The role of Janus kinases in haemopoiesis and haematological malignancy


Dr Asim Khwaja, Department of Haematology, Royal Free and University College Medical School, 98 Chenies Mews, London WC1E 6HX, UK. E-mail:


The production of blood cells is regulated by a number of protein growth factors and cytokines that influence cell survival, proliferation and differentiation. Many of these molecules bind to cell surface receptors, which belong to a family of closely related cytokine receptors that lack intrinsic catalytic activity but are intimately associated with tyrosine kinases of the Janus kinase (JAK) family. Ligand binding induces the activation of JAKs, which sit at the apex of a signalling cascade in which a key role is played by members of the signal transducers and activators of transcription (STAT) group. Congenital deficiencies in JAK–STAT signalling are associated with immunodeficiency states and acquired activating mutations and translocations are involved in the pathophysiology of haematological malignancy. The latter findings have raised hopes that drugs that target aberrant JAK–STAT signalling may be useful for the treatment of human disease.

JAKs and cytokine receptor signalling

There are four mammalian Janus kinases (JAKs): JAK1, JAK2, JAK3 and TYK2 (reviewed in Ihle et al, 1998; Ward et al, 2000; Yamaoka et al, 2004; Valentino & Pierre, 2006). Completion of sequencing of the human genome indicates that this is the full complement of mammalian JAKs. Fish also have four JAKs whereas flies have only one (Hopscotch) and worms appear to have none (Hou et al, 2002; Yamaoka et al, 2004). The first partial sequence for a JAK family member (TYK2) was obtained by screening a T-cell cDNA library by low stringency hybridisation with a c-fms restriction fragment containing the tyrosine kinase catalytic domain and the full-length sequence was later obtained by the same group (Firmbach-Kraft et al, 1990; Krolewski et al, 1990). JAK1 and JAK2 cDNAs were identified in factor dependent cell-Paterson (FDCP1) murine myeloid cells by reverse transcription polymerase chain reaction (PCR) techniques using degenerate primers to highly conserved sequences in known tyrosine kinase catalytic domains (Wilks et al, 1991). Finally, JAK3 was also cloned by a similar PCR technique (Johnston et al, 1994; Kawamura et al, 1994).

Among tyrosine kinases, JAKs have a unique domain structure with a C-terminal kinase domain (JAK homology 1, JH1) adjacent to a catalytically inactive pseudokinase domain, JH2 (Fig 1). These two adjacent ‘kinase’ domains, one active and the other inactive, give this family their name after Janus, the Roman God of gates and doors (or beginnings and endings) who is represented by a double-faced head looking in opposite directions. There is an N-terminal region (JH3-4) which resembles an SH2 domain but has ill-defined function (Radtke et al, 2005). The domain structure is completed by an amino-terminal FERM (band four point one, ezrin, radixin, moiesin) domain (JH5–7), which mediates JAK binding to cytokine receptors.

Figure 1.

 Overall scheme of Janus kinase (JAK), signal transducers and activators of transcription and suppressor of cytokine signalling family members showing important domains. The JAK2 mutation (V617F) found in myeloproliferative disorders and the key regulatory phosphorylation site (Y1007) are shown.

The link between JAKs and cytokine signalling was first made when α-interferon sensitivity to a non-responsive cell line was conferred by add back of DNA that encoded for TYK2 (Velazquez et al, 1992). Similar experiments revealed a role for JAK1 and JAK2 in γ-interferon signalling (Muller et al, 1993; Watling et al, 1993; Ihle et al, 1995). Cytokines of the haemopoietic system include interleukins (IL), colony-stimulating factors, interferons, erythropoietin (EPO) and thrombopoietin (TPO). The majority of these bind to a family of homologous transmembrane polypeptide receptors that share a number of features – these receptors can be either single chain [e.g. granulocyte colony-stimulating factor receptor (G-CSFR), EPO receptor (EPOR) and TPO receptor (TPOR)] or heterodimeric with a common signalling subunit and a unique ligand-binding chain (reviewed in Moutoussamy et al, 1998; Leonard & Lin, 2000; Rane & Reddy, 2002). The latter can be grouped into those receptors that share the common β-chain [granulocyte–macrophage colony-stimulating factor (GM-CSF), IL-3, IL-5], those sharing GP130 [IL-6, leukaemia inhibitory factor (LIF), oncostatin M, IL-11] and those that share the common γ-chain (IL-2, IL-4, IL-7, IL-9, IL-13 and IL-15). The single chain and heterodimeric group together constitute type I cytokine receptors, which share basic structural features and are characterised by the presence of four conserved cysteine residues, a WSXWS motif and fibronectin type III domains in the extracellular part of the receptor and by conserved Box1/Box2 regions in the membrane proximal intracytoplasmic domain. The type II cytokine receptors, which include those for the interferons and IL-10, lack the WSXWS motif but retain Box1/2.

Signalling via cytokine receptors is initiated by ligand binding. This is thought to either induce homodimerisation (e.g. G-CSFR) (Horan et al, 1996; Tamada et al, 2006) or heterodimerisation/oligomerisation of receptor subunits (e.g. GM-CSFR) (Hayashida et al, 1990) or, alternatively, to induce a conformational change in preformed receptor dimers (e.g. EPOR) (Livnah et al, 1999; Constantinescu et al, 2001). JAKs are constitutively associated with cytokine receptors with binding mediated by interactions between the FERM domain of the JAK and the Box1 membrane proximal intracytoplasmic region of the receptor (Pellegrini & Dusanter-Fourt, 1997; Richter et al, 1998; Haan et al, 2001, 2002). There is increasing evidence that JAK/receptor interaction is also involved in the regulation of cell surface expression, recycling and degradation of the associated cytokine receptor (Gauzzi et al, 1997; Ragimbeau et al, 2003). It has been shown that the FERM domain of JAK2 binds to the EPOR in the endoplasmic reticulum and promotes its cell surface expression (Huang et al, 2001). JAK2 or TYK2 can promote surface expression of the TPOR partly by protecting it from proteasomal degradation (Royer et al, 2005). Although JAKs are described as cytosolic tyrosine kinases, the close association with cytokine receptors results in an almost exclusive membrane distribution and it has been suggested that this forms the functional equivalent of a receptor tyrosine kinase, such as FLT3 or KIT, which has intrinsic enzymatic activity in its cytoplasmic domain (Behrmann et al, 2004).

Receptor oligomerisation or activation induces the close proximity of receptor-associated JAK molecules resulting in their transphosphorylation and activation. A number of tyrosine residues have been shown to be phosphorylated in JAK2 that have varying effects on its function. Tyrosines 1007 and 1008 in the activation loop of the catalytic domain are transphosphorylated and Y1007 phosphorylation is obligatory for kinase activity (Feng et al, 1997). Y1007 has also been shown to be the binding site for suppressor of cytokine signalling (SOCS)-1/JAB, a downregulatory mechanism (Yasukawa et al, 1999). Phosphorylation of Y813 enhances binding of SH2-Bβ which enhances the activation of JAK2 as well as JAK2-mediated phosphorylation of signal transducers and activators of transcription 5 (STAT5) (Kurzer et al, 2004; Nishi et al, 2005). Phosphorylation of Y221 (just C terminal to FERM domain) has uncertain function (Argetsinger et al, 2004; Feener et al, 2004) and of Y570 in the JH2 domain is probably inhibitory to JAK2 activity (Argetsinger et al, 2004; Feener et al, 2004). Other phosphorylations have been demonstrated but are of unknown function (Matsuda et al, 2004).

The JH2 domain is thought to play a significant role in the regulation of JAK catalytic activity. Deletion of the JH2 domain, or of subregions within it, have been shown to result in an increased basal activity of JAK2 and JAK3 (Saharinen et al, 2000); however, such deletions can also abolish the response to cytokine stimulation of JAKs indicating that, although in the absence of ligand JAK2 is autoinhibited by an intramolecular interaction between JH2 and JH1 domains, induction of maximal activation requires an intact JH2 domain (Saharinen & Silvennoinen, 2002; Saharinen et al, 2003). The crystal structures of the isolated catalytic domains of both JAK2 and JAK3 have been recently reported, in both instances in complex with a kinase inhibitor (Boggon et al, 2005; Lucet et al, 2006). However, these studies have not shed any light on the role of JH1 and JH2 interactions. Prediction modelling of the structure of the JH1/2 domains of JAK2 has suggested that there are two main sites of interaction between JH1 and JH2 domains (Lindauer et al, 2001): first, between two α-helices of the two domains and secondly, between the activation loop of JH1 (D994–E1024) and a loop between two β-strands of the N-terminal lobe of JH2 (V617–E621). This interaction would modify the catalytic activity of JH1 and, as is described below, is the site of a mutation commonly found in myeloproliferative diseases (V617F, see sections below).

Gene targeting studies of JAK family members show distinct phenotypes, which are detailed in Table I. JAK1−/− mice are runted at birth and die during the perinatal period – these mice are small and have defective lymphoid and neural development (Rodig et al, 1998). Cells from these animals show that JAK1 is required for signalling via class II cytokine receptors (all interferons) and is also involved in signalling via the common γ-chain receptor family (including IL-2 and IL-7) and via receptor complexes that utilise GP130 (such as IL-6 and LIF). The JAK2 knockout is lethal approximately mid-way through gestation (around day 12) and is characterised by a failure of definitive erythropoiesis (Parganas et al, 1998) – the phenotype is similar to that seen in mice lacking either EPO or its receptor. Fetal liver progenitors from JAK2−/− mice show defective responses to cytokines that utilise the common β-chain cytokine receptor (e.g. IL-3 and GM-CSF) as well as to TPO and γ-interferon. Fibroblast responses to interferon-α and interferon-β were normal. The JAK3 knockout reveals the intimate and virtually exclusive role of JAK3 with cytokines that signal via the common γ-chain (IL-2, IL-4, IL-7, IL-9, IL-15 and IL-21) (Nosaka et al, 1995; Park et al, 1995; Thomis et al, 1995). These animals are immunocompromised due to a severe reduction in T and B lymphocytes and natural killer (NK) cells – those lymphoid cells that remain are functionally defective. Mutations in JAK3 are associated with human immunodeficiency syndromes (see below). Mice that lack TYK2 are phenotypically normal and have a modest susceptibility to viral infection and a reduced response to IL-12 (Karaghiosoff et al, 2000; Shimoda et al, 2000). Although TYK2 is implicated in α-interferon signalling, TYK2−/− cells show an impaired response to α-interferon that is only detectable at low cytokine concentrations. TYK2 null mice are resistant to the lethal effects of lipopolysaccharide (LPS) (Karaghiosoff et al, 2003). It has been suggested that TYK2 signalling may play a role in tumour immune surveillance, probably mediated by a reduced NK and NKT anti-lymphoma response (Stoiber et al, 2004).

Table I.   Janus kinases (JAKs) and their associated cytokines and the phenotype of murine knockout models.
JAKAssociated cytokinesKnockout phenotype
  1. LIF, leukaemia inhibitory factor; IL, interleukin; OSM, oncostatin M; GM-CSF, granulocyte–macrophage colony-stimulating factor; EPO, erythropoietin; TPO, thrombopoietin; LPS, lipopolysaccharide.

JAK1Interferons (α, β and γ) γ-chain family including IL-2, 4, 7, 9, 13, 15
GP130 family including IL-6, LIF, OSM, IL-11
Defective lymphoid and neural development
Defective responses to interferons and to γ-chain and Gp130 cytokines
JAK2β-Chain family including GM-CSF, IL-3 and IL-5
EPO, TPO and γ-interferon
Failure of definitive erythropoiesis
Defective responses to β chain cytokines and to TPO and γ-interferon
JAK3γ-Chain family including IL-2, 4, 7, 9, 13, 15, 21Severe combined immunodeficiency
TYK2Interferons (α and β)Suboptimal α-interferon response
LPS resistance

The role of STATs in cytokine signalling

Investigation of the mechanisms for defective interferon responses also led to the identification of the first members of the STAT family and the discovery of a link between JAKs and STATs (Schindler et al, 1992; reviewed in Ihle, 2001; Levy & Darnell, 2002; Paukku & Silvennoinen, 2004; Yu & Jove, 2004; Valentino & Pierre, 2006). In mammalian cells there are seven STATs: STAT1, 2, 3, 4, 5a, 5b, 6 and 7 (Fig 1). STAT homologues are found in a number of species including amoebae, nematode, fruit fly and zebrafish.

In a ‘quiescent’ cell, STATs exist in an unphosphorylated monomeric form in the cytosol. Cytokine stimulation leading to JAK activation results in the tyrosine phosphorylation of the cytoplasmic domain of the receptor generating docking site(s) for STAT SH2 domains. For example, the STAT3 SH2 domain commonly binds to a phosphorylated tyrosine motif consisting of pYXXQ. Once recruited to the receptor-JAK complex, STATs are tyrosine phosphorylated at the C terminus by JAKs, which results in the formation of STAT homodimers or heterodimers via SH2 domain–phosphotyrosine interactions.

Several STATs are also regulated by serine phosphorylation of a residue in the transcription activation domain at a conserved PMSP motif located about 20–30 amino acids C-terminal of the conserved tyrosine phosphorylation site (Bromberg & Darnell, 2000; Decker & Kovarik, 2000). STAT1 and STAT3 are phosphorylated at serine 727 and STAT4 at S721. This phosphorylation can be stimulated by cytokines and a number of serine/threonine kinases have been implicated (including ERK, p38, JNK and PKC-delta) (Jain et al, 1998, 1999; Abe et al, 2001; Gartsbein et al, 2006); overall, the data suggest that phosphorylation has a positive effect on transcription (Wen et al, 1995; Varinou et al, 2003). There are reports that STATs may also be regulated by acetylation and that reversible acetylation of STAT3 can regulate its dimerisation and transcriptional effects (Wang et al, 2005; Yuan et al, 2005).

The physiological roles of STAT proteins has been investigated in a wide variety of settings and experimental models and due to the complexity of interactions between various JAKs and STATs the most clear-cut results have been from mouse knockout studies (Table II). STAT1 null mice are developmentally normal but show a marked sensitivity to viral and other microbial pathogens due to a profound defect in interferon signalling (Meraz et al, 1996). STAT1 is also involved in tumour suppression/surveillance mechanisms as deficient mice are more susceptible to chemical-induced carcinogenesis and to the development of tumours in a p53 null background (Yu & Jove, 2004). STAT2 is also involved in interferon signalling and antiviral responses and knockout mice are susceptible to viral infections (Park et al, 2000). STAT3 knockout mice die early in embryogenesis and further evaluation of its functions has required tissue-specific deletion (Takeda et al, 1997). A number of such studies have been carried out and show that STAT3 plays a critical role in a variety of tissue functions including skin remodelling, mammary gland involution, post-injury motor neuron function and hepatocyte mitogenesis (Ihle, 2001; Paukku & Silvennoinen, 2004). Perhaps surprisingly, as G-CSF is a potent activator of STAT3, conditional deletion of STAT3 in haemopoietic cells does not impair G-CSF signalling. Indeed, these animals have increased neutrophils for reasons that are unclear but may involve a reduction in the expression of the cytokine signalling inhibitor SOCS3 (Lee et al, 2002). STAT4 and STAT6 are involved in selective T-cell differentiation responses: STAT4−/− mice have an impaired response to IL-12, which is critical for controlling T-helper (Th) differentiation down the Th1 pathway (Kaplan et al, 1996). These mice are resistant to autoimmune diseases characterised by a Th1 response and also show susceptibility to infection with intracellular organisms (Paukku & Silvennoinen, 2004). STAT6−/− animals have a loss of IL-4 responsiveness leading to impaired Th2 differentiation (Takeda et al, 1996). In addition, STAT6-deficient B cells are unable to undergo class switching and produce immunoglobulin E (Shimoda et al, 1996). STAT5 proteins play essential and non-redundant roles in growth hormone (GH) and prolactin signalling (Liu et al, 1997; Udy et al, 1997). Mice that lack both STAT5a and STAT5b surprisingly do not show major haemopoietic defects in signalling via EPO and TPO, two cytokines that are potent activators of STAT5 and play critical roles in red cell and platelet production respectively (Teglund et al, 1998). STAT5a/b deletion has an unexpected effect on peripheral T-cell proliferation and activation via the T-cell receptor and IL-2 (Moriggl et al, 1999) and haemopoietic stem cells (HSC) from STAT5a−/b− mice have a repopulating defect in transplantation experiments, indicating a role for STAT5 in stem cell expansion (Bunting et al, 2002). The preceding data have utilised a knockout approach that may result in the presence of a truncated form of STAT5, STAT5deltaN. More recent data utilising STAT5a−/b− mice with a complete deletion of the STAT5 locus show that these animals die in the perinatal period and have a profound severe combined immunodeficiency (SCID) phenotype similar to that seen in cytokine receptor γ-chain knockout mice (Yao et al, 2006).

Table II.   The STAT family and and the phenotype of murine knockout models.
STATKnockout phenotype
  1. STAT, signal transducers and activators of transcription; GH, growth hormone; IL, interleukin; Th, T-helper.

STAT1Viable. Susceptibility to microbial pathogens and tumours.
Defective responses to interferons (α, β and γ)
Defective responses to interferons (α and β)
STAT3Embryonic lethal. Conditional knockouts have various defects in adult tissues in growth, survival and differentiation
STAT4Viable. Impaired IL-12 response and Th1 differentiation
STAT5aViable. Impaired mammary gland development (prolactin defect)
STAT5bViable. Impaired growth (GH defect)
STAT5a/bViable. Defects of single knockouts plus defective T-cell responses and impaired response to stem cell and erythropoietic stress
STAT6Viable. Impaired IL-4 response and Th2 differentiation

Although STAT proteins are often regarded as the key mediators of JAK signalling it should be remembered that other signalling pathways are also activated by many cytokines via the receptor-JAK complex. Normal or aberrant JAK activation can lead to signalling via the Ras/RAF/MEK/mitogen-activated protein kinase (MAPK) and PI3K/Akt/mTOR modules (Rane & Reddy, 2002). In addition, STATs may be activated by tyrosine kinases independent of JAKs (Yu & Jove, 2004). Kinases that have been implicated in STAT activation include cytosolic tyrosine kinases such as Src family members, Bmx, focal adhesion kinase, Fes and receptor tyrosine kinases such as the receptors for platelet-derived growth factor (PDGF), stem cell factor (SCF), hepatocyte growth factor and vascular endothelial growth factor (Paukku & Silvennoinen, 2004). It is not always clear how STAT phosphorylation is achieved by these receptor tyrosine kinases – it has been suggested in some instances that this is via JAKs or Src family kinases or by direct phosphorylation. STATs may also be phosphorylated by aberrant tyrosine kinases found in haematological malignancies, such as FLT3 and BCR-ABL (Carlesso et al, 1996; Mizuki et al, 2000).

These findings show that STATs are not the sole mediators of JAK signalling and that the detection of activated STAT pathways does not necessarily indicate JAK activity.

Negative regulation of JAK–STAT signalling

The magnitude and the duration of cytokine receptor signals are tightly controlled by several regulatory mechanisms to prevent excessive signalling and abnormal cellular activation that could lead to autoimmune disorders or malignant transformation. Cytokine signalling is regulated at a number of levels – by receptor internalisation; by the action of phosphotyrosine phosphatases; by the effects of members of the protein inhibitors of activated STATs family (PIAS) and the STAT-interacting LIM protein (SLIM) ubiquitin ligase; and by the induction of expression of members of the SOCS family.

Tyrosine phosphatases

The SHP1 tyrosine phosphatase is characterised by the presence of two SH2 domains that enable interaction with phosphotyrosines present on activated receptors and on JAK2. SHP1 is predominantly expressed in the haemopoietic system and mice deficient for SHP1 have expansion of myeloid cells with extramedullary haemopoiesis and splenomegaly as well as immune dysregulation resulting in autoimmune disease (reviewed in Zhang et al, 2000) SHP1-deficient cells show hyper-responsiveness to cytokines, such as GM-CSF and IL-3, and hyperphosphorylation of JAK1/2 after GH or EPO. Hypermethylation of the SHP1 promoter has been described in malignancy (see below).

CD45 or the leucocyte common antigen, is a transmembrane protein with phosphatase activity in its cytoplasmic domain. Mice that lack CD45 show hyperactivation of JAKs and downstream STATs and enhanced biological responses to a variety of cytokines acting on haemopoietic cells including IL-3, IL-4, EPO and interferons (Irie-Sasaki et al, 2001, 2003).

PTP1B (also known as PTPN1) and TC-PTP (PTPN2) are phosphatases with a high degree of sequence homology (Bourdeau et al, 2005). PTP1B is widely expressed and TC-PTP has higher expression levels in haemopoietic cells but is also found in many other tissues. PTP1B was identified by several groups as having a significant effect on leptin signalling – this hormone signals via JAK2. PTP1B−/− mice have hypersensitivity to leptin and this is due to prolonged JAK2 activation, which results from the loss of phosphatase activity (Zabolotny et al, 2002). PTP1B directly binds and dephosphorylates JAK2 and has also been shown to act on TYK2. PTP1B recognises the (E/D)-pY-pY-(R/K) motif found in JAK2 and TYK2 but does not act on JAK1 and JAK3 which lack this consensus sequence (Myers et al, 2001). In contrast, TC-PTP has been shown to selectively dephosphorylate JAK1 and JAK3 (Simoncic et al, 2002). An alternatively spliced nuclear isoform of TC-PTP has phosphatase activity against STAT1 (ten Hoeve et al, 2002), is also a STAT3 phosphatase and acts to terminate STAT transcriptional activity (Bourdeau et al, 2005). Loss of PTP1B expression or reduced activity has been described in Bcr-Abl-positive chronic myeloid leukaemia (CML)-derived cell lines and it has been suggested that this may contribute to relative resistance to imatinib (Koyama et al, 2006). However, this is likely to be due to direct effects of PTP1B on Abl kinase activity rather than a JAK effect (LaMontagne et al, 1998).


The mammalian PIAS protein family was originally identified as negative regulators of STAT signalling and has since been shown to regulate transcription factors of a much broader range including nuclear factor kappa B, SMADs and p53. There are four members: PIAS1, PIAS3, PIASx (or PIAS2) and PIASy (or PIAS4). Recent studies have shown PIAS members to have small ubiquitin-like modifier (SUMO) E3-ligase activity. The covalent conjugation of SUMO to substrate proteins is analogous to, but distinct from, ubiquitination and this reversible process can alter protein localisation and stability and protein–protein interactions, thereby modulating transcription factor activity. PIAS proteins alter STAT activity in a number of ways – first, by direct inhibition of DNA binding, e.g. PIAS1/STAT1 and PIAS3/STAT3. Secondly, by recruiting other co-regulators that repress transcription, e.g. histone deacetylases, e.g. PIASx recruits HDAC3 to repress STAT4-dependent transcription. Thirdly, PIAS proteins can promote the sumoylation of STATs, e.g. multiple PIASs can cause STAT sumoylation and this is likely to inhibit STAT signalling (Ungureanu et al, 2003, 2005).

The SLIM protein contains a LIM domain and a PDZ domain and interacts in the nucleus with tyrosine phosphorylated STATs and specifically inhibits gene transcription mediated by STATs 1 and 4 (Tanaka et al, 2005). This is achieved by promoting the ubiquitination and degradation of these STATs. Paramyxoviruses (inc mumps and measles viruses) also use the ubiquitin modification of STATs to evade the effects of interferons – at least 3 ‘V’ proteins function as E3 ubiquitin ligases with specificity for STATs 1–3 (Ungureanu & Silvennoinen, 2005).


The major inducible pathway that regulates cytokine and JAK–STAT signalling is that of the SOCS family (reviewed in Elliott & Johnston, 2004; Paukku & Silvennoinen, 2004; Wormald & Hilton, 2004; Yoshimura, 2005). There are eight members – CIS and SOCS1–7 which all have a variable N-terminal region, a central SH2 domain and a C-terminal SOCS box (Fig 1). CIS was the first member of this family to be identified and was shown to be an immediate early response gene to signalling via IL-2, IL-3 and EPO induced in a negative feedback loop via the transcriptional activity of STAT5. SOCS expression can also be induced by non-cytokine agonists such as LPS (Stoiber et al, 1999) and chemokines (Stevenson et al, 2004) and this may contribute to subsequent non-responsiveness to cytokines.

Suppressor of cytokine signalling proteins inhibit JAK–STAT signalling in a number of ways: first, by competing for binding of STAT or JAKs to the receptor complex. CIS and SOCS2 bind to phosphorylated tyrosine residues on activated cytokine receptors and are thought to compete for or sterically hinder the binding of STAT5 proteins to the receptor complex (Yoshimura et al, 2005). CIS binds to Y401 of the EPOR which is the major STAT5-binding site (Yoshimura et al, 1995). SOCS5 can bind to the IL-4 receptor and inhibit JAK1 binding (Seki et al, 2002).

Secondly, by direct binding to JAK molecules and inhibiting their activity. Both SOCS1 and SOCS3 possess a kinase inhibitory region which is believed to inhibit the effects of JAKs by behaving as a pseudosubstrate thereby preventing access of genuine substrates to the catalytic pocket (Elliott & Johnston, 2004). SOCS1 binds to the phosphorylated activation loop residue Y1007 of JAK2 via its SH2 domain whereas the SOCS3 SH2 domain has been shown to bind to Y757 of GP130, Y401 of the EPOR and Y729 of the GCSFR (Yamamoto et al, 1994).

Thirdly, by promoting the ubiquitination and subsequent degradation of JAK proteins. Degradation of a protein by the ubiquitin–proteasome pathway entails two successive events: the covalent attachment of a chain of ubiquitin moieties to the substrate protein and the ATP-dependent proteolysis of the substrate by the 26S proteasome. Ubiquitin transfer requires the activity of E1 (ubiquitin activating), E2 (ubiquitin conjugating) and E3 (ubiquitin ligase) enzymes (Willems et al, 2004). The SOCS box interacts with a variety of proteins involved in the ubiquitin pathway and CIS/SOCS members act as E3 ligases to promote the degradation of JAKs. In addition to being regulated at the transcriptional level, SOCS proteins themselves may be modified by the ubiquitin–proteasome complex – they have a short half-life that is prolonged in the presence of proteasomal inhibitors (Ilangumaran et al, 2004).

Clinical associations of abnormal JAK/STAT signalling

JAK3, STAT1 and immunodeficiency

Mutations of the cytokine receptor common γ-chain account for a high proportion of cases of SCID (reviewed in Cavazzana-Calvo et al, 2005). In the mid-1990s, mutations in JAK3 that were associated with autosomal recessive SCID were discovered: affected individuals have absent T and NK cells with relatively normal numbers of poorly functional B cells (Macchi et al, 1995) and this affects around 10% of SCID cases (Cavazzana-Calvo et al, 2005). Most patients are compound heterozygotes, having inherited a distinct mutation from each parent and some individuals are homozygous for their mutations as a result of parental consanguinity (O'Shea et al, 2004a). The majority of described mutations have a major effect on protein expression and/or stability, or less commonly, result in mutated and non-functional Jak3 protein. Missense mutations and small in-frame deletions may permit protein expression but interfere with kinase activity, binding to the cytokine receptor γ-chain or intracellular trafficking (O'Shea et al, 2004a). As there are no mutational ‘hot spots’, diagnosis usually rests on sequencing the entire gene. Treatment of JAK3-deficient individuals has relied on stem cell transplantation but gene therapy approaches, akin to the treatment of γ-chain-deficient SCID, have been proposed (Cavazzana-Calvo et al, 2005).

Mutations in STAT pathways that cause immune deficiency states have also been described. Two kindred with the same heterozygous mutation in STAT1 (L706S) causing partial dominant STAT-1 deficiency (Dupuis et al, 2001) results in reduced responses to γ-interferon with a clinical picture of susceptibility to mycobacterial infections. Complete, autosomal recessive, STAT1 deficiency has also been reported (Dupuis et al, 2003) – this abolishes cellular responses to interferons and led to premature death from viral infection (Picard & Casanova, 2004).

JAK3 as a drug target

The highly specific immunodeficient phenotype seen in JAK3 deficiency has led to the proposition that, drugs that could selectively inhibit JAK3 may prove to be valuable in the management of organ transplantation and auto-immune disease. These could complement current immunosuppressants by having a novel mode of action. Due to the restricted tissue distribution of JAK3, an effective inhibitor has the potential to have much less toxicity compared with many immunosuppressants, such as ciclosporin (inhibitor of calcineurin), rapamycin (inhibitor of mTOR activation) and glucocorticoids (O'Shea et al, 2004b). The most advanced of the JAK3 inhibitors in drug development is known as CP-690,550 which has a 50% inhibitory concentration against JAK3 of 1 nmol/l (Changelian et al, 2003). Importantly, this compares with 20 nmol/l for JAK2 and 100 nmol/l for JAK1. CP-690,550 is orally bioavailable and has been shown to be effective in a number of allogeneic solid-organ transplant models and has entered clinical trials in renal transplant patients and rheumatoid arthritis (Changelian et al, 2003; O'Shea et al, 2005). The preclinical studies have not shown significant effects on haemoglobin (Hb) levels, platelet or neutrophil counts, suggesting that in vivo anti-JAK2 effects are minimal but data from the human trials is awaited.

JAKs in oncogenesis

Janus kinase signalling is activated in haematological malignancies by a number of mechanisms including the downregulation of negative regulators of JAK–STAT pathways, amplification of the JAK2 locus and the involvement of JAK2 in chromosomal translocations and, most recently, by the identification of an activating point mutation in JAK2. Human T lymphotropic virus type I (HTLV-1) infection, the pathogenic agent for the development of adult T-cell leukaemia/lymphoma, has been shown to result in the activation of JAK and STAT signalling pathways (Migone et al, 1995; Takemoto et al, 1997). Both JAK1 and JAK3 have been shown to be activated and this may be mediated, at least partly, by the HTLV-1 tax protein transactivating the promoters of a number of cytokines and their receptors, resulting in autocrine or paracrine signalling (Grassmann et al, 2005).

Downregulation of negative regulators.  Hypermethylation of normally unmethylated CpG islands in gene promoter elements can lead to downregulation of transcription and hence protein levels, and has been widely described in cancer biology with respect to tumour suppressor genes (Baylin & Ohm, 2006). SHP1 hypermethylation has been described to occur at high frequencies in a number of haemopoietic tumours including myeloma, anaplastic lymphoma, mantle cell and follicular lymphomas, and acute leukaemias (Oka et al, 2002; Chim et al, 2004a,b; Johan et al, 2005a). In myeloma cells, treatment with the demethylating agent 5-azacytidine led to re-expression of SHP1 and a downregulation in levels of phosphorylated STAT3 (Chim et al, 2004a). It has been suggested that hypermethylation of SHP1 may be mediated by STAT3, resulting in a signalling amplification loop (Zhang et al, 2005). Hypermethylation of the SOCS3 promoter has been reported in a variety of epithelial tumours, including lung and hepatocellular carcinomas, but not so far in haematological malignancies (He et al, 2003; Niwa et al, 2005). Hypermethylation of CpG islands in the SOCS1 gene has also been described (Yoshikawa et al, 2001) and lie within what is now thought to be exon 2 of the gene and not in a promoter region (Melzner & Moller, 2003). This correlates with reduced expression of SOCS1 mRNA, which increases after treatment with demethylating agents and has been described in acute leukaemias, myelodysplastic syndrome (MDS) and myeloma amongst others (Galm et al, 2003; Watanabe et al, 2004; Brakensiek et al, 2005).

Recently, mutations in SOCS1 have been described in lymphomas. In primary mediastinal B-cell lymphomas (PMBL), there is frequent gain of genetic material at 9p24 which includes the JAK2 locus (Bentz et al, 2001) and gene expression profiling has shown increased JAK2 mRNA levels (Savage et al, 2003). Melzner et al (2005) found that JAK2 protein was not overexpressed in the MedB1 PMBL line, which has trisomy 9, but that the protein half-life was increased due to delayed degradation. They showed that these cells harbour biallelic SOCS1 mutations that abolish SOCS box activity and also found deletion mutations in nine of 20 primary PMBL tumours which result in truncated or aberrant transcripts (Melzner et al, 2006). This group has also described similar deletions in Hodgkin lymphoma cell lines and in micro-dissected Hodgkin cell samples from eight of 19 primary cases (Weniger et al, 2006). In these cases, there was a high correlation with the presence of phosphorylated STAT5 by immunohistochemistry, suggesting JAK activation in the tumour.

In the course of evaluating the mechanisms for constitutive STAT3 activation in anaplastic lymphomas, Zhang et al (2002) found that levels of PIAS3, which is an inhibitor of STAT3, were absent in the majority of anaplastic large cell lymphoma (ALCL) cell lines. The mechanism for this was not identified. Ueda et al (2003) showed that the mRNA expression of PIASy was significantly reduced in AC133+ samples from patients with advanced MDS [refractory anaemia (RA) with excess blasts, MDS-acute myeloid leukaemia (AML)] compared with samples from patients with RA.

Fusion proteins.  Three distinct fusion proteins involving JAK2 have been described –TEL/Ets translocation variant (ETV)6–JAK2, BCR–JAK2 and pericentriolar material (PCM1)–JAK2. These are relatively rare and involve different lineages and clinical phenotypes.

The TEL–JAK2, as a consequence of t(9;12)(p24;p13), has been identified in single cases of pre-B acute lymphoblastic leukaemia (ALL), atypical CML in transformation and T-ALL (Lacronique et al, 1997; Peeters et al, 1997). The fusion results in a constitutively active kinase due to the oligomerisation motifs in TEL bringing together JAK2 kinase domains. In addition, two of the breakpoints result in loss or disruption of the inhibitory JH2 domain, possibly contributing to activation (Lacronique et al, 1997). Murine bone marrow transplantation experiments with cells infected with TEL–JAK2 retroviruses generated a mixed T-lymphoid/myeloproliferative disorder (MPD) – transformation was dependent on JAK2 kinase activity and on the TEL oligomerisation domain (Schwaller et al, 1998). Although TEL–JAK2 expression results in the activation of a number of downstream signalling pathways including phosphoinositide-3 kinase (PI3K) and MAPK, STAT5 is required for transformation – STAT5ab−/− mice fail to develop lympho-myeloproliferative disease in response to TEL–JAK2 (Schwaller et al, 2000).

BCR–JAK2, associated with t(9;22)(p24;q11.2), has been described in a single case of Philadelphia-negative CML which was resistant to imatinib therapy (Griesinger et al, 2005). Recently, several groups have identified a recurrent translocation t(8;9)(p22;p24) that fuses the great majority of the human autoantigen PCM1 gene to the C terminal two-thirds of JAK2 (Bousquet et al, 2005; Murati et al, 2005; Reiter et al, 2005; Adelaide et al, 2006). This includes the JH2 and JH1 domains and is predicted to result in a constitutively active kinase due to multiple coiled-coil oligomerisation motifs in PCM1. Of 15 patients so far described, 10 had atypical MPDs and the others had AML, pre-B ALL and T lymphoblastic lymphoma.

The JAK2 V617F mutation and MPDs

The MPDs, originally grouped together by Dameshek (1951), are characterised by increased numbers of differentiated blood cells and are believed to arise in a multipotential haemopoietic progenitor. They include polycythaemia vera (PV), ET, primary idiopathic myelofibrosis (IMF) (the classic MPDs), as well as chronic eosinophilic leukaemia/hypereosinophilic syndrome (CEL/HES), systemic mastocytosis and CML. ET and PV share several features: a hypercellular marrow with overproduction and predominance of one lineage; hypersensitivity to cytokines such as EPO; presence of extramedullary haemopoiesis; progression in a significant proportion of cases to myelofibrosis and a relatively low propensity for evolution to acute leukaemia in the absence of the use of leukaemogenic cytoreductive treatment. The role of the BCR-ABL tyrosine kinase in CML pathophysiology has now been established for some years and more recently it has become apparent that aberrant tyrosine kinase activity is also associated with other MPDs (De Keersmaecker & Cools, 2006). A number of fusion partners have been described for the fibroblast growth factor receptor 1 (FGFR1) tyrosine kinase that typically result in a ‘8p11’ myeloproliferative syndrome associated with eosinophilia, lymphoblastic lymphoma and a propensity to transform to AML (Macdonald et al, 2002), although the BCR-FGFR1 fusion is associated with a CML-like phenotype (Demiroglu et al, 2001; Roumiantsev et al, 2004). Systemic mast cell disorders are frequently found to have activating point mutations in the SCF receptor, KIT (D816V) (Lennartsson et al, 2005). An interstitial deletion on chromosome 4q12 results in a fusion of the FIP1L1 gene to the PDGFRA gene and is causally implicated in the pathogenesis of HES/CEL and is also associated with mast cell disorders with associated eosinophilia (Cools et al, 2003; Tefferi & Pardanani, 2004). Translocations involving the PDGFRB locus at 5q33 have been described in a number of patients with atypical MPDs (Reilly, 2003). Diseases involving ABL, PDGFRA and PDGFRB kinases are all responsive to the inhibitor imatinib mesylate (Hannah, 2005).

Until just over a year ago, no abnormalities in TK signalling had been identified in the classic MPDs: PV, ET and IMF. Then, in a relatively short space of time, five publications detailed the identification of a somatic mutation in JAK2 at high frequency in these diseases (Baxter et al, 2005; James et al, 2005; Kralovics et al, 2005; Levine et al, 2005a; Zhao et al, 2005). This G–T mutation results in phenylalanine being substituted for valine at position 617 (V617F) in the pseudokinase/JH2 domain and results in a protein with increased kinase activity and hyper-responsiveness to cytokine signalling. Within the last 12 months, there has been a remarkable effort to characterise the biological and clinical correlates of this mutation with a large number of articles published.

The JAK2 V617F mutation was identified using a variety of approaches. The group of Vainchenker found that a kinase inhibitor of JAK2 or knockdown of JAK2 expression using small interfering RNA technology could inhibit the formation of eythropoietin-independent erythroid colonies that are a hallmark of PV (James et al, 2005). This led to sequencing of JAK2 and the detection of the mutation. Work from Kralovics et al had previously identified loss of heterozygosity (LOH) of a region on chromosome 9p in PV and identified a 6·2-Mbp region common to all 51 patients screened. As this region contained JAK2, with its known role in erythropoiesis, this was screened further for mutations (Kralovics et al, 2005). The three other groups targeted JAK2 as part of a general sequencing screen of tyrosine kinases and phosphatases in MPDs (Baxter et al, 2005; Levine et al, 2005a; Zhao et al, 2005).

A number of groups have now reported on the frequency of JAK2 V617F in MPDs (Tables III–V). The prevalence of the mutation in PV is very high and ranges in various series from 65% to 100% with an average of 82% in nearly 1000 reported cases. The differences in reported rates are likely to be due at least three reasons: first, the stringency of the criteria used to diagnose PV; secondly, the sensitivity of the method used to detect mutations and finally, the source of DNA. The majority of studies have used peripheral blood neutrophils as they are thought to be derived from the same clonal progenitor that is transformed in PV (reviewed in Tefferi & Gilliland, 2005; Tefferi & Spivak, 2005; Vainchenker & Constantinescu, 2005; De Keersmaecker & Cools, 2006; Nelson & Steensma, 2006). Direct sequencing techniques are likely to have a lower sensitivity than techniques that employ PCR amplification of the mutant allele (Campbell et al, 2006a).

Table III.   JAK2 V617F mutations in polycythaemia vera.
Mutation frequency (%)Homozygous frequency (%)Number screenedMethodReference
  1. PCR, polymerase chain reaction; AS-PCR, allele-specific PCR; ARMS-PCR, amplification refractory mutation system PCR; ND, not done.

8930 45PCR sequencingJames et al (2005)
6527128PCR sequencingKralovics et al (2005)
7425164PCR sequencingLevine et al (2005a)
9726 73AS-PCRBaxter et al (2005)
83ND 24PCR sequencingZhao et al (2005)
8133 72ARMS-PCRJones et al (2005)
9526 38PCR sequencingTefferi et al (2005b)
86ND 29PyrosequencingJelinek et al (2005)
92ND 25Real time AS-PCRPassamonti et al (2006)
10027 22PCR sequencing (cDNA)Goerttler et al (2005)
95ND 80PCR sequencingJames et al (2006)
9221 63PCR sequencingTefferi et al (2006)
6535 17PCR-SSCPLee et al (2006)
80ND 84AS-PCRVizmanos et al (2006)
8424 63ARMS-PCRVannucchi et al (2006b)
Table IV.   JAK2 V617F mutations in essential thrombocythaemia.
Mutation frequency (%)Homozygous frequency (%)Number screenedMethodReference
  1. PCR, polymerase chain reaction; AS-PCR, allele-specific PCR; ARMS-PCR, amplification refractory mutation system PCR; ND, not done.

43   21PCR sequencingJames et al (2005)
233  93PCR sequencingKralovics et al (2005)
323 115PCR sequencingLevine et al (2005a)
570  51AS-PCRBaxter et al (2005)
417  59ARMS-PCRJones et al (2005)
550  22PCR sequencingTefferi et al (2005b)
30ND  10PyrosequencingJelinek et al (2005)
53ND  19Real-time AS-PCRPassamonti et al (2006)
335  42PCR sequencing (cDNA)Goerttler et al (2005)
53ND 776AS-PCRCampbell et al (2005)
490 150PCR sequencingWolanskyj et al (2005)
576 130AS-PCRAntonioli et al (2005)
62ND 243AS-PCRVizmanos et al (2006)
641  85ARMS-PCRVannucchi et al (2006b)
Table V.   JAK2 V617F mutations in idiopathic myelofibrosis.
Mutation frequency (%)Homozygous frequency (%)Number screenedMethodReference
  1. PCR, polymerase chain reaction; AS-PCR, allele-specific PCR; ARMS-PCR, amplification refractory mutation system PCR; ND, not done.

43   7PCR sequencingJames et al (2005)
5722 23PCR sequencingKralovics et al (2005)
359 46PCR sequencingLevine et al (2005a)
5019 16AS-PCRBaxter et al (2005)
4329 35ARMS-PCRJones et al (2005)
300 10PCR sequencingTefferi et al (2005b)
95ND 19PyrosequencingJelinek et al (2005)
57ND 30Real time AS-PCRPassamonti et al (2006)
5721 14PCR sequencing (cDNA)Goerttler et al (2005)
453117PCR sequencingTefferi et al (2006)
55 152AS-PCR/sequencingCampbell et al (2006b)
68ND 22AS-PCRVizmanos et al (2006)
38  21AS-PCRJohan et al (2005b)
8139 31ARMS-PCRVannucchi et al (2006a)

The JAK2 V617F mutation has been detected in progenitors and myeloid cells including cells with HSC, common myeloid progenitor, megakaryocyte–erythroid progenitor phenotypes as well as colony-forming cells and more mature progeny, such as neutrophils and platelets (Baxter et al, 2005; Jamieson et al, 2006; Kiladjian et al, 2006a). So far, mutant JAK2 V617F has not been reported in T or B lymphocytes (James et al, 2005; Lasho et al, 2005). The mutation is somatic and has not been detected in any normal individuals or patients with secondary erythrocytosis. An initial report suggesting that JAK2 V617F could be detected in buccal samples from a small proportion of PV patients is likely to be due to contamination of such specimens with granulocyte DNA (Kralovics et al, 2005; Levine et al, 2005a). In just over a quarter of cases, only the mutant JAK2 V617F allele is found and this has been shown to be due to LOH at the JAK2 locus on chromosome 9p (Kralovics et al, 2005). LOH in this case is not due to gene deletion but to both copies being from a single parental origin due to mitotic recombination (acquired uniparental disomy) (Kralovics et al, 2002). This results in two mutant copies increasing the JAK2 V617F dosage. This feature is unusual for a gain-of-function mutation and LOH is normally associated with the loss of a tumour suppressor gene (Payne & Kemp, 2005).

The JAK2 V617F mutation is also found in IMF and ET with a prevalence of around 50%. Homozygous mutations are uncommon in ET compared with PV. The mutation has been screened for in a number of other haematological malignancies – it is found in some cases of atypical MPD (Jones et al, 2005; Steensma et al, 2005), in a subset of patients with MDS perhaps in association with 5q− (Ingram et al, 2006) and rarely in AML unless it is secondary to a previous MPD (Jelinek et al, 2005; Levine et al, 2005b; Frohling et al, 2006; Lee et al, 2006; Steensma et al, 2006). No cases have been described in lymphoid malignancies (Levine et al, 2005b) although a distinct mutation, JAK2 L611S, was discovered in one case of pre-B-ALL during a screen for JAK2 mutations using denaturing high-performance liquid chromatography (Kratz et al, 2006). The JAK2 V617F mutation is also found at similar frequencies to that seen in sporadic disease in familial MPDs. Germline transmission of JAK2 V617F has not been detected but the presence of this acquired mutation in these familial cases suggests that this can precipitate disease on the background of an underlying genetic predisposition (Cario et al, 2005; Bellanne-Chantelot et al, 2006).

Biological effects of JAK2 V617F.  As discussed above, the JH2 domain exerts an inhibitory effect on JAK activity and the V617F mutation is predicted to disrupt this inhibition (Lindauer et al, 2001). Ectopic expression of JAK2 V617F in either epithelial or haemopoietic cell lines results in autophosphorylation of mutant JAK2 [which is not seen with wild-type (WT) JAK2] and activation of downstream signalling (James et al, 2005; Levine et al, 2005b). BAF3 or FDCP cell lines expressing the EPOR and engineered to stably express JAK2 V617F showed a degree of factor independence as well as marked hypersensitivity to EPO. James et al (2005) showed that these responses were inhibited by the presence of WT JAK2, perhaps due to competition for limiting numbers of cytokine receptor subunits. Work from the group of Lodish has shown that co-expression of JAK2 V617F with a homodimeric type 1 cytokine receptor (EPOR, TPOR or G-CSFR) facilitates the transformation of cells to growth-factor independence, suggesting that the mutant JAK2 requires a receptor scaffold so as to be active (likely due to the need for proximity required for transphosphorylation and activation) (Lu et al, 2005). This contrasts with the effects of the TEL–JAK2 fusion which can readily transform cells on its own, presumably because of the strong homo-dimerisation effects of the TEL moiety (Lacronique et al, 1997). Ectopic expression of JAK2 V617F can also sensitise cells to the effects of IGF1, a common feature of PV progenitors (Staerk et al, 2005).

Expression of JAK2 V617F in murine HSC in a transplantation model has been reported. James et al (2005) showed that this resulted in erythrocytosis without a major effect on granulocyte or platelet counts. Wernig et al (2006) have recently shown that JAK2 V617F expression produces an increased haematocrit and splenomegaly in recipient mice with a variable degree of leucocytosis and marrow fibrosis depending on the strain of mice used. Although bone marrow examination showed megakaryocyte hyperplasia, no increase in platelet counts was seen. The phenotype of these mice contrasts markedly with those transplanted with other MPD-associated activated TKs, such as BCR-ABL or TEL-PDGFRA where a MPD without elevated haematocrit develops (Wernig et al, 2006). These results suggest that the presence of the JAK2 V617F may be sufficient to induce PV but the disease phenotype may be affected by other, so far unknown, genetic modifiers. Lacout et al (2006) found that JAK2 V617F expression in transplanted C57BL/6J mice resulted in polycythaemia accompanied by neutrophilia. This was followed by marrow fibrosis with resulting anaemia in a manner akin to the ‘spent phase’ and fibrosis seen in the natural history of some cases of PV. In one group of secondarily transplanted animals, where there was a relatively low level of JAK2 V617F expression, a brief transient period of thrombocytosis was seen. It has been suggested that this is a situation analogous to that seen in clinical ET where JAK2 V617F homozygosity is rare and may provide some of the explanation for this phenotype. However, the animal data so far are limited and will require further study, preferably using an approach where the mutant JAK2 is ‘knocked-in’ to the endogenous JAK2 locus.

JAK2 V617F in ET.  The frequency of JAK2 V617F is significantly lower in ET and semiquantitative analysis shows that JAK2 V617F levels may be low, suggesting the presence of both mutant clonal and residual normal haemopoietic cells (Antonioli et al, 2005; Campbell et al, 2005; Wolanskyj et al, 2005). These studies were mainly carried out with granulocyte DNA but the results are unlikely to be due to selective involvement of the megakaryocyte lineage as one report shows good concordance between platelet and granulocyte levels of JAK2 mutant (Campbell et al, 2005) and in another, V617F JAK2 was discovered in the platelets of only two of 24 patients negative by granulocyte PCR (Kiladjian et al, 2006a).

In ET, using techniques to detect X-chromosome inactivation patterns (XCIP), a significant proportion of cases have been shown to have polyclonal haemopoiesis (Harrison et al, 1999). Polyclonal cases of ET have been shown to have mutant JAK2 V617F at a similar prevalence to monoclonal cases (Antonioli et al, 2005; Kiladjian et al, 2006a; Levine et al, 2006). This has been interpreted as a failure of relatively insensitive clonality assays in picking up low-level clones that can be detected by JAK2 mutation analysis. It is perhaps surprising that the mutant clone does not come to dominate considering its potential proliferative advantage and this finding is distinct from other MPDs, such as BCR-ABL-driven CML, where the mutant cells comprise the majority of cells. Further, it is not easy to reconcile the presence of a minor JAK2 mutant population, perhaps detectable only by PCR-based techniques, with the presence of clinical disease, i.e. how does a clone that is only, say, at the 15% level cause a platelet count over 1000 × 109/l. To add to the complexity of JAK mutations in ET, some monoclonal cases of ET (by XCIP analysis) have only low levels of mutation suggesting that the JAK2 V617F is a secondary change in a subpopulation (Levine et al, 2006) and a significant proportion of monoclonal cases of ET have WT JAK2 (Antonioli et al, 2005; Levine et al, 2006). The lack of significant thrombocytosis in some murine transplantation models as well as these clinico-pathological findings indicate a more complex role for JAK2 V617F in ET which needs further clarification.

JAK2 V617F in IMF.  There are variable reports of the clinical correlates of JAK2 V617F in myelofibrosis with a link to higher white cell count but not to rate of leukaemic transformation (Mesa et al, 2006a,b) and also to an older age at presentation as well as an association with pruritus and thrombosis (Tefferi et al, 2005a). Campbell et al (2006b) have reported a reduced red cell transfusion requirement and worse survival in mutant-positive patients. The reduced red cell requirements have been interpreted as being consistent with erythroid stimulation by the mutant JAK2 but no data are given in this study as to how cases of myelofibrosis secondary to PV were excluded (e.g. red cell mass estimation). The use of JAK2 mutant status as a tool to assist in the choice of therapies such as stem cell transplantation is not appropriate at present and will require additional data from other investigators.

Other clinico-pathological correlates of JAK2 V617F.  The presence of JAK2 mutation in PV, ET and IMF has been reported to correlate with other biomarkers such as polycythaemia rubra vera type 1 (PRV1) expression and endogenous erythroid colony formation (Goerttler et al, 2005; Tefferi et al, 2005a,b). The association of PRV1 expression with JAK2 V617F in ET and IMF has since been questioned (Antonioli et al, 2005; Bellosillo et al, 2006; Vannucchi et al, 2006a). In patients with PV, homozygous JAK2 V617F is associated with an increased Hb at diagnosis and increased rate of fibrotic change, but not with thrombosis or bleeding risk or with duration of disease (Tefferi et al, 2006). In ET, a large series from the UK showed that mutation is associated with features that resemble PV, including increased Hb and neutrophils, risk of transformation to PV, low ferritin and EPO and increased risk of venous thromboses (Campbell et al, 2005). Similar results have been found by other groups (Antonioli et al, 2005) and have led to suggestions that patients with JAK2 V617F-positive ET may have a forme fruste of PV with the level of erythrocytosis influenced by genetic or acquired modifiers (Campbell et al, 2005). The presence of JAK2 V617F is also associated with lower platelet counts and a requirement for lower doses of hydroxycarbamide but not anagrelide (Campbell et al, 2005). V617F-positive patients treated with hydroxycarbamide had significantly lower rates of arterial thrombosis compared with those treated with anagrelide, an effect not seen in the JAK2-WT group.

Quantification of mutant JAK2 V617F following treatment with either imatinib or α-interferon has been reported by two groups. Jones et al (2006) reported that significant molecular responses to imatinib were uncommon; however, clinical response rates in the reported cases were also low and in the two cases with complete haematological response showed a several fold reduction in mutant level. Kiladjian et al (2006b) showed that patients treated with interferon, who had a much higher rate of clinical response, showed modest but significant decreases in mutant JAK2 level although with reduction below detection limits in only one case. These results suggest that in some cases of PV, the mutant cells are more sensitive to antiproliferative agents albeit with a modest differential compared with residual non-mutant cells.

Diagnostic implications of JAK2 V617F.  The very high frequency of JAK2 V617F in PV has implications for the diagnosis of this disease. Incorporating a mutation screen early into the algorithm for investigating a case of suspected polycythaemia could help to streamline diagnosis although clearly, the presence of a JAK2 mutation alone does not distinguish PV from IMF or ET. James et al (2006) have shown that, in the investigation of 88 patients with erythrocytosis, according to World Health Organization criteria, the JAK2 V617F mutation was found in 57 of 61 cases diagnosed as PV and 0 of 11 as idiopathic erythrocytosis. This contrasted with 43 of 45 and eight of 21, respectively, on PVSG criteria. They, and others (Tefferi & Pardanani, 2006) have suggested that mutation analysis can be used to help screen individuals and that this may reduce the need for further investigations, such as red cell mass and bone marrow biopsy. However, the utility of these types of approach remains to be demonstrated in a prospective clinical study. In ET, where differentiating a primary proliferative condition from a reactive one is notoriously difficult, the use of JAK2 mutation analysis may assist in identifying patients with a stem cell disorder. JAK2 analysis has broader applicability as clonality assays are restricted to females and are also not evaluable in older women due to the phenomenon of age-related skewing. Of course, it has to be remembered that patients without a JAK2 mutation can still have a primary MPD.

Implications for therapy.  The identification of JAK mutations in MPDs, and the role that aberrant JAK signalling plays in other haematological malignancies, has raised the exciting prospect of developing specific inhibitors for clinical use and significant efforts to screen for such compounds are under way. In the rare patients with malignant disease associated with strongly activating JAK2 translocations, such as those with PCM1-JAK2, such inhibitors, by analogy with imatinib in CML, may have significant clinical activity. In the management of chronic MPDs, there are a few outstanding questions that will need careful evaluation. First, whether it is possible to obtain an inhibitor that has preferential activity against mutant rather than WT JAK2 and which does not produce significant haematological toxicity. Secondly, as the management of many patients with PV with conventional therapies, including venesection, aspirin and hydroxycarbamide, has a reasonable outcome at relatively modest costs, a cost–benefit analysis of potentially expensive long-term targeted therapy will be needed. Thirdly, in ET and IMF, the role of JAK2 V617F as the driving force in disease pathophysiology is not as yet clear and the potential impact of JAK2-directed therapy is uncertain. With the huge effort being expended in investigating JAK signalling, we can be certain that the next few years will bring further developments in this fast-evolving field.