Clinical significance of genetic aberrations in secondary acute myeloid leukemia

Authors


  • Conflict of interest: The authors have no relevant conflicts of interest to disclose.

Abstract

The study aimed to identify genetic lesions associated with secondary acute myeloid leukemia (sAML) in comparison with AML arising de novo (dnAML) and assess their impact on patients' overall survival (OS). High-resolution genotyping and loss of heterozygosity mapping was performed on DNA samples from 86 sAML and 117 dnAML patients, using Affymetrix Genome-Wide Human SNP 6.0 arrays. Genes TP53, RUNX1, CBL, IDH1/2, NRAS, NPM1, and FLT3 were analyzed for mutations in all patients. We identified 36 recurrent cytogenetic aberrations (more than five events). Mutations in TP53, 9pUPD, and del7q (targeting CUX1 locus) were significantly associated with sAML, while NPM1 and FLT3 mutations associated with dnAML. Patients with sAML carrying TP53 mutations demonstrated lower 1-year OS rate than those with wild-type TP53 (14.3% ± 9.4% vs. 35.4% ± 7.2%; P = 0.002), while complex karyotype, del7q (CUX1) and del7p (IKZF1) showed no significant effect on OS. Multivariate analysis confirmed that mutant TP53 was the only independent adverse prognostic factor for OS in sAML (hazard ratio 2.67; 95% CI: 1.33–5.37; P = 0.006). Patients with dnAML and complex karyotype carried sAML-associated defects (TP53 defects in 54.5%, deletions targeting FOXP1 and ETV6 loci in 45.4% of the cases). We identified several co-occurring lesions associated with either sAML or dnAML diagnosis. Our data suggest that distinct genetic lesions drive leukemogenesis in sAML. High karyotype complexity of sAML patients does not influence OS. Somatic mutations in TP53 are the only independent adverse prognostic factor in sAML. Patients with dnAML and complex karyotype show genetic features associated with sAML and myeloproliferative neoplasms. Am. J. Hematol., 2012. © 2012 Wiley Periodicals, Inc.

Introduction

The continuous production of terminally differentiated blood cells in the hematopoietic system is a tightly regulated process involving self-renewal, proliferation, and differentiation of stem and progenitor cells. Disruption of this process by acquired genetic lesions may cause the dominance of stem cell clones with variable output of myeloid cells. As a consequence, the production of terminally differentiated cells may be excessive resulting in myeloproliferative neoplasms (MPN) or deficient accompanied with dysplasia, with or without the presence of blasts, resulting in myelodysplastic syndromes (MDS).

The classic BCR-ABL-negative MPNs include three disease entities—polycythemia vera (PV), essential thrombocythemia (ET), and primary myelofibrosis (PMF) [1]. In 95% of PV and approximately half of ET and PMF cases the initiation of the clinical phenotype is hallmarked by somatic mutations in the JAK2 gene [2–5], often amplified to homozygosity by uniparental disomy (UPD) of chromosome 9p [6]. MDS are a heterogeneous group of disorders and a number of genetic lesions have been implicated in their pathogenesis, including del5q [7], as well as mutations in the RNA splicing pathway [8], TET2 [9], EZH2 [10], and other genes. MPN and MDS are chronic disorders with an elevated risk of disease progression to secondary acute myeloid leukemia (sAML). The rate of leukemic transformation in BCR-ABL-negative MPN patients is 7% [11], while MDS patients transform in 30% of the cases [12]. Disease progression to sAML is characterized by the presence of >20% of blasts in the bone marrow and sequential acquisition of genetic aberrations [13, 14]. Although several studies on limited patient cohorts showed association of gene mutations and cytogenetic aberrations with leukemic transformation of MPN and MDS, the leukemogenesis process remains poorly understood. Mutations in FLT3, NRAS [15], NPM1 [16], RUNX1 [17], DNMT3A [18], IDH1, IDH2 [19], TET2 [20, 21], and TP53 [22] have been implicated in leukemic transformation, as well as several chromosomal aberrations, such as deletions of IKZF1 [23], JARID2, AEBP2 [24], and amplifications of MDM4 [22].

Studies of clonal hierarchy showed that sAML can arise on the background of the chronic phase founder clone or alternatively as an independent event, resembling dnAML [25–27]. Unlike dnAML patients who often achieve complete remission after treatment, patients who transform to sAML have very poor prognosis and die within a few months following AML diagnosis [28], suggesting that leukemogenesis must differ substantially between sAML and dnAML.

Since the genetic basis of sAML is mainly unknown, the aim of our study was to delineate genetic profiles of sAML patients using high-resolution genome-wide single-nucleotide polymorphism (SNP) arrays, as well as direct sequencing of genes known to be involved in AML pathogenesis. Furthermore, we aimed to compare the frequency of recurrent genetic lesions in sAML and dnAML, and analyze their prognostic significance in order to define which genetic aberrations account for the poor prognosis of sAML patients. We used a bioinformatic approach to study co-occurrence of genetic aberrations and their association with patient diagnosis, in order to identify new genes involved in leukemogenesis and define potential new markers that would allow better genetic stratification of AML patients.

Methods

Patient samples

A total of 203 patients were included in this study, 117 diagnosed with dnAML and 86 with sAML. Patients diagnosed with therapy-related AML were not included in the study. Patients were diagnosed as sAML according to the 2008 WHO classification, and all samples were collected at the time of sAML diagnosis. Time to leukemic transformation was measured from the time of MPN/MDS diagnosis to the date of sAML diagnosis. Clinical data were available for 184 patients (110 dnAML and 74 sAML). Classical cytogenetic analysis was performed according to routine cytogenetic procedures, using GTG-banding technique. Peripheral blood samples from patients were collected from institutions in Italy, Austria, Czech Republic, and Serbia, following the local ethical regulations. Genomic DNA was isolated from granulocytes, bone marrow, or whole blood samples, following standard protocols.

Microarray analysis

All patients' DNA samples were processed and hybridized to Genome-Wide Human SNP 6.0 arrays (Affymetrix, Santa Clara, CA) according to the manufacturer's instructions. Raw data were analyzed using Genotyping Console Version 3.0.1 software (Affymetrix, Santa Clara, CA) for quality, identification of copy number alterations, and losses of heterozygosity. Detected chromosomal aberrations (gains, deletions, and UPDs) were annotated. The criteria for annotating UPD regions were terminal location on the chromosome and size of >1 Mb. UPDs found in patients with numerous interstitial runs of homozygosity (>10 Mb) were excluded from further analysis, as they infer parental consanguinity [29]. All aberrations mapping to known copy number variation loci according to the Database of Genomic Variants (DGV Version 5, human reference genome assembly hg18) were not annotated.

Mutational analysis

Exon sequencing of all coding exons of RUNX1, TP53, as well as exon 1 of NRAS, exons 8 and 9 of CBL, exons 4 of IDH1 and IDH2 was performed using BigDye Terminator Version 3.1 cycle-sequencing kit and the 3130xl Genetic Analyzer (Applied Biosystems, Foster City, CA). Sequencher Version 4.9 software (Gene Codes Corporation, Ann Arbor, MI) was used for sequence analysis. Screening of FLT3 internal tandem duplications, FLT3 mutations at position D835 and insertions in exon 12 of NPM1 were performed as previously described [30, 31]. All primer sequences are listed in Supporting Information Table I.

Table I. Association of Recurrent (More Than Five Events) Chromosomal Aberrations with Patient Diagnosis
Chromosomal aberrationStart positionEnd positionChromosomal band% of sAML patients% of dnAML patientsTotal no. eventsTarget gene locus P P value after Bonferroni correction
  1. UPD, uniparental disomy; sAML, secondary acute myeloid leukemia; dnAML, de novo acute myeloid leukemia.

9p UPD19,000,0009p24.317.4015 <0.001<0.01
7q LOSS101,350,000102,000,0007q22.120.94.423 CUX1<0.0010.018
7q LOSS65,180,00065,830,0007q11.2113.91.7514 0.0010.043
7q LOSS148,400,000150,400,0007q36.122.16.126 0.0010.043
20q LOSS35,000,00035,100,00020q11.238.107 0.0020.086
4q LOSS106,000,00010,664,00004q2410.50.910 TET20.0050.191
20q LOSS36,000,00036,170,00020q11.239.30.99 0.0050.198
1q GAIN201,000,000204,496,0001q32.1706 0.0060.205
9q GAIN78,900,00079,650,0009q21706 0.0060.205
20q LOSS33,500,00034,000,00020q11706 0.0060.205
7p LOSS11,000,00018,000,0007p2111.62.613 0.0170.63
7p LOSS50,080,00050,600,0007p12.213.94.417 IKZF10.0210.752
7p LOSS38,200,00038,390,0007p14.112.83.515 0.0270.968
5q LOSS130,900,000139,400,0005q3118.67.925 0.0301
5q LOSS89,489,00095,350,0005q14->q1516.36.121 0.0341
Monosomy 7   8.11.79 0.0401
9p GAIN138,761,0009p2470.97 0.0441
11q UPD118,500,000134,452,38411q23.3->qter70.97 0.0441
6p LOSS5,150,0005,460,0006p25.15.80.96 0.0861
6p LOSS14,130,00015,800,0006p235.80.96 JARID20.0861
6p LOSS19,955,00020,820,0006p22.35.80.96 0.0861
18q LOSS75,100,00075,950,00018q235.80.96 0.0861
12p LOSS11,800,00012,900,00012p1310.54.414 ETV60.1591
3p LOSS71,600,00073,000,0003p131.25.37 FOXP10.2421
Trisomy 8   12.8818 0.3411
20q GAIN29,300,00030,520,00020q11.214.61.76 0.4051
17q LOSS26,000,00026,600,00017q11.29.36.115 NF10.4271
21q GAIN45,700,00046,400,00021q22.35.83.59 0.5031
21q GAIN36,300,00040,250,00021q225.84.410 0.7481
17q LOSS27,200,00027,400,00017q11.28.1715 SUZ120.7921
9q LOSS92,300,00093,850,0009q222.33.56 11
13q UPD106,100,000114,142,98013q33->qter3.54.48 11
21q GAIN13,280,00027,000,00021q11->q214.64.49 11
12q LOSS92,900,00093,400,00012q223.52.66 11
17p LOSS4,850,0005,170,00017p13.24.64.49 11
17p LOSS7,080,0008,080,00017p13.14.65.310 11

Statistical analysis

Patient characteristics, frequencies of mutations, and chromosomal aberrations were compared using the Fisher's exact test for categorical data and Mann-Whitney test for continuous data. OS was measured from the date of AML diagnosis to time of death from any cause. Patients were censored at the time of the last follow-up. In the sAML patient cohort 52 patients died during the follow-up, while 22 patients were censored. Survival curves were constructed using the Kaplan-Meier method. The comparison of OS curves was performed using the log-rank test and multivariate analysis of OS using the Cox regression model. All calculated P values are two-tailed. SPSS 20.0 (SPSS, Chicago, IL) statistical software package was used for statistical analyses. The negative interaction information analysis was performed as described in the Supporting Information methods.

Results

Patient characteristics

The study included a total of 203 patient samples, diagnosed with sAML (N = 86) or de novo AML (N = 117). The sAML patient cohort consisted of 48 patients with a previous diagnosis of MPN, 36 with MDS, and 2 with MPN/MDS overlap. The median duration of chronic phase before transformation was 5 years (range 0.2–22.8). Clinical characteristics of the patient cohort are summarized in Supporting Information Table II. Patients with sAML were treated according to different therapeutic protocols (Supporting Information Table III), however, this did not influence the outcome since most sAML patients did not respond to therapy. Nine sAML patients received allogeneic hematopoietic stem cell transplantation.

Table II. Multivariate Analysis of Overall Survival Duration of Patients with Secondary Acute Myeloid Leukemia
VariablesHazard ratio95% CI P
TP53 mutation2.671.33–5.370.006
Complex karyotype1.310.69–2.490.411
Del7q (CUX1)0.760.33–1.750.527
Del7p (IKZF1)1.270.43–3.780.659
Age2.180.97–4.900.059
Table III. Clinical and Genetic Characteristics of De Novo Acute Myeloid Leukemia Patients with Complex Karyotype
UPNSexAgeFABOverall survival (months)Complex karyotype NPM1 FLT3 IDH1/2 RUNX1 NRAS CBL TP53del17pdel7q (CUX1)del3p (FOXP1)del7p (IKZF1)del12p (ETV6)
4F30M00.5YesWtWtWtWtWtWtWtYes
11M41M53YesWtWtWtWtWtWtWt
42F30M25YesWtWtWtWtG13DWtWt
44F61M23YesWtWtWtWtWtWtWtYesYesYesYes
51F63M60.5YesWtWtWtS418PfsWtWtY220CYesYes
67M55M23YesWtWtWtWtWtWtWtYes
68F66M21YesWtWtWtWtG12CWtS215G/R337CYesyesYes
72F82M42YesWtWtWtWtG12DWtWtYesYesyesYes
86M50M13.4YesWtWtWtWtWtWtWtYesYesYes
104M60M410.6YesWtWtR132CWtWtWtG279EYesYes
121M70M04.3YesWtWtWtWtG12DR420VWtYesYes

Frequencies of mutations in genes associated with AML differ between sAML and dnAML

To investigate whether the frequency of mutations in genes commonly affected in dnAML is the same in sAML we screened for mutations in FLT3, NPM1, TP53, CBL, IDH1, IDH2, RUNX1, and NRAS (Fig. 1A) in both patient groups. Within the sAML patient cohort we did not observe any significant difference in frequency of mutations with respect to the previous diagnosis (Supporting Information Fig. 2). At least one gene mutation was found in 56.41% of dnAML and 52.38% of sAML patients. In dnAML patients the most frequent were NPM1 mutations, present in 24.79% (29 of 17) of cases, followed by FLT3 mutations found in 23.93% (28 of 117) of patients. The frequency of these mutations was significantly lower in sAML patients, where NPM1 was mutated in 4.94% and FLT3 in 7.23% of cases (P = 0.002 and P = 0.016, respectively). In contrast, TP53 mutations were found in 16.67% of sAML and only 4.27% of dnAML cases (P = 0.046). The frequencies of mutations in RUNX1, IDH1, IDH2, CBL, and NRAS were not significantly different in the two patient groups (Fig. 1A). TP53 and RUNX1 mutations mainly affected the DNA-binding domains of these proteins (Supporting Information Fig. 1). The list of all mutations in analyzed genes and their PolyPhen-2 prediction scores of functional effects are listed in Supporting Information Table IV. Furthermore, we observed that certain mutations co-occur, while others show mutual exclusivity (Fig. 1B–E). Co-occurrence of NPM1 and FLT3 mutations was observed in dnAML only, while RUNX1 and FLT3 mutations co-occurred in 4 dnAML and 2 sAML patients (Fig. 1B,C). In both patient groups, IDH1 and IDH2 mutations were mutually exclusive. Mutations in TP53 were found to be mutually exclusive with IDH1/2 in sAML (Fig. 1E).

Figure 1.

Gene mutation profiles in secondary (sAML) and de novo acute myeloid leukemia (dnAML). (A) Comparison of mutational frequencies in genes affected in myeloid malignancies shows significant bias of FLT3 and NPM1 mutations toward dnAML and TP53 toward sAML. All P values have been corrected for multiple testing using the Bonferroni correction. The co-occurrence of mutations in dnAML (B) and sAML (C) is shown with Circos diagrams [32]. The length of the arc corresponds to the frequency of mutations in the individual gene, while the width of the ribbons connecting two arcs corresponds to the number of patients carrying both mutations. Certain mutations show mutual exclusivity in both dnAML (D) and sAML (E). Each vertical line represents one patient with at least one mutation.

Microarray-based karyotyping of sAML and dnAML

To get a deeper insight into the genetic complexity of sAML and dnAML, we performed high-resolution genome-wide analysis of DNA copy number abnormalities and losses of heterozygosity on 200 samples (114 dnAML and 86 sAML) using Affymetrix Genome-Wide SNP 6.0 arrays. Comparing with only 11.4% of dnAML cases, 44.2% of sAML patients presented with a complex karyotype (P < 0.0001), defined as more than or equal to three unrelated chromosomal aberrations not included in the WHO 2008 classification criteria [13]. In 17.4% (15 of 86) of the sAML cases we could not detect any chromosomal aberration, which was significantly lower comparing with 37.7% (43 of 114) of such dnAML patients (P = 0.0017). In the remaining samples we detected a total of 669 chromosomal aberrations, mainly deletions (59%, 395 of 669), but also gains (27.95%, 187 of 669) and UPDs (13%, 87 of 669) (Fig. 2). We found 36 recurrent chromosomal aberrations, present in more than five patients in our cohort (Table I). High number of deletions allowed the fine mapping of common deleted regions (CDR) to <1 Mb on specific chromosomal arms, pointing out the possible target genes, such as CUX1, IKZF1, TET2, JARID2, SUZ12, RUNX1, TET1, and NF1.

Figure 2.

Whole-genome view of all chromosomal aberrations and gene mutations in 86 sAML and 117 dnAML patients. Each line represents the whole genome of one patient in which at least one genetic aberration was detected with Affymetrix Genome-Wide Human SNP 6.0 arrays. The size and physical position of each chromosomal aberration is represented with a colored bar. Deletions are represented in red, gains in green, and uniparental disomies in blue. Mutations in analyzed genes are shown as filled boxes. UPN, unique patient number.

We compared the frequency of recurrent chromosomal aberrations in the two patient groups and found that 18 out of 36 recurrent cytogenetic aberrations show a bias toward sAML. However, after applying Bonferroni correction for multiple testing, only four chromosomal aberrations significantly associated with sAML (Table I). The strongest association was observed for 9pUPD, as a consequence of high prevalence of this aberration in chronic phase of MPN. The second strongest association was observed between sAML and del7q22.1 mapping to tumor suppressor CUX1 locus. Interestingly, gains of chromosome 1q32.1, targeting the MDM4 locus, were found exclusively in sAML patients (N = 6).

We also compared the frequency of recurrent chromosomal aberrations in sAML patients who developed AML following MPN or MDS chronic phase, and found that besides 9pUPD which showed association with previous MPN diagnosis, all other aberrations occurred in similar frequencies in both post-MPN and post-MDS AML (Supporting Information Table V).

All patients with 13qUPD (N = 7) and 17pUPD (N = 4) were carrying homozygous mutations in FLT3 and TP53, respectively. Both TP53 alleles were affected in 13 out of 19 patients with TP53 mutations (Supporting Information Table VI). CBL mutations were amplified by 11qUPD in four cases, however, three sAML patients with 11qUPD did not carry mutations in CBL.

Association of recurrent genetic lesions

In order to analyze the association of recurrent genetic lesions, we performed Fisher's exact test on all pairs of analyzed gene mutations and recurrent cytogenetic aberrations (more than five events), provided they are located on different chromosomes. Supporting Information Table VII lists all significant aberration pairs (N = 16). Del5q was found to be associated with TP53 mutations (P = 0.0004), del12p targeting ETV6 locus (P = 0.0022), del17p (P = 0.0159), and del7q (P = 0.0251). We also found association of del18q with gain of 20q (P = 0.0033) and del4q targeting TET2 locus with del20q (P = 0.0153).

The negative interaction information analysis was performed to detect the pairs of aberrations that have different patterns of association (co-occurrence or mutual exclusivity) depending on the disease subtype. The top 100 hits obtained from this analysis are listed in Supporting Information Table VIII. We observed that del5q frequently occurs together with 9pUPD and TP53 mutations in sAML patients, and that TP53 mutations and del10q targeting TET1 gene locus are present at the same time in four sAML patients, while these aberrations never co-occur in dnAML.

Prognostic significance of recurrent genetic lesions

In order to define which genetic feature had the most adverse effect on the survival of patients with sAML, we analyzed the influence of mutations in TP53, del7q (CUX1), del7p (IKZF1), and presence of complex karyotype. These features were chosen because they were found to associate with sAML and are rare in both dnAML and chronic phase of MPN and MDS [31]. We focused our analysis on overall survival since most of the sAML patients did not respond to therapy or died early following sAML diagnosis. Patients with sAML carrying TP53 mutations showed a significantly shorter median OS (1.8 months, 95% CI, 0.8–2.8), compared with patients with wild-type TP53 (5.6 months, 95% CI, 1.5–9.7; P = 0.002) (Fig. 3A). Karyotype complexity did not show a significant impact on OS of sAML patients. The median survival of sAML patients with complex karyotype was 4 months (95% CI, 2.5–5.5), while the OS of patients with noncomplex karyotype was 8 months (95% CI, 3.0–13.0) (Fig. 3B). The effect of del7q (CUX1) and del7p (IKZF1) on OS of sAML patients was not significant (Fig. 3C,D). We included age, complex karyotype, mutations in TP53, del7p, and del7q in the model and performed multivariate analysis of OS of sAML patients. The multivariate analysis confirmed that mutated TP53 was the only independent adverse prognostic factor for OS (hazard ratio 2.67; 95%CI, 1.33–5.37; P = 0.006) in sAML (Table II).

Figure 3.

Impact of recurrent genetic lesions and karyotype complexity on overall survival (OS) of patients with secondary acute myeloid leukemia (sAML) represented by Kaplan-Meier curves. (A) OS in sAML patients with mutated (N = 14) or wild-type TP53 (N = 59). (B) OS in sAML patients with complex (N = 31) or noncomplex (N = 43) karyotype. (C) OS of sAML patients according to presence (N = 17) or absence (N = 57) of del7q targeting CUX1 locus. (D) OS of sAML patients according to presence (N = 9) or absence (N = 65) of del7p targeting IKZF1 locus.

De novo AML patients with complex karyotype exhibit genetic features of secondary AML

There were 11 dnAML patients with complex karyotype in our patient cohort with available clinical data. The median OS of these patients was 3 months (95% CI, 2.0–4.0), and all patients died during the follow-up. To gain better understanding of poor OS of patients with complex karyotype within the dnAML group we looked at genetic profiles of each of these patients (Table III) and found that TP53 was affected in 54.5% (N = 6 of 11) of these patients. TP53 mutations were present in three patients, and additional three patients carried a deletion of TP53 locus on chromosome 17p. TP53 mutations were only present in this subtype of dnAML. We also found del7p (IKZF1) and del7q (CUX1) in four and three patients, respectively. In addition, these patients carried deletions mapping to the transcription factors FOXP1 (45.4%, N = 5 of 11) and ETV6 (45.4%, N = 5 of 11) (Table III). Taken together we observed a number of genetic features present in dnAML with complex karyotype typical for either MPN or sAML.

Discussion

The objective of this study was to examine to what extent the genetic basis of leukemogenesis in post-MPN and post-MDS AML differs from dnAML, by systematic analysis of genetic profiles of sAML and dnAML patients. We grouped the patients who developed sAML after MPN or MDS chronic phase, since we did not observe any genetic difference between the two groups in the leukemic stage. The comparison of mutational frequency in genes known to be affected in myeloid malignancies showed that TP53 mutations significantly associated with sAML, whereas NPM1 and FLT3 were prominently involved in dnAML. IDH1/2, NRAS, RUNX1, and CBL mutations seem to be universally contributing to dnAML and sAML leukemogenesis. The mutual exclusivity of certain mutations such as IDH1/2 with CBL, TP53 and with NRAS confirms that distinct leukemogenic pathways are involved in sAML [33].

We extended the genetic marker analysis by karyotyping patients with high-resolution SNP arrays. The high number of chromosomal aberrations observed enabled us to fine map CDRs to several putative tumor suppressor gene loci (Table I). Compared with dnAML, we observed higher karyotype complexity in sAML and identified a number of cytogenetic lesions differentially distributed in both AML groups. Previous attempts to identify karyotype differences between sAML and dnAML were limited with the use of classical cytogenetic methods, although the association of del7q with sAML was previously reported [34]. Association of del7q was recently confirmed by SNP array karyotyping and the 7q CDR was mapped to the CUX1 locus [31]. Cux1 deficiency negatively affects the expression of ATM/p53 pathway members [35]. CUX1 deletions in sAML must have a weaker effect on DNA damage response pathways compared with mutated TP53 as we did not observe influence of CUX1 deletions on OS of patients, whereas mutated TP53 negatively impacts OS. Our study convincingly showed for the first time that mutated TP53 in sAML is a strong independent prognostic factor of poor survival. Previous reports showed worse survival of TP53 mutated dnAML patients with complex karyotype [36, 37]. The importance of p53 in the leukemogenesis of sAML is further supported by our finding that gains of 1q32.1 harboring the MDM4 (known inhibitor of p53) were found exclusively in secondary AML. They often co-occurred with del7q (CUX1) (N = 4 cases), suggesting that DNA damage response pathway defects play a crucial role in sAML pathogenesis.

A recent report of complex clonal architecture in sAML, showing various rare mutations in each patient [14] highlights the need for identification of robust genetic markers with relatively high frequency. Since monoallelic TP53 mutations are detectable in chronic phase MPN [22] and MDS [38] (3.5% and 7.5% of the cases, respectively), screening for TP53 mutations in chronic myeloid malignancies could be a useful marker predicting leukemic transformation.

The analysis of pair-wise association of genetic aberrations identified several new collaborating defects, such as del5q and del12p (ETV6), which require further functional validation. We confirmed previous reports that TP53 mutations are often found together with 5q deletions [39]. It is unclear at this point to what extent the poor prognosis associated with mutated TP53 is influenced by 5q deletions.

The high rate of karyotype complexity and its nonsignificant effect on OS in sAML suggests that many chromosomal lesions represent functionally irrelevant aberrations and that chromosomal instability is a hallmark of sAML pathogenesis. The karyotype complexity in sAML is likely due to longer clonal evolution before leukemic transformation. A recent study showed that therapies administered during chronic phase of MPN do not associate with increased risk of AML transformation, and that 25% of therapy-naive patients develop AML [40]. However, a different study with longer observation time reported increased transformation rates in patients treated with pipobroman or hydroxyurea [41]. It is likely that certain patient-specific adaptations to therapies might utilize pathways promoting leukemic transformation, such as the DNA damage response pathway. In contrast to sAML, complex karyotype significantly contributed to poor OS in the dnAML patients as previously reported [42]. Interestingly, the majority of dnAML patients with complex karyotype exhibited genetic features typical for either chronic phase MPN or sAML. These lesions included mutated TP53, presence of CUX1 deletions, and deletions of FOXP1 and ETV6 previously shown to be associated with chronic MPN [31]. This observation suggests that a number of patients diagnosed as dnAML with complex karyotype might in fact be sAML patients with a previously undiagnosed MPN or MDS either due to masked chronic phase phenotypes or absence of hematological examination preceding leukemic transformation.

We have shown that the genetic features of AML arising de novo substantially differ from post-MPN and post-MDS AML. Our data reinforce the fact that dnAML and sAML should be treated as separate AML subtypes. Furthermore, genetic resemblance of dnAML with complex karyotype and sAML indicate that these patients might constitute a relatively homogenous group for therapeutic intervention. This study demonstrates the use of genetic stratification of AML patients and suggests that despite the immense genetic heterogeneity among AML patients, certain markers have strong influence on patients' survival.

Acknowledgements

The authors thank Paola Guglielmelli, Lisa Pieri, Klaudia Bagienski, and Martin Schalling for their technical assistance and valuable contributions.

Ancillary