A selective inhibitor of the p110delta isoform of PI 3-kinase inhibits AML cell proliferation and survival and increases the cytotoxic effects of VP16.

Current therapy for acute myeloid leukaemia (AML) is suboptimal with a high incidence of relapse. There is strong evidence that constitutive phosphoinositide 3-kinase (PI3K) activity plays a significant role in the pathophysiology of AML. PI3K products are derived from the activity of a number of PI3K catalytic isoforms (class I, II and III) but the relative contribution of these enzymes in AML remains unknown. As non-isoform-selective inhibitors of PI3K such as LY294002 may produce unwanted toxicity to normal tissues, we have investigated the role of the leukocyte-restricted p110delta PI3K isoform in 14 cases of AML. p110delta was detected in all cases whereas the expression levels of the other class I PI3Ks varied more widely, and were often undetectable. The p110delta-selective compound IC87114 inhibited constitutive phosphorylation of the PI3K target Akt/PKB and reduced cell number to a mean of 66+/-5% (range 14-88%). In eight cases, the combination of IC87114 and VP16 (a topoisomerase II inhibitor) was synergistic in reducing viable cell number, and was associated with a reduction in constitutive NF-kappaB activity. IC87114 did not have direct adverse effects or enhance the activity of VP16 on the proliferation and survival of normal haemopoietic progenitors. Overall, our results identify the p110delta isoform as a potential therapeutic target in AML and support a clinical approach to use isoform-selective over broad-spectrum PI3K inhibitors.

PI3-kinase/Akt is constitutively active in primary acute myeloid leukaemia cells and regulates survival and chemoresistance via NF-kappaB, Mapkinase and p53 pathways.

The phosphoinositide 3-kinase (PI3-kinase) signalling pathway plays a key role in the regulation of cell survival and proliferation. We show that the PI3-kinase/Akt pathway is constitutively active in primary acute myeloid leukaemia (AML) cells and that blockade by the selective inhibitor LY294002 reduces survival of the total blast population (mean 52%). The ERK/MAPK module is also constitutively active and treatment with the MAPKK inhibitor U0126 reduces cell survival by 22%. In 10 of 18 samples, PI3-kinase contributes to MAPK activation as incubation with LY294002 leads to a marked reduction in its phosphorylation. PI3-kinase inhibition reduces survival of the CD34+38- AML progenitor subset by 44%, whereas MAPKK inhibition has little effect. Reporter assays in primary AML cells show that blocking PI3-kinase leads to a marked reduction of constitutive NF-kappaB activity and promotes p53-mediated transcription. This is associated with a synergistic interaction between LY294002 and Ara-C. An inducible activated form of Akt protects normal myeloid cells from Ara-C and etoposide-mediated apoptosis. These results show that blocking PI3-kinase has direct antileukaemic effects and potentiates the response to conventional cytotoxics via a number of targets including NF-kappaB, p53 and MAPK. Inhibitors of PI3-kinase and Akt may be useful in the treatment of AML.

Impact of PTEN abnormalities on outcome in pediatric patients with T-cell acute lymphoblastic leukemia treated on the MRC UKALL2003 trial.

PTEN gene inactivation by mutation or deletion is common in pediatric T-cell acute lymphoblastic leukemia (T-ALL), but the impact on outcome is unclear, particularly in patients with NOTCH1/FBXW7 mutations. We screened samples from 145 patients treated on the MRC UKALL2003 trial for PTEN mutations using heteroduplex analysis and gene deletions using single nucleotide polymorphism arrays, and related genotype to response to therapy and long-term outcome. PTEN loss-of-function mutations/gene deletions were detected in 22% (PTEN(ABN)). Quantification of mutant level indicated that 67% of mutated cases harbored more than one mutant, with up to four mutants detected, consistent with the presence of multiple leukemic sub-clones. Overall, 41% of PTEN(ABN) cases were considered to have biallelic abnormalities (mutation and/or deletion) with complete loss of PTEN in a proportion of cells. In addition, 9% of cases had N- or K-RAS mutations. Neither PTEN nor RAS genotype significantly impacted on response to therapy or long-term outcome, irrespective of mutant level, and there was no evidence that they changed the highly favorable outcome of patients with double NOTCH1/FBXW7 mutations. These results indicate that, for pediatric patients treated according to current protocols, routine screening for PTEN or RAS abnormalities at diagnosis is not warranted to further refine risk stratification.

Clonality studies in acute myeloid leukemia.

The study of X-chromosome inactivation patterns (XCIPs) to determine tumor clonality was established by Fialkow using G6PD protein isoenzymes but was limited by the low frequency of heterozygotes. Analysis was extended to most females with the demonstration of differential DNA methylation patterns on active and inactive X-chromosome alleles and uses Southern blotting or PCR of either restriction enzyme polymorphisms, eg PGK, HPRT, or variable number tandem repeat sequences, eg M27beta, HUMARA. More recently RNA polymorphisms have been reported enabling direct analysis of expressed transcripts from the two alleles. Interpretation of clonality requires knowledge of an individual's constitutive XCIP as skewing (>75% expression of one allele) occurs in a significant proportion of hematologically normal females, probably due to the stem cell pool size at the time of Lyonization. Furthermore, acquired skewing occurs with increasing age. For myeloid disorders in the young, T lymphocytes serve as a suitable control XCIP, but interpretation of imbalanced XCIPs in the elderly can be difficult. In AML, XCIPs at presentation are consistent with a clonal disorder whereas in remission comparison with normal controls suggests that true clonal remission is infrequent. Sequential analysis of samples may be helpful in some patients in order to determine evolving clonal populations.

Investigation of methylation at Hha I sites using the hypervariable probe M27 beta allows improved clonal analysis in myeloid leukaemia and demonstrates differences in methylation between leukaemic and remission samples.

The methylation-sensitive enzyme Hha I has been used to assess the differentially methylated patterns on active and inactive X-chromosomes at the DXS255 locus recognized by the hypervariable probe M27 beta. The X-chromosome inactivation ratios obtained from 37 haematologically normal females using PstI and HhaI and correlated well with results obtained using PstI Hpa II (r = 0.97), and in 19 individuals with values obtained probing for either phosphoglycerate kinase or hypoxanthine phosphoribosyl transferase (r = 0.92). At least one Hha I site was found to be unmethylated on all alleles on inactive X-chromosomes. A monoclonal or oligoclonal pattern could be obtained by digestion with Hha I in 18/22 (82%) patients with acute myeloid leukaemia who had previously shown hypermethylation of both alleles using Hpa II, although in six of these patients differences in methylation could still be demonstrated between leukaemic and remission samples.

Assessment of X-chromosome inactivation patterns using the hypervariable probe M27 beta in normal hemopoietic cells and acute myeloid leukemic blasts.

The value of the hypervariable X-linked probe M27 beta for use in the analysis of X-chromosome inactivation patterns in normal blood and bone marrow cells and in the assessment of clonality in acute myeloid leukemia (AML) blast cells has been determined. By electrophoresing samples for 30 h, heterozygosity of the M27 beta locus was demonstrable in 324/415 females (78%) and this value could be increased to 93% by electrophoresing for 50 h. Determination of the X-chromosome inactivation patterns in blood and bone marrow samples from hematologically normal females was possible in approximately 90% of heterozygous individuals. The X-chromosome inactivation ratios obtained with M27 beta were comparable with phosphoglycerate kinase or hypoxanthine phosphoribosyl transferase in 46 individuals heterozygous for one of these genes in addition to M27 beta. The results obtained were closely correlated (r = 0.89). In 21 of 35 (60%) AML blasts, however, there was hypermethylation of the M27 beta locus and clonal analysis was not possible. Hypermethylation was not related to FAB type.

Quantification of X-chromosome inactivation patterns using RT-PCR of the polymorphic iduronate-2-sulphatase gene and correlation of the results obtained with DNA-based techniques.

Most studies using X-chromosome inactivation patterns (XCIPs) to determine clonality in females have used polymorphic DNA loci, but interpretation may be complicated by the complexity of the differential methylation patterns necessary to distinguish active and inactive X-chromosomes. Recent description of polymorphisms within the transcribed region of three X-chromosome genes has enabled XCIP analysis of allele expression at the RNA level, which should circumvent this problem. We report here a quantitative RT-PCR assay for one of these loci, the iduronate-2-sulphatase (IDS) gene, using a mismatch primer to introduce a Bc/l cutting site in one of the alleles. DNA samples from 150 females were screened and 61 (41%) were found to be informative for the polymorphism. Reproducible results were obtained from clonal analysis of 56 RNA samples covering the complete spectrum of XCIP ratios. The values correlated closely with those obtained from the corresponding DNA samples analyzed using either PGK or HUMARA (r=0.98) provided that the number of amplification cycles was limited to 25 to reduce heteroduplex formation. However, sample purity is of greater importance than in the DNA-based assays as contaminating red cells and platelets will still contain the relevant transcripts and could influence the result.

Variable expression of p16 protein in patients with acute myeloid leukemia without gross rearrangements at the DNA level.

The p16 protein is an inhibitor of cyclin-dependent kinases (cdk) 4 and 6. Both cdk4 and cdk6 phosphorylate the retinoblastoma protein (pRb) and are thought to be required for cell proliferation. Mutations and homozygous deletions in the p16 gene have been found in a number of cell lines but the incidence of abnormalities in primary tumours is controversial. We have studied the p16 gene in 76 cases of acute myeloid leukemia (AML) and 18 hematologically normal controls by quantitative Southern blotting. No deletions or rearrangements were detected. Twenty-five cases also showed no abnormal band patterns by RT-PCR-SSCP analysis of the coding sequence. We analyzed 60 AML samples for p16 protein expression by Western blot analysis. A reduced level of p16 was observed in six cases, however, none of them showed methylation of the 5'-CpG island. Over-expression of p16 protein was detected in six cases. Studies in cell lines have suggested a feedback loop between p16 and pRb with a futile overproduction of p16 protein in cells containing abnormal pRb. In accordance with these findings, four of the AML cases with high levels of p16 had abnormal pRb (two alteration in band size, two absent). From our data we suggest that gross abnormalities in the p16 gene are rare in AML, but that p16 levels are variable and high levels are associated with pRb abnormalities in a subset of cases.

Quantification of X-chromosome inactivation patterns in haematological samples using the DNA PCR-based HUMARA assay.

Quantification of X-chromosome inactivation patterns (XCIPs) using PCR amplification of the human androgen receptor (HUMARA) locus is potentially valuable in a range of haematological disorders. Of 236 females screened, 203 (86%) were heterozygous. For quantitative XCIPs it was necessary to limit the number of PCR cycles to 20 to reduce preferential amplification of shorter alleles. The optimized PCR method was compared with Southern blotting results using either PGK, HPRT or M27beta in 51 haematologically normal females and blast cells from 27 patients with acute myeloid leukaemia (AML). Reproducible XCIP results were obtained in all 78 samples using digestion with Hpa II prior to amplification (median difference in duplicate values 3%, range 0-17%) and they correlated well with Southern blotting results, r=0.966. Greater variability was observed in the results using Hha I digestion (median difference 4%, range 0-48%). There were marked inconsistencies in repeated analyses of three AML samples and although the HUMARA-Hha I results correlated well overall with Southern blotting in the remaining 75 samples (r=0.922), in nine samples there were still discrepancies with > or = 20% difference between the two values. These results suggest that PCR analysis of the HUMARA locus in Hpa II-digested DNA is suitable for the quantification of XCIPs in haematological samples but results with Hha I should be treated with caution.

The beta subunit common to the GM-CSF, IL-3 and IL-5 receptors is highly polymorphic but pathogenic point mutations in patients with acute myeloid leukaemia (AML) are rare.

The intracytoplasmic tail of the granulocyte-macrophage colony-stimulating factor receptor (GM-CSFR) beta c chain is essential for the activation of ligand-mediated signal transduction pathways in myeloid cells. Alterations in this region could deregulate normal signalling processes. We have therefore used RT-PCR-SSCP analysis of the receptor tail to look for point mutations in RNA from 35 patients with acute myeloid leukaemia (AML) and 10 haematologically normal controls. Patterns differing from those of the haemopoietic cell line TF-1 were detected in 25/35 (71%) AML patients and 8/10 (80%) normal controls. A total of six base substitutions were identified by sequencing. Three were conservative for the amino acid involved, three led to amino acid differences, valine652-->methionine, glycine647-->valine and proline603-->threonine. One alteration was found only in a normal control, the other five were all found in both AML patients and normal controls suggesting that they were DNA polymorphisms. Two substitutions were particularly common with allele frequencies of 0.23 (G1972-->A, unchanged proline648) and 0.13 (C1306-->T, unchanged serine426). These results indicate that the GM-CSFR beta c chain is highly polymorphic but point mutations of the intracytoplasmic tail do not appear to contribute frequently to the pathogenesis of AML.

Analysis of mutations in the GM-CSF receptor alpha coding sequence in patients with acute myeloid leukaemia and haematologically normal individuals by RT-PCR-SSCP.

Mutations of signal transducing molecules such as Ras have been shown to confer a growth advantage in leukaemic blasts and contribute to the pathogenesis of the disease. Alterations of signal transducing molecules other than Ras may play a role in leukaemogenesis. Knowledge of such mutations could also further our understanding of the normal signalling processes. We have therefore studied the coding sequence of the GM-CSF receptor alpha chain (GM-CSFR alpha) in patients with acute myeloid leukaemia (AML) and non-AML controls using single strand conformation polymorphism (SSCP) analysis. Abnormalities were detected in 4/32 AML patients (13%) and 2/15 non-AML controls (13%). Direct sequencing of PCR products revealed five different base substitutions. Three were conservative, two caused amino acid changes. The base substitution leading to amino acid change alanine to glycine at position 17 was found in both an AML patient and a control. It lies in the signal sequence and does not affect the mature protein. The other base change altering arginine to glutamine at position 164 is unlikely to influence the receptor structure as this structural position in the chain is not well conserved in members of the cytokine receptor family. Both amino acid changes were constitutive alterations as they could be demonstrated in the patients' children. The base changes described in the AML patients thus represent polymorphisms and do not contribute to the pathogenesis of AML.

Transcription-mediated amplification and hybridisation protection assay to determine BCR-ABL transcript levels in patients with chronic myeloid leukaemia.

Detection of BCR-ABL transcripts in chronic myeloid leukaemia (CML) is used to confirm the diagnosis and to monitor residual disease. Quantitative techniques are required to predict response to therapy or early relapse. We have evaluated an assay in which transcription-mediated amplification (TMA) of BCR-ABL and ABL transcripts is achieved using reverse transcriptase and RNA polymerase. The products are quantified in the hybridisation protection assay (HPA) using acridinium ester-labelled DNA probes and chemiluminescence. The method is a single tube procedure which uses small amounts of RNA (<500 ng/triplicate analysis), is technically simple (requiring just two waterbaths and a luminometer), rapid (total assay time <4 h) and sensitive (capable of detecting one BCR-ABL-positive K562 cell in the presence of 10(4)-10(5) BCR-ABL-negative cells). BCR-ABL signals from patient RNA samples were quantified relative to known amounts of K562 RNA and normalised to levels of ABL. BCR-ABL/ABL ratios ranged from 0.15 to 1.59 (median 0.65) in RNA from diagnostic blood or bone marrow of 18 CML patients and were < or =0.0001 in 20 normal controls. Sequential samples analysed from six CML patients post-allogeneic bone marrow transplantation who relapsed and received donor lymphocyte infusions showed BCR-ABL/ABL ratios which reflected patient status or treatment. A BCR-ABL/ABL ratio of 0.01 served as a useful arbitrary indicator value, with results above and below this value generally correlating with relapse or remission, respectively.

Expression of two alternatively spliced forms of the 5' untranslated region of the GM-CSF receptor alpha chain mRNA.

The granulocyte-macrophage colony-stimulating factor receptor (GM-CSFR) is composed of at least two chains (alpha and beta). The alpha chain binds GM-CSF specifically with low affinity, and the binding is converted to high affinity when the alpha chain is associated with the beta chain. To date, there are at least six isoforms described for the GM-CSFR alpha, all involving alternative splicing at the 3' end, which alters the coding region and hence the protein produced. To detect variants at the 5' end of the GM-CSFR alpha mRNA, RNAse protection and reverse transcriptase polymerase chain reaction (RT-PCR) assays were performed using a probe spanning nucleotides 102-392 and pairs of primers covering exons 1-4. in addition to the expected full-length transcript, two mRNAs were detected, one containing a deletion of 24 nucleotides by alternative splicing at the 3' end of exon 2 (exon 2b-deleted isoform) and another in which exon 2 was completely deleted (exon 2-deleted isoform). Together, the isoforms were more highly expressed form). Together, the isoforms were more highly expressed than the full-length sequence (TF-1 cells: full-length 36 +/- 2.8% vs. exon 2-deleted isoforms 64 +/- 5.5%). These isoforms were detected in primary hematopoietic cells, blasts from patients with acute myeloid leukemia (AML), and malignant cell lines and the relative mRNA expression for the isoforms, was always similar to that of TF-1 cells. As sequences in the 5'untranslated region can be involved in the modulation of translational efficiency, translation of constructs constructs corresponding to these exon 2 deleted isoforms was assessed using an in vitro reticulocyte lysate system. Deletion of exon 2 resulted in significantly lower in vitro translation of the receptor protein relative to the full-length sequence (53, 56, and 76% in three separate batches of reticulocytes), while deletion of exon 2b resulted in higher translation of the sequence (164, 128, and 305%; p = 0.01). These data suggest a mechanism by which expression of the GM-CSFR alpha protein may be regulated by alternatively spliced transcripts with different translational efficiencies.

Analysis of the coding sequence for the GM-CSF receptor alpha and beta chains in patients with juvenile chronic myeloid leukemia (JCML).

Peripheral blood granulocyte-macrophage progenitors (CFU-GMs) from patients with juvenile chronic myeloid leukemia (JCML) are able to proliferate spontaneously in the absence of exogenous growth factors and studies have shown that these progenitors display a hypersensitive growth response to GM-colony-stimulating factors (CSFs) in culture. We screened RNA from six patients with JCML for mutations in the GM-CSF receptor (GM-CSFR) coding sequence using RT-PCR-SSCP. Altered patterns were found in five patients for the GM-CSFR alpha chain, and in five patients for the beta chain. Two nucleotide substitutions accounted for all alpha chain abnormalities; one was conservative for Val333 and the other caused an Ala17-->Gly substitution. Both had previously been detected in normal controls. Sequencing of beta chain abnormalities revealed four base substitutions. Three were previously described polymorphisms, one causing a Pro603-->Thr substitution and two conservative point mutations involving Ser426 and Pro648. A further nucleotide mutation, which resulted in a Glu249-->Gln substitution, was also found but is not thought to be of pathological significance. Polymorphisms of the GM-CSFR alpha and beta chains are common, but pathogenic point mutations of the GM-CSFR are infrequent in patients with JCML and are unlikely to be involved in the hypersensitivity to GM-CSF demonstrated by progenitor cells.

Evaluation of clonality in myeloid stem-cell disorders.

Clonality in myeloid stem-cell disorders can be determined using either indirect methods such as analysis of X-chromosome inactivation patterns (XCIPs), or detection of specific abnormalities such as the chromosomal translocations characteristic of myeloid leukemias. XCIPs are particularly useful for disorders lacking evidence of a specific marker. Most females can be studied using polymerase chain reaction (PCR) analysis of differential DNA methylation patterns in the human androgen receptor (HUMARA) or phosphoglycerate kinase (PGK) genes, and approximately 68% can be studied using transcription assays of three polymorphic genes, glucose-6-phosphate dehydrogenase (G6PD), iduronate-2-sulfatase (IDS), and p55. Studies are limited by the incidence of constitutive and acquired (age-related) skewing and results must be carefully interpreted with reference to appropriate control samples. These techniques have been applied to clonality status of hematological disorders, lineage involvement in a clonal process, and detection of clonal evolution.

Oxygen dissociation curve in Eales's disease.

The oxyhaemoglobin dissociation curve (ODC) was investigated in 15 patients with Eales's disease and 11 controls to test the hypothesis that Eales's disease is caused by retinal hypoxia secondary to impaired oxygen release from the blood. Patients with the peripheral form of Eales's disease did not have any significant difference in the P50 values when compared with the controls. However, the P50 values were significantly lower (P < 0.005) for the patients with mixed retinopathy. This degree of increased oxygen affinity is not physiologically significant but may reflect an intrinsic abnormality of the red cells in these cases.

PI3K/mTOR inhibition upregulates NOTCH-MYC signalling leading to an impaired cytotoxic response.

PI3K, mTOR and NOTCH pathways are frequently dysregulated in T-cell acute lymphoblastic leukaemia (T-ALL). Blockade of PI3K and mTOR with the dual inhibitor PI-103 decreased proliferation in all 15 T-ALL cell lines tested, inducing cell death in three. Combined PI3K/mTOR/NOTCH inhibition (with a γ-secretase inhibitor (GSI)) led to enhanced cell-cycle arrest and to subsequent cell death in 7/11 remaining NOTCH mutant cell lines. Commitment to cell death occurred within 48-72 h and was maximal when PI3K, mTOR and NOTCH activities were inhibited. PI-103 addition led to upregulation of c-MYC, which was blocked by coincubation with a GSI, indicating that PI3K/mTOR inhibition resulted in activation of the NOTCH-MYC pathway. Microarray studies showed a global increase in NOTCH target gene expression upon PI3K/mTOR inhibition. NOTCH-MYC-induced resistance to PI3K/mTOR inhibition was supported by synergistic cell death induction by PI-103 and a small molecule c-MYC inhibitor, and by reduction of the cytotoxic effect of PI-103+GSI by c-MYC overexpression. These results show that drugs targeting PI3K/mTOR can upregulate NOTCH-MYC activity, have implications for the use of PI3K inhibitors for the treatment of other malignancies with activated NOTCH, and provide a rational basis for the use of drug combinations that target both the pathways.

Impact of NOTCH1/FBXW7 mutations on outcome in pediatric T-cell acute lymphoblastic leukemia patients treated on the MRC UKALL 2003 trial.

Activating mutations in the NOTCH1 pathway are frequent in pediatric T-cell acute lymphoblastic leukemia (T-ALL) but their role in refining risk stratification is unclear. We screened 162 pediatric T-ALL patients treated on the MRC UKALL2003 trial for NOTCH1/FBXW7 gene mutations and related genotype to response to therapy and long-term outcome. Overall, 35% were wild-type (WT) for both genes (NOTCH1(WT)FBXW7(WT)), 38% single NOTCH1 mutant (NOTCH1(Single)FBXW7(WT)), 3% just FBXW7 mutant (NOTCH1(WT)FBXW7(MUT)) and 24% either double NOTCH1 mutant (NOTCH1(Double)FBXW7(WT)) or mutant in both genes (NOTCH1(MUT)FBXW7(MUT)), hereafter called as NOTCH1±FBXW7(Double). There was no difference between groups in early response to therapy, but NOTCH1±FBXW7(Double) patients were more likely to be associated with negative minimal residual disease (MRD) post-induction than NOTCH1(WT)FBXW7(WT) patients (71% versus 40%, P=0.004). Outcome improved according to the number of mutations, overall survival at 5 years 82%, 88% and 100% for NOTCH1(WT)FBXW7(WT), NOTCH1(Single)FBXW7(WT) and NOTCH1±FBXW7(Double) patients, respectively (log-rank P for trend=0.005). Although 14 NOTCH1±FBXW7(Double) patients were classified as high risk (slow response and/or MRD positive), only two had disease progression and all remain alive. Patients with double NOTCH1 and/or FBXW7 mutations have a very good outcome and should not be considered for more intensive therapy in first remission, even if slow early responders or MRD positive after induction therapy.

The importance of relative mutant level for evaluating impact on outcome of KIT, FLT3 and CBL mutations in core-binding factor acute myeloid leukemia.

Several different mutations collaborate with the fusion proteins in core-binding factor acute myeloid leukemia (CBF-AML) to induce leukemogenesis, but their prognostic significance remains unclear. We screened 354 predominantly younger (<60 years) adults with t(8;21) (n=199) or inv(16) (n=155) entered into UK MRC trials for KIT, FLT3 tyrosine kinase domain (FLT3(TKD)), N-RAS, K-RAS and c-CBL mutations and FLT3 internal tandem duplications (FLT3(ITD)) and assessed the impact of relative mutant level on outcome. Overall, 28% had KIT, 6% FLT3(ITD), 10% FLT3(TKD), 27% RAS and 6% CBL mutations. Mutant levels for all genes/loci were highly variable. KIT mutations were associated with a higher cumulative incidence of relapse but in multivariate analysis this was only significant for cases with a higher mutant level of 25% or greater (95% confidence interval (CI)=1.01-1.52, P=0.04). Similarly, only FLT3(ITD-HIGH) was a significant adverse factor for overall survival (OS; CI=1.27-5.39, P=0.004). Conversely, FLT3(TKD-HIGH) and CBL(HIGH) were both favorable factors for OS (CI= 0.31-0.89, P=0.01 and CI=0.05-0.85, P=0.02, respectively). KIT mutations were frequently lost at relapse, which is relevant to minimal residual disease detection and the clinical use of KIT inhibitors. These results indicate that relative mutant level should be taken into account when evaluating the impact of mutations in CBF-AML.