High Resolution Array-Based CGH and SNP Studies of AML Genomes.

To test if small deletions or amplifications (ie. below the resolution of cytogenetics) exist in bone marrow-derived tumor DNA from acute myeloid leukemia (AML) patients (pts), we used a dense tiling path array comparative genomic hybridization (aCGH) platform consisting of 386,165 unique oligomers spaced evenly at ∼6Kb intervals across the genome. We analyzed 144 adult de novo AML pts; 64 had normal karyotypes, and 80 had 1 or 2 clonal aberrations. Similar numbers of FAB M0/1, M2, M3, and M4 pts were included, and all samples had >30% blasts (median=72%). To generate a cancer-free control set of data, we also analyzed 23 DNA samples from normal individuals matched for age and ethnicity, and with no history of cancer. Both the tumor and cancer-free control DNA samples were co-hybridized with a pool of control DNAs from blood of 4 healthy young males. To define the sensitivity and specificity of the aCGH platform, we examined its ability to detect cytogenetically defined chromosome gains and losses. Of the 33 gains and losses present in >20% of metaphases, 29 (88%) were detected by aCGH. Of the 20 gains and losses present in ≤20% of metaphases, aCGH detected only 5 (25%). Three of 63 (4.8%) balanced translocations [t(15;17), t(8;21), t(9;11)] were detected using aCGH, indicating that breakpoints of some translocations contained small deletions. Further, we identified many previously described germline copy number variants (CNVs) in both the AML pts and cancer-free controls. To improve our ability to define even smaller somatic microdeletions and amplifications, we tested 20 AML pts using CGH arrays containing 1.5 million probes per genome (average probe spacing 1.5 Kb). To preclude detection of germline CNVs, the higher resolution CGH experiments were performed comparing tumor and skin-derived DNA from the same patient. These same sample pairs were also analyzed individually with the Affymetrix 500K SNP arrays. Using stringent criteria to define abnormal segments, we identified 64 altered loci in the 20 AML pts that were not apparent cytogenetically, and that contained ≥1 gene. SNP arrays confirmed aCGH findings in 7/9 loci >100 Kb, and in 1/55 loci <100 Kb in size. In addition, SNP arrays revealed copy number neutral loss of heterozygosity of the 11p arm in 2/20 AML pts, indicating partial uniparental disomy (UPD) involving this region. We also detected somatic deletions in the T cell receptor (TCR) (n=3/20) and immunoglobulin heavy chain (n=1/20) genes, including a homozygous deletion measuring 4.3 Kb in size. The remaining loci identified with the 1.5M oligo aCGH platform were validated using quantitative PCR with matched tumor and germline DNA. Only 5/60 putative calls were validated using this approach, and include a deletion of IGFBP2, and amplifications of CROP, CPEB4, HOMER1, and ZNF148. In summary, 13 loci containing genes have been validated by SNP arrays or qPCR. No recurrent deletions or amplifications were found in the 20 AML pts. Thus, an additional 74 AML pts are being screened for evidence of recurrence at these loci. Our data suggest that an ultra-dense platform may be required to detect the majority of somatic copy number changes in AML genomes, and that UPD is relatively rare in AML pts, occurring in ∼10% of pts, and recurrent only in the 11p region.