Use of Genomic Technologies to Identify Novel Genetic Abnormalities and Therapeutic Targets In Acute Lymphoblastic Leukemia.

With the progressive intensification of chemotherapy, the majority of children with ALL now achieve long-term survival. In parallel, a number of molecular subtypes of ALL have been identified that are associated with treatment outcomes, which are either excellent (TEL-AML1 or trisomy of chromosomes 4, 10, and 17), intermediate (MLL rearrangements or E2a-PBX), or very poor (BCR-ABL or hypodiploidy). Yet, the underlying genetic abnormalities in the majority of children with ALL, such as those with “high-risk” disease who remain resistant to current therapies, remain to be discovered. Supported by the NCI SPECS and TARGET Initiatives, the Children’s Oncology Group (COG), and The Leukemia & Lymphoma Society, we are using comprehensive genomic technologies (i.e., expression profiling, genome-wide analyses of DNA copy number abnormalities [CNAs] and germline polymorphisms, and direct gene sequencing) to develop molecular classifiers for outcome prediction that can be used to discover novel underlying genetic abnormalities and therapeutic targets in ALL. Our work has focused on a cohort of 220 children with “high-risk” ALL registered to COG Trial 9906. Using supervised learning methods on gene expression profiles, molecular classifiers predictive of relapse-free survival (RFS) and minimal residual disease (MRD) at end-induction have been developed. A 38-gene molecular risk classifier predictive of RFS (MRC-RFS) can distinguish two groups of high-risk ALL patients with different relapse risks: low (4 yr RFS: 81%, n=109) vs. high (4 yr RFS: 50%, n=98) (P< 0.0001). In multivariate analysis, the best predictor combines MRC-RFS and end-induction flow MRD, classifying children into low- (87% RFS), intermediate- (62% RFS), or high-risk (29% RFS) groups (P<0.0001). A 21-gene molecular classifier predictive of MRD can effectively substitute for end-induction MRD, yielding a combined classifier that similarly distinguishes three risk groups at pre-treatment (low: 82% RFS; intermediate: 63% RFS; and high: 45% RFS) (P< 0.0001). This combined molecular classifier was further validated on an independent cohort of 84 children with high-risk ALL registered to COG Trial 1961 (P = 0.006). Using unsupervised clustering methods, 8 distinct cluster groups based on gene expression were identified, 6 of which were entirely novel. Two of the novel clusters were associated with strikingly different outcomes (95% 4-year RFS vs 20% 4-year EFS). Novel underlying genetic abnormalities and genes that may represent novel therapeutic targets have been identified in each of these clusters. Interestingly, children of Hispanic ethnicity were disproportionately represented in the poorest outcome clusters. CNAs were revealed in genes regulating B lymphoid development in 50.2% of cases (PAX5 in 30.7% and IKZF1 in 24.9%). In addition, recurring CNAs were detected in a number of other genes known to play roles in transformation, including CDKN2A/B, RB1, BTG1, IL3RA, NRAS, KRAS, NR3C2, and ERG. CNAs in IKZF1, EBF, and BTLA were strongly associated with the poorest outcome clusters defined by gene expression profiling. Deletion of IKZF1 was particularly associated with negative outcome (p=0.002). These ongoing studies demonstrate that molecular classifiers can be used to distinguish distinct prognostic groups within high-risk ALL, significantly improving risk classification schemes and the ability to prospective identify children who will respond to or fail current therapies. These classifiers are now being integrated into the design of COG clinical trials. The discovery of novel cluster groups and underlying genetic abnormalities is being exploited to develop new therapeutic targets for this disease.