The Genomic and Epigenomic Landscapes of Blast Crisis Transformation in Chronic Myeloid Leukemia

The transition from chronic phase (CP) to blast crisis (BC) chronic myeloid leukemia (CML) is characterized by reprogramming of the CML transcriptome (Radich et al. PNAS 2006), and shortened survival. Current models propose genomic instability as causal in BC transformation with enhanced DNA damage and impaired DNA repair inducing genetic mutations (ranging from large chromosomal aberrations to point mutations), altered gene function, and eventually BC transformation (Perrotti et al. JCI 2010). Consistent with this model are the phenomena of BC clonal evolution, and the increased frequency of ABL kinase domain mutations found in BC. Because different mutational processes are associated with distinct cancer-specific mutation signatures (Alexandrov et al. Nature 2013), this model also predicts the existence of a CML-specific mutation signature. In addition, recent work has highlighted the importance of epigenetic alterations in hematologic malignancies (Shih et al., Nat. Rev. Cancer, 2012). However, we lack a complete understanding of the type or frequency of genetic alterations in BC, and the relative contribution of genetic vs. epigenetic events in reprogramming the BC transcriptome.

To address these knowledge gaps, we analyzed the CML progression genome, epigenome, and transcriptome in 12 CP/BC sample pairs. Whole-genome sequencing revealed the CML genome to be relatively stable with respect to structural variations, indels, and somatic single nucleotide variants. The average number of nonsynonymous coding mutations per BC genome was 5, placing the BC coding genome in the same mutation frequency range as AML and ALL genomes (Alexandrov et al. Nature 2013). In addition, we identified a novel mutation signature in all CML samples suggesting a CML-specific mutational process. 1175 genes were 'hit' by genomic, mostly copy number, alterations in >1 sample, and included TCR genes and Ikaros (IKZF1) among lymphoid BC pairs. Only 21 recurrently altered genes were affected by somatic SNVs or indels, with resistance-associated ABL1 mutations being commonest.

We next used DNA methylation arrays to assess the BC epigenome, and found 20,651 CpG sites (out of 455,187) to be hyper-methylated, and 3225 to be hypo-methylated in BC compared to CP. Combined methylome and transcriptome analysis demonstrated an inverse relationship between methylation and expression changes at a subset of CpG sites enriched at promoters. Genes with increased methylation/decreased expression or decreased methylation/increased expression included those involved in cell cycle control/heme biosynthesis, and molecular mechanisms of cancer/G-protein coupled receptor signaling/MAPK signaling respectively. Unsupervised methylation-based clustering segregated samples into CP, lymphoid BC and myeloid BC groups, recapitulating expression-based clustering, and further supporting a functional role for DNA methylation in BC transcriptional reprogramming.

We next performed an integrative analysis by combining the genome, methylome, and transcriptome datasets, and included data from 34 additional CML samples. Top ranking candidate genes included epigenetic modifiers, and hematopoetic differentiation- and stem cell-related genes. Functional analysis of candidate genes and epigenetic processes using genetic and epigenetic drug-based approaches are ongoing.

In summary, we conclude that: 1. The genomic and epigenomic landscapes in BC are characterized by a modest number of recurring events in the former, but consistent and striking differences in the latter, 2. The BC methylome is functionally associated with the robust gene expression changes found in BC, and 3. Epigenetic modifier drugs may be of use in reversing the gene expression changes characteristic of BC.