Microarray Analyses In A Case-Control Cohort Of T-ALL Samples Identifies Gene Signature Of Potential Prognostic Significance

Stuart S. Winter, MD¹, Hadya Khawaja², Zeyu Jiang, PhD³, Charles Kooperberg, PhD⁴, Adolfo Ferrando, MD⁵, Thomas Look, MD⁶ and Richard S. Larson, MD, PhD².

¹Pediatric Hematology/Oncology, University of New Mexico Health Sciences Center (HSC), Albuquerque, NM; ²Pathology, University of New Mexico HSC, Albuquerque, NM; ³Biostatistics, University of New Mexico HSC, Albuquerque, NM; ⁴Division of Public Health Sciences, Fred Hutchinson Cancer Center, Seattle, WA; ⁵Institute for Cancer Genetics, Columbia University, New York, NY and ⁶Pediatric Oncology, Dana Farber Cancer Institute, Boston, MA.

Risk stratification remains limited for patients being treated for T-ALL due to a lack of biologic predictors of outcome. As a consequence, treatment assignment on modern protocols has been largely achieved through random assignment. Recent observations have suggested that overexpression of specific gene(s) may provide a reliable means to risk stratify patients. We hypothesized that microarray analysis may identify gene sets that distinguish both therapeutic response and patient outcome in T-ALL. We analyzed the gene expression profiles of 45 primary T-ALL samples (24 CCR, 21 relapse) from a matched, case control study with sufficient cRNA for microarray analysis (COG #8704). We performed oligonucleotide microarray analysis using Affymetrix U133Av.2 genechips which have approximately 23,000 target genes and ESTs. Following heirarchical clustering in dChip and R Language analyses (Chiaretti et al. Blood, 2004), but using RMA normalization, we identified 37 genes that serve as reliable predictors of CCR or relapse. Leave-one-out least discriminant analysis cross-over validation further constrained our prognostic gene identifiers to 21 genes of robust significance. These 21 genes predict 87 % of CCR and 82% of relapse accurately (p<0.0001, two-tailed Fishers exact test). These results were verified by qRT-PCR. Transcriptional factors previously described as having prognostic significance were not identified in our study. Twenty-six of the 45 cases received high dose L-asparaginase (16 CCR, 10 relapse) on the companion study. As a result, we examined whether a distinct signature could be also identified that distinguishes response to dose-intensified asparaginase treatment for patients with T-ALL. Using the same approach, a 27-member gene signature was identified that accurately predicted response in 24 of the 26 cases (92 %; 2 cases of relapse were misclassified). These results have identified two sets of genes that may be further pursued as prognostic indicators in T-ALL or as predictors of response to therapy.