C-CBL mutations affecting residue 371 confer dismal prognosis in MDS and CMML

Abstract

Abstract In myeloid neoplasm (MN), c-CBL missense mutations affecting the linker region and zinc finger domains (LR/ZF, 352-420 residues) impair the ubiquitination function while preserving PI3K/AKT activation and are linked to an MPN phenotype. Among hits occurring at LR/ZF domains, X-ray diffraction analysis of c-CBL protein showed that mutations at residue 371 can result in different conformational changes, possibly affecting disease phenotypes. Furthermore, a recent study showed that tyrosine 371 has a key role in enhancing FLT3 signaling. We hypothesize that these hits may correspond to different consequences including preference or specific computational events, clinical phenotypes or outcomes. To address this stipulation we gathered a large cohort of MDS (n=3301) and MDS/MPN (n=622) from Cleveland Clinic and University of Rome Tor Vergata as well as metanalytic sources, identifying a total of 83 patients harboring LR/ZF c-CBL mutations (33 MDS [1%], 50 MDS/MPN [8%]; 93 hits). Our c-CBL mutated cohort included 11 patients with 371mut residue and 70 patients with other LR/ZF hits (2 carried both mutation types). To corroborate our hypothesis, we compared the aminoacidic localization of the missense mutations in our MN cohort vs a genomically annotated CHIP cohort derived from 200,453 healthy individuals (105 harbouring c-CBL mutations). Remarkably,we detected no differences in LR/ZF domains while no 371 residue hits were found in CHIP suggesting their stronger pathogenicity or dependence on the presence of preceding founder mutation (0 vs 13% in our case series; p=0.002). This finding prompted further exploration of this alteration. Comparing patients with 371mut residue vs other LR/ZF hits, we observed, in the first cohort, younger age (71 vs 75 years, p=0.0344) and no differences in terms of complete blood count or blast percentage. Furthermore, the first cohort was enriched in high-risk karyotypic abnormalities (-7/del(7q): 18 vs 1%, p=0.014; del(17): 9 vs 0%, p=0.014) and MPN and unfavorable mutations (CALR: 9 vs 0%, p=0.011; KIT: 18 vs 3%, p=0.029; SETBP1: 36 vs 7%, p=0.004; TP53: 18 vs 1%, p=0.006). Among 14 patients with LR/ZF mutations with evaluable response to HMA, CR rate was 22% vs 0% in 371mut patients (3 patients). Furthermore, c-CBL-371mut showed a higher VAF-based clonal hierarchy (ancestral hits in 63% vs 25%, p=0.0312). Survival analysis, including a comparator cohort of 1209 c-CBLwt MDS and MDS/MPN, showed a correlation between c-CBL-371mut and dismal outcome independent of age, IPSS-M and diagnosis (p=0.027; HR:2.744, 95%CI: 1.119-6.732). Similar results were obtained when focusing on AML-free survival (p=0.02; HR:1.311, 95%CI: 1.043-1.649). Most of the patients died with active disease (7, of whom one with AML progression) while only one patient died in remission. We then decided to investigate (in an unbiased/unsupervised fashion) these types of mutations by applying unsupervised re-clustering of c-CBL mutants using our original cohort of patients3. Among the 14 previously identified molecular clusters, ZF/LR-mutated cases were predominantly associated with cluster 6 (characterized by enrichment of SRSF2 and RAS mutations), 9 and 12 (both characterized by normal karyotype and good response to HMA), while 371mut patients were assigned to cluster 7 (p=0.009), characterized by younger age and lower response to HMA. As proof of concept, to investigate the biological underpinning of these differences, we utilized BeatAML public dataset and compared transcriptome analysis of patients harboring c-CBL 371mut(1 patient) vs LR/ZF mutation (1 patient harboring p.R420Q mutation) vs FLT3-ITD mutation. Remarkably, FLT3-ITD and c-CBL 371mut cases exhibited a similar expression of genes involved in proliferation pathways (JAK2, LYN, FYN, STAT5B, AKT1, MTOR, KRAS, NRAS), enhanced when compared to the one harboring c-CBL 420mut. Our study shows differences among c-CBL mutant MDS and MDS/MPN patients. The features of mutations affecting 371 residue vs other ZF/LR domains in terms of survival, co-mutational landscape, transcriptomic profile, incidence in CHIP population and unbiased clustering suggest different leukemogenesis processes. These differences should be considered in molecular prognostic scoring systems.