Time to Leukemia (TTL) in NOD/SCID Mice Determines Patient Outcome and Is Characterized by a 5 Genes Signature Associated with Relapse

Acute lymphoblastic leukemia (ALL) is the most frequent malignant disease in childhood. Advances in therapy, in particular stratification of patients to appropriate treatment according to risk factors improved long term survival attaining cure rates of up to 80 %. Despite the efforts achieved by the stratification strategies the majority of relapse patients are recruited from the low risk groups emphasizing the need for additional independent risk factors. In a recent study we transplanted primary leukemia samples obtained from pediatric patients with newly diagnosed B cell precursor ALL (BCP-ALL) into NOD/SCID mice. Time to leukemia (TTL) was analyzed for each patient sample transplanted from date of transplant to date of leukemia manifestation in the recipients. We demonstrated that patients whose leukemia cells engrafted rapidly leading to manifestation of the disease within 10 weeks (TTLshort) showed a clearly inferior relapse free survival in contrast to patient samples with prolonged in vivo growth (TTLlong). Interestingly, the same distinct difference in relapse free survival was observed in the low risk groups only. Multivariate analysis showed an almost 45- fold increased risk for relapse in TTLshort patients. In order to further characterize the biological properties of the leukemia cells in the two groups, gene expression profiles of samples with short versus long TTL in the xenograft model were analyzed using a human whole genome array (Affymetrix U133 Plus 2.0). Here, we used quantitative traits analysis (QTA) correlating gene expression values (relative expression) to the time from transplant to manifestation of leukemia in the NOD/SCID mice (TTL, in weeks). 14 different xenograft samples (TTLshort n= 7, mean TTL 8.14 weeks; TTLlong n= 7, mean TTL 19.71 weeks; T-test P = .03) isolated from leukemia bearing mice were investigated. All 14 patients included were stratified in low (standard or intermediate) risk groups. All patients were negative for TEL/AML1- fusion, BCR/ABL- fusion or MLL rearrangement. By QTA we identified 5 genes that were significantly correlated (Spearman correlation, for all 5 genes P < .0001) to time from transplant to leukemia manifestation in the recipient mice (TTL). Analysis of the 14 xenograft samples using the 5 genes identified showed clustering of all but one sample in one group. The 5 genes were than explored for their power to predict relapse in an independent cohort of pediatric ALL patients. For these patients expression profiles were analyzed in leukemia samples obtained at diagnosis. All patients in this cohort were also stratified in low risk groups and negative for the presence of TEL/AML1-, BCR/ABL- or MLL- fusion transcripts. 8 of the 38 patients encountered relapse, 5 at early and 3 at late time points (< or > 24 months after diagnosis). Clustering according to the 5 genes identified by QTA lead to separation of the patients into two groups. Interestingly, all relapsed patients except one clustered in the group associated with the signature characteristic for TTLshort. Most importantly, all 5 patients with early relapse gathered in this cluster. Taken together, we used a novel approach directly correlating gene expression values to the continuous variable time to leukemia (TTL) which might be reflected more accurately by this strategy than comparing two groups. We applied this gene signature characteristic for TTLshort, thereby identifying poor overall survival in a set of independent pediatric ALL patients and found clustering of all early relapse patients in one group. This new independent risk factor might identify patients with high risk for early relapse avoiding xenografting in the mouse model.