells by destabilizing microtubules d Pharmaceutical-grade rigosertib destabilizes microtubules in cells and in vitro d A rigosertib-binding mutation in tubulin alleviates rigosertibmediated toxicity

The association of DNA Ligase IV (Lig4) with XRCC4 is essential for repair of DNA double-strand breaks (DSBs) by Non-homologous end-joining (NHEJ) in humans. DSBs cytotoxicity is largely exploited in anticancer therapy. Thus, NHEJ is an attractive target for strategies aimed at increasing the sensitivity of tumors to clastogenic anticancer treatments. However the high affinity of the XRCC4/Lig4 interaction and the extended protein-protein interface make drug screening on this target particularly challenging. Here, we conducted a pioneering study aimed at interfering with XRCC4/Lig4 assembly. By Molecular Dynamics simulation using the crystal structure of the complex, we first delineated the Lig4 clamp domain as a limited suitable target. Then, we performed in silico screening of ~95,000 filtered molecules on this Lig4 subdomain. Hits were evaluated by Differential Scanning Fluorimetry, Saturation Transfer Difference - NMR spectroscopy and interaction assays with purified recombinant proteins. In this way we identified the first molecule able to prevent Lig4 binding to XRCC4 in vitro. This compound has a unique tripartite interaction with the Lig4 clamp domain that suggests a starting chemotype for rational design of analogous molecules with improved affinity.