Via developing novel methodologies, we were able achieve the above aims, as follow:
Work Package 1: Methylation-based analysis of gene-enhancer coupling. We developed a methylation-orientated strategy to explored gene-associated, cis-regulatory elements. Focusing on 125 pan-cancer or glioblastoma (GBM) driver genes, and 52 reference genes, we performed analyses of two million base-pairs (bp) centered at the promoters of the genes. Utilizing deep methylation-sequencing data of 24 capturing libraries, we analyzed the association between methylation levels of the captured sites and expression levels of the targeted genes, across GBM tumors. The analysis revealed cis-regulatory circuits (n=1,154; q<0.05; R2 > 0.3) between certain methylation sites and controlled genes. Most (78%) of the genes had multiple (2-68) circuits, averaging 8.3 circuits per gene, of them 3.5 circuits in average were positive (expression raised with methylation), and 4.8 negative. This extensive mapping data pave the way for the following stages of the study.
Work Package 2: Massive parallel function analysis of genetic and epigenetic enhancer alleles. We developed a method for systematic functional analysis of enhancer sequences before or after DNA methylation. For this experiment, an entire library of captured DNA segments was cloned into gene-reporter vectors, downstream to minimal promoters. The obtained expression vectors were inserted into GBM cells, and allowed to produce RNAs. The transcriptional effect of each segment was then examined, in windows of 500 bp. We next compared the functionalities of the captured segments under whole-methylation versus de-methylation conditions. Of the 26,152 annotated segments, 10,998 displayed activity differences of at least one and a half fold between methylated and un-methylated states. Of the methylation-sensitive elements, the majority (83.7%) reduced their original activity, or shifted to the opposite functionality (i.e. enhancers became weaker or turned silencers, and vice versa), upon methylation.
Work Package 3: Development of DNA-based expression markers. We utilized the enhancer-gene coupling and the functional data that have been produced in WPs 1-2, to develop gene-expression prediction models. Overall, significant models of inter-patient’s expression variation were developed for 81 of the genes (58 drivers and 23 reference genes). Out of these, the expression of 58 genes (39 drivers and 19 reference genes) were best described by synergic combinations of sites, that together provide better description than each of the sites alone. Of these synergic models, 23 used sites from both negative and positive sites.
Work Package 4: Targeted modification of regulatory elements. We applied genome editing to verify the uncovered gene regulatory circuits. Take together, the results reveal the internal organizations of gene regulatory domains, as well as on the regulatory role of DNA methylation. We found that cis-regulatory domains are composed of spatially-overlapped, gene-specific regulatory networks. These networks comprised multiple regulatory units, each of them provides a define, positive or negative effect on the expression of the targeted genes. Under control conditions, DNA methylation dictates the mode and the level of these effects. The sum effects of methylation variation in a small number of key methylation sites, located in positive and negative units, effectively describe the variation in the expression of cancer genes among glioblastoma patients.
