Genome wide molecular analysis of minimally differentiated acute myeloid leukemia

Abstract

Background Minimally differentiated acute myeloid leukemia is heterogeneous in karyotype and is defined by immature morphological and molecular characteristics. This originally French-American-British classification is still used in the new World Health Organization classification when other criteria are not met. Apart from RUNX1 mutation, no characteristic molecular aberrations are recognized.Design and Methods We performed whole genome single nucleotide polymorphism analysis and extensive molecular analysis in a cohort of 52 patients with minimally differentiated acute myeloid leukemia.Results Many recurring and potentially relevant regions of loss of heterozygosity were revealed. These point towards a variety of candidate genes that could contribute to the pathogenesis of minimally differentiated acute myeloid leukemia, including the tumor suppressor genes TP53 and NF1, and reinforced the importance of RUNX1 in this leukemia. Furthermore, for the first time in this minimally differentiated form of leukemia we detected mutations in the transactivation domain of RUNX1. Mutations in other acute myeloid leukemia associated transcriptions factors were infrequent. In contrast, FLT3, RAS, PTPN11 and JAK2 were often mutated. Irrespective of the RUNX1 mutation status, our results show that RAS signaling is the most important pathway for proliferation in minimally differentiated acute myeloid leukemia. Importantly, we found that high terminal deoxynucleotidyl transferase expression is closely associated with RUNX1 mutation, which could allow an easier diagnosis of RUNX1 mutation in this hematologic malignancy.Conclusions Our results suggest that in patients without RUNX1 mutation, several other molecular aberrations, separately or in combination, contribute to a common minimally differentiated phenotype.

Introduction

Minimally differentiated acute myeloid leukemia (AML-M0) is a distinct entity but heterogeneous in manifestation.1 It accounts for less than 5% of all cases of AML, is mostly seen in elderly patients and has a remarkably poor prognosis.1–3 AML-M0 blasts show low expression of myeloperoxidase, express at least one myeloid surface antigen (CD13, CD33, CD15) and have immunophenotypic characteristics of early progenitor cells such as expression of CD34, terminal deoxynucleotidyltransferase (TdT), HLA-DR and CD117.1–3 Co-expression of lymphoid-associated antigens is sometimes seen but does not preclude diagnosis.4

The incidence of abnormal karyotypes is more frequent in AML-M0 (71 to 81%) than in other subtypes.2,3,5 Complex aberrant karyotypes are detected in approximately 20% of AML-M0 cases, and unbalanced chromosomal changes involving −5/del(5q),−7/del(7q), +8, +11 and +13 are the most frequent.2,3,5 Nevertheless, unlike for other AML subtypes, no characteristic translocation has been described for AML-M0.

The most frequently recurring molecular alterations observed in AML-M0 are mutations in RUNX1 (alias AML1).6 These are mainly biallelic or dominant-negative point mutations, other than translocations, and are detected in 15 to 35% of cases.6 RUNX1 is a transcription factor essential for hematopoiesis that binds DNA through its Runt domain.7 RUNX1 is described as both a tumor suppressor gene (TSG), as in AML-M0,8 and as an oncogene, as it is frequently involved in chromosomal translocations in various hematologic malignancies.9 Other genes found to be mutated in AML-M0 include FLT3, RAS and PTPN11.10–14 However, these mutations are considered to be, rather than surrogates, collaborating abnormalities with mutations in transcription factors such as RUNX1.15

In this study we aimed to identify new molecular alterations that could explain the etiology of AML-M0. Together with cytogenetic and immunophenotypic analyses, we analyzed a unique cohort of 52 AML-M0 samples by whole genome single nucleotide polymorphism (SNP) screening to reveal regions of loss or gain that could contain putative TSG or oncogenes. We also performed extensive mutation analysis in genes known to be mutated frequently in AML.

Design and methods

Patients’ material

This study was performed with archived material from diagnostic cryo-preserved bone marrow aspirates from 52 patients (numbered from 1 to 9 and 18 to 60), classified morphologically and immunophenotypically as having AML-M0 (presented in Table 1 and in an Online Supplementary file). Material was collected from the medical centers at the Robert-Rössle-Clinic, Charité Berlin, Germany; University of Leiden, the Netherlands; University of Groningen, the Netherlands; Erasmus University, Rotterdam, the Netherlands; and University of Vienna, Austria. All samples were handled in a coded fashion and research was conducted following the medical ethical guidelines of the national organization of scientific societies (FEDERA; Code for Proper Secondary Use of Human Tissue in The Netherlands). Pure tumor cell populations were sorted by flow cytometry from mononuclear cells isolated from bone marrow or peripheral blood at the time of diagnosis.8 T cells from each sample (with the exception of those from patients 38 and 49) were expanded using previously described conditions as a source for control material.8,16 DNA was isolated from the sorted tumor cells and T cells using the QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany). RNA was isolated from the mononuclear cell fraction using the QIAamp RNA Blood Mini Kit (Qiagen). cDNA was prepared using SuperScript First-Strand Synthesis System for reverse transcriptase polymerase chain reaction (RT-PCR) (Invitrogen, Breda, the Netherlands).

Karyotyping

Cytogenetic analysis was performed on GTG-banded chromosomes and the karyotype was assigned according to the criteria of the International System for Human Cytogenetic Nomenclature (ISCN).17 A complex aberrant karyotype is defined in this study by at least five abnormalities.

Single nucleotide polymorphism analyses

SNP analysis was performed using the GeneChip Mapping 10k 2.0 array (Affymetrix, Santa Clara, CA, USA) following the manufacturer’s instructions. Briefly, 250 ng of total genomic DNA was digested with XbaI restriction enzyme and ligated to a universal adapter. The ligated fragments were then PCR-amplified using primers complementary to the universal adapters. PCR products were purified, fragmented by DNase I, labeled with biotinylated dATP and hybridized to the array. The arrays were scanned using a GeneChip Scanner (Affymetrix). Affymetrix GeneChip 5.0 genotyping software was used to examine the SNP hybridization patterns and to make SNP calls. The resulting data were analyzed with the dChip software package.18

Mutation screening

RUNX1, FLT3 internal tandem duplications (ITD) and FLT3 D835 mutation screening was performed as previously described.8,19 KIT D816V mutations (exon 17) were screened using the HinfI restriction assay.20 All patients were screened for these mutations.

Melting curve analysis was performed for all patients to detect mutations in NRAS (codons 12/13 and 61), KRAS (codon 61), JAK2 (codon 617), PTPN11 (exons 3 and 13), PTPN6 (exons 3 and 13) and SPI1 (exons 1 to 4). The above mentioned regions were amplified in reactions containing LCGreen PLUS (Idaho Technology, Salt Lake City, UT, USA) using the primers and conditions described in the Online Supplementary Appendix. Subsequently, melting curves of the PCR products were generated in a LightScanner HR 96 (Idaho Technology). Aberrant melting curves were subjected to DNA sequencing to confirm mutations, as previously described.8 Standard PCR techniques were used to amplify RUNX1 (exons 6 to 8), NRAS (codon 61), CEBPA, CEBPD, SPI1 (exon 5), TET2, PURA, PILRA, PILRB from genomic DNA and RUNX1, CSF2RA and IL3RA from cDNA (Online Supplementary Appendix). Subsequent DNA sequencing was performed as previously described.8 CEBPA and CEBPD sequencing reactions included, in addition to standard reagents, 1.6 M of betaine (Sigma-Aldrich, Saint-Louis, MO, USA).

NPM1 exon 12 and CEBPB mutations were studied by fragment analysis. After amplification (the conditions and primers are presented in the Online Supplementary Appendix), the PCR products were analyzed on a 3730 DNA Analyzer (Applied Biosystems, Foster City, CA, USA) with GeneScan 500 ROX (Applied Biosystems) as a size standard. Detected mutations were confirmed by sequencing.8

The tumor specificity of mutations detected in genomic DNA of AML tumor cells was confirmed by sequencing the respective control DNA (expanded T cells) for the absence of mutations.

Results

Minimally differentiated acute myeloid leukemia shows karyotype heterogeneity

The cohort of 52 patients was not selected for any cell-biological feature other than the diagnosis of AML-M0. The cohort had a median age of 61 years and consisted of 47 de novo cases, 1 therapy-related case and 4 cases of secondary leukemia (Table 1).

Karyotypes were available for 49 cases (Table 1). Six (12%) cases had complex aberrant karyotypes, 14 (29 %) cases were normal and 17 (35%) cases were associated with a single abnormality. The remaining 12 cases carried two or three chromosomal abnormalities. The predominant single abnormalities were trisomy 13, trisomy 8 and monosomy 7. Overall the most frequent abnormalities detected were del(5q), trisomies 8 and 13 and monosomy 7/del(7q) (Table 3).

We also detected several chromosomal abnormalities that might involve known transcription factors, including the del(16)(q22) (patient 42), frequently associated with CBFB, the inv(3)(q21q26) and t(2;3)(p23;q27) (patients 51 and 4), associated with EVI1, and the previously reported translocations involving ETV6 in cases 2, 9 and 43 (Table 1).21

Single nucleotide polymorphism array analysis revealed several potential new regions involved in minimally differentiated acute myeloid leukemia

We compared DNA isolated from flow-sorted leukemic cells to control DNA of the same patients using the Affymetrix GeneChip 10K mapping array. Loss of genetic information, detected as loss of heterozygosity (LOH) and/or copy number changes, was found in several chromosomes (Tables 1 and 2). Twenty-three regions were found to be affected in 37 AML-M0 patients. Most patients had losses or gains additional to those found by karyotype analysis. Interestingly, of the 14 patients with a normal diploid karyotype, nine displayed LOH on at least one chromosome (Table 1).

LOH was most frequently observed in chromosomes 21, 17, 7 and 5. As regards chromosome 21, copy neutral LOH (also known as uniparental disomy [UPD] or [partial] isodisomy) was found in 14 cases and a hemizygous deletion in 3 cases (Figure 1A). Homozygous deletions of the region on chromosome 21 harboring the tumor suppressor gene RUNX1 were detected in three of the 17 cases with LOH (Table 2, Figure 1A). LOH in chromosomes 5 (9 patients) and 7 (9 patients), in contrast to that in chromosome 21, was mainly due to deletions. In chromosome 17 the minimal regions of overlap in LOH between the patients comprised two separate regions, one including TP53 and the other NF1 (Table 2, Figure 1B). LOH was also detected in chromosomes 3 (5 patients), chromosome 4 (4 patients) and chromosome 12 (4 patients). LOH in chromosome 12 resulted from a hemizygous deletion that included the ETV6 locus, as previously reported.21 Microdeletions (smaller than 2 Mb) were present in three patients, affecting chromosomes 3, 4, 8 and 11 (Table 2). The remaining examples of LOH were restricted to a limited number of patients (Table 2). In addition, two cases showed gain of genomic regions in chromosomes 3 and X/Y (Table 2).

Mutation analysis of candidate genes did not reveal new targets

Following the results of the SNP analysis we screened several candidate genes, located within the minimal common regions of LOH or gain, for mutations. Genes were selected based on their potential relevance for leukemia, and included PURA, PILRA, PILRB, TET2 and CEBPD for regions of LOH, and IL3RA and CSF2RA for a region of amplification near the pseudoautosomal region of chromosomes X and Y (Table 2). In most cases, mutation screening was restricted to the samples showing LOH or copy number gain (Online Supplementary Table S2). No mutations were detected in any of these candidate genes. However, CSF2RA and IL3RA, which are involved in hematopoietic development,22 were shown to be highly up-regulated by gene expression microarray analysis (data not shown).

RUNX1 mutations in minimally differentiated acute myeloid leukemia are not restricted to the RUNX1 runt domain

Mutations in RUNX1 were detected in 18 patients (35%) (Table 1 and Online Supplementary Table S1). In concordance with the SNP analysis, 13 of these mutations were homozygous/biallelic (Table 1 and Figure 1A). Patient 38 had a homozygous mutation but without detectable LOH (Table 1). The most common RUNX1 mutations found in exons 3, 4 and 5 were base substitutions in or near to the DNA binding runt domain (Table 1). We also detected four insertions that resulted in truncated proteins as well as one insertion/duplication and one deletion. In addition, exons 6, 7 and 8 of RUNX1 were sequenced in samples for which a mutation was expected based on the LOH results and other analyses. In two samples (from patients 26 and 41) we found a frame shift mutation. In only one patient (patient 52) with UPD (Figure 1A), a RUNX1 mutation remained undetected. In addition to patients 18, 39 and 53 (Figure 1A, Table 2), RUNX1 appeared to be homozygously deleted in patient 19.8

Table 1.Molecular, clinical, immunological and cytogenetic features of the patients.

Table 2.Summary of minimal common regions of chromosomal losses, gains and uniparental disomy detected by LOH and copy number analyses.

Mutations in other acute myeloid leukemia- associated transcription factors are infrequent

We screened all patients for mutations in CEBPA, CEBPB and SPI1. We found biallelic CEBPA mutations in patient 35 (Table 1 and Online Supplementary Table S1). No mutations were detected in CEBPB or SPI1. All patients were also screened for exon 12 insertions in NPM1. Patient 59 showed a four nucleotide insertion (Table 1 and Online Supplementary Table S1). This patient did not express CD34 and had a normal karyotype, as reported in other cases with NPM1 mutation.23

RAS-related genes are frequently mutated in minimally differentiated acute myeloid leukemia

We detected 25 mutations in 22 patients (42%) in genes related to the RAS and JAK signaling pathways (Tables 1 and 3; Online Supplementary Table S1). FLT3 mutations were the most frequent mutations, i.e., nine FLT3 ITD and one FLT3 D835. Activating mutations in RAS genes were present in six patients (11.5%), two in KRAS and four in NRAS. PTPN11 mutations were observed in 6 patients (11.5 %). Activating mutations in codon 617 of JAK2 were found in three patients, two of whom had a previous history of hematopoietic disorders. In patient 23, the JAK2 mutation was homozygous, in line with UPD detected in chromosome 9 (Table 1). PTPN11 mutations coexisted with RAS, FLT3 and JAK2 mutations in one case each (Table 1).

We did not find any mutations in exons 3 or 13 of PTPN6, a PTPN11-related gene, or in KIT. A summary of the results is presented in Table 3 and detailed information on the detected mutations is given in Online Supplementary Table S1.

RUNX1 mutations are associated with terminal deoxynucleotidyl transferase expression in minimally differentiated acute myeloid leukemia

Data on protein expression of TdT were available for 40 cases (Table 1). Twenty-two (55%) cases demonstrated high TdT expression (>10% of cells). Notably, we found a strong association between high protein expression of TdT and presence of a RUNX1 mutation (one-sided Fisher’s exact test; using a 10 % expression cut-off p=0.00002 or a 20 % cut-off p=0.00001). The only exceptions to this association were patients 28 and 38 who, although carrying RUNX1 mutations, showed no or low expression of TdT, and patients 3, 22 and 43 who showed high TdT expression without having RUNX1 mutation (Table 1).

Figure 1.Single nucleotide polymorphism analysis of patients showing chromosome 17 or 21 abnormalities. (A) LOH in chromosome 21 was detected in 17 patients. LOH was related to a hemizygous deletion in four patients (2, 25, 33, 53) and UPD (copy neutral) in 13, as can be seen by the copy number call. Homozygous deletions were detected in three patients (18, 39, 53) and affected the RUNX1 locus. Chromosome representation, cytoband and gene distribution (overview) are at the left side of the panel. Candidate genes are underlined. The left heat map shows inferred LOH calls based on a hidden Markov model considering haplotype (using the paired normal). Each column represents one patient’s result. Each box represents the combined call for one or more SNP between tumor sample and respective control (T cells). The right heat map represents chromosome copy number inferred using the paired normal as reference and a median smoothing. Dark gray boxes represent two copies, light gray boxes represent one copy (deletion) and black boxes represent three copies for each chromosome locus. Homozygous deletions are represented by striped boxes (see copy number heat map). (B) LOH at chromosome 17 was detected in 11 patients. Two main regions of LOH were detected: one including the TP53 locus, shared by patients 1, 5, 7, 25, 31, and 33, the other including the NF1 locus, in patients 6, 8, 25, 45 and 54. In addition three patients showed extensive LOH of the q arm of chromosome 17. Legend as in panel (a).

Discussion

In the last years, the approach to the classification of AML has shifted from a morphological to a molecular basis. Although the cohort of patients studied here was selected based on morphological criteria, recent results using gene expression profiling have shown that AML-M0 patients cluster together as a separate group of AML-M0.24 In the present study we aimed to identify TSG and oncogenes that might contribute to the AML-M0 phenotype. Whole genome SNP analysis revealed various new regions of LOH containing known and candidate TSG. Conversely, chromosomal gains were rare. Many of the LOH regions described here could not have been found by standard cytogenetic techniques as they were UPD (Table 1 and 2). Two LOH regions containing known TSG were on chromosome 17. LOH at 17p13.1 (TP53) was clearly independent from that at 17q11.2 (NF1). This result confirms similar findings in AML.25,26 However, our results suggest that both regions have equal importance in AML-M0, since the number of occurrences for each region is comparable. TP53 has an important role in the maintenance of chromosomal stability and its deletion has been linked to AML with complex karyotype.26 In concordance, three AML-M0 patients with TP53 loss also had complex karyotypes. NF1 is involved in negative regulation of the RAS pathway (see below) and is found to be mutated in patients with hematopoietic disorders.27 Interestingly, a third region of LOH was detected in chromosome 17 implying that another TSG is present at 17q (Table 2).

In a number of minimal LOH regions without known TSG, we screened several candidate genes or transcripts for mutations, including CEBPD at chromosome 8, TET2 at chromosome 4, and others at chromosomes 3, 5 and 7 (Table 2). Candidate genes were selected based on the likelihood of them having a role in AML-M0 or, in the case of the region containing TET2, because it was described as a microdeletion in four patients.28 No mutations were found in any of the candidate genes. However, for some of the LOH regions detected, hemizygous deletions may already be sufficient for the neoplastic process, without requiring a mutation in the other allele.29 In fact, several reports suggest that haploinsufficiency of one or more genes, especially in chromosome 5 and 7, contributes to AML.26,30,31 Of note, some of the areas of LOH detected by us, in particular the ones with UPD, were too large to be screened efficiently for TSG. Importantly, genome wide SNP analysis also showed that the genomic region on chromosome 21 containing the TSG RUNX1 is the most frequently affected region in AML-M0 (Figure 1A). The complete loss of RUNX1 or biallelic RUNX1 mutations observed in the majority of patients with UPD in chromosome 21 was in line with mitotic recombination being the mechanism of homozygosity.8 On the other hand, deletion of one RUNX1 allele appears to co-exist with complex patterns of LOH (Table 1 and 2). In all, we detected 18 cases (35 %) with RUNX1 mutation of which 15 were biallelic. This number is higher than in previous studies.6 Most of the mutations affected the runt domain and are considered to result in loss of DNA binding ability.32,33 Interestingly, two of the RUNX1 mutations were found in the transactivation domain. To our knowledge, C-terminal RUNX1 mutations have not been reported in AML-M0, although a few cases in other AML subtypes have been reported and they are frequent in myelodysplastic syndrome.6,34,35

Table 3.Summary of molecular findings and main cytogenetic abnormalities.

As over 60% of the AML-M0 cases retained normal RUNX1, we screened the cohort for mutations in other hematopoietic transcription factors frequently implicated in AML. Mutations in CEBPA, CEBPB and NPM1 were rare, showing that AML-M0 is largely unrelated to these provisional World Health Organization (WHO) subgroups.36 In fact, only a few patients (patients 35, 51 and 59) classified in this study as having FAB AML-M0 would not be considered to be part of the WHO subgroup with minimally differentiated leukemia. Contrary to a study showing a high (23%) frequency of mutation in AML-M0,37 we and others found no mutations in SPI1.13,38 It seems that mutations of these transcription factors are not an alternative to RUNX1 mutation in AML-M0. However, as we previously reported, ETV6 mutations are infrequent alternatives to RUNX1 mutation in this cohort.21

FLT3, RAS and PTPN11 are genes of the RAS pathway and mutations in these genes heve been reported to collaborate with RUNX1 mutations in the pathogenesis of AML by providing a proliferative advantage to the cells.10–14 We detected a higher frequency of RAS (12%) and PTPN11 (12%) mutations in AML-M0 than in previous studies (Table 3).11–13 Mutation frequencies for FLT3 (19%) were in accordance with other published data for AML and AML-M0.10,39–41 It is possible that the number of cases involving these genes is even higher, since we sequenced only mutational hotspots. Interestingly, mutations in FLT3, RAS and PTPN11 were absent in patients showing a deletion of the NF1 region. This result is in line with the view that deletion of NF1 might be an alternative to activation of the RAS signaling pathway in AML-M0. Finally, we detected three cases with mutation in JAK2, another gene involved in cell proliferation (Table 3). JAK2 mutations occur frequently in myeloproliferative disorders and less commonly in myelodysplastic syndrome and de novo AML.42,43 Though two of our cases had a previous history of hematopoietic disorders, this result could indicate some relation between AML-M0 and JAK2 mutation (Table 1).

Contrary to previous findings, mutation of RAS pathway-related genes did not correlate with RUNX1 mutation.14,44 We also did not find a positive association between -7/del(7q) and RUNX1 mutation or a negative one between del(5q) and RUNX1 mutation as previously reported in myelodysplastic syndrome,44 although a similar trend was noticeable. In fact, the only mutation associated with mutations in RUNX1 was trisomy 13, as observed by us and others.19,34 Trisomy 13 is also correlated with higher FLT3 expression and is probably another factor contributing to proliferative advantage in AML-M0.

TdT expression is a common characteristic of CD34 immature AML and is associated with a poor prognosis.45 Remarkably, RUNX1 mutation and TdT expression were strongly correlated. TDT encodes a DNA polymerase normally expressed in pre-B and pre-T lymphocytes during early differentiation.46 Recently, we found, by gene expression profiling, that AML-M0 RUNX1 mutants showed a characteristic B-lymphocyte signature, which could explain this correlation.24 TdT expression in leukemia is frequently assessed by immunophenotyping. Thus screening for RUNX1 mutations in AML-M0 at diagnosis might become possible by measuring TdT.

In conclusion, whole genome SNP analysis confirmed our previous findings that events leading to partial UPD are a major cause of mutation homozygosity in AML-M0.8 In fact, recent reports show that UPD is common in AML, myelodysplastic syndrome and myeloproliferative disease,47–50 suggesting that this is a general mechanism leading to loss of TSG activity in hematologic disorders. The observed heterogeneity in chromosomal losses in AML-M0 without RUNX1 mutation suggests that not one but several genes may be alternatives to RUNX1 mutation. Conversely, mutations related to cell proliferation, though genetically diverse, affect mainly the RAS pathway. Combining trisomy 13 and hemizygous loss of NF1 with the mutations found in FLT3, NRAS, JAK2 and PTPN11, we detected a strikingly high frequency (63%) of cell proliferation-related mutations. Importantly, we showed a strong association between TdT expression and RUNX1 mutation suggesting that TdT expression may serve as a surrogate marker for RUNX1 mutation in AML-M0.

Acknowledgments

we would like to thank Yavuz Ariyurek, Monique van Velzen, Sacha Lind and Nicos Lakenberg for experimental support, Liesbeth Hameetman for critical reading of the manuscript and Stefan White for MLPA.

Footnotes

• The online version of this article contains a supplementary appendix.
• Authorship and Disclosures FPGS contributed to the conception and design of the study; acquisition, analysis and interpretation of data, and drafted the article; IA, BM, GB-M, HW, RV contributed to the acquisition and analysis of data; HV, EWAM, PJMV, HCK-N, WRS, W-DL and MG-G contributed to interpretation of data, critical manuscript revision, conception and design of the study. All authors revised the article critically for important intellectual content approved the final version to be published.
• The authors reported no potential conflicts of interest.
• Funding: this work was supported by grants from the Calouste Gulbenkian Foundation and the Foundation for Science and Technology (Portugal) to FPGS.

• Received March 26, 2009.
• Revision received May 4, 2009.
• Accepted June 1, 2009.