Short Abstract: The most common congenital heart disease (CHD) is the ventricular septal defect (VSD), which is in turn a subfeature of Tetralogy of Fallot (TOF) representing the most common form of cyanotic CHD. The underlying causes for the majority of CHDs are still unclear and most probably consist of combinations of genetic, epigenetic and environmental factors. DNA methylation is the most widely studied epigenetic modification and several cardiac regulators have already been shown to be differentially methylated in CHD patients. Here, we present the first analysis of genome-wide DNA methylation data (MBD-seq) obtained from myocardial biopsies of TOF and VSD patients. We found clear methylation differences between patients and controls and moreover, between both patient groups. We defined stringent sets of differentially methylated regions, which are significantly enriched for genomic features like promoters, exons and cardiac enhancers. For TOF, we linked DNA methylation with genome-wide gene expression data (RNA-seq) and found a significant overlap for hypermethylated promoters and down-regulated genes, and vice versa. Among the differentially methylated genes with distinct expression changes are the sarcomeric component TNNI1, the co-receptor TDGF1, the endothelin converting enzyme ECE2, and the cytochrome C oxidase assembly protein SCO2. Moreover, we found examples of methylation changes co-localized with novel, differential splicing events among sarcomeric genes. Finally, we demonstrate the interaction of differentially methylated and expressed genes in TOF with mutated CHD genes in a molecular network. In summary, our data suggest that DNA methylation likely contributes to the pathogenesis of CHD by modulating disease-specific gene expression profiles.

Short Abstract: Genome assembly remains a challenging issue for large plant genomes due to their high amount of repetitive regions resulting in publications of draft genomes focusing on gene space instead of the whole genome. A selection for these regions before sequencing increases the coverage and facilitates the assembly procedure compared to a WGS-approach. This can be achieved using differences in the methylation state of CG-rich regions which are highly correlated with the gene space especially in plant genomes. Hypomethylated regions are gene-rich regions whereas methylated regions are mainly repetitive. DNA methylation is a widespread epigenetic mark playing a key role in regulation of gene expression and thus is of importance in many biological processes like early embryogenesis, stem cell differentiation, genomic imprinting, stress responses, vernalization and transposon silencing mechanisms.
We present the results of an optimized methyl filtration protocol with subsequent next generation sequencing, a variant of MRE-seq. The NGS data was analyzed using different approaches illustrating the comparability of de-novo assembly and reference based mapping. Sequenced hypomethylated fragments from Oryza sativa and Picea abies were analyzed showing a substantial depletion of repetitive regions together with enrichment for transcribed regions. Furthermore a clear increase of sequencing coverage in active genomic regions can be observed. As the method targets genomic sequence not only within but also around genes, many important components are also represented including introns and potentially regulatory regions of promoters.

Short Abstract: The fig tree (Ficus carica L.) breeding programs by conventional methods such as directed crosses, in order to obtain new cultivars, are unworkable in many countries, as in Brazil. In this way, the genetic breeding, with the use of mutagenic, becomes an important research line for the improvement of culture, being necessary to gather information about this species, mainly in relation to its genetic variability, for perform propagation projects and appropriate management. Given the above, The objective of this study was to verify the existence of epigenetic variability due to DNA methylation in irradiated fig selections, with each other and when compared to the main commercial cultivar, Roxo-de-Valinhos, using MSAP and subsequent DNA sequencing, treated with sodium bisulfite, for detection of the position of the polymorphic regions, analyzed by bioinformatics tools. With the sequencing of DNA isolated of the differentially methylated sites, it was possible to verify different patterns of methylation in them by sequencing the DNA treated with sodium bisulfite, in coding regions of regulatory genes of the development and fruits ripening, besides they have been found in the mitochondrial DNA of treatments, which regulates the supply of energy in ATP form for the plants, being closely related to their development, justifying the different phenotypes found in both fruits and plant growth that suffered stress due to exposure to gamma radiation. Since the material used as control was found also methylated, a supposed demethylation of the genomic material may be responsible for phenotypic variation among treatments.

Short Abstract: ChIP-Seq experiments provide genome-wide data relating the binding of a protein of interest, or histone mark, to a genomic location. This is achieved by shearing the DNA, adding and precipitating an antibody that binds to the epigenetic mark of interest and sequencing the precipitate.

The typical ChIP-Seq control experiment, usually referred to as the 'Input', assays the available chromatin by shearing and sequencing the DNA without adding the antibody. In a typical bioinformatics workflow the Input is subtracted from the ChIP-Seq signal, or the latter is divided by the former, and the resulting signal is assumed to represent the presence of the mark of interest. This approach is unsatisfactory in many, if not all, applications. The Input is almost never sequenced to the same depth as the ChIP-Seq, and little attention is paid to the relative coverage ratios. Additionally, some data show Input signals that appear to mimic the ChIP-Seq signal remarkably closely, when considered at similar coverage levels.

Here we provide a principled approach to handling the Input, based on jointly learning functions of the Input and ChIP-Seq data from relevant genetic and epigenetic predictive data. Maximum likelihood parameter estimation for the high-dimensional, large data is achieved via a semi-stochastic optimisation approach. We treat the results in a Bayesian manner, allowing us to infer a cleaner signal for the mark of interest. The method is used to handle difficult-to-interpret ChIP-Seq data and to produce cleaner signals for protein-DNA binding analysis to aid motif discovery.

Short Abstract: Comparing histone modifications profiles across various conditions may provide insights into our understanding of cell specific properties and phenomena such as cell differentiation, disease development or effects of disease treatments. Characterizing and quantifying differential histone modifications profiles is a challenging problem. This is due to many technical aspects of the ChIP-seq protocol, which may result in different signal to noise ratio between conditions considered and different GC content bias. In addition, when comparing histone modification profiles in different cancer cells, one should consider the difference in the corresponding copy number profiles.

We present a method (HMCan-diff) for identifying differential histone modifications from ChIP-seq data applicable to cancer datasets. Our method takes into account replicates, mappability, GC-content, copy number variation and background noise. HMCan-diff can also utilize information from ChIP-seq spike in data, if available, for inter-condition normalization. HMCan-diff applies a three-state hidden Markov model (HMM) to identify regions characterized by differential histone modifications.

Short Abstract: Histone modifications associate with important regulatory regions such as promoters and distal enhancers that control the expression of genes. Time-course genome-wide maps of these epigenetic marks have become available in a growing number of biological settings including stem cell reprogramming and differentiation, adipogenesis, cardiac development, circadian rhythms, embryogenesis and lymphocyte development. However, our understanding of the underlying cellular processes remains limited, because the current bioinformatics tools often fail to utilize fully the temporal aspects of this data. Here, we present a novel computational method for systematic detection of major classes of spatio-temporal patterns of epigenetic changes. The method takes as input data from a series of chromatin immunoprecipitation followed by high throughput sequencing (ChIP-seq) experiments for a single histone mark that are performed at consecutive time points during a given biological process. The method uses a probabilistic mixture model that explicitly models the spatio-temporal nature of the data to identify regions for which the mark either expands or contracts significantly with time or holds steady. Furthermore, it incorporates information about replicate experiments at each time point, which can increase the accuracy of the method. We present applications of the method on publicly available data from T-cell development, which help in understanding the underlying regulatory dynamics during this process.

Short Abstract: In embryonic stem (ES) cells, developmental regulators have a characteristic bivalent chromatin signature marked by simultaneous presence of both activation (H3K4me3) and repression (H3K27me3) signals. These genes are thought to be in a 'poised' state for subsequent activation or silencing during differentiation, implying an important role for epigenetic modifications in directing cell fate decisions. We collected eleven pairs (H3K4me3 and H3K27me3) of ChIP sequencing datasets in human ES cells and eight pairs in murine ES cells, and predicted high-confidence (HC) bivalent promoters. Over 85% of H3K27me3 marked promoters were bivalent in human and mouse ESCs. HC bivalent promoters were enriched for developmental factors and were more likely to be differentially expressed upon transcription factor perturbation than active (H3K4me3-only) promoters. Bivalent promoters were CpG rich while H3K27me3-only promoters lacked CpG islands. Murine HC bivalent promoters were occupied by PRC1, PRC2 and RNA polymerase II and grouped into four distinct clusters with different biological functions. Using ChIP sequencing data, we demonstrated that bivalent and active promoters are enriched for binding of distinct sets of regulators. Finally, we identified a ‘TCCCC’ sequence motif specifically enriched in bivalent promoters in both human and mouse ESCs.

Short Abstract: Accumulation of tau proteins and amyloid-β peptides in the brain are two hallmarks of Alzheimer’s Disease (AD). Recent studies suggest that epigenetic mechanisms are likely to play a key role in the pathogenesis of AD. Here, we studied genome wide the active mark H3K9ac using ChIP-seq in 669 post-mortem human brain samples to detect alterations of the epigenome induced by tau. RNA-seq was performed for 500 samples to assess the effect on transcription. We considered modifications of local H3K9ac domains as well as large genomic regions and distinguished alterations primarily associated with tau from those with amyloid.

We identified 26,384 H3K9ac domains which primarily occurred at promoters (15,225) and enhancers (8,071). H3K9ac levels at promoters were positively correlated with transcription, even though H3K9ac alone was not sufficient for transcription. Tau protein loads were significantly associated with H3K9ac levels in 5,980 domains and had a much broader impact than amyloid (610 domains). Domains positively associated with tau showed a strong enrichment (p<10^-16) for binding sites of CTCF, which regulates chromatin structure. Indeed, we found large genomic regions showing concordant tau associated increases in H3K9ac. Average transcription in these regions was consistently up-regulated. Strikingly, effect sizes within the regions were highly correlated with the regions' proportion of open chromatin.

Our results demonstrate a genome wide change in chromatin structure in AD, which is mediated by tau. Tau is known to cause heterochromatin relaxation in Drosophila models. CTCF could be a key factor in the pathogenic process of chromatin opening.

Short Abstract: Epigenome-wide association studies (EWAS) provide a novel means of studying the epigenetic basis of human disease. A challenge confronting EWAS though is the assessment of tissue specificity of identified differentially methylated positions (DMPs) and regions (DMRs). For example, DMPs identified in an EWAS using heterogeneous tissue samples, such as whole blood and brain, may originate from only a distinct subpopulation of cell types. To this end, we have developed an analysis approach that determines the tissue-specific regulatory component of a set of EWAS DMPs through the detection of enrichment of overlap with DNase I hypersensitive sites (hotspots) across a wide range of tissues. Our tool, eFORGE (experimentally-derived Functional element Overlap analysis of ReGions from EWAS), is available as standalone software and provides tabular and graphical summaries of the enrichments.

This tool is derived from FORGE (http://www.1000genomes.org/forge-analysis), which analyses the tissue-specific regulatory component of SNPs in the context of genome-wide association studies (GWAS). For a given set of significant EWAS DMPs (i.e., 450K array probes), eFORGE generates 1000 randomly selected background sets, matched for gene feature and CpG island relationship, and calculates a binomial P-value for each of the cell types catalogued in NIH Epigenomics Roadmap and ENCODE datasets. eFORGE can test for both enrichment (e.g., hypo-DMPs) and depletion (e.g., hyper-DMPs) of the input set, and the type of test should be chosen with careful consideration of the underlying biology.

eFORGE (http://eforge.cs.ucl.ac.uk/) provides a user-friendly tool to investigate the tissue-specific component of epigenetic marks identified through EWAS, and has the potential to reveal mechanistic disease insights

Short Abstract: Histone proteins are subject to a number of post-translational modifications, an important mechanism for chromatin regulation. Changes in chromatin regulation lead to different diseases.
Through genome-wide measurements of these modifications, e.g., H3K4me3, H3K36me3, studies have investigated the prevalence of modifications along genes. It was shown that the abundance of histone modifications around the transcription start site (TSS) of a gene is predictive of gene's expression (Karlic et. al., PNAS 2009, 7).
New studies report that many genes also show expression in antisense direction of their TSS. However, little is known about the mechanisms of antisense expression for regulating gene expression. In this work, we are interested to gain insight into the association between genomic abundance of histone modifications and sense/antisense expression.
For this purpose, we use methods from machine learning to build predictive models of sense and antisense expression using ChIP-seq and RNA-seq data from different cell types. We suggest the use of the fused lasso approach to capture the correlation of histone signals along the genome (Tibshirani et. al., Royal Statistical Society 2005, 1). This technique is able to tradeoff fusion of coefficients against a sparse solution. We show that the fused lasso improves performance and leads to stable feature prediction.
Also, we demonstrate that the histone signals selected by the models differ between sense and antisense prediction and fit the knowledge reported in the literature, but also generate new hypotheses. We find that sense and antisense expression follows a unique histone code that is reproducible for different cell types.

Short Abstract: Human IPSCs have the unique ability to self-renew in vitro while maintaining their capacity to differentiate into derivatives of the three germ layers. Thus, they represent a valuable system for disease modelling, drug screening and ultimately cell-based therapy. The success of human IPSC cell-based therapies will depend on the development of efficient and robust methods to differentiate human IPSCs into clinically relevant and safe population of cells. However, systematic studies are currently lacking to evaluate the quality and safety of human IPSC lines generated with different reprogramming methods and from different cell types. Here, we perform extensive genome wide analyses to define the impact of these different aspects on pancreatic cell production from human IPSCs. We generated human IPSC lines from both blood and fibroblasts of 5 donors using Sendai virus and episomal plasmids (n=45). In collaboration with the HipSci initiative, the resulting lines were characterised using extensive genome wide assays including RNA-seq, ChIP-seq, exome sequencing, 450K-methylation arrays as well as M-FISH karyotyping. Furthermore, human IPSC lines were subsequently differentiated into pancreatic beta-like cells and subjected to RNA-seq and functional characterisation. These analyses reveal the impact of several confounding factors that could affect pancreatic cells production from human IPSCs including genetic variability, impact of reprogramming and somatic cell memory. Together, these results provide an approach to characterise at a deeper molecular level human IPSCs and their differentiated derivatives thereby providing new standards for their use in the clinic.

Short Abstract: DNA methylation is one of the most widely used markers for the study of epigenetic contributions to phenotypic variation and disease. There are several methods for analyzing genome-wide DNA methylation data in common use, but there has been no rigorous evaluation of their performance. We have performed a systematic assessment and comparison of four packages: MethySig, methylKit, eDMR and DSS, using an empirical dataset of 12 reduced representation bisulphite sequencing libraries (6 test, 6 control). Surprisingly, we observed very low concordance among these commonly used model-based and binomial test-based approaches: using equivalent pre-processing and filtering parameters for each method, we found that the four methods identified significant differentially methylated cytosines at a concordance rate of less than 1%. Similarly low levels of concordance were observed with identification of differentially methylated regions using tiled data. Our study highlights the need for systematic approaches to reliable differential methylation analysis via data simulation. This concept of simulation will be discussed in the context of the growing implementation of epigenomic data in human medicine.

Short Abstract: Super-enhancers are clusters of transcriptional enhancers which play a key role in cell type specific gene expression and have been found highly associated with many human diseases. Since super-enhancers are cell-type specific and less in number, it will be useful to discover and characterize super-enhancers rather than exploring thousands of enhancers operating in a cell. The recent research well demonstrates the importance of super-enhancers in regulation of gene expression. However, the identification is challenging at this stage due to limited knowledge of potential features. The master transcription factors for many cell types is not known and very limited amount of ChIP-seq data is available of Mediators, which are highly enriched for super-enhancers. Here, we developed a supervised machine learning method to predict super-enhancers from a list of enhancers. We trained our model using variety of different signatures including histone modifications, chromatin regulators, protein co-activators, Mediators and sequence motif data in mouse embryonic stem cells (mESC). After applying different feature ranking methods, we found H3K27ac, Brd4 and Med12 as a best feature subset while Brd4 is better predictor then H3K27ac. We evaluated the performance of our model using 10-fold cross validation and achieved AUC 0.93, and also using an independent test data for four different cell-types in human with average AUC 0.91. We used and compared six different state-of-the art learning models. Our model can accurately predict super-enhancers for other cell types in mouse and human genome.

Short Abstract: Genome-wide the accessibility of genomic DNA is regulated by chromatin opening and closing. This central epigenomic effect can be measured by FAIRE-seq, which is a next-generation sequencing technique that detects accessible genomic DNA via nucleosome depletion. Through the transcription factor VDR the vitamin D metabolite 1,25(OH)2D3 has a direct effect on gene regulation. In an experimental time series over 3 months, in which one human individual was challenged every 28 days with a bolus of 80,000 IU vitamin D3, we determined by FAIRE-seq genome-wide changes in open chromatin in white blood cells (PBMCs) at days 0, 1, 2, 28, 29, 30, 56, 57, 58 and 92. We applied several machine learning and data mining algorithms (self-organizing maps, k-means, and random forests) to chromatin opening time series at more than 5,000 genomic loci, to detect a small number of relevant profiles. Over 300 genomic loci were persistent at all time points, the vast majority of which show changes in chromatin accessibility. We identified several groups of chromatin accessibility profiles for the 3 months observation period. These profiles were compared with transcriptomic data from the same individual and epigenomic data from in vitro cell models, such as THP-1 human monocytes. The majority of the identified profiles reflect the epigenome-wide impact of vitamin D3. The most prominent impact on the epigenome were observed at a serum vitamin D status of more than 100 nM, which doubles the presently recommended level. Thus, vitamin D3 supplementation has a direct effect on the epigenome of PBMCs.

Short Abstract: Identification of changes in DNA-protein interactions from ChIP-seq data is an important step to unravel regulatory biological processes. The differential peak calling problem is about finding genomic regions with changes in ChIP-seq signals describing the interaction of a protein with DNA between two cellular conditions. Several approaches, so-called two-stage differential peak callers, identify such genomic regions by using a combination of single peak callers with statistical tests for detecting differential digital expression. These two-stage differential peak callers fail to detect subtle changes of protein-DNA interactions. One-stage differential peak callers are based on signal segmentation strategies. We propose HMM-based one-state differential peak callers: ODIN finds differentials peaks in pairs of ChIP-seq data, whereas THOR is able to take replicates of ChIP-seq experiments into account. Both tools perform genomic signal processing, peak calling and p-value calculation in an integrated framework. We also propose an evaluation methodology to compare ODIN and THOR with competing methods. The evaluation is based on the association of differential peaks with expression changes in the same cellular conditions as well as simulated data. Our empirical study based on several ChIP-seq experiments from transcription factors, histone modifications and simulated data shows that our approaches perform better in most scenarios compared to the considered competing methods.

Short Abstract: FoxM1 and Kaiso are two important oncogenes in non-small cell lung carcinomas, however few studies demonstrate the direct regulatory relationship between FOXM1 and ZBTB33. We have performed CHIPseq of FOXM1, H3K27ac and H3K27me3 in PTC02 (a local non-small lung cancer cells) and RNAseq between PCT02 and normal control tissues. FOXM1 genome-wide ChIP-seq data showed that ZBTB33 (Kaiso) binding motif was significantly enriched on FOXM1 binding sites in PTC02, GM12878 and ECC1. The regions with both FOXM1 binding and Kaiso motif, which were defined as F&K regions, had higher level of FOXM1 binding tags and higher ratio of sites occupied by FOXM1, compared to the regions merely with FOXM1 binding sites (defined as F region) or the Kaiso motif (defined as K region). The F&K regions were also highly bound by the Kaiso protein across five cell lines GM12878, A549, SKnsh, Hepg2, K562 with over 90% of sites occupied by Kaiso. Interestingly, higher level of H3K27ac and lower level of H3K27me3 appeared on the F&K region than F or K regions. The genes associated with F&K regions were significantly upregulated in PCT02 and 60 paired lung cancer tissues. Functional enrichment analysis showed that these F&K associated genes were highly enriched on the UBC (ubiquitin C) interaction partners. In summary, this study suggests that the FOXM1 and Kaiso might work together to upregulate ubiquitin C related genes for promoting non-small cell lung cancer. Further studies will investigate causal the relation between FOXM1 and Kaiso by knockdown of FOXM1 or Kaiso.

Short Abstract: Imprinted genes have been target of many studies, mainly in human and mouse, and lately in bovines due to the interest of understanding the epigenetic mechanisms underlying important meat quality phenotypes and the possibility of applying it in animal breeding programs in the future. Genomic DNA from 146 steers was genotyped using the Illumina BovineHD BeadChip in order to identify heterozygous individuals with known allele origin. Total mRNA from muscle was extracted and sequenced by Illumina HiScanSQ. The software ALEA was used to create a diploid genome for each individual, in which haplotype regions were reconstructed from the individual haplotypes. ALEA also maps short sequencing reads to the in silico genome constructed and detects reads that are uniquely aligned only to one of the two haploid genomes. In-house software was developed to compute the frequency of reads mapped to each allele and to perform binomial statistical tests in order to identify allele specific expression. From 742,906 SNPs contained in the Illumina BovineHD BeadChip, 419 were assigned to be imprinted based on the following criteria: heterozygous in the individual, homozygous in its sires, at least 20x RNA-Seq coverage, and p<0.05 for the statistical test. The Ensembl software VeP (Variant Effect Predictor) was used to determine the effect of these SNPs on genes, transcripts, and protein sequence, as well as regulatory regions. The VeP report together with phenotype information of the steers will be carefully examined in order to elucidate the molecular mechanisms involved with beef quality.

Short Abstract: Differential exon usage has been reported to affect the large majority of genes in mammalian genomes. It has been shown that different splice forms sometimes have distinctly different protein function. Here, we present an analysis of the Human Epigenome Atlas (version 8) to connect the differential usage of exons in various developmental stages of human cells/tissues to differential epigenetic modifications at the exon level. We found that the differential incidence of protein isoforms across developmental stages is often associated with changes in histone marks as well as changes in DNA methylation in the gene body or the promoter region. Many of the genes that are differentially regulated at the exon level were found to be associated with development and metabolism

Short Abstract: Chromatin organisation, such as the deposition of active or repressive histone modifications, plays an important role in regulating gene expression. Advances in ChIP-seq and associated bioinformatics techniques have enabled genome-wide analysis of chromatin modification and transcription regulation dynamics. The chromatin landscape of human, mouse, and several model organisms have been widely studied, but other medically and biotechnologically important species, including many fungal species, are not yet fully studied. Here we present a genome-wide chromatin landscape of an important filamentous fungal model organism, Aspergillus nidulans. Using ChIP-seq, we generated genome-wide profiles of H3, H3K4me3, H3K4me1, H3K27ac, H3K9me3, H3K9ac, PolII and transcription factors in A. nidulans under different nutrient availability and quality as a novel approach to dissect the regulation of cellular metabolic homeostasis. We used a recently developed hierarchically linked infinite hidden Markov model (hiHMM) to systematically discover and characterise the most prevalent combination of histone modifications, i.e., chromatin state.

Our analysis revealed an interesting new chromatin state that consists of both H3K9me3 and H3K27ac, and is associated with gene repression. This state is covering 15-20% of the genome, and is located most prevalently in sub-telomeric regions, covering secondary metabolism and transmembrane transport genes. In addition, 10-20% of the genome are marked by classical enhancer chromatin marks, suggesting that these previously uncharacterised regions may have potential regulatory functions. Our work represents a valuable resource in A. nidulans that opens new avenues for investigation of the dynamic chromatin organisation and gene regulation in fungi.

Short Abstract: With Hi-C, a genome-wide analysis of chromatin interaction profiles is now possible. These interactions comprise pairs of loci that are close spatially, but not necessarily close in the sequence. This spatial co-localization of different chromosomal regions (cis as well as trans) can be due to a complex combination of factors viz. specific, direct contacts between two loci, nonspecific binding as a result of the packing of the chromatin fibre or co-localization due to functional association or having the same subnuclear structure.

In this work we show that the underlying genomic sequence at these loci is predictive of the long-range chromatin interactions involving these loci. We achieve this by casting it into a binary classification problem -- using a string kernel to measure similarity between the genomic sequences and a support vector classifier to classify a given set of linearly distal regions into spatially proximal or distal with respect to a particular TSS-containing region. In the first part we use the oligomer distance histograms kernel over the sequences and train an SVM downstream, while in the second we build a more sophisticated model with multiple kernels over putative fragments of these sequences thus capturing the hypothesized combinational mechanisms of interaction. Additionally, we exploit the sequences and their putative causal fragments to further improve the classification performance by employing multiple kernel learning to obtain the optimal weighting for kernels and multiple instance learning to accommodate for the lack of knowledge on the definite causal fragment(s). Our computational experiments also provide meaningful insights.

Short Abstract: Medulloblastoma is a highly malignant paediatric brain tumour, often inflicting devastating consequences on the developing child. Genomic studies have revealed four transcriptionally distinct molecular subgroups with divergent biology and clinical behaviour. An understanding of the regulatory circuitry governing the transcriptional landscapes of medulloblastoma subgroups, and how this relates to their respective developmental origins, is currently lacking. Using H3K27ac and BRD4 ChIP-Seq, coupled with tissue-matched DNA methylation and transcriptome data, we describe the active cis-regulatory landscape across 28 primary medulloblastoma specimens. Analysis of differentially regulated enhancers and super-enhancers reinforced inter-subgroup heterogeneity and revealed clinically relevant insights into oncogenic TGFβ signaling in Group 3. Computational reconstruction of core regulatory circuitry identified a master set of transcription factors responsible for subgroup divergence and revealed candidate cellular origins for Group 4. The integrated analysis of cis-regulatory elements in primary human tumour samples reveals insights into cis-regulatory architecture, unrecognized dependencies, and cellular origins.

Short Abstract: Active regulatory regions are occupied by transcription factors (TFs) and display a defined chromatin state. While the relevance of chromatin in gene regulation is undisputed, it is controversial whether a chromatin state is a consequence of or prerequisite for TF binding. Since TFs differ in their sensitivity to chromatin (e.g. pioneer TFs), the relationship between chromatin and TF binding is factor-specific and context-dependent. We measured the effect of DNA methylation on the genomic binding landscape of TFs by using DNase I hypersensitive sites (DHS) as a comprehensive indicator of TF binding. Monitoring changes in DHS reveals that removal of DNA methylation in mouse embryonic stem cells does not globally perturb the DNase I landscape. This argues that binding of most TFs expressed in stem cells is not limited by DNA methylation. Importantly however at specific loci defined TFs reveal sensitivity and appear to compete with DNA methylation for binding. This allows to investigate the interplay between transcription factors and DNA methylation.

Short Abstract: Post-translational modifications (PTMs) of histone proteins play an important role in maintaining chromatin structure and in the dynamic regulation of DNA replication and repair, transcription of genes and propagation of epigenetic traits. One of the largest challenges in current chromatin biology is to characterize the relationships between co-existing histone PTMs, the order and hierarchy of their deposition and their distinct biological functions. We developed a work flow to analyze the co-existing marks as revealed by "middle-down" MS experiments of histone proteins. The implemented new method quantifies positive and negative interplay between pairs of methylation and acetylation marks in proteins. Many of the detected features were conserved within different cell lines, thereby revealing general rules for crosstalk between histone marks. The observed features are not only in accordance with previously reported examples of crosstalk but also revealed novel types of interplay. Integration with data coming from ChIP-seq allowed aligning the frequency of histone PTMs with MS-measured levels. Further investigation of the involved pathways reveals a well-defined hierarchy of biological functions within single and binary histone marks. These results show that we gather deeper insight into chromatin function by consideration of multiple marks e.g. measured with mass spectrometry approaches.

Short Abstract: DNA methylation is one of the best known epigenetic players involved in the regulation of gene expression. A major influence in transcription is achieved by methylated cytosines located at gene regulatory elements. Previous studies have shown that Growth arrest and DNA damage protein 45 alpha (Gadd45a) has been implicated in epigenetic gene activation by promoting site specific active DNA demethylation. Our lab has also found that Inhibitor of growth protein 1 (Ing1), a tumor suppressor, is required for targeting Gadd45a to the sites of demethylation. Both Gadd45a and Ing1 share similar functional properties such as regulating DNA repair, growth arrest, senescence and apoptosis. Gadd45a along with its cofactor Ing1 participates in the demethylation process by promoting DNA repair at specific genomic loci. However, the genome-wide target sites of Gadd45a/Ing1 mediated demethylation are still unknown.

We have carried out whole genome bisulfite sequencing (WGBS) of wild-type and Gadd45a/Ing1 double knock-out mouse embryonic fibroblasts (MEFs) at single base resolution. We observed site-specific differentially methylated CpGs. Clusters of methylated CpGs led to the identification of differentially methylated regions (DMRs). We identified considerably higher number of hypermethylated than hypomethylated regions in Gadd45a/Ing1 deficient MEFs compared to wild-type cells. Importantly, these DMRs are preferentially found at the gene regulatory regions. Using MEFs RNA-Seq data, we also found that the identified DMRs are correlated with deregulated gene expression. This study not only corroborates the role of Gadd45a and Ing1 in gene specific DNA demethylation but reveals significant insight into the role of both proteins in gene regulation.