Abstract
In order to better understand the pathogenesis of acute promyelocytic leukemia (APL, FAB M3), we sought to determine its gene expression signature by comparing the expression profiles of 14 APL samples to that of other AML subtypes (M0, M1, M2, M4, n=62) and to fractionated normal whole bone marrow cells (CD34 cells, promyelocytes, PMNs, n=5 each). We used ANOVA and SAM (Significance Analysis of Microarrays) to select genes that were highly expressed in APL cells and that displayed low to no expression in other AML subtypes. The APL signature was then further refined by filtering genes whose expression in APL was not significantly different from that of normal promyelocytes, yielding 1121 annotated genes that reliably distinguish APL from the other FAB subtypes using unsupervised hierarchical clustering, both in training and validation datasets. Fold change differences in expression between M3 and other AML FAB classes were striking, for example: GABRE 35.4, HGF 21.3, ANXA8 21.3, PTPRG 16.9, PTGDS 12.1, PPARG 11.1, STAB1 9.8. A large proportion of the APL versus other FAB dysregulome was recapitulated when we compared APL expression to that of the normal pattern of myeloid development. We identified 733 annotated genes with significantly different expression in APL versus normal myeloid cell fractions. These dysregulated genes were assigned to 4 classes:
persistently expressed CD34 cell-specific genes, repressed promyelocyte-specific genes, prematurely expressed neutrophil-specific genes and genes with high expression in APL and low/no expression in normal myeloid cell fractions.
Expression differences in several of the most dysregulated genes were validated by qRT-PCR. We then examined the expression of the APL signature genes in myeloid cell lines and tumors from a murine APL model. The bona fide M3 signature was not apparent in resting NB4 cells (which contain t(15;17), and which express PML-RARA), nor in PR-9 cells following Zn induction of PML-RARA expression, suggesting that neither cell line accurately models the gene expression signature of primary APL cells. Most of the nodal genes of the mCG-PML-RARA murine APL dysregulome (Yuan, et al, 2007) are similarly dysregulated in human M3 cells; however, the human and mouse dysregulomes do not completely coincide. Finally, we have begun investigating which APL signature genes are direct transcriptional targets of PML-RARA. The promoters of the APL signature genes were analyzed for the presence of known PML-RARA binding sites using multiple computational methods. The analyses demonstrated that several transcription factors (EBF3, TWIST1, SIX3, PPARG) have putative retinoic acid response elements (RAREs) in their upstream regulatory regions. Additionally, we examined the promoters of some of the most upregulated genes (HGF, PTGDS, STAB1) for known consensus sites of these transcription factors, and found that all have putative binding sites for at least one. These results suggest that PML-RARA may initiate a transcriptional cascade that relies not only on its own activity, but also on the actions of downstream transcription factors. In summary, our studies indicate that primary APL cells have a gene expression signature that is consistent and highly reproducible, but different from commonly used human APL cell lines and a mouse model of APL. The molecular mechanisms that govern this unique signature are currently under investigation.
Large-scale integrated analysis of ovarian cancer tumors and cell lines identifies an individualized gene expression signature for predicting response to platinum-based chemotherapy