SNP Array Analysis in Acute Myeloid Leukemia Reveals Frequent and Recurrent Acquired Genetic Alterations Linked to Prognosis: a Study of the ALFA Group

Abstract 2533

Acute myeloid leukemia (AML) is a heterogeneous disease because of different leukemogenic mechanisms and variable response to antileukemic treatment. In addition to age and leukocytosis at diagnosis, cytogenetic abnormalities are key factors to assess prognosis. Several gene alterations such as those involving CEBPa, NPM1 and FTL3 have been identified and have improved the classification of AML. However, variability remains and cannot be completely explained. In this study, we performed SNP array analysis on a cohort of 128 AML patients in order to identify new genetic alterations and potential new candidate genes involved in leukemogenesis or disease progression.

Patients were classified according to MRC cytogenetics (26 favourable; 77 intermediates, 15 adverse, 10 no informative caryotype). Patients were aged from 11 to 65 years and distributed across all French-American-British (FAB) classes except M3 (8 M0, 25 M1, 38 M2, 26 M4 and 11 M4Eo, 12 M5, 2 M6), Paired DNA was extracted from bone marrow aspirates obtained at diagnosis and after achieving complete remission (CR), and analyzed using Affymetrix Genome-Wide Human SNP Array 6.0 to distinguish acquired from constitutional genetic abnormalities. Copy number variations (CNA) were validated on a custom Agilent microarray (Human Genome CGH Microarray 105k). Data were analyzed using Affymetrix' Genotyping Console 3.0.2, Agilent's GeneSpring GX. Statistical analyses were executed using R version 12.1. The Cox proportional hazard regression model was used to relate genetic abnormalities to treatment outcome, with karyotype included as the second covariate. Time was censored at transplantation date if bone marrow transplantation was performed.

We found 210 genomic abnormalities in 74 patients: 197 CNA and 13 copy neutral losses of heterozygosity (uniparental disomy or UPD), resulting in 1.6 abnormalities on average per patient (range 0 to 17). Among CNA, deletions were more frequent than gains (130 vs. 66). CNA spanned from 8kb to 191MB (median of gains 24MB, median of losses 2MB). 116 of them had not been detected by conventional cytogenetics. UPD spanned from 23MB to 150MB (median, 33 MB). Abnormalities were located over all chromosomes except for chromosome14 and were particularly frequent on chromosomes 2, 7, 11, 16, 17, and 21 (54% of all abnormalities). We defined 72 minimal common regions which were altered in at least 2 patients. Among the 43 common regions shorter than 5Mb, 16 contain at least one gene reported in AML or cancer. CNA and UPD were distributed across all FAB subtypes. Of note, 7 of 8 patients with AML M0 had 1 or more abnormalities.

Among the 74 patients with CNA or UPD, 36 had an intermediate caryotype. We found no significant association of number of abnormalities with known gene alterations known to influence prognosis (i.e., CEBPa, FLT3itd or NPM). As expected, there were significantly more alterations in patients with a favourable or adverse caryotype (P=0.0045, Fisher's exact test). Furthermore, the number of genetic abnormalities was significantly associated to disease-free and to overall survival (P=0.010 and 0.0016, respectively). This remained significant in a multivariate analysis including karyotype (P=0.046 and 0.01, respectively)

CNA or UPD was detected in 56% of AML patients who achieved remission of these 45% have intermediate cytogenetics. We identified 43 minimal common regions shorter than 5MB, which were altered in at least 2 patients and 37% of these regions involve genes previously reported in AML or cancer. Increased genomic alterations were significantly associated with favorable and adverse cytogenetics and with disease-free and overall survival. Prognostic significance of number of abnormalities remains significant after adjustment for cytogenetics. DNA SNP array analysis may be useful to better define prognostic subgroups in addition to conventional cytogenetics and may identify candidate genes implicated in leukemogenesis or disease progression.