A neomorphic protein interface catalyzes covalent inhibition of RAS G12D aspartic acid in tumors

INTRODUCTIONMutant RAS is the primary oncogenic driver in RAS-addicted cancers. RAS proteins act as molecular switches that cycle between a guanosine triphosphate–bound “ON” state and a guanosine diphosphate–bound “OFF” state. Under normal circumstances, cell growth is tightly regulated by modulating the population of RAS in the ON versus the OFF state. In RAS-addicted cancers, mutations in RAS shift this balance towards the ON state, leading to increased RAS signaling and uncontrolled cell growth. Mutations in RAS are present in nearly a quarter of a million new cancer cases each year and are particularly frequent in pancreatic, colorectal, and lung cancers.RATIONALEHistorically, RAS-addicted cancers have been challenging to treat with targeted drug therapies. Mutant-selective inhibitors that covalently and irreversibly inactivate mutant RAS are advantageous because they target the mutant protein present only in cancer cells and can achieve sustained target engagement even in the context of variable tumor exposures that often result from typical inhibitor dosing schedules. Such irreversible inhibitors of KRASG12C, in which glycine is replaced with a reactive thiol-containing cysteine residue, have been approved for both lung and colorectal cancers, but no similar therapies are available for the most common RAS mutation, KRASG12D. This mutation introduces a carboxylate-containing aspartic acid residue that has low reactivity and high abundance on protein surfaces compared with cysteine, thereby posing substantial hurdles to the design of irreversible inhibitors that have sufficient potency and selectivity yet maintain the properties necessary to enable once-daily oral dosing.RESULTSTo create covalent inhibitors of KRASG12D, we used structure-based drug design to modify compounds that bind the abundant intracellular chaperone cyclophilin A (CYPA) and create a neomorphic protein-protein interface between CYPA and active RAS to covalently modify the D12 mutation located in the induced pocket at the interface. Precisely positioning reactive groups within this privileged environment enabled selective, enzyme-like rate enhancement of the covalent reaction between D12 and aziridine warheads that have low intrinsic reactivity. X-ray crystal structures and computational methods confirmed a role for the protein-protein interface in enabling selective reactivity. This approach yielded the investigational agent zoldonrasib (RMC-9805), currently undergoing clinical evaluation (NCT06040541), and the preclinical compound RMC-9945. These compounds efficiently covalently engaged RASG12D and potently suppressed oncogenic RAS signaling in RASG12D-mutant cancer cell lines in a CYPA-dependent manner. Consistent with the mechanism of action, the compounds exhibited low-potency noncovalent inhibition of RAS-driven proliferation in cell lines and patient-derived organoids with wild-type and non-G12D mutant RAS, and covalency conferred selectivity and durability of inhibition toward RASG12D. The chemical and metabolic stability was sufficient to enable once-daily oral dosing in mice, and zoldonrasib displayed marked antitumor activity in multiple preclinical models of KRASG12D-mutant pancreatic, lung, and colorectal cancers.CONCLUSIONCreation of a neomorphic protein-protein interface at RAS through chemical remodeling of the cellular chaperone CYPA selectively accelerated covalent bond formation between an inhibitor and an aspartic acid residue, enabling discovery of potent and irreversible inhibitors of the most common RAS mutation in human cancers. Zoldonrasib is an orally bioavailable, RAS(ON) G12D–selective covalent inhibitor that drives deep and durable tumor regressions in multiple preclinical models of KRASG12D cancers across indications. This strategy has the potential to greatly expand the repertoire of residues on cancer drivers or other proteins of therapeutic value that can be targeted by covalent warheads and may enable additional mechanisms of target modulation.