The key driver gene in the BCR-ABL1-negative MPNs PV, ET and PMF is JAK2, encoding the tyrosine kinase Janus kinase 2, a critical mediator in erythropoiesis [8,9,10]. The point mutation c.1849G>T, affecting exon 14, causes a valine to phenylalanine substitution at position 617 in the JH2 domain.

The JAK2V617F variant is the most frequent mutation in BCR-ABL1-negative MPN, detected in 97% of PV [11], 50–60% of ET, and 55–60% of PMF cases [12]. Unlike the MPNs ET and PMF, mutations in MPL and CALR are not found in patients with PV. However, one study reported two JAK2V617F-negative PV patients who harboured the CALR type 1 (52-bp del) mutation in peripheral granulocytes at diagnosis [13]. Both patients had haemoglobin above the threshold for diagnosis (including according to the WHO’s 2016 revised criteria), moderately elevated platelet counts and normal leukocyte counts, and bone marrow biopsy showed hypercellularity. No mutation was found in JAK2 exons 12, 13, and 14 or in MPL exon 10 for either patient. Nevertheless, this is only an isolated report, and subsequent studies, including the retrospective molecular diagnosis of 524 JAK2V617F-negative patients with suspected MPN, have failed to identify CALR mutations in patients with PV [14].

If the molecular analysis of JAK2V617F is negative, analysis for mutations in exon 12, such as N542-E543del, E543-D544del and K539L, is indicated [15] since mutations in exon 12 of JAK2 are responsible for the majority of the remaining JAK2V617F-negative PV patients. These exon 12 mutations consist of deletions, insertions, duplications and substitutions of bases that affect the region linking the SH2 and JH2 domains and lead to the constitutive activation of the kinase [16]. Nevertheless, a study by Skoda and colleagues found that JAK2V617F and other exon 12 mutations can coexist in two separate clones in some patients [17].

Interestingly, some studies (including those performed using mouse models) have suggested that mutations in exon 12 of JAK2 may confer a different clinical phenotype to JAK2V617F. For example, patients with exon 12 mutations were younger and frequently presented with erythrocytosis [18,19,20]. Nevertheless, in a European collaborative study, 106 patients harbouring the most frequent exon 12 mutations were reported to have similar rates of survival and constitutive symptoms to patients with JAK2V617F [16].

According to the European LeukemiaNet (ELN), the standardized technique for the detection of the JAK2V617F variant is based on allele-specific PCR, which also allows for the quantification of the allelic frequency [21,22]. The technique of digital PCR (dPCR) is increasingly being employed since it permits the absolute quantification of the JAK2V617F variant without the need for the prior use of a standard curve [23,24]. For the detection of other JAK2 variants, Sanger sequencing of genomic DNA is often employed, although it can miss mutations with allele frequencies less than 10–20% [15].

In recent years, more sensitive detection methods, such as sequencing of plasma RNA [25] and NGS [26,27], are being increasingly employed for the molecular diagnosis of patients with PV. NGS provides the advantage of potentially detecting all JAK2 variants, including rare and/or previously undescribed mutations [26].

The most recently developed technique for the genetic profiling of patients with MPN and other haematological malignancies is the isolation of cell-free circulating DNA (ccfDNA) from patient plasma (also commonly known as liquid biopsy) coupled with its analysis by NGS [28]. A recent study assessed the feasibility of mutational analysis by the NGS of 107 patients with MPN (including 33 patients with PV) comparing the use of ccfDNA from plasma versus genomic DNA isolated from granulocytes. The authors confirmed that the sensitivity and accuracy were similar for both types of genetic material tested, with a similar mutational profile identified. However, the median VAF detected was significantly higher in ccfDNA than in DNA extracted from granulocytes (29% versus 25%, respectively) [29].

Nevertheless, despite the use of more sensitive techniques, 2–3% of patients with PV are estimated to be JAK2 wild type. Approximately 2% of such JAK2-negative PV patients harbour a mutation in the gene SH2B3 (also known as LNK), which encodes an inhibitor of cytokine-dependent cell growth in hematopoietic cells. Thus, loss of function mutations in SH2B3 causes the sustained activation of JAK2 resulting in erythrocytosis [30,31]. Indeed, Lnk−/− mice models developed an MPN-like phenotype, with abnormal megakaryopoiesis and erythropoiesis, cytokine hypersensitivity, splenomegaly, fibrosis, and extramedullary haematopoiesis [32,33]. The search for mutations in SH2B3 in JAK2-negative patients suspected of having PV is recommended [34], particularly in the hotspot in exon 2, although mutations in SH2B3 have also been described in PV patients with concomitant JAK2V617F [35].

In summary, when considering the molecular analysis of patients with a suspicion of PV, it is important that (i) techniques with a sensitivity of at least 5% are employed to search for driver mutations, and (ii) such testing is not simply limited to the detection of the JAK2V617F variant or other exon 12 mutations. For JAK2V617F-negative PV patients, additional mutational screening is justified, first in other exons followed by the SH2B3 gene (Figure 1), despite the additional time and financial costs involved [34].

Driver mutations are represented in the gray boxes. The estimated frequency among patients diagnosed with PV is shown in the white boxes. In all cases, the concomitant detection of passenger mutations by NGS is recommended (blue box) for a more accurate prognosis, for example, to predict transformation to MF or AML, as well as to detect mutations associated with increased thrombotic risk, such as TET2. ¹The standardized technique according to the European LeukemiaNet (ELN) is based on allele-specific PCR, which allows for the detection of the JAK2V617F mutation as well as the quantification of the allelic frequency [21]. ²The use of next-generation sequencing (NGS) is highly recommended, or other molecular techniques with a sensitivity of at least 5%.

The JAK2 kinase interacts with the erythropoietin (Epo), thrombopoietin (Tpo), and granulocyte colony-stimulating factor (G-CSF) receptors to regulate the normal proliferation, differentiation, and survival of myeloid cells [36]. The JAK2V617F driver mutation and other gain-of-function JAK2 variants that cause the constitutive activation of the JAK2 kinase induce the proliferation of hematopoietic stem cells and provide them with a survival advantage [37,38,39]. The aberrant activation of JAK2 leads to signalling via the erythropoietin (Epo) receptor without cytokine stimulation [10,38,39] and the overactivation of several downstream signalling pathways, including JAK2/STAT5, RAS/MAPK/ERK and PI3K pathways [40]. In addition to STAT5, STAT3 is also activated in MPNs, with the phosphorylation of STAT3 and STAT5 shown to be uniformly increased in PV, in contrast to ET and PMF.

The JAK2V617F mutation has been shown to be clonal [10,38,39] and present in haematopoietic stem cells (HSCs) [40]. Several studies have demonstrated that the JAK2V617F mutant promotes genomic instability via two mechanisms: increasing levels of DNA damage and decreasing DNA damage-induced cell death. For example, aberrant JAK2/STAT5 and PI3K signalling lead to an accumulation of DNA damage via the production of high levels of reactive oxygen species and the stimulation of homologous recombination, leading to increased amounts of double-strand breaks [41,42,43,44] that in turn impede replication fork progression [45].

Indeed, Quelle et al. linked the misregulation of the JAK2 signalling pathway to the suppression of p53-mediated cell death in response to DNA damage in hematopoietic cells as early as 1998 [37]. JAK2/STAT5 activation promotes the G1-S phase transition of the cell cycle due to the increased degradation of p53 as a result of enhanced MDM2 translation [42,46]. Impairment of the intra-S checkpoint response in erythroblasts from PV but not ET patients was also shown to result in increased DNA damage due to attenuated p53 signalling [45]. Furthermore, JAK2/STAT5 pathway activation induces overexpression of the STAT5 target gene BCL2L1, which codes for the anti-apoptotic protein Bcl-xl [37,47], as well as inhibiting its deamidation [48] preventing apoptosis in response to the detection of DNA damage. Together, these mechanisms result in an accumulation of proliferating malignant cells with unresolved DNA damage.

Interestingly, treatment of mice engrafted with JAK2V617F-mutated bone marrow cells with the antioxidant N-acetylcysteine resulted in reduced levels of DNA damage and decreased splenomegaly [44], while the dietary intervention of patients with MPN, including N-acetylcysteine supplementation, led to a certain improvement in symptom burden [48,49]. These results suggest that the reduction of damaging ROS effects with antioxidants may potentially be a promising therapeutic approach.

The principal model of pathogenesis of PV and other MPN is based on genomic instability; however, increasing evidence is also supporting the consideration that the clonal evolution characteristic of MPN and other cancers may be triggered by chronic inflammation and an escape from anti-tumour immunological surveillance [50,51]. Indeed, the incidence of second cancer has been reported in up to 10% of all MPN patients [52,53], with a second solid or lymphoid tumour detected in 31 of the 1042 PV patients (3.0%) analyzed from the ECLAP study [54].

Additionally, smoking has been shown to be a significant risk factor for MPN development due in part to the persistent stimulation of myeloid cells [55], with a hazard ratio for daily smokers of 2.5 for any MPN, and of 4.3, 1.8 and 1.7 for MF, ET and PV, respectively [56].

Whether immune dysregulation can be considered an MPN disease “driver” or not remains to be fully elucidated [57], but the connection between chronic inflammation and vascular function, and atherosclerosis in particular, is indisputable [58,59]. For a comprehensive recent review on the role of inflammation in MPN, we refer readers to the article by Di Battista and colleagues published in this special edition [60].

Presence of the JAK2V617F variant, although a distinguishing feature of PV, is not a specific molecular marker for PV. Nor is it a marker for the BCR-ABL1-negative MPN, having been detected in other haematological neoplasms including AML, myelodysplastic syndrome, and chronic myelomonocytic leukaemia [61,62]. The variant has even been found at low levels in the peripheral blood of healthy individuals [63].

One of the first molecular markers identified for the diagnosis of patients with PV was the polycythemia rubra vera 1 gene (PRV1). Granulocytes isolated from the peripheral blood of patients with PV and some cases of ET had an increased expression of PRV1 together with a decreased expression of MPL, with a positive correlation observed between PRV1 expression and splenomegaly, helping to differentiate MPN from secondary erythrocytosis and thrombocytosis [64,65]. However, the use of PRV1 lost popularity when the JAK2V617F driver mutation was discovered in 2005 [10] and PRV1 was not included as a diagnostic marker for PV in the 2008 WHO classification [1]. Moreover, PRV1 expression does not discriminate PV from other BCR-ABL1-negative MPN [66].

A second molecular marker whose expression in circulating mononuclear cells was shown to be significantly increased in 87% of patients with PV but not secondary polycythaemia is the insulin-like growth factor 1 receptor (IGF-1R) [67,68]. The levels of IGF-1R in PV were significantly different in patients with PV compared with patients with ET or PMF, although expression of IGF-1R was significantly higher in patients with the other MPN compared to controls [68].

Distinguishing PV from ET can be complicated due to the shared phenotypic characteristics of the two diseases, particularly the differential diagnosis of PV from JAK2V617F-positive ET. Prior to the 2016 WHO update to the diagnostic criteria for MPN (in which haemoglobin and haematocrit diagnostic thresholds for PV were lowered to 16.5 g/dL and 49% for men, and 16 g/dL and 48% for women, respectively, and bone marrow biopsy was made mandatory, except for cases fulfilling the 2008 WHO diagnostic criteria [2]), PV patients with thrombocytosis at disease presentation were frequently misclassified as JAK2V617F-positive ET patients [69,70,71,72]. It was shown that “masked PV” [71,72] could be distinguished from ET based on red blood cell mass (RCM), with masked PV presenting an increased RCM above 125% [70,73].

This differentiation is important for the prognosis of a patient with MPN, since a higher rate of thrombosis exists among patients with PV than ET, as well as a higher rate of transformation to secondary myelofibrosis (SMF) and AML in PV than ET [71,74]. Importantly, the combined rate of transformation to SMF and/or AML for masked PV was 1.60 per 100 patient years compared to 0.97 for overt PV [74]. These inferior outcomes may possibly be due to a more aggressive form of the disease, the lower intensity of treatment for masked PV patients misdiagnosed with ET, or perhaps due to a later diagnosis [75]. For example, masked PV patients misclassified as JAK2V617F-positive ET would not be phlebotomized. Moreover, misdiagnosed younger patients without prior thrombosis would be categorized as low risk (according to current international guidelines for ET patients) and thus would not even receive cytoreductive treatment, even in those presenting with leucocytosis and thrombocytosis [75].

The differentiation between PV and ET at a molecular level can also be complicated by the presence of passenger mutations in genes common to both MPN. However, mutations in the genes KMT2A (MLL) and TP53 were more common in patients with PV (13.64% and 6.25%, respectively) than patients with PMF (3.45% and 0%, respectively) or ET (0% and 0%, respectively) [76]. Some studies have characterized the expression of several genes and proteins in an attempt to differentiate PV. For example, the overexpression of NFE2 was described in patients with PV and not in patients with ET [77], while high expression of the proteins HSP70 and calreticulin in the membrane of erythrocytes of patients with PV compared with patients with ET suggested an important role for these factors in the proliferation and erythroid function of PV, respectively [78,79]. Moreover, the expression of microRNAs (miRNA), single-stranded non-coding RNA 18–24 nucleotides in length that regulate gene expression by binding to target mRNA, was shown to be different between patients with PV and those with ET (Table 1), as well as between patients with PV and control patients [80,81]. For example, miR-145 and miR-451 were upregulated in nucleated erythroid cells cultured from progenitor cells from patients with PV compared to patients with ET, whereas miR-575 was upregulated and miR-196b was downregulated in PV patients compared to healthy controls [81].

Nevertheless, the JAK2 allelic burden is sometimes employed to differentiate patients with PV from patients with ET [82,83,84]. Homozygosity of JAK2V617F (represented by a VAF of 50% or more) or a high JAK2V617F allele burden resulting from mitotic recombination of chromosome 9 or 9p loss of heterozygosity [10,39] occurs more frequently in patients with PV while it is significantly less common in patients with ET [83,85]. This observation was confirmed in transgenic mice, in which the increased expression of JAK2V617F resulted in a change of phenotype from ET to PV [86,87].

A hereditary basis of PV (and other MPNs) was first suggested by observations of a particularly high incidence of PV among people of an Ashkenazi Jewish descent [88], although the common genetic predisposing factor among this population has still yet to be identified. However, the familial basis of PV may be more frequent than previously anticipated, with a population study finding a 5.7-fold increased risk of PV among the first-degree relatives of patients with MPN [89]. Moreover, an interview-based investigation of family history identified familial cases of MPN in 18 of 206 patients with PV (8.7%), with 13 families having first-degree relatives affected by PV [90].

Genetic studies observed that carriers of the JAK2 46/1 haplotype (GGCC) had a higher susceptibility to develop MPN [91,92]. Specifically, carriers of the 46/1 haplotype had a 3.7-fold increased risk of acquiring the JAK2V617F mutation [93], a higher susceptibility to acquire a JAK2 exon 12 mutation [19,94], and an increased disposition (odds ratio of 2.3) for the development of PV versus ET [85].

However, despite the predominance of the JAK2V617F mutation among patients with PV, earlier studies determined that this “driver” mutation was not always disease initiating [95] and might occur as a later secondary genetic event in some patients with MPN [96,97]. Familial cases of MPN, where different family members have a different driver mutation and/or a different MPN, support a pre-JAK2 event and are suggestive of an unknown germline event that predisposes to MPN (unpublished observations, [98]).

One hypothesis is that JAK2 germline mutations, such as JAK2T108A and JAK2L393V, are the genetic predisposing factor that precedes the acquisition of JAK2V617F in MPN [99]. Albeit of uncertain significance, these germline mutations were identified in PV patients as co-existing predisposing mutations with somatic JAK2V617F, and have also been reported in an adenocarcinoma cell line and in diffuse large B cell tumours, respectively, supporting the observation that they provide a proliferative advantage. Nevertheless, their effect on disease initiation has not been fully determined.

Other predisposing germline mutations have been identified among patients with familial MPN, including mutations in the RBBP6 gene, encoding an E3 ubiquitin ligase known to promote the degradation of p53. Variants such as RBBP6R1569H were reported to lie in the p53-binding domain, and thus are predicted to affect p53 function [100]. Other germline predisposition variants for MPN include the single nucleotide polymorphism (SNP) rs2736100 in the gene encoding the telomerase reverse transcriptase TERT [101]. In fact, both the TERT SNP rs2736100 and the JAK2 GGCC haplotype had an additive effect on MPN risk in the presence of an RBBP6 mutation and a higher VAF [102].

Genome-wide association studies (GWAS) have confirmed that various loci associated with telomere length, including rs7705526 and rs2853677, are predisposing factors for MPN risk [103]. Specifically, an increased risk of MPN development was observed for individuals with longer telomere length measured in leukocyte (LTL), as determined using a combined “teloscore” from 11 telomere-length-associated SNPs [104]. The authors of the study concluded that longer telomeres are a risk marker for MPN development.

Nevertheless, a ground-breaking study, presented at this year’s American Society of Hematology (ASH) Annual meeting, demonstrated that driver mutations, such as JAK2V617F were frequently acquired in childhood or even in utero, many decades before an MPN diagnosis [105]. The authors observed a mean time of 34 years between driver mutation acquisition and MPN presentation in patients where JAK2V617F was the only or the first driver mutation. In 4 of the 10 patients analyzed, DNMT3A single-nucleotide variants (SNVs) were the driver event, followed by the acquisition of JAK2V617F, although the DNMT3A mutations were also acquired in childhood or in utero. The authors concluded that the observed latency between driver mutation acquisition and MPN presentation suggests that other factors, including germline, must impact the selection of clones harbouring driver mutations and favor their clonal expansion. Indeed, while normal hematopoietic stem cells accumulated approximately 18 somatic mutations per year, clones harbouring driver mutations such as JAK2V617F accumulated 1.5–5.5-fold more mutations per year and were also found to have shorter telomeres.

Genetic analyses have led to the description of diverse non-driver mutations in patients with PV, particularly as a result of NGS studies. These include SNVs in genes that encode proteins that regulate epigenetic processes, such as TET2, DNMT3A, ASXL1, EZH2, and IDH1/2, as well as proteins implicated in the splicing of messenger RNA, such as SRSF2, SF3B1 and U2AF1 [12,106], with non-driver mutations detected by NGS in 53% and 75% of PV patients according to two recent studies [107,108].

The effect of passenger mutations on the prognosis of the disease is still not well characterized, although some studies have associated SNVs in ASXL1, SRSF2, and IDH1/2 with poor survival [107,109,110]. Mutations in two novel genes were reported at this year’s ASH Annual meeting to have prognostic value in MPN. SNVs in the gene NFE2 (nuclear factor erythroid 2), with a role in the maturation and differentiation of erythroid cells, were found to be independently associated with shortened time to blastic transformation (with a hazard ratio of 9.9) and overall survival. Such mutations were most common in PV patients compared to PMF and ET, detected in 7.3%, 5.3% and 3.6%, respectively [111]. Similarly, SNVs in the gene STK11 (serine/threonine kinase 11), encoding a kinase with tumour suppressor function, were found to promote leukemic progression in patients with MPN [112].

However, the clinical relevance of some passenger mutations in patients with MPN is still a matter of debate. For instance, the presence of somatic mutations in hematopoietic tissue is positively correlated with age and confers a selective advantage leading to clonal haematopoiesis of indeterminate potential (CHIP, with an established cut-off of ≥2% VAF of in peripheral blood) [85,113,114,115,116,117,118,119].

The most frequently mutated CHIP-associated genes are TET2, DNMT3A, ASXL1 (commonly referred to as the DTA genes), and JAK2. Recent studies have found an association between the accumulation of DTA SNVs and a higher risk for developing myeloid neoplasms, including BCR-ABL1-negative MPN and AML, as well as cardiovascular pathologies and a higher prevalence of developing derived complications from diabetes [119].

TET2 SNVs are particularly common in PV, detected in between approximately 10% and 36% of patients [12,106,108,120,121], and are frequent in both JAK2V617F-positive and -negative MPN. Experiments in mice suggested that mutated TET2 confers a clonal advantage to cells mutated in JAK2 that accelerates their proliferation [122,123]. Moreover, the acquisition of a mutation in DNMT3A followed by mutation in JAK2 was shown to lead to an ET phenotype, whereas the acquisition of the JAK2 mutation first was shown to result in a PV phenotype [124].

Nevertheless, there remains some controversy over whether the presence of CHIP mutations is sufficient for the clonal expansion that results in the emergence of MPN since DTA mutations (as well as the JAK2V617F mutation) are often identified in healthy individuals [116]. An alternative hypothesis is that such mutations simply predispose individuals to acquire mutations in other oncogenes or other genomic abnormalities necessary to drive clonality [105].

Sequencing studies and colony genotyping are helping to determine clonal evolution [125,126] and the order of mutation acquisition. For example, mutations including NFE2, TP53, NRAS, and PPM1D were found to be acquired significantly later in MPN disease [85]. A seminal study by Ortmann et al. investigated the influence of the order of mutations in JAK2 and TET2 in patients with MPN [127]. Firstly, the authors observed that among MPN patients with mutations in both genes, the prior acquisition of TET2 mutations diminished the proliferative effect of the JAK2V617F mutation. Secondly, they found that PV was more common among JAK2-first patients compared to ET or PMF. Finally, they concluded that the order of mutation acquisition influenced the age at which a patient presented with an MPN, with JAK2-first patients presenting at a younger age compared with TET2-first MPN patients. This observation was supported by a recent study which concluded that the order of mutation acquisition was associated with time to MPN presentation [105].

Although quantification of the JAK2V617F allelic burden is not obligatory for the diagnosis and follow-up of patients with PV according to the WHO criteria [1,2], patients homozygous for JAK2V617F were shown to be more symptomatic than patients who were heterozygous. Specifically, a positive correlation was described between a high allelic frequency of mutated JAK2 and clinical presentation including pruritus [131,132,133], myelopoiesis [11,132], splenomegaly [11,131,132,134,135], and a negative correlation with the platelet count [11,132,135,136].

Moreover, determination of the JAK2V617F VAF at diagnosis can provide important prognostic information. For example, in PV, a higher mutant allele burden has been associated with an increased incidence of progression to SMF [11,83,131,137].

Both of these observations are suggestive of a “dosage” effect (reviewed in [98]), where a high JAK2V617F burden results in a stronger activation of JAK2/STAT (and possibly other) signalling pathways, leading to the stronger activation of downstream genes [138], and thus a more aggressive phenotype. In support of this hypothesis, the expression level of several JAK2 target genes—such as PRV1 [138] and NFE2 [139]—was shown to be dose dependent for the mutant allele burden.

A high JAK2V617F allele burden in patients with PV has also been associated with a risk of thrombotic events [140,141,142]. A study of 1306 patients with MPN including 397 patients with PV, observed an incidence of thrombosis at a diagnosis of 21% for homozygous and 15% for heterozygous JAK2V617F [142], with JAK2V617F homozygosity at diagnosis identified as an independent risk factor for major vascular events in a multivariate analysis [141]. A second study reported a 7.1-fold increased risk of major vascular events for patients with a mutant allele of 75% and above compared to less than 25% VAF during follow-up [130]. However, one study associated a significantly increased risk of venous thromboembolism with a much lower mutant allele burden, of just 20% or more [131]. Despite the differences in the VAF threshold reported to have a negative effect on thrombotic risk, it is clear that patients with a higher JAK2 mutant burden are at risk of both vascular events and fibrotic transformation.

Nevertheless, one study’s findings contradict the prognostic impact of the JAK2V617F allelic burden in patients with PV. Gene expression profiling of CD34+ cells isolated from 19 patients with PV led to the identification of a group of patients with a more aggressive disease—with higher rates of blastic transformation and worse overall survival—and a group with a less aggressive form despite similar JAK2V617F allele frequencies in both groups. Patients in the group with more aggressive PV showed differential expression of the NOTCH and SHH pathways as well as inflammatory cytokines and histone genes [143]. Interestingly, this study also revealed gender-specific differences in the expression profile in that men with PV had significantly more differentially regulated genes than women with PV.

Cytogenetic abnormalities are detected in approximately 10–20% of PV patients at diagnosis, with some of the most common alterations including gain of chromosomes 8 and 9, and deletion of (1p), (13q) and (20q) [144,145,146].

Although some studies have not shown a prognostic difference based on cytogenetic characteristics [146], several groups have reported, as association between a higher risk of fibrotic progression and the presence of chromosome 12 abnormalities [147], the gain of (1q) [148], and an abnormal karyotype. Moreover, the majority of investigations, including one by the International Working Group for Neoplasms Research and Treatment (IWG-MRT), have reported an association between an abnormal karyotype and transformation to myelodysplastic syndrome or AML [146,148,149,150] and poorer overall survival [145,148,150].

Interestingly, a retrospective study of 422 PV patients with cytogenetic information available at diagnosis observed dynamic changes associated with progression to AML, including an increased frequency of abnormal karyotype, increasing from 20% to 90% in the blast phase, as well as a different distribution of cytogenetic abnormalities [149]. For example, complex karyotype as well as deletion of (5q)/chromosome 5, 7q/chromosome 7, and (17p)/chromosome 17/i(17q) were more common in the blast phase, whereas the gain of chromosomes 8 and 9 were more common in the chronic phase of the disease. Using this information, the authors stratified patients into low (normal karyotype; +8, +9; or presence of one other alteration), intermediate (del(20q), +(1q) or presence of two other alterations) and high-risk (complex karyotype) groups according to their cytogenetics at diagnosis [148].

The BCR-ABL1-negative MPNs PV, ET and PMF are associated with a high frequency of haemorrhages and thrombosis, including myocardial infarction, ischemic stroke, deep vein thrombosis, and thrombo- and pulmonary embolisms. Notably, vascular events are the major cause of morbimortality among patients with PV, with the incidence of events estimated between 6% and 17% over three years [151,152].

Thrombotic risk prediction tools have been developed for each BCR-ABL1-negative MPN subtype. However, thrombotic events are frequent among patients with PV, even among those classified as low-risk according to the current ELN recommendations [21,153], as well as patients who receive cytoreductive or anti-aggregant treatment [154].

Several common thrombophilia markers are associated with an increased risk of developing vascular events in patients with PV (reviewed in [155]), including high levels of C-reactive protein in blood serum and SNVs in the factor V-Leiden (F5) gene [156,157,158]. Higher factor VIII levels, a known risk factor for venous thrombosis and coronary artery disease [159], were identified in PV patients compared to healthy subjects, with levels of 141 versus 98 IU/dL, respectively [160]. The PV patients were also found to have significantly higher VWF antigen and activity which, by multivariate analysis, was predicted for by JAK2V617F allelic burden. Nevertheless, the ELN does not currently recommend thrombophilia testing as part of routine clinical practice in PV or MPN patients [21,153].

Other prognostic biomarkers reported for PV include an abnormal karyotype, identified as a risk factor for venous thrombosis [145], and leucocytosis, a risk factor for arterial thrombosis, although its impact on venous thrombosis remains inconclusive [161,162].

The International Prognostic Score of thrombosis in World Health Organization-essential thrombocythemia (IPSET-thrombosis) was the first prognostic algorithm to incorporate the JAK2V617F mutation to better predict the risk of thrombosis among ET patients [163], although the influence of JAK2V617F in PV patients as well as the clinical significance of the allelic burden on thrombotic risk remain to be determined. Some studies have found that a higher JAK2V617F allelic burden was associated with increased risk of thrombosis in PV and ET [164]. For example, JAK2V617F allele burden cut-offs for PV > 25.7% and > 90.4% were established for arterial thrombosis and for venous thrombosis, respectively [165].

Moreover, elevated levels of the inflammatory biomarker C-reactive protein in peripheral blood were associated with the PV phenotype (vs. essential thrombocythemia), older age, cardiovascular risk factors and a JAK2V617F allele burden over 50% [59,166], further supporting suggestions of a possible link between inflammation and atherosclerosis [58,59,60]. A seminal study from 2017 identified an association between mutations in the DTA and JAK2 genes and coronary disease [119]. Mechanistically, mutations to DTA genes are suggested to alter methylation patterns and thus cause an increased transcription of pro-inflammatory genes triggering atherosclerosis [119,167,168]. To support this theory, Fuster and colleagues studied the effect of the expansion of TET2-deficient cells in atherosclerosis-prone, low-density lipoprotein receptor-deficient mice and observed that the partial reconstitution of the bone marrow with TET2-deficient cells was sufficient to generate clonal expansion, which was also associated with a marked increase in the size of atheromatous plaques and the increased secretion of pro-inflammatory cytokines such as IL-1β [167].

In accordance with these results, Jaiswal et al. found that individuals with clonal hematopoiesis had a 1.9-fold higher risk of developing coronary disease and a 4-fold higher risk of myocardial infarction, rising to 12.1-fold in individuals with the JAK2V617F mutation [119]. In support of the link between constitutive JAK2 activation and increased risk of vascular events, the rs3184504 SNP in the SH2B3 (LNK) gene was also associated with an increased risk of coronary heart disease [169]. This variant, encoding the TT genotype (R262W), caused an increased platelet count, leukocytosis and hypertension. At the molecular level, SH2B3 mutation increased megakaryopoiesis via the upregulation of AKT signalling, shown to be specifically upregulated in platelets and to promote prothrombotic and proatherogenic aggregates [170]. Specifically, our group recently showed that DTA mutations were associated with increased thrombotic risk in patients with PV, with a specific interaction identified for pathogenic TET2 mutations [108].

Activated leukocytes (including neutrophils and monocytes) appear to play an important role in the development of platelet aggregates, markers of a prothrombotic state. In addition, there are reports that leuko-platelet aggregations directly correlate with platelet and leukocyte counts [171,172], adding weight to leukocytosis as a thrombotic risk factor. Interestingly, and as previously mentioned, the expression of CD177 (PRV-1), a GPI-linked cell surface glycoprotein with a role in neutrophil activation, was increased in PV patients, and to a lesser extent in ET and PMF patients [64,65]. Indeed, higher CD177 expression was associated with an increased risk of thrombotic and bleeding complications due to increased circulating neutrophils [173,174]. For a complete and recent review on the role of neutrophils in thrombosis, we refer readers to the article by Ferrer-Marín and colleagues published in this special edition [175]. It is clear that we need to further elucidate the molecular mechanisms of leukocyte–platelet interactions in order to design better prophylactic treatment plans for the prevention of thrombotic complications in patients with PV.

Progression to fibrotic disease (post-PV MF) or blastic disease, both associated with dismal survival rates, is estimated to occur in 10% and 15% of PV patients, respectively [3]. Molecular indicators of fibrotic progression from PV are not well characterized. Some evidence exists suggesting an association between chromosome 12 abnormalities and an increased risk of developing post-PV MF [176]. Evidence also exists to support a significantly higher JAK2V617F allelic frequency [11,83,131,137,176] or a progressive increase [177] in patients who transform to SMF. Unlike in PMF, the presence of additional high-risk mutations (ASXL1, EZH2, IDH1, IDH2 or SRSF2, as single or multiple mutations) was not correlated with reduced leukaemia-free survival or indeed overall survival, with the exception of mutations in SRSF2, associated with reduced survival [178,179].

As the molecular profile of post-PV MF was shown to be considerably different from PMF [180], the Myelofibrosis Secondary to PV and ET Prognostic Model (MYSEC-PM) was specifically developed to predict survival in post-PV/post-ET MF (Table 4) [181].

The score was shown to more effectively assess prognosis in post-PV MF patients, with poor agreement for risk classification between the DIPSS and MYSEC-PM scores when applied to the same post-PV patients [180,182]. Therefore, the MYSEC-PM is the most appropriate prognostic score for post-PV MF patients rather than the International Prognostic Scoring System (IPSS) or Dynamic IPSS (DIPSS) scores, both of which were developed for PMF.

Interestingly, in a retrospective study of 376 patients with SMF, including 188 post-PV MF, monosomal karyotype (observed in 8.5% of the SMF patients studied) was associated with significantly worse survival, with a median of 2 years, independently of the MYSEC-PM risk group [183].

In terms of blastic transformation, it is also clear that the genetic profile of patients with post-PV AML differs from that of de novo AML. For example, mutations in FLT3 and NPM1 are very uncommon and the JAK2V617F-positive clone is frequently lost in transformation to AML [184,185]. Moreover, abnormal karyotype is a risk factor for leukemic transformation in PV and ET [145,146,147,150], estimated to occur in 80% of PV cases that progress [146].

In some cases, the transformation from PV to AML is associated with the loss of function of TP53 due to an acquired mutation in the TP53 gene leading to clonal dominance [85,177,186,187,188]. In fact, in a mouse model, the expression of JAK2V617F in combination with the loss of TP53 was demonstrated to cause an AML phenotype [188]. This result was reflected in a later NGS study, which observed that post-PV/-ET AML patients with a TP53 mutation had a survival rate at 12 months of just 18% compared to 48% for patients without a TP53 mutation [187].

Other mutations associated with a higher risk of leukemic transformation include ASXL1, EZH2, RUNX1 and SRSF2 [85,130,187,189]. Indeed, a recent French study that investigated the time between diagnosis and leukemic progression in 49 patients with post-PV/ET AML (including 24 post-PV) observed that mutations in IDH1/2, RUNX1, and U2AF1 were associated with a shorter time to transformation, while mutations in the genes TP53, NRAS, and BCORL1 were associated with a longer time to transformation [187]. Such mutations are often detectable at low allelic frequencies in earlier stages of the disease, while others may appear later in the progression [186,187,188].

TET2 was also observed in some studies to be a gene whose mutation is associated with the leukemic transformation of PV patients. One study that investigated mutations in 63 patients with AML secondary to an MPN found that mutations in TET2 were acquired at transformation in 43% of cases [190], and the acquisition of mutations in TET2 in patients with PV JAK2V617F was associated with transformation to AML and reduced survival [120]. Nevertheless, the prognostic impact of TET2 mutations, and indeed other mutations associated with CHIP, is not entirely clear. For example, when the impact of additional mutations on the risk of leukemic transformation was investigated, the presence of >1 additional SNV was associated with a higher risk of leukemic transformation and this association was even higher when TET2 SNVs were removed [187]. Moreover, mutations in ASXL1 were observed at similar allelic frequencies before and after transformation, suggesting that they may not contribute to the dominance of the emerging leukemic clone [188].

One of the largest sequencing studies performed to date, on 2035 MPN patients (including 356 PV patients), identified eight genetic subgroups [85]. The authors developed a sophisticated prognostic model based on 63 clinical and genomic variables to estimate a patient’s probability of leukemic transformation. These findings have now been adapted into a personalized risk calculator, available online (https://cancer.sanger.ac.uk/mpn-multistage/, accessed on 26 November 2020).