Integration of Molecular Information in Risk Assessment of Patients with Myeloproliferative Neoplasms

Philadelphia chromosome-negative myeloproliferative neoplasms (MPN) are clonal disorders of a hematopoietic stem cell, characterized by an abnormal proliferation of largely mature cells driven by mutations in JAK2, CALR, and MPL. All these mutations lead to a constitutive activation of the JAK-STAT signaling, which represents a target for therapy. Beyond driver ones, most patients, especially with myelofibrosis, harbor mutations in an array of “myeloid neoplasm-associated” genes that encode for proteins involved in chromatin modification and DNA methylation, RNA splicing, transcription regulation, and oncogenes. These additional mutations often arise in the context of clonal hematopoiesis of indeterminate potential (CHIP). The extensive characterization of the pathologic genome associated with MPN highlighted selected driver and non-driver mutations for their clinical informativeness. First, driver mutations are enlisted in the WHO classification as major diagnostic criteria and may be used for monitoring of residual disease after transplantation and response to treatment. Second, mutation profile can be used, eventually in combination with cytogenetic, histopathologic, hematologic, and clinical variables, to risk stratify patients regarding thrombosis, overall survival, and rate of transformation to secondary leukemia. This review outlines the molecular landscape of MPN and critically interprets current information for their potential impact on patient management.


Introduction
In 1951, William Dameshek coined the term "myeloproliferative disorders" to include what are currently known as classic Philadelphia chromosome-negative myeloproliferative neoplasms (MPN) [1]. The latest 2016 World Health Organization (WHO) classification of myeloid neoplasms includes four diseases: polycythemia vera (PV); essential thrombocythemia (ET); and myelofibrosis, either overt or prefibrotic primary myelofibrosis (PMF) [2]. The original breakthrough, Dameshek's argument was that they might represent a relatively homogeneous family of clinically heterogeneous disorders that were expression of, and were caused by, an abnormal proliferation capability of a stem/progenitor cell in the bone marrow. Indeed, each has unique clinical and hematologic presentation, yet they are largely phenotypically overlapping, especially at presentation, suggesting that they might represent a continuum of a single disease. Not formally recognized in the WHO diagnostic criteria are those forms of MF that develop as progression of prior PV and ET (post PV-MF and post ET-MF, also collectively known as "secondary" MF). All MPN can transform into secondary acute myeloid leukemia, referred to as MPN-blast phase (BP), which is typically refractory to conventional chemotherapy and has dismal prognosis.
MPN are considered rare hematologic neoplasms, with an incidence that varies from 0.1 to 2.8/100,000 European inhabitants per year [3]; prevalence remains difficult to deter-mine, but it may be anticipated to be progressively rising. In fact, the survival of patients with MPN is slowly but definitely improving in the last few decades, thanks to earlier diagnosis (nowadays facilitated by the inclusion of driver MPN-associated mutations among the WHO diagnostic criteria); better risk stratification (largely thanks to the knowledge of the genomic structure, including but not limited to driver mutations); and introduction of novel therapies, including JAK inhibitors, as well as allogeneic hematopoietic stem cell transplant (allo-HSCT).
There has been a tremendous accumulation of information deriving by the sequencing of DNA of MPN patients, which started in 2005 with the description of a point mutation in JAK2 (V617F), marking more than 60% of all MPN, soon followed by discovery of JAK2 exon 12, MPL, and CALR mutations. Their high frequency among MPN (overall, > 85%) and relative selectivity compared to other hematologic neoplasia warranted a revision of the diagnostic criteria, ultimately leading to current 2016 WHO classification. Then, not without surprise, the appreciation of an unexpected degree of genomic abnormalities complexity in MPN arose following the extensive use of next-generation sequencing, and opened the way to the generation of novel, more powerful risk stratification systems, particularly in MF. The appreciation of the uniform involvement of the JAK-STAT signaling pathway in MPN promoted the development of a new class of drugs, the JAK inhibitors, which although lacking selectivity against the mutated protein have nonetheless dramatically changed the life of patients, particularly MF and PV, and may be also carrying some advantages in terms of survival (MF) and reduction of thrombotic events, the leading cause of morbidity and mortality in PV.

Driver Mutations in JAK2, MPL, and CALR
Constitutive activation of the JAK-STAT signaling is the hallmark of all MPN and is sustained by somatic mutations in driver genes including JAK2, MPL, and CALR. The JAK-STAT pathway is critically involved in the regulation of cytokine-and growth factor receptor-mediated effects, as well as cell growth, survival, and differentiation of hematopoietic and immune compartments. Dysregulation of the JAK-STAT pathway confers cytokine independence and/or hypersensitivity to mutated cells, resulting in a survival advantage [4]. Somatic mutations in JAK2, MPL, and CALR are referred as "driver mutations" on the basis of their role in driving the MPN clinical phenotype and are crucial for the diagnosis along with selected laboratory and histopathological features. These mutations, documented in up to 85% of MPN patients, are mutually exclusive, but anecdotal cases of co-occurrence have been described in some reports [5,6]. Conversely, patients with features of MPN without any drive mutations are referred to as "triple negative" (TN) [7].
Two different types of JAK2 mutations are described in MPN. The first type, discovered in 2005 by four different groups, is a single nucleotide variant at codon 617 (in exon 14) resulting in a valine to phenylalanine substitution [8][9][10][11]; the second type, described in 2007, comprises different changes in exon 12, particularly in-frame insertions or deletions in the region between codons 536 and 544, collectively defined as exon 12 mutations [12,13]. JAK2V617F is detected in approximately 95% of polycythemia vera (PV) and 50-60% of essential thrombocythemia (ET) and primary myelofibrosis (PMF), whereas JAK2 exon 12 mutations are found exclusively in 2-3% of PV patients lacking the more common JAK2V617F. Patients with exon 12 mutations are characterized by erythroid-dominant myeloproliferation with a consequent lower leukocyte and platelet count at diagnosis, subnormal serum erythropoietin levels, a subtle tri-lineage hyperplasia in the bone marrow, and younger age at diagnosis in comparison to those with V617F mutation. However, there are no substantial prognostic differences between JAK2 exon 12 and JAK2 V617F-mutated PV patients [14,15].
The JAK2V617F mutation is capable of driving all the different MPN through ligandindependent activation of erythropoietin (EPO), thrombopoietin (TPO), and granulocytecolony stimulating factor (G-CSF) tyrosine-kinase receptors. This constitutive activation results in an abnormal trilinear myeloproliferation [16][17][18]. Functionally, the inhibitory effect of the JAK2 pseudokinase domain is abrogated and the autonomous activation of intracellular signaling promotes cell proliferation, differentiation, and survival through downstream signaling molecules, including signal transducers and activators of transcription (STATs) [19], mitogen-activated protein kinases (MAPKs), phosphoinositol-3-kinase (PI3K), and mammalian target of rapamycin (mTOR) [20]. Although the question how JAK2V617F may induce different MPN is still incompletely answered to, the different clinical phenotypes probably reflect different host characteristics, different progenitor cell stage at which the mutation arises, the presence and order of additional variants, and the influence of bone marrow microenvironment. Moreover, JAK2V617F allele burden has been associated with different phenotype of MPN; homozygous status or a high mutant allele burden is found in most PV patients, and is associated with increased risk of thrombosis and fibrotic evolution [21,22], whereas ET and PMF patients tend to have a lower burden [23].
In 2006, mutations in the TPO receptor MPL, which specifically regulates megakaryopoiesis and platelet production through the JAK-STAT pathway, were discovered [24]. Mutations in MPL are found in 3-8% of ET and PMF and consist in gain of function variants at tryptophan 515 (W515) in exon 10 [25]. The most common point mutations are W515L and W515K. Other rare mutations at the same position include W515R, W515A, and W515G [26]. Although the MPLS505N variant was identified previously as a germline mutation in cases of familial thrombocythemia [27], it can also be acquired as a somatic event. Recently, many second-site mutations that enhance S505N-driven activation were described [28].
Finally, in 2013, two independent groups reported a breakthrough discovery by identifying a novel driver mutation in a large part of JAK2 and MPL wild-type MPN patients. By applying whole exome sequencing, both groups identified somatic mutations in CALR encoding for calreticulin, a 46-kDa endoplasmic reticulum (ER) chaperone protein.
CALR has a key role in the quality control of glycoprotein folding and maintenance of cell calcium homeostasis given its negatively charged C-terminus [29]. To date, more than 50 different mutations in CALR exon 9 have been described in approximately 20-25% of ET and 25-30% of PMF cases but not in PV patients [30,31]. The two most frequent CALR mutations, accounting for approximately 80% of all the mutations, include a type-1 52-bp deletion (c.1092_1143del; p.L367fs*46) and a type-2 5-bp insertion (c.1154_1155insTTGTC; p.K385fs*47). Atypical mutations are grouped as type 1-like and type 2-like in relation to structural similarities. All CALR mutations lead to a +1 frameshift of the open reading frame, resulting in a mutant protein lacking the ER-retention motif (KDEL) and containing a novel C-terminal chain of positively charged amino acids instead of the negatively charged amino acids of wild-type CALR. CALR subtypes contribute to determine clinical phenotype and outcomes in MPN. In PMF, type-1/1-like mutations are more prevalent than type-2/2-like (70% vs. 13%), while in ET they are more equally distributed (51% vs. 39%) [32]. In ET patients, type-1/1-like mutations were associated with a fibrotic phenotype and a significantly higher risk of myelofibrotic transformation, whereas type-2/2-like mutations were associated with higher platelet count and a lower risk of thrombosis [33]. In a large series of ET patients, both CALR variants correlated with a higher platelet count, lower hemoglobin, and leukocyte count compared to mutant JAK2 [34]. Moreover, type-1/1-like mutations have been associated with a favorable survival in PMF [35]. Intense research efforts lead to elucidation of the mechanism underlying the oncogenic activity of mutated CALR in MPN. Mutant CALR binds to MPL within the ER, resulting in the constitutively activation of the JAK-STAT signaling [36][37][38][39]. Moreover, the MPL-CALR complex has been detected on cell surface, and its translocation appeared essential for the oncogenic activity [40,41]. More recently, a novel intriguing mode of action of mutated CALR was described suggesting an interleukin-6-dependent and TPO-MPL-independent mechanism [42].

Triple-Negative MPN
Mutations in JAK2 exon 14 and exon 12, CALR exon 9, and MPL exon 10 account for over 90% of all MPN cases. These mutations are mutually exclusive, although rare cases of coexisting driver mutations are reported [5]. However, approximately 15% of ET and less than 10% of PMF patients lack a known driver mutation and are accordingly referred to as TN [43]. These cases of MPN require a particularly diligent diagnostic workup since reactive causes mimicking MPN-like disorders as well as alternative myeloid malignancies need to be carefully excluded. Moreover, it was reported that some TN cases tested positive for canonical somatic driver mutations at low mutant burden when re-sequenced using methodologies with a higher sensitivity [44][45][46].
At least 10% of TN ET and PMF patients harbor mutations outside of the canonical MPL and JAK2 hotspots. Non canonical MPL mutations include T119I, S204F/P and E230G in the extracellular domain and Y591D/N in the intracellular domain [45], whereas non canonical JAK2 mutations include V625F, F556V, R683G, and E627A [46]. As reported in functional studies, most of these genomic variants lead to a constitutive activation of the JAK-STAT signaling [45,46]. Some of them are exclusively described as somatic events, while others also represent germline variants, with the possibility that these patients may have a non-clonal erythrocytosis or thrombocytosis. These findings impose important implications for the diagnosis and management of TN patients, including unique therapeutic approach and eventually the screening of family members. Triple negative ET is an indolent disease with low incidence of vascular events and progression to MF, and is usually diagnosed in young, female patients. On the contrary, TN MF is a more aggressive disease with an adverse prognosis and a high rate of progression to blastic phase (BP).

Somatic Additional Mutations in Genes Frequently Involved in Myeloid Neoplasms
In last years, high-throughput next-generation sequencing (NGS) approaches resulted in the identification of a considerable number of additional somatic mutations with a prognostic and therapeutic value. At least one additional mutation can be found in more than 50% of patients, and the prevalence increases with patient age. These co-occurring mutations are not restricted to MPN, but occur also in other myeloid malignancies including acute myeloid leukemia (AML), myelodysplastic syndromes (MDS), and MDS/MPN overlap diseases, as well as in healthy individuals in the context of age-related clonal hematopoiesis (ARCH/CHIP) [44,47]. Additional genomic abnormalities affect genes involved in DNA methylation (TET2, DNMT3A, IDH1, and IDH2), histone modification (ASXL1, EZH2), mRNA splicing (SFRB1, SRSF2, U2AF1, and ZRSR2), signaling pathways (LNK/SH2B3, CBL, NRAS, KRAS, and PTPN11), and transcription factors (RUNX1, NFE2, TP53, and PPM1D).
The TET2 gene encodes the second member of the TET proteins family, which are involved in the stepwise oxidation of 5methylcytosine (5mC) to 5-hydroxyymethylcytosine (5hmC), an important process in stem cell gene regulation. Loss of function mutations in TET2 are mostly heterozygous and occur in 10-15% of MPN patients and 20-25% of BP. They are the most common co-occurring mutations in JAK2V617F-mutated MPN, although without a clear prognostic effect [48][49][50]. Recently, it has been reported that the order of acquisition of JAK2 and TET2 mutations may influence MPN phenotype, with "JAK2-first" more commonly detected in PV and "TET2-first" in ET patients [51]. DNMT3A belongs to a family of highly conserved DNA methyltransferases involved in either de novo DNA methylation or maintenance of pre-existing methylation patterns throughout cell division. DNMT3A mutations are frameshift/nonsense, resulting in reduced methyltransferase activity, and occur in 5-15% of MPN patients, with lower frequency in PV and ET compared to PMF and BP [48][49][50]. Similarly to TET2, MPN patients are more likely to present with PV when JAK2V617F is acquired before DNMT3A mutation, while patients who first acquired DNMT3A mutations more commonly develop an ET phenotype [52]. The IDH1 and IDH2 genes encode two additional epigenetic regulators involved in both DNA methylation and histone modification. Mutations in isocitrate dehydrogenase genes (IDH1 and IDH2) occur mostly as point missense mutations at residues R132 in IDH1 and R140 or R172 in IDH2. The mutant protein acquires the ability to convert alpha-ketoglutarate (a-KG) to 2-hydroxyglutarate (2-HG), favoring leukemogenesis through epigenetic dysregulation of a number of other genes. IDH1 and IDH2 mutations have been associated with worse prognosis in PMF with a higher risk of transformation to BP. They are reported in up to 6% of MPN and 30% of BP [53,54].
The ASXL1 and EZH2 genes encode for two important chromatin-binding protein of the polycomb family involved in the epigenetic regulation of gene expression through the interaction with polycomb repressive complex 1 (PRC1), polycomb repressive complex 2 (PRC2), and other transcription activators and repressors [55]. Loss of ASXL1 function results from nonsense and frameshift mutations in exon 12 and are more common in PMF and BP (18-37%) compared to PV and ET (5-10%) [54,56]. ASXL1 mutations are associated with a worse prognosis in PV and PMF patients [48,57,58], and this unfavorable impact is not overcome by allogenic hematopoietic stem cell transplantation (allo-HSCT) [59]. Of interest, it has been recently reported that ASXL1 mutations are frequently acquired during ruxolitinib treatment as part of clonal evolution of the disease [60]. EZH2 mutations are less frequent than those in ASXL1, being found in up to 10% and 15% of MPN and BP patients, respectively. EZH2 seems to act as a tumor suppressor in MPN, and a high number of loss of function mutations have been identified that synergizes with JAK2V617F in initiating MPN and promoting myelofibrosis [61]. In PMF, EZH2 mutations correlate with a higher leukocyte count, blast count, and larger spleen size at diagnosis. Moreover, EZH2 mutations are an independent poor prognostic factor for reduced overall survival [62].
Acquired mutations in genes encoding for spliceosome proteins and other regulatory splicing factors are reported in all myeloid malignancies, particularly MPN and MDS, and mainly affect SRSF2, U2AF1, ZRSR2, and SF3B1 genes. Spliceosome-affecting mutations are typically mutually exclusive and mostly occur as heterozygous missense mutations that confer a dominant negative activity affecting RNA splicing [63]. The SRSF2 gene is the most frequently mutated, with variants mainly occurring at the Proline 95 residue. While being infrequent in PV and ET patients, SRSF2 mutations occur in 8-22% of PMF and BP, where predict for poor overall and leukemia-free survival [56,[64][65][66][67]. U2AF1 mutations, mainly at the hotspot residues Serine 34 and Glutamine 157, occur in 5-15% of PMF and BP and are associated with a significantly shorter overall survival [68]. SF3B1 mutations affect exons 14-16, particularly the Lysine 700 amino acid, and are more commonly found in PMF and BP (up to 10%). Although rare in PV and ET, U2AF1 and SFRB1 mutations were associated with an inferior myelofibrosis-free survival compared to the with wild-type counterpart [48]. Finally, ZRSR2 mutations are very rare in MPN and are more frequently identified in PMF, without a clear impact on prognosis.
LNK/SH2B3 is an adaptor protein that inhibits signaling through cytokine and tyrosine kinase receptors, including JAK2 [69,70]. Mutations affecting LNK/SH2B3 are mostly missense substitutions and can be found in up to 10% of MPN and BP patients [48][49][50]. Although somatic LNK mutations detected in MPN patients are sparce through the gene, the LNKE208Q variant is the only one that has been identified both as germline variant in idiopathic erythrocytosis and acquired mutation in MPN, either alone or in association with a driver mutation [71]. Germline LNKE208Q mutation has been reported also in cases of familial MPN [72,73].
Mutations in oncogenic RAS pathway genes (NRAS, KRAS, CBL, NF1, PTPN11) are overall rare in MPN patients. CBL encodes for a multifunctional adaptor protein with ubiquitin ligase activity, and mutations are associated with stabilization of receptor tyrosine kinases, resulting in constitutive activation of signaling pathways, cytokine hypersensitivity, and autonomous cell proliferation [74]. CBL mutations are mostly missense substitutions and can be detected in up to 6% of PMF and BP patients [49,50]. Missense mutations in NRAS/KRAS, particularly in codons 12, 13, and 61, lead to constitutive activation of growth signaling [75], are rare in PMF (up to 5%), and are reported in up to 15% of BP [48,49]. In a recent paper, CBL/NRAS/KRAS mutations in PMF patients were associated with adverse clinical features, shorter overall survival, and poor response to JAK inhibitor therapy [76]. PTPN11 mutations are found in up to 8% of BP and are associated with a reduced overall survival [50].
Among transcription factors, RUNX1 and NFE2 are the most commonly mutated genes in MPN. RUNX1 mutations, result in protein inactivation, occur in up to 10% of BP, and are associated with a shortened survival [44]. The NFE2 gene encodes for an important regulator of megakariocytopoiesis and/or erythropoiesis. Mutations are heterogenous but mainly represented by frameshift variants and deletions [77]. Although overall rare, NFE2 mutations are enriched in PV (8% of cases). In one study the authors failed to find meaningful hematological and clinical correlates, nor a clear prognostic value [78], while in another study including a larger cohort of MPN patients, NFE2 mutations adversely affected prognosis, increasing the risk of BP progression independently of age and other co-occurring mutations [79]. TP53 is a transcription factor with a key tumor suppression function in response to cellular stress and DNA damage. TP53 mutations are frequent in BP, where biallelic loss of TP53 has been reported in up to 35% of cases, with a dismal outcome [80]. Recently, very low burden of TP53 mutations has been reported in patients with chronic phase MPN, even many years before diagnosis, with an association with older age [81]. The significance of these very low mutation burdens in relation to the risk of evolution into BP, remains however largely unknown. PPM1D mutations are more frequent in cases of therapy related AML and MDS, particularly after cytotoxic agents [82], with an intrinsic chemoresistance to standard therapies [83]. Mutations in PPM1D, a regulatory inhibitor of TP53, were recently described in 1.9% of patients with MPN, both within the driver clone and as an independent clone [44]. In Table 1, there is a summary of somatic additional mutations in MPN with their frequency and clinical impact.

Polycythemia Vera and Essential Thrombocythemia
For many years, thrombosis prediction in PV and ET relies only on two clinical variables: older age (>60 years) and prior history of thrombosis. More recently, several studies highlighted the possibility that genetic factors contribute to the assessment of thrombotic risk. Unlike patients with PV who almost exclusively carry mutations in JAK2, in ET the effect of driver mutations has been extensively evaluated comparing JAK2V617F mutated and unmutated patients. Overall, these studies showed that the presence of JAK2V617F is associated with a higher risk of thrombosis [88][89][90]. In 2012, on the basis of these pieces of evidence, the International Prognostic Scoring of Thrombosis in ET (IPSET-thrombosis) was developed by adding the JAK2 mutated status and cardiovascular risk factors (2 and 1 points, respectively) to the classic variables older age (>60 years) and prior thrombosis (1 point each) [91]. A re-analysis of the original IPSET-thrombosis data led to the exclusion of cardiovascular risk factors with the development of a revised IPSET [92] ( Table 2) that was subsequently validated in a large independent cohort of ET patients [93]. On the basis of the score, it is suggested that young ET patients without a prior history of thrombosis and lacking the JAK2V617F mutation may not need aspirin because of their very low risk of thrombosis. Moreover, it was reported that in this category of ET patients, aspirin could eventually increase the risk of bleeding without reducing the risk for thrombosis [92,94]. Recently, the prognostic relevance of additional somatic mutations was investigated in two large cohorts of PV and ET patients [48]. Upon age-adjusted multivariate analysis of their impact on overall, leukemia-free, and myelofibrosis-free survival, molecular variants associated with adverse outcomes included ASXL1, SRSF2, and IDH2 in PV, and SH2B3/LNK, SF3B1, U2AF1, TP53, IDH2, and EZH2 in ET. Specifically, the presence of at least one of these mutations was associated with inferior overall survival in both PV (median survival, 7.7 vs. 16.9 years) and ET (median survival, 9 vs. 22 years). On the basis of the results obtained from the analysis of 906 molecularly annotated patients from Mayo Clinic and University of Florence cohorts, the authors incorporated genetic data into two new prognostic models: the Mutation-Enhanced International Prognostic Scoring System for PV (MIPSS-PV) and ET (MIPSS-ET) ( Table 3) [95]. MIPSS-PV included age >67 years, leukocyte count ≥15 × 10 9 /L, thrombosis history at diagnosis, and the presence of SRSF2 mutation. Conversely, MIPSS-ET included male sex; age >60 years; leukocyte count ≥11 × 10 9 /L; and the presence of mutations in SRSF2, SF3B1, U2AF1, or TP53. A recent sub-analysis of the ET cohort included in the MIPSS-ET demonstrated that ASXL1/RUNX1/EZH2 mutations were associated with a decreased risk of arterial thrombosis, suggesting a different underlying biology [96]. These interesting and promising results require further studies to determine their practical role in clinical management. Table 3. Clinical-molecular prognostic scores in polycythemia vera and essential thrombocythemia.

Primary and Post-PV/ET Myelofibrosis
Currently, two prognostic models that include exclusively hematologic and clinical variables are employed to stratify PMF patients into risk categories with significant differences in overall survival: the International Prognostic Scoring System (IPSS) [97], applicable at the time of diagnosis, and the Dynamic IPSS (DIPSS) [98]. These scores can be applied at any time during the clinical course of PMF, using five clinical variables that independently predict inferior survival: age > 65 years, hemoglobin < 10 g/dL, leukocyte count > 25 × 10 9 /L, circulating blasts ≥ 1%, and constitutional symptoms. Subsequently, the DIPSS was revised to the DIPSS plus [99], including red blood cell transfusion need, platelet count < 100 × 10 9 /L, and unfavorable karyotypes (complex karyotype or sole or 2 abnormalities that included +8, −7/7q-, I (17q), inv (3), −5/5q-, 12p-, or 11q23 rearrangements) [100]. As outlined previously, in addition to driver mutations, more than 80% of patients with PMF harbor genomic variants in myeloid genes, often in multiple combinations [49]. Moreover, in PMF patients, driver and additional mutations have been shown to influence overall and leukemia-free survival, independent of IPSS and DIPSS/plus [35,56,57]. The prognostic contribution of driver mutations supports the distinction between presence or absence of type 1 CALR mutations [101,102], whereas additional abnormalities in ASXL1, SRSF2, EZH2, and IDH1/IDH2 were defined as a high molecular risk (HMR) variable, with a prognostic relevance amplified by the number of those mutations in individual patients [49,54,56]. At the light of these observations, the Molecular Enhanced International Prognostic Score Systems (three-tiered MIPSS70 and four-tiered MIPSS70-plus) were developed using a cohort of patients aged ≤70 years, potentially eligible for allo-HSCT, recruited from multiple Italian centers and the Mayo Clinic, Rochester, USA [103]. The MIPSS70-plus score additionally included unfavorable karyotype defined by any abnormal karyotype other than normal karyotype or sole abnormalities of 20q2, 13q2, + 9, chromosome 1 translocation/duplication, -Y, or sex chromosome abnormality excluding -Y. Both models predicted LFS as well as OS, but MIPSS70-plus seemed to have the best performance in identifying a very high-risk category of patients age ≤70 years, 23% of whom developed acute leukemia likely as a consequence of additional cytogenetic abnormalities. A further revision termed MIPSS70-plus v2.0 [84] incorporated the U2AF1Q157 variant as an additional HMR mutation on the basis of previous findings [68]. Additional refinement of risk categories was provided by defining new sex-and severity-adjusted hemoglobin thresholds for anemia (severe for hemoglobin levels of <8 g/dL and <9 g/dL, and moderate for hemoglobin levels of 8 to 9.9 g/dL and 9 to 10.9 g/dL, respectively, for women and men) [104], and by integrating a refined three-tiered cytogenetic risk distribution with the introduction of very high risk (VHR) group; the latter included patients with single/multiple abnormalities of −7, I (17q), inv (3)/3q21, 12p-/12p11.2, 11q-/11q23, or other autosomal trisomies not including +8/+9 [105].
In 2018, all these information regarding the prognostic role of genetic and cytogenetic alterations in PMF prompted the development of a Genetically Inspired Prognostic Scoring System (GIPSS). The latter was exclusively based on molecular (absence of type-1 CALR mutations and presence of ASXL1, SRSF2, and U2AF1Q157 additional somatic mutations) and cytogenetic variables [106]. GIPSS was not inferior to DIPSS and MIPSS70-plus in discrimination ability and prediction accuracy. Moreover, these data were validated in a large independent cohort of PMF patients [107].
Notably, all the above models were developed exclusively for patients with a diagnosis of primary MF. When applied to secondary MF, among IPSS, DIPSS, and DIPSS-plus, DIPSS was found to be the most accurate risk stratification model. Further, a four-tiered Myelofibrosis Secondary to PV and ET-Prognostic Model (MYSEC-PM) was specifically developed for secondary MF. This clinical-molecular score included the following variables: hemoglobin level, peripheral blasts, platelet count, age, presence of constitutional symptoms, and CALR mutational status [108]. The MYSEC-PM was demonstrated to perform better than IPSS.
In order to accurately predict patients' outcome following allo-HSCT, the Myelofibrosis Transplant Scoring System (MTSS) was recently formulated for patients with primary and post-PV/ET myelofibrosis. The score variables include: age (≥57 years), performance status, platelet and leukocyte count prior to transplantation (<150 × 10 9 /L and >25 × 10 9 /L, respectively), HLA-mismatched unrelated donor, and CALR/MPL and ASXL1 mutational status [109]. Finally, a recent large study based on sequencing of 69 genes in more than 2000 patients with MPN identified a prognostic role for CBL, NRAS, RUNX1, TET2, P53, GNAS, IDH2, and U2AF1 in both OS and LFS. This study led to the creation of a personalized predictive individual model for disease progression and death, integrating a high number of demographics, clinical variables, and genomic variables [44]. However, due to its complexity, clinical application is limited at present.

Impact of Mutational Landscape on Therapeutic Decisions
The remarkable knowledge accumulated in recent years concerning mutational landscape of MPN opened the possibility of developing targeted therapies aiming at specifically interact with cell pathways demonstrated to be dysregulated. The identification of the specific dependence of MPN on JAK/STAT pathway dysregulation, irrespective of the underlying driver mutation, led to the development of small-molecule inhibitors of the JAK family of tyrosine kinases (JAKi). To date, several compounds have been developed that differ in structure, mechanism of action, potency, and kinase selectivity, all being type I inhibitors. It is important to underline that JAKi target the ATP-binding pocket of the JAKs without being selective against mutant JAK protein, thus explaining the clinical efficacy also in JAK2-unmutated patients. Ruxolitinib was the first JAK1/2 inhibitor that received approval for myelofibrosis treatment on the basis of the results from the COMFORT-I/II studies [111,112]. Ruxolitinib was effective in reducing spleen volume and alleviating constitutional symptoms, with possible effects on OS. Long term follow-up studies suggested modest reduction of JAK2V617F allele burden, with rare cases of molecular remission [113,114]. Subsequently, ruxolitinib was investigated in the RESPONSE [115] and RESPONSE-2 [116] studies that enrolled high-risk PV patients resistant or intolerant to hydroxyurea (according to European Leukemia-Net (ELN) criteria [117]) with and without splenomegaly, respectively. A modest progressive decline in JAK2 mutant allele burden in patients under ruxolitinib was documented also in the RESPONSE trial, although without clear clinical correlations [118].
Unfortunately, most patients with myelofibrosis on ruxolitinib eventually become resistant to therapy with progression of symptoms and splenomegaly, worsening cytopenias, or evolution to BP. In the COMFORT-II study, among patients responsive to JAKi treatment, less than 50% had chance of maintaining response at five years [113]. Intriguingly, in a recent paper, Newberry et al. [60] demonstrated that 35% of patients treated with ruxolitinib had a clonal evolution after ruxolitinib discontinuation, defined by the acquisition of at least one additional mutation. ASXL1 mutations was the most frequent, followed by TET2, EZH2, and TP53. The main downside of this study was the lack of a control patients' cohort [119]. Subsequently, another study, which included a control group of 25 MF patients treated with hydroxyurea, confirmed these observations but also demonstrated that clonal progression is independent of the treatment [120].
The impact of the mutational landscape on treatment outcomes with JAKi has been addressed by a few studies with overall inconsistent findings. A retrospective analysis of the COMFORT-II trial reported that spleen and anemia responses were not correlated with either driver or HMR mutations [111]. Conversely, JAK2V617F allele burden ≥50% was associated with higher spleen response rates to ruxolitinib [121]. In another study evaluating 95 ruxolitinib-treated patients, patients with ≥3 mutations had ninefold lower odds of spleen response and shorter time to treatment discontinuation [122]. Similarly, mutations in ASXL1 and CBL as well as an HMR profile correlated with shorter time to treatment failure [123]. Moreover, loss of spleen response was associated with HMR mutations, whereas the absence of ASXL1 mutations and > 20% reduction in JAK2V617F allele burden at any time during treatment correlated with long-term spleen response [120]. Altogether, these data suggest that although no individual mutations can predict the response to JAKi, particular variants and higher mutational burden may influence the duration of response and hence treatment failure, likely as a consequence of a more aggressive disease. More recently, mutations in the RAS pathways genes (NRAS, KRAS, CBL) were identified as independent predictors of reduced symptom and spleen responses to JAKi [76].
Allo-HSCT remains the only potentially curative therapy for MF patients; despite many improvements, outcomes of allo-HSCT are still burdened by substantial morbidity and high transplant-related mortality and only a minority of patients are eligible for such an intensive procedure. Consequently, insights into molecular mechanisms of MPN pathogenesis have spurred drug development, searching for more effective treatments. In recent years, interferon-α (IFNα), especially in the better tolerate pegylated forms (Peg-IFNα), has emerged as a promising approach in MPN, particularly in PV and ET [124][125][126][127][128][129]. Although IFNα is not currently approved for treatment in ET and PV, consensus guidelines recommend interferon as an option for first-line cytoreduction, particularly in younger or pregnant patients [117]. As suggested by the high rate of reduction of JAK2V617F allele burden under Peg-IFNα compared to mutated CALR, the treatment appears to selectively target the mutant JAK2 clone [130]. Enhanced sensitivity of JAK2V617F-mutated cells to IFNα may be related to high expression and phosphorylation levels of STAT1 [131]. Interestingly, the presence of concomitant non-driver mutations is associated with smaller mean decreases in JAK2V617F allele burden with Peg-IFNα treatment [125]. At this regard, IFNα may have decreased ability to eradicate TET2 positive clones even when JAK2V617Fmutant clone is markedly reduced, indirectly suggesting that the presence of additional mutations can predict responses to IFNα treatment [132]. However, also the dose of the drug seems to impact on the response to treatment; at this regard IFNα at a higher dose in a cohort of 31 CALR-positive ET patients induced hematologic responses in all patients with a median reduction of CALR mutated allele burden from 41% to 26% [133]. Similar to findings in JAK2 mutated patients, the presence of TET2, ASXL1, IDH2, and TP53 additional mutations was associated with poorer molecular responses. In a long-term 83-month followup of ET and PV patients treated with Peg-IFNα, median duration of hematologic and molecular response was 66 and 53 months, respectively [134]. Additionally, although in most patients JAK2V617F allele burden increased after the first 2 years of treatment, three patients had a complete molecular remission even after discontinuation of therapy. Recently, it was demonstrated that a diplotype spanning the coding region of the IFNL4 gene influences molecular response to IFNα in PV patients [135]. Nonetheless, disease evolution remained comparable to historical data in patients treated with other therapies, with an 8% transformation rate to MF or BP [134].
Imetelstat, a 13-mer lipid-conjugated oligonucleotide that targets the RNA template of human telomerase reverse transcriptase (TERC), was tested both in MF and ET patients. In 2015, Mayo Clinic investigators reported on 33 patients with intermediate-2 and high-risk MF according to DIPPS-plus score treated with imetelstat (2 h of intravenous infusion; starting dose, 9.4 mg/kg every 1 to 3 weeks) and observed a complete (n = 4; 12%) or partial response in seven patients (21%). Remarkably, all four patients with complete remission experienced reversal of bone marrow fibrosis, and a molecular response occurred in three of these four patients. Responses to imetelstat were correlated with the presence of JAK2V617F, SF3B1, or U2AF1 mutations and the absence of ASXL1 mutation [136]. In another pilot study included 18 ET patients treated with imetelstat, a partial molecular response was detected in seven out of eight JAK2V617F-mutated patients. Overall, JAK2V617F allele burden was reduced by a median of 71% 3 months after the initiation of treatment, along with allele burden reductions of mutated MPL and CALR (15-66%) [137]. In the latter cohort of ET patients, additional somatic mutations significantly reduced the depth of response and had an impact on duration of response. Among acquired mutations, ASXL1, EZH2, and U2AF1 were responsive to imetelstat, unlike SF3B1 and TP53 mutations [138]. Recently, in a phase II study of two imetelstat doses, VAF of JAK2V617F, CALR, or MPL driver mutations by at least 25% was observed in 42.1% of patients in the 9.4 mg/kg arm, and 5.6% of patients in the 4.7 mg/kg arm. Additionally, patients who achieved ≥20% VAF reduction demonstrated higher rates of spleen response, symptom response, and BM fibrosis improvement, along with a longer median OS [139].
Mutations in IDH1/IDH2 and TP53 are overall uncommon in chronic phase MPN but occur more frequently in BP disease. These findings may have a clinical impact according to the favorable results from studies evaluating ivosidenib and enasidenib (anti-IDH1 and IDH2, respectively) in patients with relapsed/refractory IDH-mutated AML [140,141], as well as high response rate of TP53-mutated AML to 10-day decitabine [142]. In a mouse model of JAK2 and IDH2 co-mutated MPN, combined inhibition of JAK2 and IDH2 (with ruxolitinib and enasidenib, respectively) normalized reduced disease burden to a greater extent than JAK inhibition alone by reversing aberrant gene expression and metabolite perturbation in the hematopoietic stem cell compartment [143]. Murine double minute 2 (MDM2), a key downregulator of TP53, is overexpressed in patients' CD34+ cells, and treatment with idasanutlin, a second-generation inhibitor of TP53-MDM2 interaction, has been shown to target MPN stem and progenitor cells both alone and in combination with IFNα [144]. On the basis of these data, researchers evaluated idasanutlin in a phase I trial showing hematologic, symptomatic, pathologic, and molecular responses in some patients [145]. Regrettably, the subsequent phase II trial of idasanutlin in HU-resistant/intolerant PV was prematurely terminated due to the high rate of gastrointestinal toxicities. Another MDM2-inhibitor, KRT-232, is being evaluated in patients with TP53 wild-type MF who are relapsed or refractory to JAKi (NCT03662126).
Anti-apoptotic proteins represent another potential target for MF therapy. At this regard, a promising option is represented by venetoclax, an oral, selective, potent inhibitor of the BCL2 antiapoptotic protein, used in combination with low-dose cytarabine or hypomethylating agents. In a recent trial designed for patients newly diagnosed with AML ineligible for standard induction chemotherapy, azacitidine plus venetoclax was superior to azacitidine alone, also in the subset of IDH1/IDH2and TP53-mutated patients [146]. Venetoclax in combination with azacytidine/decitabine has also been tested in BP-MPN. In a multicenter series of 32 consecutive cases, complete remission was achieved by 44% of patients and was more likely to occur in the absence of pre-leukemic PV/post-PV MF, complex karyotype, and NRAS/KRAS mutations, with no correlation with neither TP53 nor IDH1/IDH2 mutational studies. Importantly, 6/14 patients with complete remission subsequently received allo-HSCT [147].
Although these therapies are promising, BP patients carries a dismal prognosis, with a median survival of ≈6 months; the only possibility of long-term survival is offered by allo-HSCT in the minority of patients who are able to achieve complete remission, or return to chronic phase, with therapy before transplant. Further studies are needed, but these target therapies may represent a valid therapeutic approach also as a bridge to transplantation.
Finally, the prognostic effect of somatic mutations in outcomes following allo-HSCT remains poorly defined. Among 133 patients with PMF or secondary MF receiving HSCT, the presence of CALR mutations was associated with better 4-year OS, non-relapse mortality (NRM), and a trend toward lower cumulative incidence of relapse [148]. In survival analysis, patients with mutated CALR had the best better prognosis following allo-HSCT, those with JAK2 or MPL mutations had an intermediate prognosis, and TN patients had the worst prognosis. Interestingly, these finding recapitulate those in non-transplant setting [35]. In another retrospective study of 169 MF patients, CALR-mutated patients were found to have lower NRM and improved PFS and OS; ASXL1 and IDH2 mutations were associated with lower PFS; whereas no impact was observed for TN, SRSF2-, or EZH2mutated patients [59]. The presence of somatic mutations in driver and non-driver genes in most MF patients offers the opportunity to use these markers as indicators of minimal residual disease (MRD) after allo-HSCT. Two large studies reported that the persistence of JAK2V617F following allo-HSCT was associated with a higher incidence of relapse and a poor OS [149,150]. More recently, in a series of 136 patients, Wolschke et al. demonstrated that patients with detectable driver mutations after allo-HSCT at either day + 100 or day + 180 had a significantly higher risk of relapse at 5 years compared to those in molecular remission [151]. These studies strongly recommended the monitoring of molecular MRD in MF patients after allo-HSCT, helping overall in patient management.

Conclusions
The summary presented in this review of the molecular abnormalities that harbor patients with MPN, and how these are progressively raising the bar of knowledge of disease pathophysiology as well as improving the management of patients, stand for an exceptional last decade or so of scientific achievements. MPN driver mutations (JAK2, CALR, and MPL) activating JAK-STAT signaling are crucial for MPN pathogenesis. Moreover, additional somatic mutations are detected in more than 50% of MPN cases, particularly in MF, and are associated with disease progression. While in MF many clinical-molecular integrated prognostic models have been developed and routinely used in clinical practice, in PV and ET, the prognostic significance of concomitant somatic mutations is starting to be explored. Yet, there is still much work to do. The category of patients with triple negative MPN is an unmet diagnostic need: why is prognosis so poor for triple-negative MF? What genetic abnormalities do they hide? Are these MPN, or some other disease, for example, are they more akin to myelodysplastic syndrome? The phenomenon of loss of sensitivity to ruxolitinib is a challenging therapeutic dilemma and is known to associate with clonal complexity detected at the time of loss of response. Will we be able to predict it in advance and select the best candidates for alternative treatments? Moreover, the events that promote progression to acute leukemia are largely unknown in terms of what prevents development of appropriate surveillance and early diagnostic criteria. These few examples will hopefully reinforce the interest and promote efforts of the scientific community to reach a full understanding and satisfactorily treatment of MPN.
Author Contributions: G.G.L. and G.C. made substantial contribution to conception, design, and writing of the review; P.G. and A.M.V. contributed to writing and editing of the manuscript. All authors have read and agreed to the published version of the manuscript.

Conflicts of Interest:
A.M.V. has received personal fees for the advisory board from Novartis, Incyte, Bristol-Myers Squibb, and AOP, as well as speaker bureau from Novartis, CTI, and Celgene. P.G. received personal fees for the advisory board from Novartis. G.G.L. and G.C. declare no conflict of interest.