AUTHORS
Catherine Cargo I, PhD, FRCPath, MB, MBBCh, Marco Roncador, MD, MSc2, Jacqueline Ferrari, MD3, Michela Bardelli4, Martina Boldini, MD5, Virginia Camilotto6, Martina Gandossini7, Paul Glover, MSc8, Daniel Painter9, Sara Pozzi10, Martina Sarchi, MD, PhD11, Suzan van Hoppe12, Helen Warren13, Anna Gallì, Bsc14, Simon Crouch15, Chiara Elena, MD16, Alexandra Smith, PhD17, Luca Malcovati, MD18
AFFILIATIONS
1 Haematological Malignancy Diagnostic Service, St James's University Hospital, Leeds, United Kingdom
2 Unit of Molecular Hematology and Precision Medicine, Fondazione IRCCS Policlinico San Matteo & University of Pavia, Pavia, Italy
3 Unit of Molecular Hematology and Precision Medicine, Fondazione IRCCS Policlinico San Matteo & University of Pavia, Pavia, Italy
4 Unit of Molecular Hematology and Precision Medicine, Fondazione IRCCS Policlinico San Matteo & University of Pavia, Pavia, Italy
5 Unit of Molecular Hematology and Precision Medicine, Fondazione IRCCS Policlinico San Matteo & University of Pavia, Pavia, Italy
6 Unit of Molecular Hematology and Precision Medicine, Fondazione IRCCS Policlinico San Matteo & University of Pavia, Pavia, Italy
7 Unit of Molecular Hematology and Precision Medicine, Fondazione IRCCS Policlinico San Matteo & University of Pavia, Pavia, Italy
8 Haematological Malignancy Diagnostic Service, St James's University Hospital, Leeds, United Kingdom
9 Epidemiology and Cancer Statistics Group, Department of Health Sciences, University of York, York, United Kingdom
10 Unit of Molecular Hematology and Precision Medicine, Fondazione IRCCS Policlinico San Matteo & University of Pavia, Pavia, Italy
11 Unit of Molecular Hematology and Precision Medicine, Fondazione IRCCS Policlinico San Matteo & University of Pavia, Pavia, Italy
12 Haematological Malignancy Diagnostic Service, St James's University Hospital, Leeds, United Kingdom
13 Haematological Malignancy Diagnostic Service, St James's University Hospital, Leeds, United Kingdom
14 Unit of Molecular Hematology and Precision Medicine, Fondazione IRCCS Policlinico San Matteo & University of Pavia, Pavia, Italy
15 Epidemiology and Cancer Statistics Group, Department of Health Sciences, University of York, York, United Kingdom
16 Unit of Molecular Hematology and Precision Medicine, Fondazione IRCCS Policlinico San Matteo & University of Pavia, Pavia, Italy
17 Epidemiology and Cancer Statistics Group, Department of Health Sciences, University of York, York, United Kingdom
18 Unit of Molecular Hematology and Precision Medicine, Fondazione IRCCS Policlinico San Matteo & University of Pavia, Pavia, Italy
ABSTRACT
Introduction
Myeloid precursor states (pre-MN) are now defined entities in the revised classifications of myeloid neoplasms (MN). Although disease-defining genetic lesions allow diagnosis of MN irrespective of marrow dysplasia, the distinction between pre-MN and early MN stages still relies heavily on morphologic criteria, which are limited by poor specificity and significant inter-observer variability. The aim of this study was to identify additional molecular signatures with high specificity and predictive value that inform a more accurate segregation of pre-MN and MN.
Methods
This study included a cohort of 1601 patients with CCUS (clonal cytopenia of undetermined significance; n=513), CMUS (clonal monocytosis and undetermined significance; n=37) and early-stage MN, including MDS (myelodysplastic neoplasms/syndromes; n=749) or MDS/MPN (MDS/myeloproliferative neoplasms; n=302) without excess blasts, investigated between 2000 and 2024, at two reference centers for myeloid malignancies, the IRCCS Policlinico San Matteo Foundation & University of Pavia in Pavia, Italy, and the Haematological Malignancy Diagnostic Service, Leeds, UK. Patients were diagnosed according to international recommendations and revised classifications and underwent targeted myeloid gene panel sequencing alongside gold standard diagnostic techniques. Correlation was made with presenting clinical features, disease progression and overall survival. Supervised molecular assignment was complemented by unsupervised Bayesian network-based mixture modelling (clustNet); candidate solutions with 5–18 clusters were evaluated using the Akaike information criterion, without including survival and taxonomy classes for model fitting.
Results
We found that high-risk pre-MN patients defined according to the Clonal Hematopoiesis Risk Score (CHRS) and the Clonal Cytopenia Risk Score (CCRS) had clinical outcomes comparable to those with confirmed MN, supporting the notion that selected pre-MN may be more appropriately classified as MN.
We then applied the MDS molecular taxonomy classification across the total cohort and calculated the positive predictive value (PPV) for MN of each molecular taxonomy class. Besides the known disease-defining genetic lesions, the -7/SETBP1, EZH2-ASXL1, and AML-like classes exhibited the highest PPVs for MN (> 80%). The vast majority of cases within these clusters had a confirmed MN suggesting a strong clinical expressivity for bone marrow dysplasia. The remaining molecular classes were more broadly distributed across pre-MN and MN and for these clone metrics/hematologic variables were compared to identify molecular and clinical differences. For the classes defined by splicing factor genes (SF3B1, SRSF2, U2AF1, ZRSR2) and biallelic-TET2 (bi-TET2) or TET2-SRSF2 the pre-MN and MN cases within each class had comparable clone metrics and clinical outcomes, suggesting that a uniform classification as MN is more consistent irrespective of the presence of significant dysplasia. In contrast the CCUS-like class, defined by mutations in TET2, DNMT3A, TP53 monoallelic or loss of Y had the lowest specificity for a MN and showed distinct differences in clinical features and patient outcomes between those with pre-MN and MN.
Finally, unsupervised clustering analysis performed across the cohort consistently recognized splicing factors, TET2-SRSF2 and EZH2-ASXL1 as molecular signatures with homogeneous clinical features and outcomes irrespective of marrow dysplasia. Cluster assignments were significantly concordant with the molecular taxonomy and every cluster was significantly enriched for its corresponding taxonomy class (P<.001). Importantly, these clusters spanned both pre-MN and MN, indicating that the class-defining signatures, rather than the pre-MN/MN boundary, drove cluster assignment.
Conclusion
In conclusion, our study provides evidence that selected molecular signatures are more robust than morphologic dysplasia to define homogeneous classes across the spectrum of pre-MN and early MN. In particular, EZH2-ASXL1, splicing factor mutations and TET2-SRSF2, in the context of unexplained cytopenia or monocytosis supports a diagnosis of MN irrespective of morphologic criteria. Conversely, in cases with other molecular patterns, BM dysplasia remains a meaningful criterion. Together, these findings advance the molecular diagnosis of myeloid malignancies and their precursor states, improving the accuracy of class assignment in everyday practice.
DISCLOSURES
Catherine Cargo I, PhD, FRCPath, MB, MBBCh: Astellas, GSK, Servier - Honoraria