Genome-Wide Profiling Reveals Recurring Molecular Defects in Relapsed Childhood Acute Lymphoblastic Leukemia (ALL): A Children’s Oncology Group Study.

Despite high remission rates (>99%) following initial treatment of pediatric ALL, at least 15–20% of children with this disease will relapse. An improved understanding of the characteristics of relapsed ALL is needed to better refine initial and salvage regimens. To this end, we used the Affymetrix 500K single nucleotide polymorphism (SNP) mapping array to interrogate genomic copy number variation (CNV) in a cohort of 21 patients with relapsed ALL. For each case, we studied DNA from normal blood, diagnostic ALL bone marrow, and relapsed ALL bone marrow. There were 13 boys, 8 girls; 76% were of European Caucasian ancestry by germline SNP markers; and the median interval between diagnosis and relapse was 21.7 months (range 7.6 to 52.2 months). Copy number for each SNP was inferred based on normalized signal intensities using DNACopy, a binary segmentation algorithm that scores each genomic segment based on the number of consecutive SNPs and their signal intensities. Compared to conventional cytogenetics (available in 16 of 21 patients), SNP analysis confirmed all (100%) regions of gain or loss determined by karyotyping, but also allowed detection of a total of 485 cryptic lesions as small as 12kb throughout the genome, with a median of 3 (diagnosis) and 3 (relapse) gains and 8.5 (diagnosis) and 8 (relapse) losses per case. Relapse samples in 21 of 21 cases (100%) shared at least one major gain or loss with the diagnostic sample, all of which were verified not to be due to germline copy number variation. In both diagnostic and relapse marrows analyzed, the most common abnormalities were deletions of CDKN2A and CDKN2B at 9p21.3 (∼57% of cases at diagnosis), in agreement with prior observations [Nature 2007, 446:758]. The most frequently amplified genes were BACE2 and TIAM1 at 21q22 (∼43% of cases at diagnosis). The majority of recurring genetic alterations are shared between the diagnostic and the matched relapse marrow. In fact, across the 21 patients analyzed, a large portion (i.e. 69%) of genomic losses and (72%) gains present at diagnosis were preserved at relapse, suggesting that the relapse clone and the initially presented leukemic cells are likely derived from a common cellular origin. Of more importance, however, is the identification of genetic aberrations that were observed at relapse but not in the matched diagnostic sample, or vice versa. Twenty of the 21 cases displayed at least 1 unique genetic alteration at relapse that was not detectable at diagnosis. Conversely, in 17 of the 21 cases, abnormalities clearly present at diagnosis were not present at relapse. These findings underscore the clonal diversity of ALL at both initial diagnosis and relapse. In total, 44 novel gains and 79 novel losses arose at relapse, and 23 gains and 83 losses were lost between diagnosis and relapse. Notably, relapse-specific deletions of ARHGAP27 and LOC201175 were observed in 3 (14%) of 21 and relapse-specific gains of CIITA, HTATIP, and RNASEH2C were each observed in 3 of 21 cases (14%). By genome-wide profiling of copy number variation, we observed unequivocal molecular evidence of clonal evolution of ALL from initial presentation to subsequent relapse, and multiple recurring and unique genomic abnormalities that provide further insight into the mechanism of relapse.