Sequencing Acute Myeloid Leukemia Genomes with “Next Generation” Technologies.

For most patients with a sporadic presentation of acute myeloid leukemia (AML), neither the initiating nor the progression mutations responsible for disease are known. Recent attempts to identify key mutations with directed sequencing approaches, or with array-based genomic studies, have had limited success, suggesting that unbiased whole genome sequencing approaches may be required to identify most of the mutations responsible for AML pathogenesis. Until recently, whole genome sequencing has been impractical due to the high cost of conventional capillary-based sequencing and the large numbers of enriched primary tumor cells required to yield the necessary genomic DNA for library preparation. “Next Generation” sequencing approaches have changed this landscape dramatically. Using the Solexa/Illumina platform, we have now sequenced the genomic DNA of highly enriched tumor cells and normal skin cells obtained from a carefully selected patient with a typical presentation of FAB M1 AML. We obtained 98.2 billion bases of sequences from the cytogenetically normal tumor cell genome (32.7 fold haploid coverage), and 41.8 billion bases of sequence from the normal skin genome (13.9 fold coverage). Using these data, we detected diploid sequence coverage of 91% of 46,320 heterozygous SNPs, defined in the tumor genome (by array-based genotyping), and 83% diploid coverage of the skin genome. Of 2,647,695 well-supported single nucleotide variants in the tumor genome, 2,588,486 (97.7%) were also detected in the patient’s skin genome, defining them as inherited. From the remaining variants, 8 have been fully validated as somatic mutations by conventional capillary sequencing using PCR-generated amplicons. We also detected somatic mutations in the FLT3 (ITD) and NPM1 genes (a classic NPMc mutation). Based on deep read-count data of the novel variants on a 454 sequencer, we hypothesize that all of the mutations are in virtually all of the tumor cells, and all were retained at relapse 11 months later, suggesting that a single dominant clone contained all of the mutations. None of the novel mutations has previously been detected in AML cases (and none were found in any of 187 additional AML cases studied here). A number of additional potential somatic mutations in regions lying near genes (but not altering coding sequences) are currently being validated and tested for recurrence in other AML samples. Whole genome sequencing of a second M1 AML genome is now underway. These results demonstrate the power of unbiased whole genome sequencing approaches to discover cancer-associated mutations in novel candidate genes.