MLL-AF9 and MLL-ENL Induce Deregulation of Similar Downstream Candidate Genes: An Analysis across Experimental Protocols and Laboratories.

Expression profiling has become an important tool for understanding gene deregulation in MLL-fusion leukemias. However, the results of gene profiling experiments are difficult to interpret when applied to leukemia cells because (i) leukemias arise in cells that differ greatly in their gene expression profiles, and (ii) leukemias most often require secondary genetic events in addition to the MLL fusion gene. Two principal model systems have been used to understand the direct effects of MLL-fusion genes. Knock-in models have the advantage of the fusion gene being under control of the physiologic promoter. On the other hand, conditional expression systems offer the ability to conduct short term experiments, permitting the analysis of direct effects on downstream genes. In the present combined-analysis, we used the Affymetrix U74Av2 oligonucleotide microarray to evaluate the effects of the MLL-fusion gene in vivo and in vitro respectively using two closely related MLL fusion genes - MLL-AF9 for knock-in and MLL-ENL for conditional expression. In the MLL-AF9 study, we compared gene expression profiles of bone marrow cells from MLL-AF9 knock-in mice (C57Bl/6, MLL-AF9^+/−) to those of age-matched wild type mice (Kumar et. al. 2004, Blood). We used a t-test (p<0.05) to selected genes that showed significant changes in expression levels. In the MLL-ENL study, we transformed murine primary hematopoietic cells with a conditional MLL-ENL vector (MLL-ENL fused to the modified ligand-binding domain of the estrogen receptor) such that the fusion protein was active only in the presence of tamoxifen. We then studied the downstream effects of the fusion protein by comparing gene expression profiles of the cells in the presence and absence of tamoxifen. We used a pair-wise comparison analysis to select genes that showed a change in expression level of 1.5 fold or greater in at least two of three experiments (Zeisig et. al. 2004, Mol. Cell Biol.). Those genes that were up-regulated in both datasets were then compiled together. This list included Hoxa7, Hoxa9 and Meis1. The results for these 3 genes were confirmed by quantitative RT-PCR in both the MLL-AF9-knock-in and the MLL-ENL-conditional-expression systems. The remaining candidate genes in the common up-regulated gene set (not yet tested by quantitative RT-PCR) include protein kinases (Bmx, Mapk3, Prkcabp, Acvrl1, Cask), RAS-associated proteins (Rab7, Rab3b), signal transduction proteins (Notch1, Eat2, Shd, Fpr1), cell membrane proteins (Igsf4), chaperones (Hsp70.2), transcription factors (Isgf3g), proteins with unknown functions (Olfm1, Flot1), and hypothetical proteins. The results of the combined analysis demonstrate that these over-expressions are (i) a direct and sustained effect of the MLL-fusion protein, (ii) are independent of secondary events that might be involved in leukemogensis, and (iii) are independent of the two partner genes that participate in these fusions. The over-expression of a few genes in both the -in vitro and in vivo experimental systems makes these molecules very interesting for further studies, to understand the biology of MLL-fusion leukemias and for development of new therapeutic strategies.