Mezigdomide overcomes CRBN mutations emerging post IMiD therapy

Abstract Background: Cereblon (CRBN), a substrate receptor for the CRL4CRBN E3 ligase complex, is the molecular target of immunomodulatory agents (IMiDs) and next generation CELMoDs™, whose mechanism involves recruitment of neosubstrates for ubiquitination and subsequent proteasomal degradation. Their activity results in significant clinical benefit to multiple myeloma (MM) patients. Several mechanisms of resistance to IMiDs/CELMoDs have been described including genomic dysregulation of the CRL4CRBN E3 ligase pathway by copy number aberrations, alternative splicing, and mutations in CRBN. Here we describe the landscape of CRBN mutations in patients post IMiD treatment and through modeling, structure, biochemical and functional assays, we probe the consequences and their potential impact to IMiD and CELMoD agent activities pre-clinically and clinically. Results: CD138+ sorted samples from MM patients across several clinical studies from pomalidomide (POM; CC-4047), iberdomide (IBER: CC-220), and mezigdomide (MEZI; CC-92480) based regimens were profiled by whole genome sequencing (WGS) from patients with >2-5+ prior lines of therapy. Somatic mutations were called using the mutect2 best practices pipeline with a paired normal germline sample. Target WGS depth was 60x allowing detection of approximately 5% VAF. Landscape of CRBN mutations across the combined relapsed/refractory cohort (n=483) identified a total of 34 unique variants across the entire span of the coding sequence including missense (n=26), indels (n=4), frameshift (n=2), and nonsense (n=2). No hotspot mutations were identified. There were many subclonal mutations with lower VAFs, with the lowest identified of 8% VAF (20% CCF). Mutations were then modeled in silico using Rosetta ddG calculations on the open and closed conformation of cereblon to assess their potential effect on protein stabilization/destabilization on either state. Experimentally determined structures of cereblon in the open and closed states were used with ligands (POM, IBER or MEZI), either from their crystallographic poses in the closed state or modeled into the open state. Based on this 3D modeling, CRBN mutations were classified as having a potential of no impact, disruption of the N-terminal belt, dissociation from DDB1, drug recruitment and protein destabilization. Metadynamics simulations were carried out on mutants of particular interest to dive deeper into potential conformational preferences of CRBN mutants compared to the wildtype apo state. Computational predictions helped to prioritize which mutants to focus on for additional experimental studies. One missense mutation, A365D, was selected to study functional activities of POM, IBER and MEZI with analyses of conformational stabilities, cryo-EM structure, substrate ubiquitination and degradation, and cell autonomous activity by mutant overexpression in cereblon-null MM cell lines. In silico metadynamics simulations with CRBN with A365 mutation in complex with MEZI indicated stabilization of the active, closed conformation over the open conformation. On the other hand, apo state or binding of POM stabilizes the open conformation of A365D mutant which lacks degradation capability. Structural studies showed the potential for MEZI to overcome this deficiency, securing a closed conformation of A365D which can additionally recruit and target Ikaros via reconstituted ubiquitination. MEZI was further confirmed to induce degradation of IKZF-family members in MM cell lines despite the clinical A365D mutation. Functional anti-proliferative and pro-apoptotic activities in cells harboring the A365D mutation shows cellular sensitivity with either IBER or MEZI, but not POM or LEN. Lastly, a patient harboring the A365D mutation was shown to have a partial clinical response to a MEZI-based regimen. Conclusions: Somatic CRBN missense mutations in MM patients who have relapsed on prior IMiD-based therapies are rare and not recurrent. In silico modeling through Rosetta ddG and metadynamics can predict the potential for IMiD or CELMoD induced conformational changes associated with an open/closed state. MEZI has the potential to overcome mutations by interacting with additional moieties outside of the TBD that may affect the stabilization of cereblon or the CRL4CRBN E3 ligase complex resulting in clinical responses whereas POM does not.