Identification of Novel Cluster Groups in High-Risk Pediatric B-Precursor Acute Lymphoblastic Leukemia (HR-ALL) by Gene Expression Profiling: Correlation with Clinical and Outcome Variables. a Children’s Oncology Group (COG) Study.

Children with HR-ALL are traditionally defined by NCI Risk Criteria (age and white blood cell count) and comprise a highly heterogeneous group of ALL cases that have not been well characterized. In an effort to shed light on the genetic diversity of HR-ALL and identify new therapeutic targets in this resistant form of disease, we previously analyzed 207 patients enrolled on COG HR-ALL Trial P9906. These analyses revealed discrete gene expression profiles (Affymetrix U133 Plus2) that distinguished 8 distinct cluster groups. Two of these groups clustered patients with known underlying genetic abnormalities (E2A-PBX1/t(1;19) or MLL rearrangements) while the remaining 6 clusters were novel and the underlying genetic lesions remain to be identified. Two of the novel clusters were associated with extremely good (94.7% 4 year RFS) or very poor outcomes (20.9% 4 year RFS). In order to validate these findings and increase the size of the study cohort to enhance statistical power, we performed gene expression profiling in an additional 283 children with HR-ALL enrolled on COG Trial AALL0232. Patients enrolled on AALL0232 met NCI high risk criteria similar to those enrolled on P9906 but additionally included some HR-ALL patients with BCR-ABL or TEL-AML1 translocations, hypodiploidy, and favorable trisomies of chromosomes 4 and 10. As shown in the table below, the cluster designation and relative composition were very similar between the two cohorts. Two notable exceptions were the decreased number of Group 1 (MLL rearranged) members and the significantly elevated number of TEL-AML1 positive patients on AALL0232. In addition to identifying again the 6 novel clusters initially identified in the P9906 cohort and the TEL-AML1+ patients, AALL0232 contained another novel cluster group with a distinct gene signature. Expression profiles from P9906 and AALL0232 were merged to yield a cohort of 490 HR-ALL patients. When the same clustering methods were applied to the merged cohort, a similar set of clusters were identified with virtually identical group membership despite being independent experimental cohorts and clinical trials. The merged cohort permitted a more rigorous definition of gene signatures distinguishing each cluster, identifying the top 50 rank order genes associated with each cluster, and a more rigorous statistical analysis of the association of cluster group with outcome and clinical variables. Although the outcome data for the AALL0232 samples is not yet mature, the high degree of correlation of this cohort with certain P9906 clusters permits us to make predictions about outcome that will allow for validation with longer follow up. Furthermore, this larger cohort allows us to continue to identify potential new therapeutic targets and underlying genetic abnormalities in HR-ALL.