Genomic Signature Predicts Resistance To Busulfan In AML Cell Lines

The alkylating agent Busulfan (Bu) is commonly used in the pre-transplant conditioning regimen for patients with acute myeloid leukemia (AML). Relapse after transplant is in part due to resistance to Bu. Our study was aimed at identifying a genomic signature that could predict resistance to Bu in AML cells.

In order to obtain a list of candidate genes which may correlate with increasing resistance to Bu we performed a linear regression analysis controlled for cancer type on NCI60 Stanford cDNA data and GI50 values. This produced a list of 111 genes correlating with increasing Busulfan GI50 (p<0.01, FDR<50%). The ability of these 111 genes to predict GI50 was subsequently tested based upon their expression in a second independent gene expression platform (Affymetrix U133 plus 2.0). This analysis yielded 6 genes (KCNH2, CD74, CD53, HCLS1, ERC2 and HLA-DQB2) predicting higher GI50 values (p<0.05). We then tested the ability of these genes to predict Bu GI50 in a panel of 6 AML cell lines (K562, HL60, HL60-MX1, KG1, THP1 and NB4). To obtain GI50 values on AML cell lines we treated the cells with doses of Bu ranging from 0-200mcg/ml. At 24 hours, cells were washed and resuspended in fresh medium. Proliferation of cells was measured by standard 3H-thymidine uptake assay at 48 hours. Sigmoidal dose response curves and GI50 values were then calculated. GI50 values varied from 13.9 - 70.4 micromoles/ml. RNA quantitation for the 6 identified genes was performed using the Quantigene assay (Panomics, Santa Clara, CA). Expression was normalized to control gene levels (RPLP0). Two genes (ERC2 and HLA-DQB2) showed low or undetectable expression and were excluded from further analysis. Logistic regression analysis showed that a combined overexpression of any 2 of the remaining 4 genes was significantly associated with increasing resistance to Bu GI50 in AML cells (p=0.006). Of the 6 cell lines tested, the 2 most resistant (THP1 and K562) both displayed this pattern of gene overexpression.

To our knowledge, this is the first attempt to obtain a list of genomic biomarkers to predict Bu resistance in AML using a genome wide approach. This data provides a rationale for validation of these 4 genes in a clinical dataset with the potential future application of predicting relapse after Bu based high dose therapy.