Genetic and Epigenetic Complexity in Myeloproliferative Neoplasms

Clonal hematopoiesis and cytokine hypersensitivity are fundamental features of BCR-ABL–negative myeloproliferative neoplasms (MPNs) and have played an important role in distinguishing these disorders from reactive conditions. In routine practice, clonality is usually established by the finding of an acquired mutation or cytogenetic abnormality, although diagnosis of each specific subtype, polycythemia vera (PV), essential thrombocythemia (ET), and primary myelofibrosis (PMF), critically depends on additional clinical, laboratory, and morphological information. Because there is no strong evidence for substantial genetic differences in the makeup of PMF and post-PV/ET myelofibrosis, these entities are referred to collectively as myelofibrosis (MF) in this review. Many of the abnormalities described below are not unique to MPNs, but are also seen in other myeloid malignancies such as myelodysplastic syndrome, myelodysplastic syndromes (MDS), myelodysplastic/myeloproliferative neoplasms (MDS/MPN) and acute myeloid leukemia (AML).

Direct or indirect dysregulation of JAK2 signaling by somatically acquired mutations has emerged as a central theme in the pathogenesis of MPNs.¹  JAK2 is a nonreceptor tyrosine kinase that plays an essential role in transducing signals from several class I cytokine receptors that are critical for normal myelopoiesis, notably the erythropoietin receptor (EpoR), the thrombopoietin receptor (TpoR, encoded by the MPL gene), and the granulocyte colony-stimulating factor receptor (G-CSFR). In the normal, unstimulated situation, JAK2 adopts an inactive conformation despite being noncovalently bound to class I receptors via its N-terminal FERM domain. Binding of specific ligands to their cognate receptors (eg, erythropoietin to the erythropoietin receptor) induces a structural change that activates JAK2 signaling, at least in part by relieving the negative regulation of catalytic activity imposed by the JH2 domain (also known as the pseudokinase domain). Activated JAK2 phosphorylates specific tyrosine residues on itself (autophosphorylation) and other proteins (transphosphorylation), and these phosphorylated tyrosines serve as specific docking sites for the recruitment and assembly of downstream signaling proteins. This in turn results in the activation of specific signaling cascades involving MAPK, PI3K, and STAT proteins. Signaling is attenuated by activation of multiple negative feedback mechanisms that may involve protein tyrosine phosphatases, suppressors of cytokine signaling (SOCS) family members, casitas B-lineage lymphoma (CBL) family members, protein inhibitors of activated STATs (PIAS), LNK, and other proteins.

By far the most prominent mutation is the G → T change in exon 14 of JAK2 that results in substitution of the normal valine residue at position 617 by phenylalanine.² JAK2 V617F is seen in > 95% of cases of PV and slightly more than half of cases of ET and MF (Table 1). Mutations of JAK2 exon 12 are seen in roughly one-third of cases of V617F-negative PV, many of which present with apparently isolated or idiopathic erythrocytosis. These mutations are usually complex insertion/deletion events affecting a region predicted to link the SH2 and JH2 domains (residues 537-543), although a single K539L substitution has been reported.³ 

Both V617F and exon 12 mutations result in constitutive activation of JAK2 in the absence of class I receptor stimulation. Although the physical structure of JAK2 is not available, homology-based modeling suggests that valine 617 is located at a predicted contact point between the catalytic domain and the JH2 domain of JAK2; substitution with phenylalanine is believed to relieve the negative regulation of kinase activity imposed by the JH2 domain. Similarly, the exon 12 mutation target region is predicted to be close to the same contact point and presumably deregulates JAK2 in a similar way.

Indirect dysregulation of JAK2 signaling can arise by several different mechanisms. Probably the most common are activating mutations of MPL, seen in a minority of cases of ET and MF but not PV. MPL mutations are located in exon 10 and most commonly result in amino acid substitutions at tryptophan 515 of the TpoR such as W515L, W515K, or W515A, although other variants at this residue have been reported.^4,5  W515 is located within a small amphipathic motif at the junction between the transmembrane and cytoplasmic domains that prevents TpoR autonomous activation⁶ ; mutation of this residue abrogates this function, thereby constitutively activating JAK2 and downstream signaling. Mutations resulting in an S505N change are seen occasionally and also result in TpoR activation. Other rare variants in MPL and JAK2 have been described, but their significance is unclear.⁵ 

Inactivating mutations of SH2B3, which encodes the signaling adaptor LNK, have been reported in a small proportion of cases for all MPN subtypes, but the mutation frequency increases to 10% upon transformation to the blastic phase.^7,8 SH2B3 mutations are localized mostly to the region encoding the pleckstrin homology (PH) domain. LNK negatively regulates JAK2 signaling and, although the precise mechanistic details remain to be established, it appears that LNK is constitutively bound to the receptors such as TpoR and is phosphorylated upon ligand-induced activation in a JAK2-dependent manner. Phosphorylated LNK binds strongly to JAK2, and this interaction is presumably important for the negative regulation of signaling. LNK can bind to and attenuate signaling from JAK2 V617F and TpoR W515 mutants in cell line models; however, it is unclear to what extent this happens under normal physiological conditions in human disease.⁹  Usually, mutations that activate JAK2 signaling are mutually exclusive, implying functional redundancy, but several JAK2 V617F–positive cases with SH2B3 mutations have been described, suggesting that the mutations may cooperate.⁸  However, the finding of multiple mutations in a single patient does not necessarily mean that they are in the same clone. SH2B3 mutations are usually heterozygous, but it is unclear whether they are haploinsufficient (ie, loss of one copy gives rise to the phenotype) or if the mutations act in a dominant-negative fashion to inhibit the activity of the wild-type allele.

Finally, CBL mutations have been reported in approximately 10% of cases of MF.¹⁰  CBL is a well-characterized protein that plays both positive and negative regulatory roles in tyrosine kinase signaling. In its positive role, CBL binds to activated signaling complexes and serves as an adaptor by recruiting downstream signal transduction components. However, the CBL RING domain has E3 ligase catalytic activity and ubiquitinylates activated target proteins on lysine residues. This negative regulatory role of CBL is best characterized for receptor tyrosine kinases, in which lysine ubiquitinylation serves as a signal that triggers internalization of the receptor/ligand complex and subsequent recycling or degradation in endosomes. However, CBL also targets other proteins for degradation, most notably STAT5, a key downstream component of JAK2 signaling. CBL mutations are almost always missense changes in exons 8 and 9 that inactivate the E3 ligase activity. The loss of catalytic activity but retention of other functions gives rise to a gain of function and hypersensitivity to multiple cytokines.^10,11 

JAK2 V617F is remarkably recurrent, a fact that merits some consideration. Although it is conceivable that something about the sequence or environment of JAK2 exon 14 makes it particularly prone to somatic acquisition of this mutation, no clear evidence for mutability has been described and no mechanism has been proposed.

In vitro mutagenesis studies have demonstrated that only large, nonpolar amino acid substitutions at V617 activate JAK2, and only one change, V617W, is as strongly activating as V617F. The fact that this change has not been described in human disease is explicable from a purely genetic perspective: the V617W substitution requires mutation of at least 3 bases, whereas V617F requires only a single base change and is thus is far more likely to occur.¹²  Nevertheless, the fact that point mutations are not commonly seen at other residues suggests that V617F has very specific consequences. From a physical perspective, V617F creates the opportunity for π-stacking of local phenylalanine residues and thus may confer a more stable structural change than other substitutions.¹³ 

From a functional perspective, 2 noteworthy gains of function have been attributed to V617F, which demonstrate that this mutation is not simply a constitutively active form of JAK2. First, SOCS3 is known to be a strong negative regulator of erythropoietin (EPO) signaling; it is thought to bind to the catalytic loop of JAK2 and target the kinase for degradation. JAK2 V617F escapes this negative regulation and, remarkably, its activity is actually enhanced by SOCS3. This is probably a consequence of an acquired ability of JAK2 V617F to hyperphosphorylate and stabilize SOCS3, thereby prolonging signaling.¹⁴  Therefore, JAK2 V617F not only escapes normal negative regulation, but may even exploit SOCS3 to potentiate its myeloproliferative capacity.

Second, JAK2 V617F (and also JAK2 K539L) acquires the ability to phosphorylate PRMT5, impairing its ability to methylate histone H2A and H4 on specific arginine residues. Knockdown of PRMT5 in human CD34⁺ cells increased colony formation in methylcellulose and erythroid differentiation in liquid cultures. Conversely, overexpression of wild-type PRMT5, but not an inactive mutant, decreased hematopoietic colony formation. These data indicate that JAK2 mutant–mediated abrogation of PRMT5 activity may also make an important contribution to the MPN phenotype.¹⁵ 

Epigenetics refers to changes in phenotype or gene expression that are heritable through cell division but are caused by mechanisms other than changes in the underlying DNA sequence. Mechanistically, epigenetics generally refers to DNA methylation of CpG dinucleotides, a modification associated with gene silencing, or a growing range of histone modifications that are associated with transcriptional activation or repression. Mutations in genes known or suspected to encode proteins involved epigenetic regulation are common in MPNs (Table 1), but their functional consequences are poorly understood.

Epigenetic mutations in MPNs are most commonly seen in TET2, which encodes a member of a family of 3 enzymes that convert 5-methylcytosine (5mC) to 5-hydroxymethylcytosine (5hmC). Inactivating TET2 mutations are seen in all MPN subtypes.^16,17  Whether 5hmC is simply an intermediate in the removal of 5mC from specific CpG dinucleotides or whether it confers specific properties is currently a matter of debate, but analysis in embryonic stem cells has suggested that the balance between 5mC and 5hmC is strongly linked with the balance between pluripotency and lineage commitment.¹⁸  Furthermore, depletion of Tet2 in mouse hematopoietic precursors skewed their differentiation toward monocyte/macrophage lineages in culture.¹⁹ 

Mutations in isocitrate dehydrogenase 1 and 2 (IDH1 and IDH2) are seen occasionally in chronic-phase MPNs, but are more frequent in blastic phase disease.²⁰  IDH1/2 normally catalyzes the oxidative decarboxylation of isocitrate to α-ketoglutarate (the third step in the Krebs cycle), but IDH mutants exhibit a neomorphic gain of function characterized by aberrant production of 2-hydroxyglutarate (2HG). TET2 function depends on α-ketoglutarate and is impaired in IDH1/2 mutant cells; in AML, IDH1/2 and TET2 mutations are mutually exclusive and associated with similar epigenetic defects. Expression of mutant IDH1/2 or Tet2 depletion in mouse cells impaired hematopoietic differentiation and increased stem/progenitor cell marker expression, suggesting a shared pro-leukemogenic effect.²¹ 

Mutations of the polycomb group genes ASXL1 and EZH2 are most frequently seen in MF, but are also seen occasionally in other MPN subtypes.^22,23  The precise function of ASXL1 is poorly understood, but it is involved in the control of chromatin structure and is believed to be a component of a complex that deubiquitinates histone H2A lysine 119. EZH2 is a relatively well-characterized gene that encodes the catalytic subunit of the polycomb repressive complex 2 (PRC2), the highly conserved histone H3 lysine 27 methyltransferase that influences stem-cell renewal by epigenetic repression of genes involved in cell-fate decisions. EZH2 mutations in MPN are inactivating, but their impact on myelopoiesis remains to be established. Other mutated genes include the DNA methyltransferase DNMT3a in chronic-phase MPNs and deletion of IKZF1 in blastic-phase MPNs.^24–26  In general, signaling mutations tend to be mutually exclusive and, based on work in AML, it seems likely that IDH1/2 and TET2 mutations will also show mutually exclusivity in MPNs. No other clear associations have yet emerged; however, in general, mutational complexity appears to increase upon progression of PV/ET to MF and blastic phases.

Aberrant methylation of several SOCS family members and other genes has been described in MPNs, but there is no clear consensus with regard to methylation frequency in different disease subtypes, so the significance of these findings remains unclear.^27–29  Other epigenetic abnormalities relate to JAK2 itself: as described above, V617F acquires the ability to phosphorylate and inactivate PRMT5. PRMT5-mediated arginine methylation of histone H4 has been shown to trigger the recruitment of DNMT3A, thus providing a direct link between histone and DNA methylation in gene silencing. These findings suggest the possibility of specific epigenetic consequences of constitutive V617F signaling that are independent of the classical JAK2/STAT cascade.¹⁵  Interestingly, JAK2 plays an apparently independent role in the nucleus by phosphorylating histone H3 tyrosine 41 and thereby excluding heterochromatin protein 1α (HP1α) from chromatin. HP1α has potential tumor-suppressive functions, and it has been suggested that its unregulated displacement from chromatin as a consequence of constitutive JAK2 signaling in MPN may contribute to genetic instability.³⁰ 

Full karyotype analysis is not generally performed routinely for PV and ET, but reports in the literature suggest that chromosome abnormalities are seen in 20%-30% and 10% of cases, respectively. However, it is likely that these estimates have been made on highly selected cases and the true incidence of chromosome abnormalities in these disorders is much lower.³¹  In MF, abnormal karyotypes have been reported in 32%-48% of cases, and a recent analysis has demonstrated that patients with a monosomal karyotype have a particularly poor prognosis.³²  Unfavorable karyotype has been included in the refined Dynamic International Prognostic Scoring System (DIPPS plus) for PMF.³³ 

Gain of chromosome 9 is common in PV and is associated with copy number gain of JAK2 V617F. Interstitial deletions of 13q and 20q are seen in all MPN subtypes, suggesting the presence of one or more tumor-suppressor genes in these regions. Of these, 20q deletion is much more frequent and may, at least in part, target the polycomb family member L3MBTL1. Depletion of L3MBTL1 results in DNA replicative stress, DNA damage, and genomic instability, suggesting that it may have the properties of a tumor suppressor.³⁴  Other recurrent abnormalities in MPN for which the molecular basis is unknown include +8 and partial trisomy for 1q, the latter sometimes seen as a consequence of a der (6) t(1;6)(q23-25;p21-22) in patients with MF.³¹ 

Although not visible by conventional cytogenetic analysis, many mutations associated with MPN progress to homozygosity due to acquired uniparental disomy (UPD), also known as copy number neutral loss of heterozygosity (CNN-LOH). Acquired UPD arises as a consequence of mitotic recombination and may be detected using single nucleotide polymorphism arrays; abnormalities commonly associated with this phenomenon are JAK2 V617F and mutations in MPL, TET2, CBL, and EZH2.^10,16,23,35 

Unexpectedly, several cases have been described that are positive for BCR-ABL and JAK2 V617F, most often by the finding of residual ET after treatment for chronic myeloid leukemia (CML). Other cases have been reported with JAK2 V617F in combination with MPL W515, KIT D816V, or JAK2 exon 12 mutations.^36,37  Furthermore, acquisition of at least 2 independent JAK2 mutations have been described in ∼ 1%-3% of JAK2 V617F–positive cases.³⁸  The probability of acquiring 2 independent rare mutations is remote, which suggests an underlying genetic instability in many cases. Genetic instability is also suggested by the unexpected finding that progression of V617F-positive MPN often results in V617F-negative AML, suggesting transformation of a stem cell that had not acquired the JAK2 mutation.³⁹  In principle, such genetic instability might be inherited or acquired somatically.

Analysis of single-cell–derived hematopoietic colonies has demonstrated that signaling mutations and cytogenetically visible changes are usually in the same clone in cases that have both abnormalities. In contrast, most cases with more than one acquired signaling abnormality are biclonal, meaning that the 2 mutations are not present in the same clone. This suggests that any genetic instability must have preceded the mutations in these cases. Moreover, in 2 informative female patients, the clones were genetically unrelated by X-chromosome inactivation studies, which is consistent with the notion that the putative instability may have been constitutional.³⁷  However, there is also evidence that instability may be acquired: in particular, JAK2 V617F has itself been shown to increase homologous recombination, ploidy abnormalities, and mutation rates at reporter loci.⁴⁰  In addition, as mentioned above, L3MBTL1 associated with del(20q) has been implicated in the maintenance of genome integrity. Overall, it appears that genetic instability may arise by several different routes.

Given the critical importance of JAK2 in myelopoiesis and the fact that V617F as a sole agent can induce an MPN-like disease in mouse models, one might have expected that JAK2 mutations initiate events in the human disease. Remarkably, this appears not to be the case. Initial challenges to the idea that JAK2 V617F initiated disease development came from clonality studies, the demonstration that V617F is never inherited in MPN families but is independently acquired by many affected individuals, and the finding that V617F-positive MPNs may transform to V617F-negative AML. Most convincingly, analysis of single cell–derived colonies reveals highly diverse clonal hierarchies between patients with no strict temporal order in the acquisition of mutations. For example, TET2 mutations often but do not always precede JAK2 V617F and, similarly, del(20q) may appear before or after V617F.^16,41–43  Currently, it is not known whether the initiating mutation or mutations in MPNs remain to be discovered or if these diseases really can develop in several different ways.

The reason that some V617F-positive patients develop PV whereas others develop ET or other myeloid disorders has been a central puzzle in the field since the initial finding of JAK2 mutations. One clear association appears to be the strength of the JAK2 signal: homozygosity for V617F is common in PV and MF but uncommon in ET. This association has been noted in many studies and is even stronger after analysis of single cell–derived colonies.⁴⁴  In mouse models, there are some data suggesting that higher levels of JAK2 signaling favor erythrocytosis, whereas lower levels favor thrombocytosis.⁴⁵  It is apparent, however, that JAK2 signal strength cannot be the whole story, because many PV patients have relatively low V617F allele burdens. In principle, disease differences may result from acquisition of JAK2 V617F at different stages of hematopoietic development, differences in genetic background, or somatically acquired mutations that have not yet been identified. Recently, the balance between STAT1 and STAT5 activation has been shown to distinguish PV and ET. Increased STAT1 activity in normal CD34⁺ progenitors produced an ET-like phenotype, whereas down-regulation of STAT1 activity in JAK2 V617F heterozygous ET progenitors produced a PV-like phenotype.⁴⁶  Whether differences in STAT activity between ET and PV cases are acquired or inherited remains to be established.

Molecular tests are used to help establish clonality and thereby assist in the diagnosis of the disease. Testing for JAK2 V617F is generally performed early during diagnostic workup of suspected MPN; a positive result indicates the presence of a myeloid neoplasm but does not indicate which subtype. Conversely, the failure to detect V617F does not exclude the possibility of an MPN. Using quantitative mutation assays, it is apparent that levels of V617F are highly variable between patients and in some cases are very low. This begs the question of what is the minimum level of V617F that should be used to diagnose an MPN. Setting a cut-off too high may lead to some MPN cases not being correctly diagnosed whereas setting the cutoff too low runs the risk of picking up false positives. The latter may arise from technical limitations of some assays or, more interestingly, the finding of low-level JAK2 V617F in ∼ 1% of otherwise hematologically normal individuals.⁴⁷  In practice, it is not possible to define an absolute cutoff for the diagnosis of an MPN, but it is generally thought that assays should be capable of detecting mutant allele burdens of 1%-2%. However, because of the possibility of finding JAK2 mutations in healthy individuals, any individual that tests positive for V617F should be considered carefully in the light of other diagnostic investigations.

In the absence of V617F, testing for MPL exon 10 mutations may be considered for cases suspected of having ET or MF, and testing for JAK2 exon 12 mutations in cases where there is a strong suspicion of PV (eg, low erythropoietin levels, raised red cell mass, or in cases with erythropoietin-independent erythroid colony formation). Testing for JAK2 exon 12 mutations is complicated by the fact that the clone size may be low in some cases, making these heterogeneous abnormalities difficult to detect. Testing for mutations in other genes is generally not performed on a routine basis, but will probably be incorporated into future diagnostic algorithms as testing prices decrease with the advent of new technologies.

Testing candidate ET cases for BCR-ABL by cytogenetics, FISH, or RT-PCR is often performed on a routine basis to exclude the possibility of CML. The yield of positive cases, however, is < 1% and therefore the value of this practice is questionable in the absence of other signs of CML such as basophilia or BM morphology. For MF, conventional cytogenetics can provide useful prognostic information, but in PV and ET, full karyotype analysis is of limited value.^31,32 

Families with 2 or more individuals affected by MPNs have been known for many years, but the genetic basis for disease predisposition remains unclear.^48,49  The failure of linkage studies to clearly identify candidate genomic regions probably reflects the intrinsic difficulty in studying the genetics of late-onset diseases, but is also consistent with the suggestion that the underlying abnormalities may be heterogeneous and incompletely penetrant. Certainly, germline mutations do not cause disease on their own, because development of the MPN phenotype depends on the acquisition of somatic mutations, most often JAK2 V617F, which is typically acquired independently by several affected family members. Activating MPL mutations, notably S505N, may occasionally be inherited through the germline, giving rise to familial thrombocytosis. Despite being nonclonal and therefore not a true MPN, affected individuals have been reported to have a high risk of developing splenomegaly and BM fibrosis, which are typical myeloproliferative features.⁵⁰ TET2 mutations are generally acquired, although one case has been described with a constitutional truncating mutation that may have been inherited.⁴²  Mutations in CBL are usually acquired, but may be inherited, in which case they are associated with Noonan syndrome–like developmental abnormalities and predisposition to juvenile myelomonocytic leukemia (JMML). Interestingly, just one mutant CBL allele is inherited, but the cells are clonal and homozygous for the CBL mutation as a consequence of acquired UPD. Remarkably, the disease spontaneously resolves in many cases, and although affected individuals are prone to developing vasculitis, it is unclear if there is also predisposition to late-onset hematological malignancy.⁵¹ 

Large-scale epidemiological studies have shown a 5- to 7-fold increased risk of MPNs among first-degree relatives of MPN patients, suggesting the presence of common susceptibility genes in the population at large.⁵²  Remarkably, a major component of this population-level predisposition turns out to be JAK2 itself, specifically a haplotype referred to as 46/1 or GGCC, which includes the JAK2 gene.^38,53–55  How 46/1 predisposes to the acquisition of JAK2 and MPL mutations is not known, but must presumably be a consequence of increased mutation rates (hypermutability hypothesis) or a functional difference that confers a selective advantage when JAK2 is mutated (the “fertile ground” hypothesis). Although mechanistically interesting, the penetrance of 46/1 is extremely low and consequently this haplotype does not contribute to familial MPN⁵⁶  and cannot be used to predict whether any individual will develop an MPN. Whether 46/1 status has any impact on hematological features or prognosis is currently a matter of debate.^57–60 

The unexpected genetic and epigenetic complexity of MPNs (Figure 1) may contribute, along with other factors, to the relatively modest clinical efficacy of JAK2 inhibitors compared with targeted therapy in CML. It is generally expected that additional mutations will be identified in MPNs, and perhaps novel disease-initiating mutations will be found. Acquired mutations and familial germline predisposition variants will most likely be identified by exome or whole-genome sequencing in the near future, and their presence, along with known mutations, will need to be related to comprehensive epigenetic profiles in appropriate cell types. Further common predisposition variants may be identified by genome-wide association analysis. Working out how these abnormalities fit together within the clonal hierarchy, how they deregulate progenitor cell kinetics, and how they may be effectively targeted will be major challenges in the next few years.

The author apologizes to all researchers whose work could not be acknowledged due to space restrictions.

Conflict-of-interest disclosure: The author declares no competing financial interests. Off-label drug use: None disclosed.

Professor N. C. P. Cross, Wessex Regional Genetics Laboratory, Salisbury NHS Foundation Trust, Salisbury SP2 8BJ, United Kingdom; Phone: (44) 1722-429080; Fax: (44) 1722-331531; e-mail: ncpc@soton.ac.uk.