Recurrent deletions of IKZF1 in pediatric acute myeloid leukemia

Abstract

IKAROS family zinc finger 1/IKZF1 is a transcription factor important in lymphoid differentiation, and a known tumor suppressor in acute lymphoid leukemia. Recent studies suggest that IKZF1 is also involved in myeloid differentiation. To investigate whether IKZF1 deletions also play a role in pediatric acute myeloid leukemia, we screened a panel of pediatric acute myeloid leukemia samples for deletions of the IKZF1 locus using multiplex ligation-dependent probe amplification and for mutations using direct sequencing. Three patients were identified with a single amino acid variant without change of IKZF1 length. No frame-shift mutations were found. Out of 11 patients with an IKZF1 deletion, 8 samples revealed a complete loss of chromosome 7, and 3 cases a focal deletion of 0.1–0.9Mb. These deletions included the complete IKZF1 gene (n=2) or exons 1–4 (n=1), all leading to a loss of IKZF1 function. Interestingly, differentially expressed genes in monosomy 7 cases (n=8) when compared to non-deleted samples (n=247) significantly correlated with gene expression changes in focal IKZF1-deleted cases (n=3). Genes with increased expression included genes involved in myeloid cell self-renewal and cell cycle, and a significant portion of GATA target genes and GATA factors. Together, these results suggest that loss of IKZF1 is recurrent in pediatric acute myeloid leukemia and might be a determinant of oncogenesis in acute myeloid leukemia with monosomy 7

Introduction

Acute myeloid leukemia (AML) is a heterogeneous disease characterized by dysregulated hematopoiesis initiated by recurrent non-random genetic aberrations.21 Despite major progress in improving treatment over the past decades, AML remains a life-threatening malignancy in children. The 5-year event-free survival (EFS) rates of pediatric AML are currently approximately 50%–60% and the overall survival (OS) rate approximately 70%.43 Treatment stratification is mainly based on early response to treatment and prognostic relevant genetic abnormalities. Although many driving recurrent genetic aberrations causing AML have been defined, in approximately 20% of the pediatric AML patients the driving genetic events remain unidentified.3

The DNA-binding protein IKAROS family zinc finger 1 (IKZF1) is a zinc finger containing transcription factor that is encoded in 8 exons on chromosome 7 (7p12.2). IKZF1 functions upon homo-dimerization and acts as a chromatin remodeling protein involved in several chromatin remodeling complexes, such as nucleosome-remodeling and histone deacetylation (NuRD).5 Alternative splicing of IKZF1 pre-mRNA gives rise to several isoforms.6 The functional differences among the various isoforms of IKZF1 are mostly due to alterations of zinc finger structures required for DNA binding (4 zinc fingers in the N-terminal domain) or protein-protein interactions (2 zinc fingers in the C-terminal domain).76 Some of these splice variants lack the ability to properly bind DNA and act as dominant-negative repressors when dimerized with functional isoforms. Loss of Ikzf1 function in mice homozygous for an Ikzf1 null allele or homozygous for a dominant-negative isoform of Ikzf1 results in an early arrest of lymphoid differentiation with lack of T and B lymphocytes and NK cells, as well as their earliest identifiable progenitors.98 In heterozygous Ikzf1 mutant mice, lymphocytes appear normal immediately after birth, but proliferation increases and differentiation arrest occurs rapidly, leading to the development of leukemia and lymphoma within three months after birth.10 In humans, heterozygous loss of IKZF1 is found in 15% of B-cell precursor acute lymphoblastic leukemia (BCP-ALL) patients; in 70%–80% of BCR/ABL1 (t(9;22)(q34;q11)) positive BCP-ALL patients and in 33% of BCR/ABL1 negative BCP-ALL patients.1511 In BCR/ABL1 ALL patients, heterozygous loss of IKZF1 associates with a poor clinical outcome. In T-ALL heterozygous loss of IKZF1 is found in only 4%, but it remains unclear if this has any prognostic value.1211

While the role of IKZF1 in lymphoid differentiation and ALL is firmly established, the role of IKZF1 in myeloid differentiation is less clear. However, there are some strong indications that IKZF1 is also involved in myeloid differentiation. In early myeloid progenitors, loss of IKZF1 function prolongs cell survival.16 During erythropoiesis, IKZF1 supports survival of the erythroid lineage, and promotes erythrocyte differentiation at the expense of granulocyte and monocyte differentiation.1917 In early megakaryopoiesis, IKZF1 represses differentiation by inhibiting genes associated with the NOTCH pathway and later on in differentiation towards megakaryocytes by controlling transcriptional regulators, such as GATA1 and RUNX1.2018 Although IKZF1 is not required for the initial differentiation towards granulocytes and monocytes, IKZF1 represses differentiation of the basophilic granulocyte lineage and promotes early maturation and survival of the neutrophil granulocyte lineage.222117

In pediatric AML, monosomy 7 is found in only 4%–5% of patients. The 5-year OS and EFS of pediatric AML patients with monosomy 7 with or without additional cytogenetic or chromosomal aberrations are poor.23 In adult AML, monosomy 7 is the most frequent single monosomy and has a similar poor 4-year OS as compared to other single autosomal monosomies.24 Deletions of the short arm of chromosome 7, 7p (del7p) were recurrently found in adult de novo AML and secondary AML developed from myelodysplastic syndrome (MDS) or myeloproliferative neoplasms (MPN).2725 Furthermore, 7p deletions were mapped to the IKZF1 locus and it was shown that loss of IKZF1 is acquired during the progression of MPN to secondairy AML, indicating that loss of IKZF1 contributes to the transformation from MPN to AML.25

Together, these studies demonstrate that IKZF1 may play a role in myeloid differentiation and that loss of IKZF1 may contribute to myeloid oncogenesis. We, therefore, hypothesized that, in pediatric AML patients with monosomy 7, IKZF1 may be an important player. In this study, we analyzed the frequency of IKZF1 deletions in a pediatric AML cohort, and studied the difference in gene expression of cases with focal IKZF1 deletions, and cases with monosomy 7.

Methods

Patients’ samples

Samples were provided by the Dutch Childhood Oncology Group (DCOG, the Netherlands), the AML–Berliner-Frankfurt-Münster Study Group (Germany and Czech Republic), the Saint-Louis Hospital (France), and the Royal Hospital for Sick Children (UK). Isolation of genomic DNA and total cellular RNA was performed using Trizol reagent. Each study group provided morphological and cytogenetic classification and clinical data. Institutional review board approval for these studies had been obtained in the participating centers.

Detection of deletions and mutations of IKZF1

Multiplex ligation-dependent probe amplification (MLPA) was performed using the p335-B1;ALL-IKZF1 kit (MRC Holland, Amsterdam, the Netherlands; data available at http://www.mlpa.com). The data were analyzed with GeneMarker v. 1.85 (SoftGenetics, State College, Pennsylvania, USA). Data were normalized to reference probes and control samples. A deletion was defined as a peak ratio below 0.75; an amplification was defined as a peak ratio above 1.25.

Array comparative genomic hybridization (Array-CGH) was performed using the human genome CGH Microarray 105K (Agilent Technologies, Santa Clara, California, USA) according to the manufacturer’s protocol; data were analyzed with Genomic Workbench v. 5.0.14 (Agilent Technologies, Santa Clara, California, USA).

Direct sequencing was used to analyze mutations or frameshifts in exons 4, 5 and 8, which are found to be hotspot areas of mutations in pediatric BCP-ALL (primers are listed in Online Supplementary Table S1).2813

Cytogenetic and molecular characterization

Reverse-transcription-PCR was performed on hotspot areas as previously described by Balgobind et al.,29 and to analyze the frequency of the BCR/ABL1 translocation (primers are listed in Online Supplementary Table S1).

Gene expression profiling

Gene expression profiling (GEP) data were available from earlier studies.30 Quantitative reverse-transcription-PCR (qRT-PCR) analysis was performed on a subset of patients to analyze the correlation between GEP data and qRT-PCR on relevant genes in patients of different cytogenetic subgroups and a variety of expression measured with GEP. The mRNA expression was determined using the average cycle threshold (Ct) in comparison to expression levels of GAPDH, using the comparative cycle threshold method.

Statistical analysis

Statistical analyses for disease outcome and correlation were performed with IBM SPSS 21 (IBM, Armonk, New York, USA). To assess clinical outcome, complete remission (CR; defined as less than 5% blasts in the bone marrow, with regeneration of tri-lineage hematopoiesis, plus absence of leukemic cells), probability of event-free survival (pEFS; defined as the time between diagnosis and first event including non-responders calculated as an event on day 0), and probability of overall survival (pOS; defined as the time between diagnosis and death) were analyzed. Both pEFS and pOS were estimated by the Kaplan-Meier method and differences compared using log rank tests. The cumulative incidence of non-response or relapse (pCIR; defined as time between diagnosis and relapse, and with non-responders included as an event on day 0) was analyzed by the Kalbfleisch and Prentice method, and compared with the Gray’s test. P<0.05 was considered significant.

Differential expression analyses between groups were conducted using the R-package ShirnkBayes31 at the probe level. In contrast to commonly used two-sample tests, this test is proven to be powerful for small sample sizes. Bayesian false discovery rate (BFDR)<0.05 was considered statistically significant.

Results

Identification of IKZF1 deletions in pediatric AML

The cohort of newly diagnosed pediatric AML patients with available peripheral blood or bone marrow samples taken at initial diagnosis included in this study has been previously described by Hollink et al. (n=293, age 0–18 years).30 Of 293 well-characterized pediatric AML cases, 258 with available high quality DNA were analyzed with MLPA using 250ng of DNA. Cytogenetic aberrations in this group were representative for the frequencies seen in pediatric AML.323 In addition, 14 healthy control samples as negative controls, and one pediatric ALL sample known to harbor an IKZF1 deletion as positive control were analyzed. A subsequent MLPA was considered if all reference probes had a peak ratio between 0.75 and 1.25, except where a deletion or amplification of a reference probe could be explained by an aberration visible in the karyotype. The P335-B1;ALL-IKZF1 MLPA kit contains a probe in each of the 8 exons of IKZF1. All probes on chromosome 7 were analyzed individually, including reference probe 15370-L13762 located on 7q11.23 in P450 cytochrome oxidorectase (POR). Additional high-resolution array-CGH was performed on patients with an IKZF1 deletion detected by MLPA, but without knowledge of the existence of monosomy 7, and one additional case of monosomy 7 from which no material was available for MLPA analysis. Mutational screening of exons 4, 5 and 8 of IKZF1 was performed on the same cohort of 258 patients. This revealed no patient harboring a frameshift mutation, which has been described as damaging in BCP-ALL. We did identify 3 patients with a heterozygous point mutation, p.V382M, which was predicted as tolerated by SIFT analysis and as benign by PoyPhen analysis, and p.G158S and p.L188P, which were predicted as deleterious by SIFT analysis and probably damaging by PolyPhen analysis, but not resulting in a truncated protein. Seventy-seven of 258 patients presented with synonymous SNP rs61731355, 17/258 with synonymous SNP rs61731356, and 1/258 with non-synonymous SNP rs376657964 (p.T333A), which is in line with the presence of these SNPs in the normal population (21%, 4% and 1%, respectively).33

Table 1.Patients’ characteristics of IKZF1 deleted pediatric AML cases (n=11).

In total, 10 patients carried a heterozygous IKZF1 deletion as identified by MLPA, and one additional case as identified by array-CGH (Table 1). As determined by MLPA, of 258 patients, 9 cases showed an IKZF1 deletion in all 8 exons, and one case showed a deletion of exon 1–4 with a normal peak ratio in the remaining exons (Figure 1). Four of these 10 cases showed a monosomy 7 with conventional karyotyping (Table 1). Using array CGH on 7 patients, we identified 2 cases with monosomy 7 in patients with an unknown karyotype, 2 patients with monosomy 7 of whom the karyotype was (wrongly) defined as cytogenetically normal, and 3 cases with a focal 0.1–0.9Mb deletion of 7p12.2, where the IKZF1 gene is located (Table 1 and Figure 2).

Figure 1.Heterozygous IKZF1 deletions in pediatric AML. (A) Multiplex ligation-dependent probe amplification results of a patient with heterozygous deletion of exons 1–8 of IKZF1 including the reference probe in POR located on 7q11.23. (B) MLPA results of a patient with heterozygous deletion of exons 1 through 4 of IKZF1. 1–8 indicates IKZF1 exons 1–8; D: reference D-fragment 96, a control probe for denaturation during the procedure; P: POR (7q11.23); X: X-chromosome specific fragment, a reference probe for gender; Y, Y-chromosome specific fragment, a reference probe for gender; Z: ZFY (Yp11.31), a reference probe for gender.

Figure 2.Array comparative genomic hybridization analysis of monosomy 7 and focal IKZF1 deletions. DNA isolated from pediatric AML samples taken at initial diagnosis and containing IKZF1 deletion based on MLPA results were analyzed using array CGH. Representative samples for monosomy 7 (patient #4, upper panel) and IKZF1 focal del7p12.2 deletions (patient #2, #3 and #10, lower panel) are shown.

Combining the data gathered from MLPA, karyotype and array CGH, we defined 11 patients with an IKZF1 deletion, 8 of which showed complete loss of one copy of chromosome 7 (monosomy 7), and 3 of which showed a focal IKZF1 deletion. As shown in Figure 2, patient #10 demonstrated a focal deletion (~0.1Mb) including exon 1–4 of IKZF1. The focal deletion in patient #2 included the IKZF1 gene and part of the C7orf72 gene (~0.3Mb), whereas patient #3 showed a focal deletion (~0.9Mb) including the IKZF1 gene and the genes for VWC2, ZPBP, C7orf72 and DDC. Together, these results map the deletions in these patients to the IKZF1 gene locus as the common deleted region.

Characteristics of IKZF1 deleted cases in pediatric AML

IKZF1 deleted cases did not differ significantly from the other pediatric AML cases in terms of age at diagnosis [median age 9.1 years (range 0.7–14.1) vs. 9.5 years (range 0.1–18.5) respectively, P=0.41]; nor in terms of sex (55% female vs. 42% female; P=0.41), and white blood cell count at diagnosis (median 30.2×10/L vs. 47.5 ×10/L; P=0.24) (Online Supplementary Table S2). All AML patients were treated with intensive collaborative group cytarabine-anthracycline-based pediatric AML treatment protocols. There was no significance difference in clinical outcomes between IKZF1 focal deleted, monosomy 7 and other AML cases. The 3-year pOS in IKZF1 deleted cases (n=11) was 70%±14% versus 63%±3% (P=0.82) in cases with 2 IKZF1 alleles (n=231); the 3-year pEFS was 36%±15% versus 46%±3% (P=0.87), and 3-year pCIR was 64%±16% versus 36±3% (P=0.09), respectively (Online Supplementary Figure S1).

No specific morphology subtypes based on the French–American–British (FAB) classification were related to IKZF1 deletions. IKZF1-deleted samples showed none or various additional somatic mutations, most frequently activating the RAS pathway with mutations in NRAS or PTPN11 (n=4) (Table 1). As in BCP-ALL, IKZF1 deletions are associated with BCR-ABL1 fusions. Therefore, in addition we screened the IKZF1-deleted cases for the BCR/ABL1 fusion and all were negative.12

IKZF1 haploinsufficient AML cells display an AML-specific gene expression signature

Original gene expression data are available in the Gene Expression Omnibus repository (http://www.ncbi.nlm.nih.gov/geo;accession GSE17855). Focal IKZF1 deletions are also found in mixed phenotype acute leukemia.34 To ensure the included cases were AML patients and not patients with a mixed ALL and AML phenotype, we used a previously defined AML versus ALL classifier.35 Patients with focal IKZF1 deletions (dark green triangles) and patients with monosomy 7 (light green squares) clustered amongst the other AML patients (red circles) and away from the ALL patients (blue circles), based on the expression signatures of the AML-specific genes (Figure 3A), the ALL-specific genes (Figure 3B), and the combined AML/ALL-specific genes (Figure 3C). This indicates that the IKZF1 haploinsufficient AML cells of included patients indeed displayed pure AML cell gene expression profiles. Only of one of the 3 patients with a focal IKZF1 deletion immunophenotypical data was available (patient #2); this patient was classified as PML/RARA negative AML FAB M3V (TdT/MPO/cyCD79a/CD34/HLADR/CD19/CD10/CD7/CD13/CD33).

Figure 3.IKZF1 haploinsufficient acute myeloid leukemia (AML) cells display an AML-specific gene expression signature. AML (n=297, red) and B-cell acute lymphoblastic leukemia (B-ALL) (n=107, blue) patients were projected onto the first two principal coordinates, obtained via multidimensional scaling of expression values of AML specific genes (A), ALL specific (B) genes and the combination of these (C) (see Thomas et al.³⁵). Focally deleted IKZF1 (dark green triangles) and monosomy 7 (light green squares) AMLs were located among other AMLs and away from B-ALLs.

Similarity between gene expression profiles of monosomy 7 and focal IKZF1-deleted cases

Loss of IKZF1 function, a well-known tumor-inducing event in ALL, is conceivably one of the reasons for the recurrent heterozygous loss of chromosome 7, as this would include losing the IKZF1 locus on 7p12.2. We hypothesized that, if loss of IKZF1 were to be one of the driving leukemic events in monosomy 7, the gene expression signature of monosomy 7 AML cells may resemble those that only have a focal IKZF1 deletion. However, we anticipated that this similarity might be quite weak because of the heterogeneity across the samples in our cohort, both derived from somatic and germ-line genetic variation and the small sample size of monosomy 7 and IKZF1-deleted cases. Therefore, we performed genome-wide differential gene expression analyses using a robust Bayesian method implemented in the R-package ShrinkBayes.31 Probe sets were limited to those targeting unique genes or having a selectivity score of more than 0.8 (www.geneannot.com). Further statistical analyses were performed in R v. 3.0.0. (R scripts are available upon request.) We first compared gene expression profiles of monosomy 7 samples (n=8) to those of IKZF1-non-deleted samples (n=247). We detected 244 probe sets representing 198 genes differently expressed (BFDR<0.05) between these two groups (Figure 4A). As might be expected, a large proportion (44%) of these genes are located on chromosome 7, and are down-regulated (Figure 4B and Online Supplementary Table S3).

Figure 4.Monosomy 7 and focal IKZF1 deletion associate with gene expression profiles with significant similarity. (A) Analysis of differences in gene expression between monosomy 7 samples (n=8) and IKZF1-non-deleted samples (n=247). Genes are ranked on Bayesian false discovery rate (BFDR) (y-axis) and binned in their chromosome of origin (x-axis). Using a BFDR of 5% (dashed line), a total of 198 genes are detected differentially expressed, of which 111 are not encoded on chromosome 7. (B and C) Mean differences per gene between monosomy 7 and non-IKZF1-deleted samples versus those of IKZF1 focal deletions and non-IKZF1-deleted samples for chromosome 7 encoded genes (B) and non-chromosome 7 encoded genes (C). Pearson’s correlation coefficients (r) are indicated in the plot, together with accompanying P values. (D) Bootstrap analyses of the correlation between monosomy 7 cases (n=8) and 3 randomly picked IKZF1-non-deleted samples (n=247). The red dashed line indicates correlation coefficient from the analyses displayed in Figure 4C. (E) Top correlating genes [dashed elipse in (C)] between monosomy 7 and focal IKZF1 deletions. Highlighted in red are those reported to be positively acting on leukemic cell self-renewal or cell cycle.

Next, we compared the mean differences in expression of these 198 genes in monosomy 7 samples (n=8) to those in samples with IKZF1 focal deletions (n=3) (Figure 4C). For non-chromosome 7 genes, highly significant positive correlation was observed in gene expression differences (r=0.50, P=3e–10). A less significant positive correlation (r=0.33, P=5e–4) was observed for chromosome 7 genes (Figure 4B). To further test that the correlation of gene expression changes between monosomy 7 and focal IKZF1 deletions is not based on chance, we performed a bootstrap analyses taking 3 random patients and comparing them to the 8 patients with monosomy 7. None of 100,000 comparisons displays a correlation approaching the value of 0.5 found for the patients with focal IKZF1 deletions (Figure 4D).

We next examined which non-chromosome 7 genes were responsible for the similarity between monosomy 7 and focal IKZF1 deletion gene expression profiles. Genes that substantially contribute to the correlation between monosomy 7 and IKZF1-only deleted gene expression differences are up-regulated in both groups (Figure 4C). Interestingly, the gene most clearly up-regulated in both monosomy 7 and IKZF1-focal deleted samples was Hemogen (HEMGN), the overexpression of which has been described to induce expansion of myeloid progenitor cells in a murine model.36 Other genes with known pro-leukemic functions in the top-correlating list are Four and a half LIM domains 2 (FHL2), Frizzled 6 (FZD6) and SET binding protein 1 (SETBP1) (Figure 4E). Very strong correlations (r=0.9117–0.9562, P<0.0001) were found between the GEP data and qRT-PCR analyses for HEMGN and FHL2, and strong correlations (r=0.7913–0.7962, P<0.0001) were found for SETBP1 and FZD6 (n=30) (Online Supplementary Figure S2).

As upregulation of HEMGN in IKZF1 hemizygous cells is relative to a fairly heterogeneous group, we stratified the control group in various more homogenous cytogenetic AML subgroups (Figure 5A). Median expression of HEMGN in each of those subgroups is clearly lower than in the IKZF1-deleted group. The same is true for FHL2, FZD6 and SETBP1 (Figure 5B–D), although it should be noted that FZD6 and SETBP1 are also highly expressed in patients with MNX1-ETV6 (t(7;12)). When excluding the patient harboring both an IKZF1 deletion and MNX1-ETV6 (t(7;12)), FZD6 and SETBP1 remain significantly more highly expressed in the IKZF1-deleted cases compared to the other pediatric AML subgroups (Figure 5E and F). These data indicate that upregulation of these genes in IKZF1-deleted samples is not confounded by specific cytogenetic aberrations in pediatric AML.

Figure 5.HEMGN, FHL2, FZD6 and SETBP1 expression in IKZF1-deleted patients as compared to other cytogenetically classified AML groups. (AD) Comparison of mRNA abundances in IKZF1-deleted samples to major cytogenetic AML subgroups revealed that the median expression of HEMGN, FHL2, FZD6 and SETBP1 in the IKZF1 deleted group is clearly higher than in each of the subgroups, except for FZD6 and SETBP1 in comparison to t(7;12)/MNX1-ETV6. (E and F) When excluding the patient harboring both the IKZF1 deletion and t(7;12)/MNX1-ETV6, FZD6 and SETBP1 remain significantly higher expressed in the IKZF1 -deleted cases compared to the other pediatric AML subgroups (P=0.003 and P=0.005, respectively).

Gene expression profiles of the IKZF1-deleted samples demonstrate enrichment in up-regulated GATA targets

HEMGN, the gene most clearly up-regulated in both monosomy 7 and IKFZ1-focal deleted samples, is transcriptionally activated by GATA1 in the myeloid cell line K562.37 We, therefore, investigated whether other genes that demonstrated a correlation between monosomy 7 and focally deleted IKFZ1 samples are also GATA1 targets in K562. For this purpose, we used a GATA1 ChIP seq dataset in K562 cells.38 From the 23 genes significantly up-regulated in the IKZF1-deleted samples, 11 genes had a GATA1 peak ratio larger than 25, indicating the gene being a GATA1 target (Figure 6A). Further analysis showed that whereas 13% of genes on our expression platform is a GATA1 target, 55% of genes that are up-regulated in IKZF1-deleted cases are GATA1 targets (P=3.5e–7) (Figure 6B). While GATA transcription factors exhibit diverse functions in cellular proliferation, differentiation and gene regulation, only GATA1, GATA2, and GATA3 are expressed in developing blood cells and are critical for hematopoiesis.4039 GATA1 (P=0.07) and GATA3 (P=0.006) were up-regulated in the IKZF1-deleted samples (n=11) as compared to the rest (n=247) (Figure 6C). Direct interaction of IKZF1 with GATA proteins has been previously demonstrated.41 Together, these data emphasize the similarity in gene expression between monosomy 7 and focal IKZF1-deleted pediatric AML patients.

Figure 6.Genes commonly up-regulated in IKZF1 focally deleted and monosomy 7 cases are enriched in myeloid GATA1 targets. (A) GATA1 chromatin interaction values (GATA1 peak value) in K562 cells of commonly up-regulated genes (n=23) in monosomy 7 and focal IKZF1 deletion (ΔIKZF1). (B) Proportion of all matching genes (n=22564) or ΔIKZF1 up-regulated (n=23) displaying a GATA1 chromatin interaction in K562 cells. P value from Fisher’s exact test. (C) mRNA expression of GATA1, GATA2 and GATA3 in IKZF1 deleted samples (red) versus IKZF1 WT samples (gray). P values according to Mann-Whitney tests.

Discussion

In the last ten years, IKZF1 has been widely studied in the context of B-cell differentiation and acute lymphoblastic leukemia, both in adults and children. There is growing evidence to indicate that IKZF1 also plays a role in various stages of myeloid differentiation. So far, the best indication that loss of IKZF1 may contribute to myeloid leukemogenesis are deletions of the short arm of chromosome 7 associated with MPN-preceded secondary AML in adults, where the commonly deleted region is mapped to the IKZF1 locus.25 Our finding of recurrent focal deletions that map to the IKZF1 gene locus in pediatric AML patients cells now provides additional support for a role for IKZF1 in myeloid leukemia.

Reduced IKZF1 gene function is a well-known recurrent event in BCP-ALL. In pediatric BCP-ALL, the overall frequency of focal IKZF1 deletions is approximately 15%, but is greatly enriched in BCR/ABL1 positive BCP-ALL at 70%–80%.4215 Here we report that focal IKZF1 deletions are also a recurrent event in pediatric AML, although with much lower frequency. In a cohort of 258 patients, representing all cytogenetic AML subgroups, we found 3 cases of a focal inactivating deletion of this gene (1.2%). In addition, 3 patients were identified with a single nucleotide variation leading to amino acid changes. In contrast to mutations found in BCP-ALL, no frame-shift mutations were detected, and the functional status of these SNV containing alleles remain to be determined.

In adult and pediatric BCP-ALL, in addition to focal IKZF1 deletions, monosomy 7 is a recurrent chromosomal aberration. In BCR/ABL1 positive BCP-ALLs, 16% (adult) and 13% (pediatric) of the cases of IKZF1 deletion can be attributed to monosomy 7.4214 In pediatric AML, monosomy 7 is a recurrent chromosome loss, in particular in myeloproliferative diseases, such as juvenile myelomonocytic leukemia that have progressed towards AML (40% of cases), but also in primary AML (4%–5% of cases).4323 In our pediatric de novo AML cohort, we found 8 cases (3.1%) of monosomy 7, consistent with earlier estimates. Across these monosomy 7 AML cases, we find various additional somatic mutations, most prominently PTPN11 (n=2) and NRAS mutations (n=2), adding up to half of the cases presenting with an activated RAS pathway.

We hypothesized that, in pediatric AML, as presumed in BCP-ALL, an important biological determinant for leukemogenesis in monosomy 7 patients is loss of the IKZF1 gene.44 Evidence for this hypothesis was obtained by comparing gene expression profiles of monosomy 7 and focal IKZF1 deletions. We found a statistically significant correlation on non-chromosome 7 genes. We attribute the changes in chromosome 7 genes to bystander effects of deletion of the entire chromosome 7, leading to a widespread reduction in chromosome 7 encoded genes, many of which can be assumed to no impact on the leukemogenic process. In contrast, the effects on non-chromosome 7 genes can be expected to be more enriched for leukemia-related changes in the genetic program due to loss of one or more chromosome 7 encoded driving oncogenic mutations. Therefore, the similarity of gene expression signatures between AML cases with monosomy 7 and those with a focal IKZF1 deletion suggests that IKZF1 is one of the driving oncogenic events in monosomy 7 pediatric AML. Further support for this hypothesis is the observation that pediatric AML patients harboring monosomy 7 have a worse disease outcome compared to those harboring 7q deletions.23 Our findings may also explain the prevalence of monosomy 7 and focal IKZF1 deletions in mixed phenotype acute leukemia, as loss of IKZF1 now appears to be able to contribute to both acute lymphoblastic and myeloid leukemogenesis.34

Several genes up-regulated in both monosomy 7 and focal IKZF1-deleted AML cases are previously implicated in leukemogenesis. In a large cohort of adult BCP-ALL, comparison between focal IKZF1 deletions and IKZF1 wild-type cases revealed upregulation of genes involved in the cell cycle, JAK-STAT signaling and stem cell self-renewal.45 We find upregulation of HEMGN, (a transcription factor implicated in myeloid self-renewal36), FHL2 (enhanced expression of which leads to enhanced cell cycle entry), FZD6 (shown to be involved in the pathologically reactivation of the Wnt/Fzd-mediated self-renewal signals that are enlisted during B-cell development and may be pathologically reactivated in the neoplastic transformation of mature B cells), and SETBP1 (overexpression of which promotes the self-renewal of myeloid progenitors4846). It is interesting that similar cellular processes are affected in the IKZF1 deleted cases, while none of the genes deregulated in IKZF1 deleted adult BCP-ALL are found deregulated in this pediatric IKZF1 deleted AML cohort.45

It is worthy of note that we find a strong enrichment of myeloid GATA1 targets among genes commonly up-regulated in focal IKZF1 deleted AML cases and monosomy 7 AML cases. We interpret this as a confirmation that the genes commonly up-regulated in IKZF1 focal deletions and monosomy 7 did not present themselves by chance. However, it is tempting to speculate that IKZF1 haploinsufficiency in pediatric AML is influencing GATA transcription factor function. Supporting this idea, IKZF1 overexpression in transgenic mice represses GATA1 expression, implying that loss of IKZF1 function may increase GATA1 function.21

Taken together, we find evidence for IKZF1 as a tumor suppressor gene in pediatric acute myeloid leukemia and suggest that IKZF1 deletion might be one of the driving events of monosomy 7 in various myeloid diseases.

Footnotes

• ↵* JDEdR and EB contributed equally to this work as 1st authors
• ↵† CMZ and MF contributed equally to this work
• The online version of this article has a Supplementary Appendix.
• FundingJdR was funded by Kinder Oncologisch Centrum Rotterdam (KOCR). AO was funded by KIKA project #109. MF was funded by KWF project EMCR 2012-5546. The authors declare no conflict of interest.
• Authorship and DisclosuresInformation on authorship, contributions, and financial & other disclosures was provided by the authors and is available with the online version of this article at www.haematologica.org.

• Received January 23, 2015.
• Accepted June 5, 2015.