DNA Sequencing of a Murine Acute Promyelocytic Leukemia (APL) Genome Using Next Generation Technology.

Abstract 3965

Poster Board III-901

We have recently established that whole genome sequencing is a valid, unbiased approach that can identify novel candidate mutations that may be important for AML pathogenesis (Ley et al Nature 2008, Mardis et al NEJM 2009). Acute promyelocytic leukemia (APL, FAB M3 AML) is a subtype of AML characterized by the t(15;17)(q22;q11.2) translocation that creates an oncogenic fusion gene, PML-RARA. Our laboratory has previously modeled APL in a mouse in an effort to understand the genetic events that lead to the disease. In our knockin mouse model, a human PML-RARA cDNA was targeted to the 5' untranslated region of the mouse cathepsin G gene on chromosome 14 (mCG-PR). The targeting vector was transfected into the RW-4 embryonic stem cell line, derived from a 129/SvJ mouse. The transfected RW-4 cells were injected into C57Bl/6 blastocysts, and chimeric offspring were bred to C57Bl/6 mice. F1 129/SvJ x C57Bl/6 mice were subsequently backcrossed onto the B6/Taconic background for 10 generations before establishing a tumor watch. About 60% of the mCG-PR mice in the Bl/6 background develop a disease that closely resembles APL only after a latent period of 7-18 months, suggesting that additional progression mutations are required for APL development. Array-based genomic techniques (expression array studies and high resolution CGH) have revealed some recurring genetic alterations that may be relevant for progression (i.e. an interstitial deletion of chromosome 2, trisomy 15, etc.), but gene-specific progression mutations have not yet been identified. To begin to identify these mutations in an unbiased fashion, we sequenced a cytogenetically normal, diploid mouse APL genome using massively parallel DNA sequencing via the Illumina platform. Since the tumor arose in a highly inbred mouse strain, we predicted that 15x coverage of the genome (approximately 40 billion base pairs of sequence) would be necessary to identify >90% of the heterozygous somatic mutations. We generated 2 Illumina paired-end libraries (insert sizes of 300-350 bp and 550-600 bp) and generated 59.64 billion base pairs of sequence with 3 full sequencing runs; the reads that successfully mapped generated 15.6x coverage. The sequence data predicted 87,778 heterozygous Single Nucleotide Variants (SNVs) compared to the mouse C57Bl6/J reference sequence, and 23,439 homozygous SNVs. Of the predicted heterozygous SNVs, 695 were non-synonymous (missense or nonsense, or altering a canonical splice site). Thus far, 80 of these putative non-synonymous SNVs have been further analyzed using Sanger sequencing of the original tumor DNA vs. pooled B6/Taconic spleen DNA and pooled129/SvJ spleen DNA as controls. 37/80 were shown to be false positive calls, and 37 were inherited SNPs from residual regions of the129/SvJ genome. 6/80 were present only in the tumor genome, and were candidate somatic mutations. These 6 were screened in 89 additional murine APL tumor samples derived from the same mouse model. Mutations in the Jarid2 (L915I) and Capns2 (N149S) genes occurred only in the proband, and are therefore of uncertain significance. 4/6 mutations were found in additional samples; 3 of these mutations were derived from a common ancestor of the proband and the other affected mice, and were therefore not relevant for pathogenesis. The other recurring mutation was in the pseudokinase domain of JAK1 (V657F), and was identified in one other mouse that was not closely related to the proband. This mutation is orthologous to the known activating mutation V617F in human JAK2, and is identical to a recently described JAK1 pseudokinase domain mutation (V658F) found in human APL and T-ALL samples (EG Jeong et al, Clin Can Res 14: 3716, 2008). We are currently testing the functional significance of this mutation by expressing it in bone marrow cells derived from young WT vs. mCG-PR mice. In summary, unbiased whole genome sequencing of a mouse APL genome has identified a recurring mutation of JAK1 found in both human and mouse APL samples. This approach may allow us to rapidly identify progression mutations that are common to human and murine AML, and provides an important proof-of-concept that this mouse model of AML is functionally related to its human counterpart.