The Genetic Basis of Primary Myelofibrosis and Its Clinical Relevance

Among classical BCR-ABL-negative myeloproliferative neoplasms (MPN), primary myelofibrosis (PMF) is the most aggressive subtype from a clinical standpoint, posing a great challenge to clinicians. Whilst the biological consequences of the three MPN driver gene mutations (JAK2, CALR, and MPL) have been well described, recent data has shed light on the complex and dynamic structure of PMF, that involves competing disease subclones, sequentially acquired genomic events, mostly in genes that are recurrently mutated in several myeloid neoplasms and in clonal hematopoiesis, and biological interactions between clonal hematopoietic stem cells and abnormal bone marrow niches. These observations may contribute to explain the wide heterogeneity in patients’ clinical presentation and prognosis, and support the recent effort to include molecular information in prognostic scoring systems used for therapeutic decision-making, leading to promising clinical translation. In this review, we aim to address the topic of PMF molecular genetics, focusing on four questions: (1) what is the role of mutations on disease pathogenesis? (2) what is their impact on patients’ clinical phenotype? (3) how do we integrate gene mutations in the risk stratification process? (4) how do we take advantage of molecular genetics when it comes to treatment decisions?


Mutation and Disease Pathogenesis
The so-called "phenotypic driver mutations", which are those able to drive the myeloproliferative phenotype in PMF, occur in JAK2, CALR, or MPL genes. These mutations were initially described as mutually exclusive, but can indeed coexist in a small proportion of cases [1]. Evidence that mutations in JAK2, CALR, or MPL are sufficient to engender an MPN phenotype has been provided by mouse models, where expression of each mutation alone accurately re-capitulate distinctive features of human disease [2][3][4][5].
In 2005, a single point mutation (V617F) was identified in half of patients with PMF [6][7][8][9]. JAK2 is intimately associated with the cytoplasmic portions of receptors for key hematopoietic cytokines, such as erythropoietin (EPO), thrombopoietin (TPO), and granulocyte colony-stimulating-factor (GCSF). Cytokine receptors are activated by ligand binding, the JAK proteins consequently transphosphorylate one another, which attracts STAT proteins, that are, in turn, phosphorylated by the JAKs. The STATs dimerize and translocate to the nucleus, where they function as a transcription factor to modulate the expression of key genes that regulate proliferation, differentiation, or survival. Normal JAK2 functions to activate intracellular signaling pathways following ligand binding; however, JAK2 V617F is rendered especially when older than 65 years, advocating caution in interpreting the results of genotyping for diagnosis in the absence of overt clinical and hematological abnormalities [24]. Then, the identification of these mutations in triple-negative patients should be carefully interpreted, taking into account also the other diagnostic criteria (clinical phenotype and blood cell count, together with bone marrow abnormalities), since they may be signs of an underlying, antecedent clonal hematopoiesis or may be shared with other myeloid neoplasms, as myelodysplastic syndromes (especially those with fibrosis) and low blast count AML [20]. Finally, the mutation profile in triple-negative patients might help in distinguishing true clonal MF from autoimmune MF. Differentiating between a clonal disease and a secondary phenomenon is, indeed, relevant from a therapeutic standpoint, since the latter is sensitive to steroid treatment [25]. Palpable splenomegaly • Lactate dehydrogenase (LDH) level increased to the above-upper-normal limit of the institutional reference range † In the absence of any of the three major clonal mutations, the search for the most frequent accompanying mutations (e.g., ASXL1, EZH2, TET2, IDH1/2, SRSF2, SF3B1) are of help in determining the clonal nature of the disease.

Mutation and Clinical Phenotype
The type of mutation, mutation allele burden, order of acquisition, and combinations of genetic events may have an impact on the disease phenotype.
JAK2 V617F mutations in PMF are associated with older age, higher HB level, leukocytosis, and lower platelet count [26,27]. Conversely, CALR mutations in PMF have been shown to associate with younger age, higher platelet count, less frequent leukocytosis, anemia, and transfusion requirements, fewer spliceosome mutations, and lower dynamic international prognostic scoring system-plus (DIPSS-plus) scores compared with JAK2-mutated disease [11,28,29]. Some studies specifically aligned CALR type 1 variants with the aforementioned characteristics and phenotypically clustered type 2 variants with JAK2 mutants [29]. Direct comparison of type 1 and type 2 CALR mutations on the basis of hematological and clinical variables has yielded inconsistent findings. CALR type 2 mutations have been shown to correlate with higher risk DIPSS-plus scores, marked leukocytosis, and higher circulating blast percentage compared with type 1 variants [29], whereas other studies failed to observe any substantial phenotypic differences [17,30], Triple-negative PMF patients have been shown to be older, with lower Hb levels, platelet and leukocyte counts, and higher IPSS risk [28]. The risk of thrombosis is higher in JAK2 mutated PMF compared to CALR mutated PMF, despite the fact that calreticulin mutated patients display higher platelet counts, both in essential thrombocythemia and in PMF [28,31]. Thus, JAK2 V617F appears to be thrombophilic also in patients with PMF, as in patients with PV and ET.
Driver mutation allele burden may be another factor that influences the clinical course in PMF. Concerning survival, earlier studies identified a low JAK2 V617F allele burden as a risk factor for shortened survival in PMF [32,33], as compared to patients whose burden was higher than 50%, suggesting a "burn out phase" of the disease, prone to bone mallow failure". Otherwise, newer investigations have found older age and high JAK2 allele burden to be associated with elevated plasma C-reactive protein levels and a pattern of progressive disease, supporting a potential relationship between allelic frequency, inflammation, and disease evolution [34]. Though debated, a high allele burden has also been associated with a greater likelihood of achieving a spleen response under JAK-inhibitor treatment [35]. Concerning the vascular risk, PMF patients with JAK2 V617F allele burden higher than 75% are high-risk patients as they are prone to develop thrombo-hemorrhagic complications during the disease course. A link between CALR mutation allelic burden and disease phenotype has been reported in the literature as it was demonstrated for JAK2 mutation. A high CALR allele burden is more frequent in PMF than in ET [36]. Patients with high CALR-mutant burden have acquired copy neutral loss of heterozygosity (LOH) of chromosome 19p, involving the transition from heterozygosity to homozygosity for the CALR mutation [11]. Anyway, 19pLOH appears to be a relatively uncommon event [37], in contrast with 9pLOH in JAK2-mutant MPN and 1pLOH in MPL-mutant MPN [8, 38,39]. It has been suggested that the clonal expansion in CALR-mutant MPN is faster than that observed in JAK2-mutant MPN [40].
Somatic mutations in MPN do not seem to be acquired in a predetermined order as happens in other malignancies, but occur rather randomly [41]. The order of somatic mutation acquisition can be inferred from the genotypes of detectable subclones. For instance, if some tumor cells have JAK2 V617F, and other from the same patient bear JAK2 V617F with an additional somatic mutation, then this indicates that JAK2 V617F came first. The order of acquisition influences clinical presentation as it has been demonstrated for ET and PV: mutations in either DNMT3A or TET2 are associated with an ET phenotype when acquired prior to JAK2 V617F; by contrast, acquisition of JAK2 V617F prior to mutation of DNMT3A or TET2 is associated with PV [42,43]. Data regarding the impact of mutation order on clinical phenotype in myelofibrosis are still lacking, though we know that in patients with multiple mutations, JAK2 V617F was generally an earlier event in those with PMF [21].
Finally, also the germline genetic background might influence the molecular and cellular consequences of nascent PMF clones after their acquisition of phenotypic driver mutations [44].
Targeted sequencing of gene sets in myelofibrosis will be increasingly used in the near future because it will allow profiling of the mutational landscape of each patient and to better define the role of single or multiple, interacting mutations on clinical phenotype and prognosis [21].

Mutation and Prognosis
Over the past 10 years, prognostic scoring systems have been developed for PMF, including the International Prognostic Scoring System (IPSS) in 2009 [45], the Dynamic International Prognostic Scoring System (DIPSS) in 2010 [46] and the DIPSS-plus in 2011 [47]. They are essentially based on clinical and hematologic parameters, the only genetic parameter being chromosomal abnormalities in the DIPSS-plus.
IPSS was subsequently modified into the DIPSS, which uses the same prognostic variables but assigns 2 points instead of 1 to hemoglobin <10 g/dL and can be applied at any time during the disease course [46]. DIPSS was further refined by incorporating three additional DIPSS-independent risk factors: unfavorable karyotype, PLT count lower than 100 × 10 9 /L, and transfusion requirement. The new prognostic model, DIPSS-plus, identifies four risk groups with median survival ranging from 15.4 to 1.3 years [47].
With the increasing knowledge of the molecular basis of PMF, it is becoming evident that both driver and non-driver mutations have relevance in terms of prognostic significance. PMF carrying CALR mutations [28], especially CALR type-1 mutations [29], have the best prognosis, whereas "triple-negative" PMF have the worst prognosis [28,48]. However, different mutations may interact with each other and modulate prognosis, as it has been shown for CALR and ASXL1: the best survival has been, indeed, reported for those patients harboring a CALR mutation with wild type ASXL1, while patients with the opposite genotype (CALR wild type and ASXL1 mutated) displayed a significantly worse prognosis. A third risk category could be identified, including those patients with both mutations or wild type for both genes [49]. This is somewhat similar to the AML scenario, where both NPM1 and FLT3 mutational status do influence prognosis [50].
Mutations in ASXL1, EZH2, SRSF2, or IDH1/2, called high molecular risk (HMR) mutations, have been shown to represent unfavorable prognostic factors, as these genetic lesions identify PMF patients who are at risk for premature death or leukemic transformation independently of the conventional prognostic scoring systems [27]. Moreover, the number of detrimental mutations represents an additional unfavorable prognostic factor, with two or more mutations being associated with shortened leukemia-free survival [51]. The frequency of HMR mutations and their prognostic consequences seems to be somewhat different between the two PMF subtypes: indeed, an HMR status has been less frequently reported in pre-PMF (27.0%) than in overt PMF (44.4%), while the negative prognostic effect of harboring more than one HMR mutations seems to be higher in pre-PMF, though statistically significant in both entities [52]. The presence of HMR mutations is also associated with a shorter time to treatment failure in patients treated with JAK inhibitors [53].
Together with the above-mentioned HMR mutations, genetic events occurring in the RAS/MAPK pathway genes have been recently associated with high-risk clinical features and worse prognosis, compared with wild type patients [54].
Then in 2018, a clinical/molecular prognostic model that includes both driver mutations and detrimental co-operating mutations was created. The Mutation-Enhanced International Prognostic Score System (MIPSS-70) includes the following risk factors: hemoglobin <10 g/dL, abnormal leukocyte count (either <4 or >25 × 10 9 /L), platelet count <100 × 10 9 /L, circulating blasts ≥2%, constitutional symptoms, bone marrow fibrosis grade, IPSS/DIPPS-plus category, driver mutations, absence of CALR type 1-like mutation, individual HMR mutations, HMR category, and the presence of ≥2 HMR mutations. Each parameter has a weight of 1 point, except for leukocytosis, thrombocytopenia, and the presence of ≥2 HMR mutations, which are each weighted 2 points. Patients are stratified into three categories: low-risk with a median survival of 27.7 years, intermediate-risk with a median survival of 7.1 years, and high-risk with a median survival of 2.3 years. In the MIPPS-70 plus that includes cytogenetic information, four risk categories are delineated with median survival ranging from 1.7 to 20 years [55]. The primary aim of this new score is to better select patient candidates for allogeneic bone marrow transplantation.
Later two new prognostic systems for PMF were introduced: the MIPSS70+ version 2.0 (mutationand karyotype-enhanced international prognostic scoring system) and the GIPSS (Genetically Inspired Prognostic Scoring System).
GIPSS, which is based exclusively on mutations and karyotype, represented the first attempt to model PMF survival on genetic features that fully replace traditional clinical variables, ads VHR karyotype (2 points), unfavorable karyotype (1 point Table 2 and represented in Figure 1.

Impact on Response to JAK Inhibitors or Interferon
The clinical efficacy of ruxolitinib in myelofibrosis seems not to be affected by the underlying driver mutations [60,61], but may be influenced by non-driver genetic events, especially when present in a high number. Patients with ≥3 mutations from a customized, myeloid neoplasms-associated gene panel have a shorter time to treatment discontinuation and shorter overall survival than those with fewer mutations [62]. Moreover, drug efficacy, in terms of both symptoms and spleen response at 6 months, has been recently reported to be lower in patients harboring mutations in the RAS/MAPK pathway, as NRAS, KRAS, or CBL [54].
Furthermore, drug failure has been associated with signs of clonal evolution in a not negligible percentage of cases, ranging from 17% in a cohort of real-life treated patients to 33% in a clinical trial setting, with a consistently negative impact on patients' survival [63,64].
In addition, the response to other JAK-inhibitors, such as momelotinib or imetelestat, may be influenced by the presence of ASXL1/SRSF2 or ASXL1 mutations, respectively [65,66]. The clinical efficacy of interferon (IFN) in myelofibrosis seems not to be affected by the underlying driver mutations but may be influenced by HMR mutations [67,68].

Impact on Transplant Outcome
Different mutations may have an impact on the outcome after hematopoietic bone marrow transplantation (HSCT). Early smaller reports suggested a benefit for JAK2 positive patients [69], whereas later studies demonstrated a positive impact for CALR positive patients and a negative impact on patients' prognosis of triple-negative status [70]. However, in those studies, only driver mutations were investigated. To determine the impact of molecular genetics on outcome after HSCT, the group of Kroger screened 169 patients affected with myelofibrosis for 16 frequently mutated genes [71]. CALR mutation was an independent factor for lower non-relapse mortality, improved progression-free survival (PFS), and overall survival. ASXL1 and IDH2 mutations were independent risk factors for lower PFS, whereas no impact was observed for triple-negative patients. Authors concluded that molecular genetics, especially CALR, IDH2, and ASXL1 mutations, may thus be useful to predict outcomes independently from known clinical risk factors after HSCT. Very recently, driver and subclonal mutations were incorporated in the myelofibrosis transplant score (MTSS), a scoring system that includes clinical-molecular and transplant-specific factors, with the aim of predicting posttransplant outcome. Multivariable analysis on survival identified age ≥57 years, Karnofsky performance status <90%, platelet count <150 × 109/L, and leukocyte count > 25 × 109/L prior to transplantation, HLA-mismatched unrelated donor, ASXL1 mutation and non-CALR/MPL driver mutation genotype being independent predictors of outcome [72].

Minimal Residual Disease (MRD)
The presence of myelofibrosis-specific molecular markers offers the possibility to monitor the depth of remission after HSCT and to start a preemptive therapy at the first, minimal reappearance of disease [73][74][75]. The importance of the clearance of JAK2/CALR/MPL mutation was recently demonstrated in a cohort of 136 patients who received alloHSCT. Patients with detectable mutations at day +100 or at day +180 had a significantly higher risk of clinical relapse at 5 years than molecular-negative patients (62% vs. 10%, p < 0.001) [76]. Minimal residual disease (MRD) detection is clinically relevant as it may prompt a reduction of immunosuppressive agents or a donor-derived leukocyte infusion (DLI). The latter, in detail, is more effective when used for a molecular rather than for a clinical relapse [77]. The sensitivity of the method to detect MRD is pivotal. For JAK2 and MPL, a sensitivity of 10 −5 has been described by PCR technique [73,74].
The use of CALR mutation as a marker of MRD is more problematic due to the variety of indels reported in the CALR gene and the scarce quantification accuracy of fragment length analysis and Sanger sequencing. Both a sensitive qPCR approach and a digital PCR assay that allows the accurate determination of CALR allelic burden for the most frequent alterations (type 1 and type 2) have been developed [81,82]. In addition, very sensitive digital PCR assays have been described for the detection of low JAK2 and CALR mutant burdens [83].
Subclonal mutations, if present, could be extremely useful to monitor MRD, both in triple-negative patients or as an additional tool for those who have a traditional driver event. In this regard, next-generation sequencing (NGS) approaches for MRD monitoring might be hopefully developed in the future.

Conclusions
The rapidly expanding field of molecular genetics for myelofibrosis has opened new horizons in the diagnosis, prognosis, treatment decision-making, and monitoring of this disorder. Recommendations for molecular testing in MPN has been recently provided [84] and include both driver and non-driver genetic events. The latter may aid in the diagnostic process of patients who tested negative for the three driver mutations, in which an extended mutation panel including at least ASXL1, EZH2, IDH1/2, and SRSF2 is recommended. The same extended panel is also appropriate for prognostic prediction, in particular for patients classified as intermediate risk, for which treatment allocation, counseling and/or frequency of monitoring may be refined according to the individual genetic background. High-sensitivity quantitative monitoring of JAK2 V617F and CALR mutations, or quantitative evaluation of another clonal marker in triple-negative patients, are recommended in all patients after HSCT for MRD assessment.
While comprehensive gene sequencing of hematological cancer patients is becoming an increasingly accessible event for routine clinical use, many uncertainties still remain. As an example, how many times should PMF patients be tested? In fact, MPNs are dynamic diseases, so that high-risk genomic features may be gained over time, requiring treatment intervention (or HSCT consideration).
Genome profiling will likely become the basis for a personally-tailored prognosis prediction, but we acknowledge that clinical translation of recent scientific advances is still a work in progress.
Author Contributions: E.R. and E.S. conceived and wrote the review, C.T., D.V., I.C.C., and D.P. contributed to finalizing the manuscript. All authors have read and agreed to the published version of the manuscript.
Funding: This work was supported by a grant from the Italian Ministry of Health for young researchers (GR-2016-02361272) to E.R.

Conflicts of Interest:
The authors declare no conflict of interest.