Receptor tyrosine kinase mutations in myeloid neoplasms.

Abstract

The receptor tyrosine kinases (RTKs) are transmembrane enzymes involved in ligand binding and signal transduction at the cell surface. They function in nearly all biological systems and have a broadly conserved molecular topology, which permits activation of intracellular protein kinase activity upon ligand binding. Mutations involving RTK genes occur in both acute and chronic myeloid leukaemias. In this review, we outline the phylogeny and biochemistry of RTKs with reference to myeloid ontogeny, classify the types of RTK mutation seen in myeloid malignancies, and discuss the prognostic and potential therapeutic significance of these acquired genetic changes. Of the many mechanisms for signal transduction across the cell membrane, the activation of cell surface receptors with inherent tyrosine kinase activity by growth factors is well understood. Tyrosine kinases are enzymes which catalyse the transfer of the γ-phosphate of ATP to tyrosine residues of protein substrates. There are at least 17 defined families of RTKs (Robertson et al, 2000). Each has a conserved structure with an extracellular ligand-binding domain, a transmembrane (TM) domain and an intracellular tyrosine kinase domain. RTK families differ primarily in their extracellular domains (Table I, Fig 1). Fig 1. Diagrammatic representation of receptor tyrosine kinase families (left) and key to domains (right) (after Robertson et al, 2000). EGF, epidermal growth factor; PDGF, platelet-derived growth factor; FGF, fibroblast growth factor; VEGF, vascular endothelial growth factor; HGF, hepatocyte growth factor; TRK, tropomyosin receptor kinase; AXL, anexelekto (Greek word for uncontrolled); LTK, leukocyte tyrosine kinase; ALK, anaplastic lymphoma kinase; TIE, tyrosine kinase with Ig and EGF homology domains; ROR, receptor tyrosine kinase-like orphan receptor; DDR, discoidin domain receptor; RET, rearranged during transformation; KLG, kinase-like gene; RYK, related to tyrosine kinase; MuSK, muscle-specific receptor tyrosine kinase. *The LTK gene was initially cloned on account of its homology with the intracellular domains of the insulin receptor. Subsequent studies have shown the locus to give rise to several differentially spliced isoforms with alternative extracellular domains (Toyoshima et al, 1993). The majority of RTK molecules are monomers. Exceptions include members of the closely related insulin receptor and ltk families, which exist as α2β2 heterotetramers, and members of the hgf family, which are αβ heterodimers (Fig 1). In all such cases, the receptor molecules are translated from a single precursor transcript and then undergo post-translational cleavage and modification in order to generate the final protein. RTK families differ primarily in their extracellular domains (Fig 1). Several structural motifs have been described (reviewed by Robertson et al, 2000). TM domains share a common alpha helical structure which not only anchors the molecule in the cell membrane, but also appears to play a critical role in receptor activation. The juxtamembrane region separates the TM domain from the cytoplasmic kinase domain. It is divergent between different RTK families, but highly conserved within them. The kinase domains are the most conserved components of the RTKs. Sequence alignments of a variety of receptor and non-receptor tyrosine kinases have defined 12 subdomains (I–XII) of high homology, containing at least nine invariant amino acids that are predicted to form the active site (Hanks et al, 1988). Crystallographic studies of the active and inactive conformations of the kinase domains of several RTKs have been made, and a conserved structure comprising eight alpha helices and eight beta strands has emerged (Hubbard et al, 1994, Johnson et al, 1996). All tyrosine kinases contain three large flexible protein domains, each bearing some of the invariant residues mentioned above. These are the activation loop (A-loop), whose conformation regulates kinase activity, the nucleotide-binding loop and the catalytic loop. The conformation of the A-loop (which comprises subdomains VII and VIII) is controlled by the phosphorylation of specific tyrosine residues within it. In the inactive state it blocks either the substrate or the ATP binding site of the enzyme. When phosphorylated, it is repositioned to contact residues in the C-terminal domain of the molecule (Mohammadi et al, 1996), thereby permitting enzyme activity. The nucleotide-binding loop contains hydrophobic residues responsible for binding the adenine moiety of ATP such that the γ-phosphate is in the correct position for catalysis. It is the least conserved part of the tyrosine kinase moiety and has been a major target in the development of receptor-specific tyrosine kinase inhibitors. The catalytic loop of protein kinases contains an invariant aspartate residue that serves as the catalytic base in the phosphotransfer reaction (Johnson et al, 1996). This is part of the highly conserved sequence His-Arg-Asp-Leu-Ala-Ala-Arg-Asn, present in the catalytic domains of several RTKs (McTigue et al, 1999). Many RTKs contain an insert of variable length and sequence in their kinase domains. This is known as the kinase insert domain (KID). Mutational analyses of several kinases have shown that the KID is not necessary for intrinsic kinase activity but that it contains tyrosine residues that are sites for autophosphorylation (Heidaran et al, 1991). The C-terminal tail lies downstream of the kinase domain and also contains target sites for autophosphorylation. Phosphorylation allows interaction with proteins containing src homology 2 (SH2) and phosphotyrosine-binding (PTB) domains that are critical for downstream signalling (Shewchuck et al, 2000). This process can be considered to occur in three sequential steps: ligand binding, signal transduction and kinase activation. Receptor activation is triggered by ligand binding, and is marked biochemically by tyrosine autophosphorylation at specific intracellular residues. The binding of a ligand to its RTK induces the dimerization of the monomeric receptors, the tetramerization of hepatocyte growth factor (HGF) family receptors, or a rearrangement of the tertiary and quaternary structures of heterotetrameric receptors, which in turn leads to receptor activation. The mechanism of receptor dimerization varies. Some ligands, e.g. platelet-derived growth factor (PDGF), are disulphide-linked dimers, while others, e.g. epidermal growth factor (EGF), are monomeric but have two receptor binding domains per molecule. Ligand binding can induce receptor homodimerization or heterodimerization. The mechanism by which ligand binding to and dimerization of RTKs drives tyrosine autophosphorylation is becoming clearer. While essential for receptor activation, ligand-induced receptor dimerization does not directly lead to increased kinase activity. Further conformational changes are required. Because of steric constraints, autophosphorylation probably represents the reciprocal transphosphorylation of the two receptor molecules within the ligand receptor complex (Schlessinger, 2000). Ligand binding is thought to lead to the apposition of the kinase domains of dimerized receptors by a mechanism critically dependent upon TM domain sequences. Only once this has occurred can transphosphorylation occur. It is known that the TM domains of several RTKs are functionally interchangeable and have conserved hydrophilic residues situated within the alpha helix such that they lie on the same face of the domain (Sternberg & Gullick, 1990; Ullrich & Schlessinger, 1990). It is thought that ligand binding leads to the coupled rotation of the paired RTKs within a receptor-ligand complex. Critically spaced hydrophilic residues located in the TM (and other) domains of each receptor are then able to form hydrogen bonds that hold them in a fixed orientation. This in turn brings the kinase domain of each receptor into contact with that of its partner, so permitting transphosphorylation (Bell et al, 2000). There are two hypotheses as to how receptor transphosphorylation might be initiated. One states that the kinase domains of unphosphorylated receptors have inherent low-level activity, with the highly mobile A-loop oscillating between an active and an inactive conformation. Ligand binding merely increases the local concentration of kinase domains (i.e. enzyme and substrate), thus increasing the likelihood of transphosphorylation. The other suggests that dimerization transiently stabilizes the A-loop in the active conformation and so substrate binding and phosphotransfer can occur. Once phosphorylation has occurred, the A-loop remains in the active conformation. There are no convincing data to distinguish between the two, but the target residues for phosphorylation are known in a variety of RTKs (Hubbard et al, 1998). Tyrosine phosphorylation of activated RTKs both promotes intrinsic kinase activity and generates sites of interaction for the various downstream phosphotyrosine-binding signalling proteins. All activated RTKs perform each of these functions to a greater or lesser extent. In keeping with this, there are two classes of tyrosine residue that are phosphorylated on receptor activation (Hubbard et al, 1998). Most RTKs contain up to three conserved residues in their A-loops which, when phosphorylated, stabilize an active conformation and permit catalysis. The other type of residue is less conserved and is located in non-catalytic positions. Phosphorylation here creates docking sites for downstream signalling molecules. There is a bewildering amount of data regarding intracellular signalling proteins. Classic genetics has predicted independent linear cascades of messenger activity downstream of each receptor, but molecular biology has found considerable overlap in the systems involved by both activating and inhibitory extracellular signals. The many signalling pathways can be considered to form a single homeostatic network capable of co-ordinating the many biological responses of a cell to its environment. There have been several recent and authoritative reviews in this area to which the reader is directed (Schlessinger, 1994, 2000; Pawson & Scott, 1997). During mouse development, the earliest intraembryonic cells with haematopoietic potential can be identified by the expression of vascular endothelial growth factor receptor 2 (VEGFR2)/flk1 protein (Shalaby et al, 1995). These primitive mesoderm cells ultimately give rise to definitive haematopoietic stem cells, which express c-kit. At an intermediate stage lies a c-kit+tie2+ cell, which is thought to represent a bipotent haematopoietic/endothelial precursor (Hamaguchi et al, 1999). Despite the general biological utility of RTK signalling and the interest in RTKs as potential therapeutic targets, there is a paucity of expression data for RTK proteins in normal human blood and bone marrow. Table II shows the RTKs whose expression has been documented at the protein level in non-neoplastic human haematopoietic cells. These studies have relied upon flow cytometric analysis and in situ hybridization of bone marrow trephine biopsy sections although, in some cases, functional assays have been performed. Using the former approach, it is clear that considerable heterogeneity of RTK expression exists within the CD34+ population in normal human bone marrow (Buhring et al, 1999). Expression of axl at the RNA level has been reported in normal marrow CD34+ cells and peripheral blood monocytes (Neubauer et al, 1994); and met RNA has been similarly found in activated monocytes (Beilmann et al, 1997). Ryk expression has been detected using a specific monoclonal antibody in normal maturing myeloid precursors in the mouse (Simoneaux et al, 1995). The expression of various RTKs has gained increasing utility both as a diagnostic tool and as an indicator of prognosis in neoplastic myeloid disorders over the last decade. Many studies have detected apparently aberrant expression of various RTK proteins or RNAs in cell lines. However, the most informative work has concerned the analysis of leukaemic cells sampled directly from patients. The majority of data has come from the analysis of acute myeloid leukaemia (AML) patients, which may reflect the relative ease in isolating neoplastic cells in these cases. The importance of c-kit as a marker for AML is widely accepted and protein expression is found in approximately 65% of cases in most series (Sperling et al, 1997). Flt3 may be found in an even higher proportion of cases (Rosnet et al, 1996), but it is also found in many acute lymphoblastic leukaemias (ALL), thereby diminishing its diagnostic utility (Birg et al, 1992). Studies of other RTKs have been less frequent and the numbers of cases examined is small. TrkA protein has been found in 44% of cases of adult AML (Kaebisch et al, 1996) and CSF1R/c-fms protein in 30% of cases of childhood AML (Ashmun et al, 1989). There have been case reports of met (Jucker et al, 1994) and ephB4 (Muroi et al, 1998) protein expression in AML. Some studies have used reverse transcription polymerase chain reaction (RT–PCR) to detect expression of RTK RNA. Using this approach, VEGFR3/flt4 mRNA has been found in 34% of all AML (Fielder et al, 1997) and axl mRNA in 35% of AMLs (Rochlitz et al, 1999). In this series, the authors concluded that axl expression was associated with reduced overall survival. Another study has shown axl mRNA in 60% of 66 cases of myeloproliferative disorders including both chronic and acute leukaemias (Neubauer et al, 1994). Some preliminary data has shown evidence for axl protein in CD34+ AML blasts (Neubauer et al, 1997). In the studies outlined to date, RTK expression has not been found to correlate consistently with any FAB subtype of AML. Only two small series have detected a bias in this respect. Kukk et al (1997) examined tie1 protein expression in AML and found it to correlate with an erythroblastic/megakaryocytic phenotype. In a larger study, ret protein and mRNA expression was found in 60% of AMLs and a correlation with monocytoid differentiation was noted (Gattei et al, 1999). In addition to their expression in myeloid tumours, there is accumulating data regarding specific mutations in RTK genes in these disorders. Some studies of AML blasts known to express c-kit or flt3 proteins have failed to demonstrate a proliferative response to the appropriate ligand in culture experiments (Stacchini et al, 1996). This may reflect a constitutive activation of these molecules on account of acquired genetic mutations. Malignant transformation by RTK molecules could be predicted by the observation that many RTK genes have constitutively active retroviral homologues which bear subtle sequence differences. Although aberrant expression of RTKs is observed in some myeloid neoplasms, there is as yet no clear evidence for RTK overexpression per se leading to a clonal myeloid disorder. Acquired mutations have been observed in RTKs of the PDGF receptor family and in fibroblast growth factor receptor 1 (FGFR1). The spectrum of observed alterations encompasses most types of coding mutations including missense, insertion events and the generation of fusion proteins by specific chromosomal translocations(Tables III, IV, V, VI and VII). Generally, the effects of mutant RTKs are thought to be dominant over the function of the un-mutated proteins and, in most cases, mutant genes are found to be heterozygous with their wild-type counterparts. In some cases these genetic changes are associated with defined myeloproliferative syndromes. There is little data regarding non-coding mutations and searches for mutations involving other RTK genes have not been reported. It is likely that these data will emerge in the coming years. Rearrangement of the PDGFRβ gene with formation of a fusion gene was first recognized in association with t(5;12)(q31;p12), the fusion gene formed being ETV6–PDGFRβ (Golub et al, 1994). Subsequently, PDGFRβ was found to contribute to three other fusion genes in patients with myeloproliferative disorders: HIP1–PDGFRβ in association with t(5;7)(q33;q11.2), H4/D10S170–PDGFRβ in␣association with t(5;10)(q33;q11–21) (Anastasiadou et al,␣1999), and Rabaptin–PDGFRβ in association with t(5;17)(q33;p13) (Magnusson et al, 2000). In addition, a CEV14–PDGFRβ fusion gene has been described in association with t(5;14)(q33;q32) occurring as a second event at the time of relapse of acute myeloid leukaemia (Abe et al, 1997). The cases described to date are summarized in Table III. Of these disorders, the most frequently observed is that associated with t(5;12)(q31;p12). The majority of cases are best described as either atypical chronic myeloid leukaemia (aCML) with eosinophilia or as chronic myelomonocytic leukaemia (CMML) with eosinophilia. However, three cases are more accurately characterized as chronic eosinophilic leukaemia (Keene et al, 1987; Yates & Potter, 1991; Granjo et al, 2000). Overall, 15 of the 16 patients for whom adequate information was available had increased eosinophils in the peripheral blood, bone marrow or both (Table III). It seems clear that all patients with t(5;12) and ETV6–PDGFRβ fusion genes have the same disease, despite a certain amount of variation in the haematological features. It would seem more satisfactory for such cases to be recognized as a specific entity rather than being aggregated with other myeloproliferative disorders with different cytogenetic and molecular genetic characteristics. One of the most notable features of this syndrome is the very striking male predominance. Only a single case, one that was not characterized at a molecular level, has been reported in a female (Pellier et al, 1996). The six patients described with PDGFRβ contributing to other fusion genes are a somewhat more heterogeneous group, but eosinophilia was again noted in de novo myeloproliferative disorders in association with HIP–PDGFRβ, H4/D10S170–PDGFRβ and Rabaptin–PDGFRβ fusion genes. As for t(5;12)-associated myeloproliferative disease, all five reported patients were male. The diagnoses were either CMML or aCML, with eosinophilia being a feature in the four cases described in sufficient detail for assessment.␣In addition, when CEV14–PDGFRβ/t(5;14)(q33;q32) appeared at relapse of acute myeloid leukaemia in a young␣woman, it was associated with the appearance of eosinophilia. Mis-sense mutations at codon 969 in the c-terminal tail encoding region of c-fms have been reported in several myeloproliferative and myelodysplastic disorders (Table IV). The haematological features are quite heterogeneous but, in the myelodysplastic syndrome (MDS) group at least, the mutation is associated with a poor outcome. The mutation Leu301Ser affecting the fourth immunoglobulin domain of the molecule has been found in occasional cases. This amino acid is conserved across all species. Interestingly, the viral transforming protein found in the feline McDonough sarcoma virus, v-fms, carries the same genetic alteration (Woolford et al, 1988). Allelic loss of c-fms has been detected in both AML and MDS (Ridge et al, 1990; McGlynn et al, 1997). There is neither a clear phenotypic association nor consensus regarding its prognostic significance. The association of the activation loop mutation Asp816Val in c-kit with adult mastocytosis has been recognized for some time (Table V, Bain, 1999). The link with childhood mastocytosis, which represents a broader spectrum of disorders, some of which are clearly reactive, is weaker. The mutation can be detected almost universally in DNA made from lesional skin biopsies, but, when studies are performed with peripheral blood or bone marrow DNA, the frequency of detection is much lower, consistent with the mutation being restricted to mast cells. Other missense mutations involving Asp816 and mutations affecting other domains of c-kit have been reported in mastocytosis (Table V), but these have generally been small series or sporadic cases reported by individual groups. Of interest among these is the Glu839Lys kinase domain mutation seen in lesional skin biopsies from childhood urticaria pigmentosa (Longley et al, 1999). In vitro evidence suggests that this mutant protein is not constitutively activated, unlike other RTK mutants. Given the high percentage of expression of c-kit in AML, the finding of mutations might be expected in this condition. The Asp816Val mutation has been found in some cases of AML (Table V). A few of these have been associated with mast cell involvement and it is not clear if the AML represents an evolution of the original mast cell disorder. Other cases have been associated with CBF leukaemias␣involving the characteristic inv(16)(p13q22) or t(8;21)(q22;q22) chromosomal rearrangements. It is not clear whether the presence of c-kit mutations is associated with any morphological phenotype or affects the outcome of the disease. A recent study has identified a heterogeneous group of mutations, which all lead to loss of Asp419 in the fifth immunoglobulin domain of c-kit present in AML (Gari et al, 1999). A possible association with CBF leukaemias was again noted, but the prognostic significance is unclear. An important goal in identifying novel mutations in AML has been to further stratify likely clinical phenotypes and hence tailor treatments. Recent studies of mutations in flt3 may satisfy this aim. In frame internal tandem duplications (ITD) affecting the JM domain of the flt3 protein are found in blast cell genomic DNA from approximately 20% of adult AML and across all French–American–British (FAB) classification subtypes, usually in the absence of detectable cytogenetic abnormalities (Table VI). A slightly lower frequency has been observed in adult MDS and in childhood AML. In all cases of AML, these mutations are associated with reduced overall survival. In children at least there is often a failure to achieve a first remission (Xu et al, 1999). Mutations involving Asp835 in the activation loop of the flt3 molecule are found in 7% of adult AML. These mutations are analagous to the Asp816 changes affecting c-kit. Similarly they are found in all FAB subtypes and appear to have no bearing on prognosis (Yamamoto et al, 2001). Rearrangements of the FGFR1 gene have been described most often in association with t(8;13)(p11–12;q11–12), with formation of a ZNF198–FGFR1 fusion gene (Xiao et al, 1998). The same fusion gene was described simultaneously by two other groups – being designated FIM–FGFR1 (Popovici et al, 1998) and RAMP–FGFR1 (Smedley et al, 1998a) respectively. A total of 16 patients have been described with t(8;13) or a related complex chromosomal rearrangement, and a syndrome comprising a myeloproliferative disorder with eosinophilia, T-lymphoblastic lymphoma (sometimes not fully characterized and designated T-non-Hodgkin's lymphoma), acute myeloid leukaemia, granulocytic sarcoma and, less commonly, B-lineage acute lymphoblastic leukaemia (Table VII). The myeloproliferative disorder has not always been fully characterized. Sometimes it has met the criteria for CMML and sometimes those for aCML. Eosinophilia has been a prominent feature, being reported in 15 of the 16 patients. Some patients have shown neutrophilia, basophilia, monocytosis and the presence of immature granulocytes. These cases clearly represent a distinctive syndrome and should be regarded as a disease entity rather than being categorized with other cases of aCML or CMML with eosinophilia. A smaller number of patients have been described with t(6;8)(q27;p11), leading to the formation of a FOP–FGFR1 fusion gene (Chaffanet et al, 1998; Popovici et al, 1999), or t(8;9)(p11;q34), which leads to formation of a CEP110–FGFR1 fusion gene (Chaffanet et al, 1998). Features of these patients have been very similar to those of the disorder associated with t(8;13) and they may reasonably be regarded as part of the same syndrome (Macdonald et al, 1994). They also have a myeloproliferative disorder, generally with eosinophilia and sometimes with AML, granulocytic sarcoma or T-lymphoblastic lymphoma (Table VII). Other patients have been described with acute leukaemia associated with rearrangement of the FGFR1 gene. Such cases include those reported by Sohal et al (2000) – AML associated with t(8;11)(p11;p15), two patients with a myeloproliferative disorder with eosinophilia and T-lymphoblastic lymphoma, associated with t(8;12)(p11;q15) and ins(12;8)(p11;p11p21), respectively, and Ph-negative chronic myeloid leukaemia with malignant mast cell disease associated with t(8;17)(p11;q25). A further patient, reported by Mugneret et al (2000) had AML and probably an underlying myeloproliferative disorder associated with t(8;19)(p12;q13.3). The clinical features in at least two of these patients are clearly very similar to those associated with t(8;13) but description of a larger number of patients is required to determine if they should be regarded as part of the same syndrome. There is much interest in the study of the transforming potential of the mutant RTK proteins that have been discovered in human myeloid malignancies. Such information would provide insights into the pathogenesis of disorders in which they are found, and may define potential therapeutic targets. A plausible paradigm for how a mutated RTK may acquire oncogenic potential is through the acquisition of constitutive kinase (and hence signalling) activity. Biochemically, this may manifest as constitutive receptor dimerization and phosphorylation. There are several levels at which mutant RTKs may be studied. In vitro, constitutive tyrosine phosphorylation can be readily detected using monoclonal antibodies directed against phosphotyrosine moieties. Constitutive receptor dimerization may be shown by chemical cross-linking analysis. Simple transfection studies can provide evidence for the transformation of cell lines to factor-independent growth; generally such analyses show the mutant gene products to act dominantly over the wild type.␣Finally, studies involving transgenic animals can provide in␣vivo evidence for the transforming␣potential␣of␣a mutant RTK. The ETV6–PDGFRβ fusion product is the most studied (Golub et al, 1994). The breakpoint is such that the fusion protein contains the TM and intracellular domains of PDGFRβ, but its extracellular domains are replaced by the amino-terminal 154 amino acids (aa) of the ETV6 transcription factor (Carroll et al, 1996). This component of the ETV6 protein is known to contain a 65aa homotypic oligomerization domain, the pointed (PNT) motif (Jousset et al, 1997). The fusion protein is constitutively phosphorylated and oligomerized in vitro, and the fusion gene transforms haematopoietic cells to factor-independent growth (Carroll et al, 1996). Further analyses have shown the transforming ability of the fusion gene to be dependent upon a functional kinase domain and the oligomerization to be dependent upon the ETV6 PNT sequences. It appears that the kinase domain of the fusion protein is constitutively activated on account of ligand-independent dimerization induced by the ETV6 PNT domain. The fusion protein has been shown to directly phosphorylate and activate the latent cytoplasmic transcription factors STAT1 and STAT5 (Wilbanks et al, 2000), bypassing conventional second messenger systems. The expression of the fusion gene under the control of an immunoglobulin enhancer in transgenic mice leads to the development of T- and B-lineage lymphoblastic lymphomas at a median age of 4 months, indicating its non-lineage-restricted transforming ability (Tomasson et al, 1999). A CMML-like disorder could be induced in transgenic mice expressing the fusion gene under the control of the CD11a promoter (Ritchie et al, 1999). In the HIP1–PDGFRβ fusion protein found in association with t(5;7)(q33;q11.2), all but the 18 carboxyl-terminal aa of huntingtin interacting protein1 (HIP1) replace the extracellular domains of PDGFRβ (Ross et al, 1998). HIP1 is a widely expressed membrane-associated protein of uncertain function, which is capable of homopolymerization (Wanker et al, 1997). The fusion protein is constitutively phosphorylated and oligomerized, and it has been shown to transform haematopoietic cells to factor-independent growth associated with constitutive phosphorylation of STAT5 (Ross & Gilliland, 1999). Mutational studies indicate that the amino-terminal 55aa TALIN domain of HIP1 (essential for homopolymerization) is the minimum requirement for fusion protein oligomerization and autophosphorylation, but that additional sequences from HIP1 itself are essential for the full transforming capacity to be realized (Ross & Gilliland, 1999). A similar model for constitutive fusion protein dimerization and autophosphorylation leading to transforming potential to that for the ETV6–PDGFRβ fusion product has been proposed. Studies of the fusion proteins involving H4 and rabaptin5 are less complete, but both are ubiquitously expressed proteins with domains necessary for homopolymerization. Interestingly an H4–RET fusion protein is created by a paracentric inversion of chromosome 10 in some papillary thyroid carcinomas. It is constitutively oligomerized on account of motifs in the H4 component of the molecule (Tong et al, 1995). Little work has been published regarding this group of mutations. The Tyr969Phe mutant has been shown to induce factor-independent growth in haematopoietic cells (McGlynn et al, 1998). These are the best studied of the c-kit mutant proteins associated with myeloid neoplasms. Most functional studies have used the analagous mutations at Asp814 in the mouse gene. However Longley et al (1999) were able to demonstrate ligand-independent autophosphorylation of human Asp816Val, Asp816Tyr and Asp816Phe in vitro. Mouse Asp814Val mutant c-kit proteins are not constitutively dimerized in the abs