Examining disease boundaries: Genetics of myelodysplastic/myeloproliferative neoplasms

Abstract Myelodysplastic/myeloproliferative neoplasms (MDS/MPN) are clonal myeloid malignancies that are characterized by dysplasia resulting in cytopenias as well as proliferative features such as thrombocytosis or splenomegaly. Recent studies have better defined the genetics underlying this diverse group of disorders. Trisomy 8, monosomy 7, and loss of Y chromosome are the most common cytogenetic abnormalities seen. Chronic myelomonocytic leukemia (CMML) likely develops from early clones with TET2 mutations that drive granulomonocytic differentiation. Mutations in SRSF2 are common and those in the RAS‐MAPK pathway are typically implicated in disease with a proliferative phenotype. Several prognostic systems have incorporated genetic features, with ASXL1 most consistently demonstrating worse prognosis. Atypical chronic myeloid leukemia (aCML) is most known for granulocytosis with marked dysplasia and often harbors ASXL1 mutations, but SETBP1 and ETNK1 are more specific to this disease. MDS/MPN with ring sideroblasts and thrombocytosis (MDS/MPN‐RS‐T) most commonly involves spliceosome mutations (namely SF3B1) and mutations in the JAK‐STAT pathway. Finally, MDS/MPN‐unclassifiable (MDS/MPN‐U) is least characterized but a significant fraction carries mutations in TP53. The remaining patients have clinical and/or genetic features similar to the other MDS/MPNs, suggesting there is room to better characterize this entity. Evolution from age‐related clonal hematopoiesis to MDS/MPN likely depends on the order of mutation acquisition and interactions between various biologic factors. Genetics will continue to play a critical role in our understanding of these illnesses and advancing patient care.


INTRODUCTION
Much of our initial understanding of these diseases was extrapolated from studies of the relatively common MDS and MPNs. Subsequent studies, including both large collaborative and small single center ones, further advanced the knowledge of genetic pathways and their pathological consequences. In this review, we will discuss these findings and their impact on disease prognosis and pathophysiology, then discuss implications for disease characterization.

CMML
The most common of the MDS/MPNs, CMML is defined as a myeloid stem cell neoplasm characterized by presence of greater than 1000 monocytes/µl in the peripheral blood (PB), with monocytes comprising at least 10% of PB white blood cells (WBCs), evidence of dysplasia in at least one lineage, and less than 20% blasts [1]. Monocyte partitioning by flow cytometry may distinguish reactive from classical monocytes and aid in diagnosis [5]. It is often diagnosed incidentally or in the context of cytopenias or antecedent autoimmune illness [6,7].
Patients may have a dysplastic phenotype characterized by low counts or a proliferative one characterized by leukocytosis (≥13,000 WBC/µl) and hepatosplenomegaly. A notable characteristic observed in vitro is hypersensitivity to granulocyte-macrophage colony-stimulating factor (GM-CSF) [8], prompting study of the anti-GM-CSF monoclonal antibody lenzilumab [9].
Biallelic mutation of TET2 has been frequently recognized [3,11]. Mutations in the RAS pathway (e.g., NRAS, KRAS, casitas B-lineage lymphoma [CBL], PTPN11) are also frequently observed, in up to 40% of patients [12]. Karyotype abnormalities exist in less than a quarter of CMML patients [3,4] and typically involve chromosomes 7 and 8. Patients with therapy-related CMML may have more frequent cytogenetic abnormalities but carry a similar mutation burden [13].
TET2 plays an important role in DNA methylation balance by performing hydroxylation of methylated cytosine residues, facilitating demethylation of these nucleotide residues. Mutations cause a loss of function, resulting in aberrant DNA methylation, dysmyelopoiesis, and clonal expansion [14][15][16]. One model of CMML pathogenesis suggests that early clonal dominance of TET2-mutant (TET MT ) clones promotes granulomonocytic differentiation of immature progenitors [17], whereas without this early dominance, additional mutations or other stochastic factors cause evolution toward other phenotypes.
Given the increased presence of mutations associated with agerelated clonal hematopoiesis in CMML [18], notably TET2 and ASXL1, CMML appears to stem from hematopoietic cells that stochastically acquire culprit mutations over time. Ancestral TET2 mutations are most common [19] and are highly associated with biallelic TET2 mutation and TET2-SRSF2 co-mutation [3]. Additionally, ASXL1 has been found to be a common ancestral event [3,19].
The sequence of driver mutation acquisition and gene interactions likely skews disease toward either a more proliferative or dysplastic phenotype [20] with RAS pathway mutations responsible for the former [12] and SF3B1 occasionally associated with the latter [21].  [22]. These risk groups were later incorporated into the CMML-specific prognostic scoring system (CPSS) [23]. Published the same year, the Groupe Français des Myélodysplasies prognostic score included ASXL1 nonsense and frameshift mutations because they conferred poor prognosis in multivariate analysis [10], confirmed later in a larger study [24]. Subsequently, ASXL1, NRAS, RUNX1, and SET-binding protein-1 (SETBP1) were independently linked to worse OS and increased risk of leukemic transformation. These mutations were integrated with the established cytogenetic risk categories into a CMML-specific genetic score, which was incorporated into the clinical/molecular CPSS (CPSS-Mol) [25].
Although a smaller study found that patients with TET2 MT CMML-1 had adverse prognosis [11], this has not been consistently observed.
In one large, multi-institutional study, not only was TET2 MT disease associated with a significant survival advantage, but truncating or multiple TET2 mutations were particularly advantageous [26]. Additionally, the impact of ASXL1 was found to be ameliorated by co-mutation with TET2 in this study [26]. Finally, another study found mutations in STAG2 and U2AF1 correlated with worse survival [3].
There have been multiple attempts to use genetic and epigenetic signatures to predict response to therapy. While somatic mutations did not predict response to hypomethylating agent (HMA) therapy in one international study, methylation profiling of mostly nonpromoter regions distinguished responders from non-responders [27]. Gene profiling of the responders found that expression of cell cycle-related genes were upregulated. In other evaluations, mutations in RUNX1 and CBL predicted worse OS in patients undergoing HMA therapy, but TET2 MT and ASXL1-wildtype (ASXL1 WT ) genotypes had better outcomes [28,29]. Interestingly, HMA therapy does not appear to alter disease mutation burden or be cytotoxic to cancer cells but instead alters gene expression through epigenetic changes [30]. Finally, genetics impact response to hematopoietic stem cell transplantation: patients with high overall mutation burden (>10 mutations) were found to be at higher risk of relapse [31]. In a recently developed prognostic model, ASXL1 and NRAS mutations independently worsened survival in CMML patients posttransplant [32].

JMML
Unlike CMML, JMML is a disease of infancy and early childhood characterized by organ infiltration of malignant cells resulting in lymphadenopathy, hepatosplenomegaly, cutaneous lesions, as well as constitutional symptoms and infectious sequelae. Its canonical feature is hypersensitivity to GM-CSF [33], as seen in CMML, thought to be related to its nature as a RASopathy. Indeed, most cases arise from somatic or germline mutations in genes involving the RAS-MAPK pathway: PTPN11, N/KRAS, NF1, or CBL are seen in about 90% of cases [34]. Mutations involving other signal transduction pathways, splicing, transcription, and polycomb repressive complex 2 have also been described [35]. Two-thirds of patients will have normal karyotype, but the remaining will have monosomy 7 (26%) and 10% with other aberrancies [36].

aCML
This entity is characterized by neutrophilia with severe dysgranulopoiesis, along with bone marrow findings of hypercellularity and granulocytic dysplasia [37]. Perhaps due to significant clinical, morphologic, and genetic overlap with the MPN chronic neutrophilic leukemia (CNL) [38], WHO guidelines advise that granulocyte precursors (promyelocytes, myelocytes, and metamyelocytes) should comprise at least 10% of PB WBCs [1]. To exclude CML and CMML, there must be no or minimal absolute basophilia or monocytosis as well as no evidence of BCR-ABL [1]. This disease is considered relatively aggressive, with a median OS of about 11-14 months [37,39,40] and a known propensity for leukemic transformation.
Evaluation of the most aCML-specific mutations has yielded insight into the pathophysiology of this disease. For instance, SETBP1, found on the long arm of chromosome 18, plays a role in cell proliferation by producing a nuclear protein that binds SET and inhibits the tumor suppressor gene PP2A [42,48]. Overexpression of this gene is associated with poor outcome in AML, and it may represent a unique mechanism in leukemogenesis [48,49].
In 2015, heterozygous mutation of ETNK1 was recognized as an abnormality largely unique to aCML [43]. As this enzyme is the first in a specific phospholipid biosynthesis pathway, its loss of catalytic function may impact cell membrane architecture, cytokinesis during cell division [50], and optimal respiratory chain performance in the mitochondrial inner membrane [51]. In fact, a recent study found that the metabolic product of ETNK1 attenuates mitochondrial activity, with loss of function driving increased production of reactive oxygen species (ROS) and mutagenesis. In vitro experiments suggested a normal phenotype can be restored by administering the underproduced enzyme product or the antibiotic tigecycline [52], which is known to inhibit synthesis of mitochondrial proteins that enable ROS production.
Despite evidence that SETBP1 confers worse prognosis in myeloid malignancies in general [53,54] and aCML specifically [42], its impact has not been found deleterious in all studies [3,40]. The frequent coexistence of mutations in ASXL1 and CBL with SETBP1 suggests an association with worse outcomes; [41,42] CBL mutation alone suggests more aggressive disease [44]. Notably, TET2 mutation is independently associated with worse OS in aCML [40], although may be mutually exclusive with mutations in SETBP1 [41]. Finally, SRSF2 mutations may correlate with better OS while those in RUNX1, NRAS, and CUX1 suggest poor outcomes [3].

MDS/MPN-RS-T
Only upgraded from a provisional entity in the 2016 WHO classification of myeloid neoplasms [1], MDS/MPN-RS-T is essentially the epitome of overlap between these two disease spectrums. As its name suggests, greater than 15% RSs are found on marrow iron stain [1].
These erythroid precursors have a perinuclear ring of blue granules on Prussian blue stain, reflective of mitochondrial iron overload resulting in dyserythropoiesis [55]. Proliferative mutations drive elevated platelet counts, arterial and venous thrombosis [56,57], vasomotor symptoms, and bleeding from acquired von Willebrand syndrome [58].
Similar to its phenotype, the genotype of this disease is a melding of stereotypical dysplastic and proliferative genetic aberrations.
Founder mutations are frequently in genes involved in RNA splicing: SF3B1 is found in about 90% of patients [4], and SF3B1 WT patients typically harbor other spliceosome mutations [59]. in about only 10%-20% of patients [3,4,55]. Abnormal karyotype and mutations in ASXL1 and SETBP1 are associated with worse OS in this population [3,64]. These mutations were found in 29% and 13% of patients in one cohort, respectively [64]. EZH2 also correlates with worse OS [3].  with mono-or biallelic mutations in TP53 segregated into a fourth group (13%), characterized by worse anemia, higher bone marrow blast percentage, and worse prognosis [3]. Finally, the remaining 26% of patients more frequently had mutations in JAK2, U2AF1, and ASXL1 and were more likely to have thrombocytosis. Survival in these "other" patients trended worse than CMML-like patients but was better than aCML-like patients.

DISCUSSION
It was not until 2001 that the WHO re-classified CMML from a subset of MDS to the newly created MDS/MPN category [67]. Since then, there have been significant advancements in understanding this group of neoplasms (Table 1) and an increase in clinical trials targeting CMML patients in particular. There remains a need to better characterize MDS/MPNs by genetic features-to more accurately understand the natural history of the disease and ensure patient eligibility for clinical trials and FDA-approved treatment options [68]. Additionally, genetics plays a critical role in disease prognostication; for instance, the worse outcomes seen among men with MDS/MPN as compared to women may be explained by gender-related differences in somatic mutation profiles [69]. Our review highlights the advances in our understanding of the genetics of these diseases.

SUMMARY AND FUTURE DIRECTIONS
The MDS/MPNs are rare and heterogeneous, making systematic study more difficult relative to other myeloid malignancies. Advances in gene sequencing technology along with large multi-institution collaborations have shed light on these illnesses: We now know that most have signaling gene mutations responsible for proliferative phenotypes, and many have gene mutations affecting epigenetic regulation, causing dysplastic features. These syndromes likely often evolve out of preexisting clones that mature into neoplasms with pheno-types based on the order of mutation acquisition and interaction between epigenetic and genetic factors. Genetic studies will likely aid in revising WHO criteria so as to more holistically classify the overlap syndromes.

CONFLICT OF INTEREST
Tania Jain serves as a consultant for Targeted