The basic leucine zipper (bZIP) transcription factor CCAAT/enhancer-binding protein-α (C/EBPα), encoded by the CEBPA gene, is essential for myeloid development.1 In acute myeloid leukemia (AML), interference with C/EBPα function may occur through several mechanisms.2 Mutations in the CEBPA coding region itself are observed in approximately 8% of AML. These mutations either involve N-terminal frameshift abnormalities, leading to increased translation of a 30 kDa isoform with dominant-negative properties, or in-frame insertions or deletions disrupting the bZIP domain located in the C-terminus.2 In many cases of AML, both types of mutations are found in the same leukemic cell, frequently involving both alleles, that is N-terminal mutation of one allele and C-terminal disruption of the other. In a previous gene expression profiling (GEP) study of 285 cases of de novo AML, two expression clusters were highly associated with CEBPA mutations.3 The majority of CEBPA mutant tumors in these clusters carried both an N-terminal and a bZIP mutation. Interestingly, in three cases only one of the two types of mutations was present, while a wild-type allele appeared to be absent. Here, we combined fluorescence in situ hybridization (FISH) experiments with single nucleotide polymorphism (SNP) array hybridization to investigate two potential mechanistic explanations for the existence of homozygous CEBPA mutations in these particular cases: deletion of the wild-type allele, or uniparental disomy (UPD) as a result of mitotic recombination.4
CEBPA mutational analysis by PCR and nucleotide sequencing has been described.3 These analyses revealed 17 mutant specimens, of which 3 carried homozygous mutations, in the cohort of 285 cases of AML.3 Two leukemias harbored homozygous mutations in the CEBPA bZIP region, whereas in the other case the mutation was found in the N-terminus (Table 1).
To investigate whether homozygous CEBPA mutations correlated with loss of the wild-type allele through deletion, we carried out dual color FISH using three BAC clones RP11-270I13, RP11-547I3 and RP11-1150B17; Supplementary Figure 1) located on 19q13.11 as probes. In all three specimens with homozygous mutations, signals for both homologues were detected in more than 95% of cells (data not shown), excluding deletion as the underlying mechanism for homozygosity.
We next used Affymetrix (Santa Clara, CA, USA) Mapping 250K NspI SNP arrays to assess the CEBPA locus in more detail. We investigated 14/17 AML cases with CEBPA mutations, including the 3 AMLs with homozygous CEBPA abnormalities. Data were imported into SNPExpress, a software tool developed in our department (Sanders et al., submitted). Mapping of probe sets to genes was done according to data provided by Affymetrix. On the arrays used, none of the probe sets mapped directly to the CEBPA gene. Probe set SNP_A-2236362, associated with neighboring gene CEBPG (Supplementary Figure 1), was therefore used as a marker for the CEBPA locus. We first determined copy numbers of chromosome 19q13.11 for each patient based on these SNP arrays using dChipSNP. This analysis confirmed the results obtained by FISH, as the calculated copy number of the region containing the CEBPA gene fluctuated closely around 2 in all 14 leukemias (Figure 1). We subsequently assessed potential regions of loss of heterozygosity (LOH) using a hidden Markov model. Strikingly, we found extensive regions of LOH of chromosome 19, including the CEBPA locus, in the three samples with homozygous CEBPA aberrations. In all three leukemias, LOH included the majority of the 19q arm, and stretched until the telomeres (Figure 1). In contrast, no LOH of the CEBPA locus was apparent in the 11 samples carrying heterozygous CEBPA mutations. These results demonstrate that copy number neutral LOH of 19q13.11 is associated with homozygous CEBPA mutations, which is indicative of UPD through mitotic recombination in this region.
UPD has recently emerged as a mechanism involved in malignancy, including AML.5, 6, 7 Genome-wide studies using high-resolution SNP arrays have indicated that approximately 20% of AMLs with a normal karyotype show either interstitial or terminal UPD.4, 5, 6, 7, 8 Acquired UPD is associated with removal of wild-type alleles of genes commonly mutated in AML, including RUNX1, FLT3 and WT1.5 Previously, a single case of AML with a homozygous CEBPA bZIP mutation and UPD has been reported.5 Our results imply that terminal UPD involving mutant CEBPA alleles is a recurrent mechanism.
In a previous unsupervised GEP study, the three AMLs described here clustered together with other CEBPA mutant AML cases in two distinct expression clusters.3 The common finding in these clusters was that the CEBPA mutant AMLs carried mutations in both alleles and consequently did not express wild-type CEBPA. Although the functional implication of replacement of wild-type CEBPA remains elusive, its absence, resulting either from biallelic mutations or from homozygous mutations, seems to be critical for leukemic development of those cases. It is currently unclear whether homozygous CEBPA mutations have a specific differential significance as opposed to biallelic heterozygous aberrations. It will be interesting to investigate whether loss of wild-type alleles of additional genes on chromosome 19 plays a role in leukemogenesis in these cases.
Another interesting question is whether the initially mutated allele in the AMLs with homozygous CEBPA mutations was acquired, or was already present in the germline. This would particularly be of interest because familial N-terminal mutations have been described. Unfortunately, we were not able to address this issue, as normal or remission control material was not available for any of the three patients.
In conclusion, we describe the association of homozygous CEBPA mutations in three cases of AML with mitotic recombination involving the CEBPA locus. Our findings are consistent with recent reports suggesting the frequent occurrence of UPD in AML, and imply that this mechanism is responsible for a substantial proportion of cases with homozygous CEBPA aberrations.
We thank Roel Verhaak for assistance in data analysis. This work was supported by a grant from the Dutch Cancer Society ‘Koningin Wilhelmina Fonds’.