Translocations involving the MLL
gene result in both AML and ALL in humans. MLL-AF9
are the most common translocations resulting in AML and ALL respectively. Although a number of retroviral and genetically engineered mouse models of MLL-fusion mediated AML have been developed, models recapitulating MLL-AF4
mediated ALL have been more elusive. Ectopic expression of MLL-AF4 through retroviral transduction has been difficult and other genetically engineered mouse models have resulted in myelodysplasia or mature B-cell lymphomas. We report the generation of a conditional knock-in model in which the MLL-AF4 fusion product is expressed within the context of the endogenous MLL
locus. Upon conditional activation, mice develop AML or ALL, the latter with an immunophenotype and gene expression profile consistent with acute B-precursor cell leukemia. The gene expression signature of murine Mll-AF4 ALL cells was found to be highly enriched in the gene expression profile of human ALL samples with MLL
rearrangements as compared to ALL samples with germline MLL
. Therefore, this murine model recapitulates human MLL-AF4 mediated ALL both in terms of disease phenotype, as well as overall patterns of gene expression. We note that our mice have a propensity toward AML development which differs from the strong association between t(4;11) and ALL in humans. Future studies will determine if this is due to expression of Mll-AF4 in different cells of origin or microenvironmental influences as recently shown for MLL-AF9 (Wei et al., 2008
). Also, the genetic background may influence the leukemia incidence and phenotype. The development of a faithful model of MLL-AF4
ALL will allow detailed characterization of the mechanisms of MLL-AF4 mediated leukemogenesis, including the mechanisms of transformation, cell(s) of origin, the phenotypes of cancer stem cells in leukemias expressing various lineage markers, and mechanisms associated with the development of mixed myeloid/lymphoid (mixed lineage) disease. Also, a faithful model of MLL-AF4
ALL will allow for assessment of putative therapeutics in a well-defined model system.
Recent studies have identified association of multiple MLL-fusion partners including AF4, AF9, and AF10 with DOT1L, a histone H3K79 methyltransferase (Bitoun et al., 2007
; Mueller et al., 2007
; Okada et al., 2005
; Zeisig et al., 2005
; Zhang et al., 2006
). We used our murine model to ask whether abnormalities in H3K79me2 were present in Mll-AF4 mediated ALL. Genome-wide assessment revealed enhanced H3K79me2 (in comparison to matched normal cells) at many loci across the genome, including across the HOXA
cluster. This prompted a similar analysis in human ALL samples, where we confirmed widespread abnormalities in H3K79me2 as a characteristic that distinguished MLL
-rearranged from MLL
-germline ALLs. These results demonstrate that in both murine and human disease mediated by MLL
rearrangement, there are widespread changes in H3K79me2, including on genes that are critical for leukemogenesis such as HOXA9
. Future studies will determine which subset of these genes are direct MLL-AF4 target genes. The results described here further underscore the utility of this murine model system for providing mechanistic insight into human disease.
In both human and murine ALL, we found that ectopic H3K79me2 was highly associated with increased mRNA expression. Wildtype MLL possesses H3K4 methyltransferase activity, a modification associated with transcriptional “priming” of the promoters of genes important for appropriate developmental cell-fate decisions (Guenther et al., 2007
). Since the methyltransferase domain of MLL is invariably lost in MLL-fusion proteins, including MLL-AF4, the question arises as to how MLL-fusions promote enhanced expression of its target genes such as HOXA
genes. We found that H3K4me3 is still associated with HOX
genes in our murine MLL-AF4 model and preliminary experiments suggest this to also be the case in human ALL with MLL-fusions (A.K. unpublished data). In both human disease and our murine model, the second MLL
allele remains germline, and thus wildtype MLL protein, or perhaps other H3K4 methyltransferases, may continue to regulate H3K4me3 at these loci. However, H3K4me3 is insufficient to fully active transcription, which appears to require the presence of other histone modifications such as H3K79me2 for transcriptional elongation to proceed (Guenther et al., 2007
). This leads to a model where MLL-AF4 recruits DOT1L to MLL target genes, and promotes methylation of H3K79 at loci with existing H3K4 methylation (i.e., by wildtype MLL or other H3K4 methyltransferases) thus stimulating transcriptional elongation of genes that are normally primed but not fully transcribed.
Prior studies have established that various cancer sub-types can be distinguished on the basis of specific oncoproteins or global mRNA expression patterns. In this study, we demonstrate that widespread differences in an epigenetic histone modification can likewise distinguish different cancer subtypes, and malignant vs. normal cells. The ability to distinguish cancer sub-types on the basis of epigenomic profiles does not necessarily offer any advantages over mRNA profiling in terms of disease diagnosis. However, the tight linkage between oncoprotein (MLL-AF4), specific epigenetic changes (H3K79me2) and altered gene expression has important therapeutic implications. Specifically, many oncoproteins are DNA binding proteins, and are generally not readily amenable to targeting by small molecules or biologics. Pharmacologic inhibition of transcription factor function, including disrupting protein-protein or protein-DNA interactions, remains a largely experimental undertaking with few examples of success. A common strategy to bypass this impasse has been to look for downstream transcriptional targets that may be more druggable (e.g., enzymes, receptors). Since oncoproteins may alter the expression of hundreds or thousands of genes, in most cases inhibiting the activity of a single druggable downstream target may not fully reverse the malignant phenotype. Inhibition of epigenetic modifying enzymes offers an alternative approach that is facilitated by the fact that enzymes are generally more amenable to drug development. Targeting the epigenetic link between an oncoprotein and downstream gene expression changes may more broadly impact the multitude of gene expression changes that in aggregate contribute to the malignant phenotype. In the case of diseases driven by MLL-fusion proteins, development of a DOT1L inhibitor may impact expression of a diversity of critical genes including HOXA9, MEIS1, FLT3, and BCL2. Indeed our data demonstrate that suppression of DOT1L expression leads to decreased expression of HOXA cluster genes. Furthermore, a DOT1L inhibitor might be applicable to other diseases characterized by ectopic H3K79me2. In this regard, the ability to identify disease sub-types that have widespread ectopic H3K79me2 would have important implications for patient selection and stratification in testing a DOT1L inhibitor. In as much as epigenetic modifications are likely to be linked to the maintenance of gene expression in many human malignancies, these strategies may be applicable to a wide variety of cancer subtypes.