Acute myeloid leukaemia (AML) is a molecularly, morphologically and phenotypically heterogeneous disease. Multiple recurrent chromosome translocations and gene mutations have been identified, and these alterations have been correlated to biological and clinical features of disease, resulting in delineation of prognostically distinct categories of AML.
Acute myeloid leukaemia would result from several classes of mutations: a class I mutation, which confers a proliferative signal (for example, a receptor tyrosine kinase (RTK)), and a class II mutation, which impairs haematopoietic differentiation, such as the core binding factor fusion genes. Finally, RTK class III mutations have been also linked to other haematological malignancies.
c-Kit is an important member of type III RTK: its activation is important in erythropoiesis, lymphopoiesis and megakaryopoiesis. Indeed, c-Kit acts as a key signalling molecule for a number of cell types, including haematopoietic stem cells and mast cells. Gain-of-function mutations in c-Kit have been described in 86% of erythroleukaemia and in 17–46% of other AML subtypes, especially those carrying core binding fusion genes, where they predict a worse outcome.1 The c-Kit gene contains 21 exons, and two alternative splicing sites have been described. One of these alternative splicing sites is located at the 3 end of exon 9 and results in the expression of two isoforms, designated GNNK− (Kit) and GNNK+ (KitA). The two isoforms have different biological activities and the GNNK− isoform is the main transcript in several human malignant tumours.2
In this retrospective study, we quantitatively tested the expression of the Kit and KitA isoforms in 63 adult patients affected by AML evaluated at the diagnosis, in order to evaluate a possible prognostic role of these isoforms in terms of complete remission rate, overall survival (OS) and progression-free survival (PFS).
Sixty-three bone marrow samples from patients affected by AML (35 men and 28 women), with a median age of 45 years (range 19–72 years), were tested at baseline. Cases were selected on the basis of the availability of bone marrow samples at time of diagnosis and of bone marrow blast percentage higher than 80%. All cases were classified according to FAB criteria, excluding acute promyelocytic leukaemia. CD34 and other myeloid surface markers (CD33, CD13, CD14, HLA-Dr) were analyzed by flow cytometry (FACScan, Becton Dickinson, Mountain View, CA, USA). Cytogenetic analysis was performed on bone marrow aspirates following the recommendations of the International System for Human Cytogenetic Nomenclature. Risk groups were defined as follows: (i) low risk: t(8;21) and inv(16); (ii) intermediate risk: patients with normal karyotype or without abnormalities classified as at low or high risk; (iii) high risk: −5/del(5q),−7/del(7q), abn (abnormality) 12p, abn 11q, +11, +13, +21, +22, t(6;9), t(9;11), t(3;3), hypodiploid karyotypes or more than three independent aberrations. All cases were also screened for FLT3/ITD and FLT3-D835 mutations, and 13 cases also for the NPM1 mutations, as previously reported by our groups.3, 4
Patient characteristics are shown in Table 1. Eleven patients received idarubicin, 35 doxorubicin, 12 daunorubicin plus etoposide and 5 mitoxantrone in the induction therapy. Fourteen cases underwent autologous and 17 allogeneic stem cell transplantation; these cases have been censored before transplantation procedure. Complete response (CR), partial response and resistance were defined according to the bone marrow percentage of <5, 5–25 and >25 of residual blast cells after induction therapy, respectively.
The expression level of the two isoforms of c-Kit was detected by fluorescent PCR amplification using specific primers flanking the splicing site, to amplify both the isoforms at the same time.5 In the amplification protocols, 28 amplification cycles were performed, in order to avoid the achievement of the plateau phase. To detect the PCR FAM-labelled products, capillary electrophoresis was performed on an ABI Prism 310 apparatus, according to the manufacturer's instructions. Size curves and fluorescence intensity were analyzed and quantified using the Genescan software (Applied Biosystems, Milan, Italy).
Two peaks were detected in 40 AML cases (63.5%); the positions of these peaks corresponded to the expected lengths of both isoforms of the c-Kit transcript. The median GNNK−/+ ratio was 5.19±3.93 (range=0.34–19.66). Twenty-three cases (36.5%) showed only the GNNK− expression. Results were at first computed after clustering AML cases as ‘low’ and ‘high’ on the basis of median GNNK−/+ ratio. Clinical features recorded at diagnosis (FAB subtype, immunophenotype, WBC and PLT counts, Hb levels, blast percentage, age, sex, cytogenetic risk) did not differ between the two subgroups.
Overall, 34 cases (54%) achieved a complete remission, 14 (22.2%) a partial remission and 10 (15.8%) were resistant, with 5 (8%) deaths during the induction treatment. The CR rate was significantly conditioned by the cytogenetic profile (31.3% for cases at high risk, 68.8% for those with standard risk vs 100% of cases at low risk, P=0.01). Analogously, age >60 years negatively influenced the achievement of CR (17 vs 63% for the younger individuals, P=0.04). On the contrary, the quality of response was not significantly influenced by the GNNK−/+ ratio.
The 3-year OS of the entire series was 28%. The advanced age and the achievement of complete remission after induction phase (chemosensitive disease) were the factors really able to influence length of survival (P=0.003 and 0.0001, respectively). On the contrary, OS was not different between cases with low and high GNNK−/+ ratio. The 3-year PFS was 34%; it was significantly influenced by the advanced age and the achieved haematological remission after induction therapy (P=0.024, P=0.0001). GNNK−/+ ratio values did not condition the PFS length.
The same analyses were then performed clustering the 63 cases in ‘one isoform’ (GNNK−) and ‘two isoforms’ (GNNK− plus GNNK+) carriers. Complete remission rate was not different between the two groups; nevertheless, patients showing the GNNK− isoform alone presented both prolonged 3-year OS (57 vs 13%, P=0.003) and PFS (58 vs 17%, P=0.016) (Figure 1). The advantage in the OS retained its statistical significance in the younger cohort (57% for cases GNNK− vs 19% for cases GNNK− plus GNNK+, P=0.04). The statistical significance was lost when PFS was considered, notwithstanding a persistent trend to significance (58% for cases with GNNK− vs 25% for cases showing both GNNK− plus GNNK+, P=0.12). However, the absence of GNNK+ did highly condition both OS and PFS in the high cytogenetic risk subgroup (3-year OS 100% for cases GNNK− vs 6% for those carrying both isoforms, P=0.01; PFS 66 vs 8%, P=0.05).
Finally, all data about c-Kit (both ratio and GNNK+ absence) were analyzed in comparison with FLT3 and NPM1 mutations. In the whole series, 24% of cases showed an FLT3 mutation (65% FLT3-ITD and 35% D835); both the GNNK−/+ ratio and the GNNK+ absence did not correlate with the FLT3 mutational status. Four cases were NPM1-mutated and nine wild type; notwithstanding the small number of tested cases, in our series the c-Kit status did not correlate with NPM1 mutations.
In conclusion, in this study, we demonstrated that a PCR assay detecting the two c-Kit GNNK isoforms represents a cheap, feasible and rapid molecular test able to identify AML cases with a worse prognosis. Our results show that this finding is independent of other already known prognostic factors, such as advanced age, high cytogenetic risk, chemoresistance, as shown also by the multivariate analysis (advanced age: P=0.41, high cytogenetic risk P=0.06, not CR P=0.0001, GNNK+ detection P=0.001).
The oncogenic potential of the two c-Kit isoforms is still a matter of debate: Caruana et al.2 did not observe difference in ligand-binding affinity between the two isoforms. On the other hand, other authors demonstrated a more rapid internalization and a ubiquitin-mediated degradation of the GNNK− form.6 If this observation implies a more rapid switch-off of the c-Kit activity, this phenomenon could be in accordance with our observations. The role of c-Kit in AML is particularly appealing for therapeutic options, considering the availability of several c-Kit inhibitors, such as imatinib and sorafenib.
Low-dose Ara-C plus imatinib did not appear to be inferior in older AML patients in comparison with historic controls receiving myelosuppressive therapy, with median OS of 138 days and 20% of patients alive after 600 days.7
On the other hand, sorafenib (BAY 43-9006, Nexavar, Bayer, Germany) is a multikinase inhibitor with activity against Raf kinase and several receptor tyrosine kinases, including vascular endothelial growth factor receptor 2, platelet-derived growth factor receptor, FLT3, Ret and c-Kit, the D816V mutants excluded. Pronounced apoptosis was recently observed in blasts from patients with acute myeloid leukaemia.8
On the basis of the above reported data, it would be useful to screen all AML patients for GNNK− and GNNK+ expression, in order to design ab initio a more aggressive therapeutic strategy, perhaps including transplantation options, for cases carrying both GNNK− and GNNK+ c-Kit isoforms.