It is now 20 years since a molecular classification of acute myeloid leukaemia (AML) was initiated through the recognition of a number of leukaemia-specific cytogenetic abnormalities and their role as independent prognostic factors (Bloomfield et al, 1984). The intensive study of some of these markers, in particular the fusion transcripts from chromosomal translocations, has provided considerable insight into the underlying disease pathogenesis through characterization of the resulting aberrant gene products and their biological and clinical consequences. This has now been incorporated into the recent World Health Organization (WHO) classification of haemopoietic and lymphoid neoplasms in which AML patients with four well-defined recurring cytogenetic abnormalities have been put together as a subgroup (Harris et al, 1999). In general, however, patients are divided into three different risk groups based on cytogenetics: those with favourable, intermediate or standard, and poor risk disease. In some centres, the response to initial therapy is also taken into account (Wheatley et al, 1999).
This classification has paved the way for a move from the indiscriminate use of high-dose chemotherapy for all patients to a more risk-adapted treatment approach. Patients with favourable cytogenetics, i.e. t(15;17), t(8;21) and inv(16), have particularly benefited from an improved understanding of the molecular pathology of their disease through identification of potential therapeutic targets. For example, the addition of all-trans retinoic acid (ATRA) during induction chemotherapy for acute promyelocytic leukaemia (APL) patients with the PML/RARα (promyelocytic leukaemia/retinoic acid receptor alpha) fusion gene has increased the 5-year survival a further 20–30% compared with chemotherapy alone (Tallman et al, 1997; Burnett et al, 1999). Patients with poor risk disease have adverse survival, which hardly exceeds 20% at 5 years (Grimwade et al, 1998). They often have complete or partial loss of genetic material, e.g. −7, −5, del(5q), del(3q) or complex karyotypes, abnormalities that are currently less suitable candidates for targeted therapy. Approximately two-thirds of newly diagnosed AML patients, however, have intermediate/standard risk disease; their remission rate is similar to that of patients with favourable disease, but their outcome is hampered by an increased relapse rate (Grimwade et al, 1998). Although some of these patients have identifiable cytogenetic abnormalities that may allow the development of novel therapies, the majority of patients, ≈ 50% of all patients, have a normal karyotype.
There has therefore been much interest in identifying further molecular markers of leukaemia that could help to improve the prognostic stratification of patients. In addition, with clear evidence for a multistep pathogenesis from both mouse models of leukaemia and the variable outcome for patients within defined cytogenetic groups, knowledge of co-operating mutations may further assist in improving therapy. Mutations in growth factor receptors and their downstream signalling molecules have long been obvious candidates for causing dysregulation of the delicate balance between proliferation and differentiation in haemopoietic cells. Many years of searching have now started to bear fruit with the demonstration that acquired mutations in the tyrosine kinase receptor gene, FLT3, are common in AML and have a major impact on prognosis. This review will outline the current knowledge on these mutations and their biological and clinical significance in leukaemia.
FLT3 (fms-like tyrosine kinase 3), also known as stem cell tyrosine kinase-1 (STK1) or fetal liver tyrosine kinase-2 (FLK-2), is one of the class III tyrosine kinase (TK) receptors that share sequence homology and structural characteristics. The latter include five immunoglobulin-like domains in the extracellular region, an intracellular juxtamembrane (JM) domain, two TK domains interrupted by a kinase insert and a C-terminal tail (Agnes et al, 1994) (Fig 1). The gene is located at chromosome 13q12 and consists of 24 exons (previously reported as 21) (Rosnet et al, 1991; Abu-Duhier et al, 2001a).
FLT3 is predominantly expressed on haemopoietic progenitor cells in the bone marrow, thymus and lymph nodes (Rosnet et al, 1993), but is also found on other tissues such as placenta, brain, cerebellum and gonads (Maroc et al, 1993). Interaction with its ligand (FL) results in receptor dimerization, autophosphorylation and the subsequent phosphorylation of cytoplasmic substrates that are involved in signalling pathways regulating the proliferation of pluripotent stem cells, early progenitor cells and immature lymphocytes (Lyman, 1995). This interaction is influenced by other cytokines such as Kit ligand (KL). In fact, when primitive human progenitor cells are stimulated in vitro with either FL or KL alone, they show little or no proliferative response, but both ligands together synergistically enhance growth (Hannum et al, 1994). Further evidence for the importance of FLT3 in early haemopoiesis has come from FLT3 knockout mice. They are healthy with normal peripheral blood counts, but have reduced numbers of bone marrow early B-cell precursors, plus a defect in primitive cells as measured by long-term competitive repopulation assays and a reduced ability to reconstitute B-cell, T-cell and myeloid lineages when transplanted into irradiated hosts (Mackarehtschian et al, 1995).
Flt3 expression in leukaemias
The expression of FLT3 has been reported at the mRNA and/or protein level in 93% of AML patients, 87% of T-cell acute lymphoblastic leukaemia (ALL) and up to 100% of B-cell ALL patients (Drexler, 1996). It has not been detected in chronic myeloid leukaemia in the chronic phase, but appears to be highly expressed during disease transformation, irrespective of the phenotype (Birg et al, 1992). In leukaemic cell lines, it is expressed in ≈ 90% of preB-cell lines, but less frequently (40–80%) in myeloid and monocytic cell lines (Drexler, 1996). Studies have also demonstrated very high levels of FLT3 mRNA and protein in AML patients, leading to the suggestion that its overexpression may play a role in the survival and proliferation of leukaemic cells (Carow et al, 1996). This is in line with the finding that FL can induce proliferation in some FLT3-expressing primary leukaemic cells and cell lines (Dehmel et al, 1996). Furthermore, the chronic exposure of mice to FL by transplantation with primary haemopoietic cells constitutively expressing the FL gene has been shown to induce leukaemia with a long latency period, indicating a possible autocrine mechanism for maintaining the survival of a leukaemic clone (Hawley et al, 1998).
Flt3 mutations in leukaemia and their biological consequences
Internal tandem duplications (ITDs)
The first FLT3 mutations to be identified were serendipitously detected during an investigation into the incidence and distribution of FLT3 mRNA in samples from adult AML and childhood ALL patients (Nakao et al, 1996). Unexpectedly long fragments were detected in the polymerase chain reaction (PCR) products of the JM domain in five out of 30 AML patients. They were also found using genomic DNA from the same patients, excluding the possibility of aberrant alternative splicing. Further analysis showed that they all contained a tandemly duplicated sequence, sometimes with insertion of additional nucleotides. The duplicated region was variable in both size and location in different individuals but always fell within the JM domain encoded by exons 14 and 15 (previously 11 and 12) (Fig 1). The resulting transcripts were always in frame and would therefore be expected to produce functional FLT3 chains. Since that first description, numerous other studies have confirmed and extended these findings to the extent that FLT3/ITDs are currently the most frequent single mutation described in AML, with a reported incidence between 13·2% and 32% in adult patients (Table I). They have also been detected in 3% of patients with myelodysplastic syndromes (Horiike et al, 1997; Yokota et al, 1997; Xu et al, 1999) and occasional patients with ALL (Xu et al, 1999; Nakao et al, 2000), although some of the latter patients had biphenotypic characteristics. They have not been found in patients with chronic myeloid leukaemia, chronic lymphoid leukaemia, non-Hodgkin's lymphoma or multiple myeloma (Yokota et al, 1997), or in normal individuals (Ishii et al, 1999; Kottaridis et al, 2001).
Table I. Incidence and biological features in adult AML patients with a FLT3/ITD mutation.
No. of patients
Incidence of FLT3/ITD (%)
Association with FAB type
Frequency in different cytogenetic subgroups
Associated clinical features
All French–American–British (FAB) classification type M3.
All intermediate and favourable risk disease.
All normal karyotype.
BCR3, breakpoint cluster region 3; LDH, lactate dehydrogenase; BM, bone marrow; PB, peripheral blood.
Preliminary in vitro analysis of FLT3/ITDs transfected into Cos7 cells showed that they induced ligand-independent receptor dimerization and phosphorylation, irrespective of the location and length of the ITD, and led to phosphorylation of wild-type (WT) FLT3 expressed in the same cell (Kiyoi et al, 1998). They have been shown to confer growth factor independence on factor-dependent cell lines such as Ba/F3 and 32D cells, and to induce constitutive activation of downstream signalling molecules such as signal transducer and activation of transcription 5 (STAT5), mitogen-activated protein kinase (MAP kinase), Akt, Src homology 2 domain-containing (SHC) transforming protein 1, Cbl, Vav and SH2-containing protein tyrosine phosphatase 2 (SHP2) (Hayakawa et al, 2000; Mizuki et al, 2000; Kiyoi et al, 2002; Tse et al, 2002). However, comparable constitutive activation is not always observed in primary leukaemic blasts. Of 27 AML samples studied by Fenski et al (2000), 18 had ligand-dependent FLT3 phosphorylation, and three of these had a FLT3/ITD. Conversely, three samples had constitutive FLT3 phosphorylation, but only one had a FLT3/ITD. Similarly, Birkenkamp et al (2001) reported that, of 12 samples with both FLT3 and STAT5 constitutive phosphorylation, only eight had a FLT3/ITD and, of five samples with constitutive STAT5 but no FLT3 phosphorylation, two had a FLT3/ITD. It is likely that, at least to some extent, these results indicate the redundancy of signalling pathways within the cell and the multiple ways in which they can become activated.
A striking feature of the FLT3/ITDs identified in AML patients is their diversity in both size and location. In general, the length of the duplication varies between 12 and 204 bp, but it has been reported to be as short as 3 bp and as long as > 400 bp (Schnittger et al, 2002a). Most ITDs occur at the 5′ end of the JM domain, in exon 14, and the TK1 domain is minimally involved, but they differ in the precise sequence duplicated and its starting point. Several inserted sequences of unknown origin varying between 9 and 36 bp have also been detected (Kiyoi et al, 1998; Frohling et al, 2002; Thiede et al, 2002a). There is no sequence that is common to all reported duplications, but the involved region does contain a tyrosine-rich stretch of sequence, and most ITDs include at least one of the tyrosines 589, 591, 597 or 599, which form part of the motifs YFYV and YEYDLK. As the latter are homologous to autophosphorylation sites in the JM domains of other TK receptors, e.g. the platelet-derived growth factor beta receptor (PDGFβR), it was initially thought that the mechanism by which duplication might lead to enhanced growth was through the gain of additional Src homology binding domain-2 (SH2) binding domains (Yokota et al, 1997). However, neither substitution of all four tyrosines with phenylalanine nor deletion of between one (ΔY599) and four (ΔY589–599) tyrosines alters the in vitro constitutive phosphorylation state and ability to induce growth factor independence of a FLT3/ITD, although these residues are clearly important for ligand-dependent activation of the WT receptor (Kiyoi et al, 2002). Instead, current models suggest that the JM domain has a negative regulatory role that is disrupted by the ITD elongation (Gilliland & Griffin, 2002; Kiyoi et al, 2002). In the WT receptor, the JM domain takes up an α-helical conformation, which blocks activation of the kinase and may inhibit self-dimerization. Ligand binding overcomes the inhibitory effect by inducing a conformational change and/or phosphorylation of key tyrosine residues. The ITDs may therefore prevent the protective association between the JM domain and kinase, exposing the latter to constitutive activation. They may further allow recruitment of molecules that could stabilize this conformation or alter downstream signalling.
The in vivo tumorigenic potential of a FLT3/ITD has been demonstrated by injection of 32D cells carrying a FLT3/ITD into syngeneic mice, which led to the rapid development of a leukaemia-like disease (Mizuki et al, 2000). Furthermore, AML cells with an ITD showed an increased ability to repopulate bone marrow in non-obese diabetic (NOD)/severe combined immunodeficient (SCID) mice (Rombouts et al, 2000). However, although transplantation of bone marrow cells retrovirally transduced with FLT3/ITDs into recipient mice led to an oligoclonal myeloproliferative disorder, it was insufficient to cause leukaemia (Kelly et al, 2002a). This indicates the requirement for additional co-operating mutations for a fully transformed phenotype in this model. Further evidence for this co-operation has come from transplantation of cells carrying a FLT3/ITD into PML/RARα transgenic mice, which considerably shortened the latency and increased the penetrance for developing an APL-like disease (Kelly et al, 2002b). It can be hypothesized that, in such murine models, only ‘two hits’ are required to generate leukaemia, the mutation in the transcription factor (e.g. PML/RARα) producing a block in differentiation, and the mutant growth factor receptor providing a proliferative or survival signal (Deguchi & Gilliland, 2002).
Two recent publications have presented data suggesting that FLT3/ITD mutations may also contribute to a block in differentiation. First, the expression of a FLT3/ITD in 32D cells inhibited the granulocyte colony-stimulating factor (G-CSF)-induced expression of myeloid maturation markers such as myeloperoxidase, lysozyme and CCAATT/enhancer binding protein ε (C/EBPε) (Zheng et al, 2002). Secondly, microarray expression analysis of 32D cells transfected with either WT or ITD FLT3 demonstrated that certain markers associated with myeloid differentiation, e.g. Spi-1 (PU-1) and C/EBPα, were suppressed in FLT3/ITD-expressing cells (Mizuki et al, 2003). It should be noted, however, that reduced differentiation is frequently a feature of an increased proliferation rate. In the factor-dependent cell progenitors (FDCP)-mix cell line, for instance, high concentrations of interleukin 3 (IL-3) block or reduce the usual differentiation induced by other growth factors (Heyworth et al, 1990).
Mutations in the second tyrosine kinase domain (TKDs)
Mutations at aspartic acid 835 (D835) and isoleucine 836 (I836) in exon 20, in the second TK domain, were first independently identified in AML patients by Yamamoto et al (2001) and Abu-Duhier et al (2001b) and have now been reported by many other groups (Table II). They include at least six different substitutions within the D835 codon leading to missense mutations, predominantly tyrosine and histidine, less frequently valine, glutamate and asparagine, and mutation of I836 to methionine (Fig 1). The complete deletion of I836, insertion of nucleotides and complex changes have also been detected, but the sequence always remains in frame. The incidence varies between 6·4% and 7·7% in unselected adult AML patients (Table II). Occasional cases have been reported in ALL (2·8%) and MDS (3·4%) (Yamamoto et al, 2001). The mutations were found to cause constitutive tyrosine phosphorylation of the receptor when transfected into Cos 7 cells, and to confer IL-3-independent growth on 32D cells, indicating their gain-of-function property (Yamamoto et al, 2001). The amino acids are part of the activation loop that blocks the access of ATP and the substrate to the kinase domain when the receptor is in an inactive state. Ligand-induced activation leads to phosphorylation within the loop causing it to take up an active configuration and allow kinase activity. The TKD mutations are thought to mimic the latter by interfering with the inhibitory effect of the loop, resulting in constitutive kinase activation. Their effect is therefore similar to that of FLT3/ITD mutations in disrupting the autoinhibitory mechanisms that normally protect the cell from unregulated signalling through FLT3. However, possible differences in the functional consequences of the various TKD mutations have not been investigated.
Table II. Incidence and biological features in AML patients with a FLT3/TKD mutation.
No. of patients
Incidence of mutant TKD (%)
Association with FAB type
Frequency in different cytogenetic subgroups
Associated clinical features
All normal karyotype.
All French–American–British (FAB) classification type M3.
A further constitutively activating FLT3 mutation in exon 20 has been reported recently by Spiekermann et al (2002), who detected an additional 6 bp leading to the insertion of glycine and serine between amino acids 840 and 841 in two out of 359 (0·5%) de novo AML patients. Functional analysis demonstrated that the mutant receptor was constitutively phosphorylated and induced factor independence in Ba/F3 cells. Like the D835/I836 mutations, it is probable that this mutation also interferes with the inhibitory role of the activation loop.
Frequency of flt3 mutations in aml patients and their biological features
The overall frequency of FLT3/ITDs reported in adult patients, irrespective of age and French–American–British (FAB) or cytogenetic classification, is ≈ 24%, with a range between 13·2% and 34%. A common finding is the association between a FLT3/ITD and leucocytosis, and an increased percentage of blast cells in the peripheral blood and bone marrow (Table I). Increased lactate dehydrogenase (LDH) levels have been reported in two studies, one in FAB type M3 (Kiyoi et al, 1997) and the other in patients with a normal karyotype (Frohling et al, 2002), but this has not been widely evaluated. Although found across the spectrum of FAB types, certain associations have been observed. Several studies have reported a high frequency in AML M3, in particular variant M3, and in M5, especially M5b; conversely, low frequencies have been found in M6 and M7 (Kottaridis et al, 2001; Noguera et al, 2002; Schnittger et al, 2002a; Thiede et al, 2002a) (Table I). Although studies of paediatric patients are much smaller, the incidence is generally lower, being ≈ 12% overall excluding patients with M3, range 4·3–22·2%(Table III). As with adult patients, a FLT3/ITD is associated with leucocytosis, and the frequency is higher in M3, particularly variant M3 (Kondo et al, 1999; Arrigoni et al, 2003).
Table III. Incidence, biological features and clinical outcome in paediatric AML patients with a FLT3/ITD mutation.
Association between FLT3 mutations and other acquired abnormalities
The association of FLT3 mutations with specific cytogenetic categories or other acquired mutations is of considerable biological interest as it may reveal clues to co-operating abnormalities leading to the transformed phenotype. As noted above, a high frequency of FLT3/ITDs (30–39%) has been observed in patients with t(15;17) and, when combined with TKD mutations, nearly half the APL patients studied carry a FLT3 mutation (Noguera et al, 2002; Shih et al, 2002a; D. Grimwade and colleagues, unpublished observations). The ITDs in particular appear to be associated with M3 variant morphology (65–80% are ITD+), together with hyperleucocytosis and the presence of the short breakpoint cluster region 3 (BCR3) PML/RARα isoform (Gallagher et al, 2002; Noguera et al, 2002; Palumbo et al, 2002; Shih et al, 2002a; Arrigoni et al, 2003; D. Grimwade and colleagues, unpublished observations), indicating that these features may have a common causative pathology. A recent study found no difference in the incidence of FLT3/ITDs between patients with or without a C/EBPα mutation (Preudhomme et al, 2002).
A possible association of FLT3 mutations with abnormalities of the MLL gene has also been noted. Although MLL tandem duplications (MLL/TDs) are only observed in 3–4% of de novo AML patients (Schnittger et al, 2000), co-duplication of both MLL and FLT3 genes was first noted in two AML patients out of just 13 studied by Jamal et al (2001). Frohling et al (2002) found no difference in FLT3 mutant status (WT, ITD or TKD) between MLL/TD patients. However, two studies reported in abstract form have suggested that there may be an association; Steudel et al (2002) found a FLT3/ITD in 33% (16/48) MLL/TD-positive patients; Libura et al (2002) found a FLT3 mutation (either ITD or TKD) in 50% (8/16) of patients with an MLL/TD and in 63% (5/8) with a topoisomerase II break, but in only 14% (6/42) of patients with an MLL translocation. Both duplications are more frequent in patients with a normal karyotype. The mechanism of duplication of each gene is likely to be different, for example there is no evidence that Alu-mediated homologous recombination, thought to be responsible for the generation of MLL/TDs (Strout et al, 1998), plays a role in the production of FLT3/ITDs; nevertheless, their joint presence may indicate exposure to genotoxic agents leading to DNA breakage (Libura et al, 2002). Of interest, although FLT3/ITDs are uncommon in AML patients with 11q23 abnormalities, a high level of expression of FLT3 mRNA has been reported in ALL patients with MLL translocations (Armstrong et al, 2002), and a high level of TKD mutations (20%) has been found in 40 infant ALLs with an MLL rearrangement (Taketani et al, 2002).
Kiyoi et al (1999) were the first to conduct a large study of 201 adult patients with de novo AML, excluding M3, and to demonstrate an adverse outcome in those with a FLT3/ITD. Furthermore, multivariate analysis showed that a FLT3/ITD was the most significant prognostic factor for predicting survival in patients below the age of 60 years, followed by cytogenetics. Since then, many studies have evaluated the impact of these mutations on clinical outcome.
As a general rule, the presence of an ITD in adult patients seems to have little or no impact on the ability to achieve complete remission (CR), and only one smaller study found a slightly reduced CR rate (Table IV). In children, however, reduced CR has been reported in three studies (Kondo et al, 1999; Meshinchi et al, 2001; Zwaan et al, 2002) (Table III). The most significant impact of an ITD is its association with increased relapse risk (RR), decreased disease-free survival (DFS) and overall survival (OS), which has been reported in most studies of children and adults less than 60 years of age (Tables III and IV). In addition to Kiyoi et al (1999), several other groups have found that an ITD is the most significant factor predicting an adverse outcome in multivariate analysis (Kottaridis et al, 2001; Meshinchi et al, 2001; Frohling et al, 2002). The incidence of a mutation in infantile AML is extremely low (Xu et al, 2000) and, therefore, larger series will be needed to address the potential clinical impact in this group. In a single study of older patients (> 55 years of age), a FLT3/ITD had no impact on clinical outcome (Stirewalt et al, 2001), although this is not surprising as patients in this group generally have unfavourable disease, with an OS at 5 years no greater than 10%, and any effect of the mutation would be difficult to detect.
Table IV. Clinical outcome in adult AML patients with a FLT3/ITD mutation.
No. of patients
All normal karyotype.
All de novo AML, < 60 years and intermediate risk cytogenetics.
All intermediate risk cytogenetics.
All FAB (French-American-British Classification) type M3.
Two studies are of particular interest as they failed to show a FLT3/ITD as an independent prognostic factor predicting for disease outcome. In a study of 640 patients, Thiede et al (2002a) found that, despite an increased RR and reduced DFS, there was only a trend for reduced OS with an ITD (P = 0·09). Similarly, Schnittger et al (2002a) reported that an ITD affected EFS in their study of 563 patients, but not OS. The difference in outcome in these two studies may have been influenced by a number of factors, such as the short follow-up of less than 1 year and the heterogeneity of patients studied with respect to age and the type of disease (de novo and secondary AML and antecedent MDS). They do, however, raise the possibility that different treatment regimens might have contributed towards a better outcome in their patients, especially the inclusion of very high-dose cytarabine. In our study of patients treated according to the UK Medical Research Council (MRC) protocols with a cumulative dose of cytarabine up to 11·6 g/m2, a FLT3/ITD was highly predictive of increased RR, reduced DFS and OS (Kottaridis et al, 2001). In contrast, in the two studies mentioned above, the cumulative cytarabine dose was up to eightfold higher, ranging from 22·4 to 92·8 g/m2 (Schnittger et al, 2002a; Thiede et al, 2002a). Nevertheless, this may depend on the subgroup of AML, as Frohling et al (2002) found that intensive chemotherapy, including post-remission cytarabine up to doses of 42 g/m2, was of no benefit in FLT3/ITD patients with normal cytogenetics. This group also fared badly in the study by Boissel et al (2002) comparing three different reinforced induction regimens, although they did find some suggestion of drug intensification improving outcome in FLT3/ITD patients. It is clear that evaluation of treatment protocols in patients with a FLT3/ITD requires further investigation, preferably prospectively within randomized trials, in order to refine further risk-adapted therapy.
Clinical outcome in relation to the number and level of mutation(s). In the one study examining the presence of more than one FLT3/ITD, this appeared to increase further the adverse outcome of ITDs in AML and was associated with a trend towards a higher white count and percentage of bone marrow blasts at diagnosis (P = 0·07) and a worse OS (P = 0·04) (Kottaridis et al, 2001). However, several studies have shown that patients with a proportionally high level of mutant compared with WT FLT3 have a worse outcome (Kottaridis et al, 2001; Whitman et al, 2001; Thiede et al, 2002a). These patients have higher white counts and percentage of bone marrow blasts and a shorter OS and DFS. Two of the studies, on patients with normal cytogenetics or intermediate risk disease, found that only the loss of WT FLT3 was an independent prognostic factor for worse outcome, not the presence of a FLT3/ITD per se (Whitman et al, 2001; Thiede et al, 2002a). The incidence of patients with loss of the WT allele in this cytogenetic group does, however, vary in different studies. Whitman et al (2001) found that, in 35% of their ITD+ patients (8/23), the level of mutant was greater than WT, whereas Frohling et al (2002) found this in only 1·4% (1/71) of their ITD+ patients. To some extent, this may reflect where the cut-off value for loss of WT allele has been set; nonetheless, the functional significance of these findings is of interest. As a mutant FLT3 in a mutant/WT heterodimer can trans-phosphorylate the WT chain (Kiyoi et al, 1998, 2002), it implies that a mutant homodimer has some gain-of-function more than simply activating the kinase. Alternatively, formation of the homodimer may reflect an underlying mechanism of genetic instability that has other unknown genomic consequences that may, in turn, influence clinical outcome.
In general, studies analysing the presence of a FLT3/TKD mutation alone, in the absence of an ITD, have found no significant impact on clinical outcome for either remission rate or survival (Abu-Duhier et al, 2001b; Yamamoto et al, 2001; Frohling et al, 2002; Schnittger et al, 2002b) (Table V). Thiede et al (2002a) found that TKD mutations were associated with a trend for a reduced OS and DFS for their entire group of patients (P = 0·09), but this was statistically significant only for OS in younger patients (≤ 60 years) with intermediate risk disease (P = 0·015). The low frequency of TKD+ patients means that only small numbers (maximum 36) have been evaluated in these studies and therefore, at present, the true prognostic significance of the mutations remains unclear. Furthermore, different mutations may have widely differing outcomes. The results are, however, intriguing as they possibly imply a difference in the clinical outcome between two events, ITDs and TKD mutations, that supposedly have the same functional consequence, i.e. loss of autoinhibitory regulation leading to constitutive FLT3 kinase activation. This might reflect different levels of constitutive activation leading to differential activation of specific signalling pathways, or indicate that the ITD is just a marker for a genomically unstable cell carrying other (unknown) dysregulating mutations.
Table V. Clinical outcome in adult AML patients with a FLT3/TKD mutation.
No. of patients
All normal karyotype.
All de novo AML, < 60 years and intermediate risk cytogenetics.
Several studies have now been published investigating FLT3 mutations in paired presentation and relapse samples (Nakano et al, 1999; Hovland et al, 2002; Kottaridis et al, 2002a; Schnittger et al, 2002a; Shih et al, 2002b) (Table VI). Their impetus was to evaluate the potential use of FLT3 mutations for markers of minimal residual disease (MRD), particularly important in patients with normal cytogenetics, but the results have raised important questions regarding the role of FLT3 mutations in the pathogenesis of AML. Most patients (86% overall from three of the studies) maintained the same FLT3 status at presentation and relapse. Patients that were FLT3/ITD+ at presentation often showed an increased mutant level at relapse, usually with evidence for the loss of WT alleles and, where more than one mutation had been detected at presentation, usually only one was predominant at relapse (Kottaridis et al, 2002a; Schnittger et al, 2002a; Shih et al, 2002b). However, a significant proportion of patients either gained (10·1%) or lost (3·9%) a FLT3 mutation at relapse, and several patients have been reported with loss of the presentation ITD and gain of a completely different mutation at relapse. These results not only restrict the use of these mutations as markers of MRD, they also imply that FLT3 mutations are secondary events, arising in an already transformed clone, which induce the outgrowth of a subclone as a result of the additional proliferative or survival advantage they confer. Further support for this secondary role of FLT3 mutations comes from the quantitative analysis of FLT3/ITDs, indicating that the mutation is frequently only present in a subpopulation of the leukaemic cells. For example, in our own study, mutant levels ranged between 0·5% and 97% of total FLT3, and 64% of patients had < 40% mutant (Kottaridis et al, 2001). In the majority of AML patients, it thus appears that FLT3 mutations are a late event, and this is not analogous to the ‘two hit hypothesis’ derived from murine models.
Table VI. Studies on FLT3 status in paired presentation and relapse samples from AML patients.
No. of patients
FLT3 status at presentation/relapse
* One patient in each study was ITD+ at both presentation and relapse but with different mutations.
A small study has been reported on the outcome of therapy in 68 relapsed patients (Boissel et al, 2002). Although FLT3 status was not evaluated at the time of relapse, survival in the 17 patients who had been FLT3/ITD+ at presentation was significantly worse than in the 51 FLT3/WT patients (0% versus 27% at 3 years, P = 0·04). Whether the presence of FLT3 mutations at diagnosis renders patients unsalvageable at relapse with current therapy should therefore be explored in larger cohorts of patients.
Treatment of patients with flt3 mutations
Bone marrow or stem cell transplantation
The optimal treatment for patients with what appears in general to lead to adverse outcome is currently unknown, and alternative strategies such as bone marrow or stem cell transplantation and more experimental therapies require evaluation. Although the benefit of transplantation in standard or poor risk disease remains unclear, there is some indication that, overall and within risk groups, it leads to a reduction in RR (Burnett, 2002). Preliminary data from our own group on 293 patients who received a transplant in first CR found that it did not negate the poor prognostic outcome of a FLT3/ITD (Kottaridis et al, 2002b). This was more obvious in the autologous transplant group, where the relapse rate was 51% versus 32% in ITD-positive and -negative patients respectively. In the allogeneic group, the analysis was confounded by an unexplained, high level of transplant-related mortality in the recipients who had a FLT3/ITD. However, there was a suggestion that the relapse rate was no higher in FLT3/ITD patients than in FLT3/WT patients. Thiede et al (2002b) examined outcome in 175 transplanted patients and found no difference in the OS and DFS between FLT3/ITD-positive and -negative patients in the autologous and unrelated allogeneic groups, but an increased rate of relapse in the FLT3/ITD-positive patients receiving a related donor transplant. This issue clearly requires further investigation as it is important to determine whether high-dose therapy or a graft-versus-leukaemia effect can overcome the poor prognostic significance of a FLT3/ITD.
Tyrosine kinase inhibitors
The success of the TK inhibitor imatinib mesylate (STI571) in the treatment of chronic myeloid leukaemia demonstrated the potential for therapy targeted at aberrant leukaemia-specific gene products. Although STI571 does inhibit FLT3, the concentration required to achieve this is toxic (Druker & Lydon, 2000). Consequently, considerable activity is being directed towards developing suitable small-molecule inhibitors that specifically target mutant FLT3 (Sawyers, 2002). To date, at least nine candidate compounds have been studied using the in vitro analysis of cells carrying FLT3/ITDs and in vivo treatment of mice transplanted with cells carrying a FLT3/ITD (Table VII), and a few have entered clinical trials.
Table VII. Studies on FLT3-targeted tyrosine kinase inhibitors.
Tyrphostins were the first compounds to be developed in the late 1980s with specificity for TK inhibition (Yaish et al, 1988), and two, AG1295 and AG1296, have been shown to be specifically cytotoxic for transfected cells and primary AML cells harbouring a FLT3/ITD (Levis et al, 2001; Tse et al, 2001, 2002). The effect was dose dependent, and maximal killing of ≈ 40% (AG1295) was observed. In Ba/F3 cells transfected with an ITD, AG1296 inhibited autophosphorylation of the receptor and suppressed constitutive phosphorylation of downstream signalling targets such as STAT5A, STAT5B and extracellular signal-related kinase (ERK) (Tse et al, 2002). The indolinone compounds, SU5416 and SU5614, have been shown selectively to induce growth arrest and apoptosis of Ba/F3 and AML cell lines expressing a constitutively activated FLT3 and to inhibit phosphorylation of FLT3 and downstream signalling targets (Yee et al, 2002; Spiekermann et al, 2003). Similarly, in vitro studies showing potent FLT3 inhibition have been reported for two bis(1H-2-indolyl)-1-methanone derivatives, D-64406 and D-65476 (Teller et al, 2002). In vivo efficacy of several compounds has also been reported using mouse models of transplanted leukaemia. The indolocarbazole derivative CEP-701, the piperazinyl quinazoline compound CT53518 and the protein kinase inhibitor PKC412 have all been shown to prolong the disease latency period, suppress tumour growth and give significant survival benefit (Kelly et al, 2002c; Levis et al, 2002; Weisberg et al, 2002).
These results have been sufficiently encouraging to indicate the potential clinical benefit of these inhibitors, and several are currently being tested as single agents in phase I and II trials in patients with relapsed/refractory AML. Overall, the compounds appear to be well tolerated, and reductions in blast cell count have been observed (Mesters et al, 2001; Stone et al, 2002). Over the next year, some indication of their potency should become apparent. Nevertheless, as FLT3 mutations are clearly secondary abnormalities in at least some patients and are often restricted to a subpopulation of leukaemic cells, it is likely that such targeted therapy will need to be carried out in combination with proven cytotoxic regimens.
The presence of FLT3 mutations in such a high proportion of patients with AML indicates the important role of this receptor in haemopoietic progenitor cells and the potential disruption caused by dysregulated tyrosine kinases. The mutations are useful molecular markers as they can be detected by straightforward techniques, requiring only the PCR and gel electrophoresis for the ITDs to identify longer length products, and the inclusion of a restriction digest for the TKD mutations. Routine screening of AML patients for diagnostic purposes is therefore feasible. The optimal treatment for such patients is unclear and requires careful evaluation in prospective trials, in particular whether patients with an ITD should be considered for early transplantation or more experimental therapy. Whether the development of FLT3-specific small-molecule inhibitors will lead to an improved outcome in these patients must await the results of large randomized trials.