Acute leukemias are a heterogeneous group of tumors, the molecular basis of which remains poorly understood. Gene expression profiling (GEP) studies have provided insight into molecular pathogenesis and can classify patients into cohorts, which nonetheless still display considerable biological heterogeneity. However, GEP provides only a snapshot of mRNA levels at a given timepoint, without any information regarding gene regulatory status - i.e. their ability to be expressed or activated. Moreover, many important gene regulatory events are not accompanied by expression changes detectable by expression microarrays and are lost in the “noise”. We predicted that these limitations could be overcome by interrogating the epigenetic regulatory status of genes in addition to their mRNA levels. For this purpose we developed a method for detection of genome-wide DNA methylation called HELP (HpaII tiny fragment Enrichment by LM-PCR). HELP is based on comparative Msp1 and HpaII isoschizomer digestion, size-specific amplification of DNA fragments and co-hybridization to custom high-density oligonucleotide arrays designed specifically for this method and that provide data of uniform quality throughout the genome. We demonstrate here that HELP yields quantitative, reproducible (R>0.98) and robust signals in leukemia patient samples, as shown by bisulfite pyrosequencing validation. To compare and contrast the ability of HELP and GEP to identify molecular signatures and differentially regulated genes we performed a proof of principle study in a small cohort of AML and ALL patients, where GEP on high density oligonucleotide arrays and HELP were performed in triplicate for each patient. We found that
unsupervised clustering of DNA methylation data more accurately classified acute leukemias according to lineage than GEP;
Specific methylation signatures distinctive of AML vs. ALL could be readily identified by performing correspondence analysis;
promoter methylation analysis yielded twice as many informative loci than GEP (600 differentially methylated genes vs. 300 differentially expressed genes with False Discovery Rate = 1%). While 90% of the differentially expressed genes were also detected by HELP, the latter further identified another 300 genes that were missed by GEP but that are differentially methylated in ALL vs. AML. Genes discovered by HELP but missed by GEP were mainly regulators of cellular proliferation, hematopoietic differentiation, apoptosis and intracellular signaling.
examples of candidate genes differentially methylated between AML vs. ALL but not detected by GEP include: TNF receptor family members, Caspase 9, DNA fragmentation factor, Methylenetetrahydrofolatereductase (MTHFR), and the tumor suppressor PAX7.
These results suggest that DNA methylation signatures accurately classify disease subtypes and identify a greater number of genes with potential biological importance than GEP. Rigorous epigenetic signature studies by HELP might thus provide a more accurate and comprehensive method for molecular classification and characterization of acute leukemias and other tumors.
Disclosure: No relevant conflicts of interest to declare.