Molecular and clinical analyses of c-CBL mutations affecting tyrosine 371 in myelodysplastic and proliferative disorders

CBL proteins constitute a family of RING finger E3 ubiquitin ligases which modulate proliferative signaling pathways via the ubiquitination and subsequent degradation of protein tyrosine kinases (PTKs), including but not limited to KIT, CSF-1, FLT3, and PDGF [ 1 ]. We previously identified recurrent missense mutations affecting the zinc finger and linker region domains (ZF/LR, 352-420 residues) of c-CBL in myelodysplastic/myeloproliferative (MDS/MPN) syndromes, such as in chronic myelomonocytic leukemia (CMML) and juvenile myelomonocytic leukemia (JMML) [ 2 , 3 ]. Indeed, c-CBL mutations correlate with myelomonocytic proliferation, resulting in CMML and JMML (10-15%, both) [ 1 , 2 , 3 ], as prototypic phenotypes, followed in frequency by MDS and acute myeloid leukemia (AML, 1–2% both) [ 4 ], in which c-CBL hits represent late event in the process of secondary leukemogenesis [ 4 , 5 ]. Although in AML prognostic models (e.g., ELN 2022) the inclusion of c-CBL mutations [ 6 ] is recommended as part of diagnostic test, no prominent role is assigned to these mutations, while, in IPSS-M, c-CBL is included in the group of 16 prognostic genes [ 7 ]. Recent studies showed that mutations occurring in ZF/LR domains not only disrupt c-CBL’s ubiquitin ligase activity but also augment PI3K/AKT pathway activation via phosphorylation of c-CBL at tyrosine residues Y700, Y731, and Y774 by LYN and FYN kinases, thus sustaining RAS pathway signaling [ 8 ]. Furthermore, X-ray diffraction analysis of c-CBL protein showed that mutations at residue Y371 can result in distinct conformational changes, possibly affecting disease phenotypes [ 9 ]. Moreover, c-CBL 371 appears to exhibit altered sensitivity to degradation by proteases [ 10 ] and may play a key role in enhancement of FLT3 signaling [ 11 ]. In light of the above recent data, we here plan to dissect whether c-CBL 371 may have distinctive impact on clinical phenotype, co-mutational associations, and outcomes compared with other ZF/LR hits. To that end, we screened a large population of myeloid neoplasia (MN) ( n = 12,084) from Cleveland Clinic, University of Rome Tor Vergata, and published sources [ 12 ], and identified 312 c-CBL mutated cases. We then performed a sub-analysis on MDS and MDS/MPN cohorts of 83 patients harboring ZF/LR c-CBL missense mutations. Only a single splice site mutation was identified (c.1096-1 G > C) and was excluded from the analysis for consistency. All the mutations have been described in Table S1 . We also interrogated two large publicly available datasets [ 13 , 14 ] to assess c-CBL hits prevalence as drivers for clonal hematopoiesis of indeterminate potential (CHIP). Information on genetic screening is available in Supplementary Materials and Table S2 . Our c-CBL mutant cohort ( n = 83) included 11 patients with Y371 mut residue and 70 patients with other ZF/LR hits (2 carried both mutation types). Consistent with the literature, Y371 ( n = 13) hits were also common together with R420 ( n = 14) and C404 ( n = 13) (Fig. 1A, B ; Fig. S1 ). Except for younger age (median age 71 vs 75; P = 0.034; Table S3 ), no differences in terms of disease phenotype, gender, complete blood count, and bone marrow blast percentage were found when compared to carriers of other ZF/LR mutations. Fig. 1: Study population and evidence on CHIP subjects. Full size image A Representation of the overall population, including 3,301 MDS and 622 MDS/MPN patients. 83 cases harbored c-CBL ZF/LR mutations (33 MDS, 50 MDS/MPN). Among them, 11 patients carried Y371 mutations, 70 carried other ZF/LR mutations, and 2 had both. B Schematic protein structure of c-CBL highlighting major domains (tyrosine kinase binding, linker region, zinc finger domain, and ubiquitin-associated domain). Recurrent mutations are mapped along the ZF/LR hotspot, with Y371 representing one of the most frequently mutated residues. For the aim of this research, only missense variants were studied. The lower histogram shows the frequency distribution of patients harboring mutations at each residue. C Genovese et al. cohort included 12380 individuals, 12325 healthy subjects (HS) with no prior history of malignancies, and 55 with a previous hematologic malignancy (HM). A single c-CBL Y371 mutation was detected in HM, while 2 ZF/LR mutations were detected in HS population ( P < 0.0001). D Kar et al. investigated the presence of CHIP in the UK Biobank cohort, including 200453 HS. Comparing the distribution of c-CBL mutations in UK Biobank cohort with the one of our MN case series, we confirmed that Y371 mutations were absent in CHIP carriers, while other ZF/LR variants occurred in both CHIP and MN. CHIP: clonal hematopoiesis of indeterminate potential, MDS myelodysplastic neoplasm, MDS/MPN myelodysplastic/myeloproliferative neoplasm, MN myeloid neoplasia, ZF/LR zinc finger/linker region. To gain insights in the relationship between c-CBL variants and CHIP, we first interrogated the database from Genovese et al. [ 13 ], including 12380 individuals. Most of the cases had no prior malignancies (healthy subjects, HS) while others had a history of hematologic malignancies (HM) ( n = 55). In the latter group a higher incidence of CHIP compared with HS was observed (25% vs 2%). Among CHIP cases ( n = 279) 3 c-CBL mutations were detected (one Y371H and two R420Q). Notably, the mutation affecting Y371 was detected in the HM cohort, while other ZF/LR domain mutations were found in HS (7.1% vs 0.7%; P = 0.01; Fig. 1C ). Of note is that the patient harboring Y371 variant died of MDS [ 13 ]. In the second database, comprising 200,453 individuals from the UK Biobank, CHIP was identified in 5.4%, including 105 (0.5%) c-CBL -driven cases; however, no Y371 mutations were detected, in contrast to our MN cohort (13/93 cases; P = 0.0002) (Fig. 1D , Fig. S1 ) [ 14 ]. This finding possibly suggests that while other c-CBL mutations can be found in CHIP, the more virulent Y371 variant advances quickly to manifest MN skipping the CHIP stage. Figure 2A shows the genetic landscape of the study population. When we compared the cohort of Y371 vs other ZF/LR mutations (Fig. 2B , Fig. S2 ), we detected an enrichment in high-risk karyotypic abnormalities (del(7): 18 vs 1%, P = 0.014; del(17): 9 vs 0%, P = 0.014) and MPN-related and unfavorable mutations ( CALR : 9 vs 0%, P = 0.011, lnOR 3; KIT : 18 vs 3%, P = 0.029, lnOR 2.02; SETBP1 : 36 vs 7%, P = 0.004, lnOR 2; TP53 : 18 vs 1%, P = 0.006, lnOR 2.73) in Y371 cohort. Figure 2C recapitulates the main genomic differences between Y371 and ZF/LR cohorts. Two unique patients presented both types of mutations; in both cases, c-CBL hits co-occurred with high-risk mutations and presented dismal outcome (Table S4 ). When stratifying patients according to IPSS-M [ 7 ], we observed a trend of clustering toward high-risk class in Y371-mutated patients ( P = 0.06; Fig. S3 ). We then analyzed the clonal hierarchy of the mutations observing a similar distribution between the two groups: the Y371-mutant cohort included 6 cases with ancestral/co-dominant mutations (54%) and 5 with sub-clonal (46%) hits, while ZF/LR cohort included 30 cases with ancestral/co-dominant mutations (45%; among them, 4 had only c-CBL mutation) and 37 with sub-clonal hits (55%; P = 0.1281; Table S3 , Fig. S4 ). Among cases with sub-clonal c-CBL , the genetic drivers belonged mainly to the spliceosome complex ( SF3B1, SRSF2, U2AF1, ZRSR2 ) and epigenetic regulators ( ASXL1, EZH2, TET2, IDH1/2 ) (Fig. S5 ). Fig. 2: c-CBL co-mutational landscape. Full size image A Oncoplot representing karyotype abnormalities (upper box) and somatic mutation (lower box) in the study cohort divided by the type of c-CBL mutation. B Correlation plot: the colorimetric scale indicates the strength of correlation, while the size of the dots the significance, as shown by the legend next to the plot. C Histogram showing significant differences among karyotypic abnormalities/ genetic mutations between Y371 and ZF/LR hits. D Kaplan-Meier curve for OS. E Kaplan-Meier curve for LFS. aCML atypical chronic myeloid leukemia, CMML chronic myelomonocytic leukemia, LFS: leukemia-free survival, MDS myelodysplastic neoplasm, MDS/MPN-NOS myelodysplastic/myeloproliferative neoplasm- Not otherwise specified, OS Overall survival, ZF/LR zinc finger/linker region. To assess potential differences in terms of outcome, we performed a survival analysis including a comparator of 1209 c-CBL WT MDS-MDS/MPN. As shown in Fig. 2D, E we observed a poor outcome in patients harboring Y371 mutations both in terms of OS ( P = 0.0018) and LFS ( P = 0.01). Given the correlation between high-risk features, we ran a multivariable analysis for survival including age, diagnosis (MDS vs CMML) and IPSS-M score confirming the independent role of Y371 mutations as predictor of inferior OS ( P = 0.011, HR: 2.673, 95%CI: 1.249–5.72; Fig. S6A ) and LFS ( P = 0.036, HR: 2.248, 95%CI: 1.052–4.803; Fig. S6B ). Since c-CBL mutation is included (as a secondary role mutation) as variable in IPSS-M risk score, we assessed the model for possible redundancy among the prognosticators, observing, via collinearity analysis and regression coefficient analysis, no overlap between the variables (Tables S5 – 7 ). Among Y371 mutant cases, 8 patients died, most of them with active disease (7, of whom one with AML progression) while only one patient was in remission of the disease at last follow-up (Table S8 ). Then, we explored the response to therapies in our population: among 14 patients with ZF/LR mutations with evaluable response to HMA, complete remission rate was 22% vs 0% in Y371 mutant cases (3 patients; P = ns; Fig. S7 ). To substantiate our molecular associations, we applied an unbiased/unsupervised machine learning-based molecular clustering algorithm to identify distinct combinatorial molecular signatures [ 12 ]. Among the 14 previously described molecular clusters (MC) c-CBL ZF/LR mutant cases grouped mainly to MC6 (15/70; characterized by enrichment of SRSF2 and RAS mutations), and MC9 (13/70; typified by normal karyotype and good response to HMA). Patients with Y371 mut were mainly grouped to MC11 (2/11), characterized by enrichment of del(7), RAS and TET2 mutations, dismal outcome and lower response to therapies, MC7 (2/11), characterized by younger age and low response to treatments, and MC12 (2/11), characterized by enrichment in high-risk MDS and CMML, high IPSS-M and high risk mutations, such as ASXL1 , SRSF2 , RUNX1 hits (Fig. S8 , Table S9 ). Our study showed that c-CBL 371 hits present a highly virulent nature, incompatible with CHIP phenotype, and a peculiar genetic signature, characterized by both high-risk karyotypic abnormalities and genetic mutations [del(7), del(17), TP53, SETBP1 ] but also genetic variants ( CALR, KIT ) correlating with MDS/MPN phenotype. Remarkably, regardless of the co-mutational landscape, Y371 has an independent role in predicting both inferior OS and LFS. One possible cause of this dismal prognosis might be the poor response to HMA, which needs to be confirmed in larger studies. Limitations of the present study include a small sample size, due to the rarity of the variants object of the research, which prevent precise comparisons, such as unbiased clustering analysis of different c-CBL variants and inferences from CHIP cohorts. However, these findings might pose a significant issue in the interpretation of the results coming from trials specifically evaluating c-CBL mutated MN (NCT04493138) with azacitidine in combination with quizartinib [ 15 ]. Furthermore, we conclude that the impact of c-CBL Y371 mutations may not be adequately reflected by IPSS-M risk score. Future studies are warranted to confirm the differential processes of leukemogenesis between hits occurring at Y371 and other ZF/LR variants.