A dissertation submitted to the Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, in fulfillment of the requirements for the degree in Master of Science (Medicine) in the Division of Human Genetics / Epigenetic mechanisms regulate gene expression, a particularly important activity during foetal development. DNA methylation contained within promoter and regulatory intergenic regions influence gene activity. In utero alcohol exposure as a result of maternal consumption during pregnancy has been associated with disruption of foetal DNA methylation and gene expression, leading to neurological dysfunction, growth retardation and facial anomalies. While similar phenotypes in offspring have been associated with chronic preconception paternal alcohol exposure, the mechanisms underlying these effects remain largely unexplored. This study aimed to: (1) validate significant changes in sperm DNA methylation in a list of ten candidate genes in male mice chronically exposed for ten weeks to ethanol (n=10) compared to a calorie-equivalent sucrose solution (n=10); (2) validate significant changes in gene expression in candidate genes in the brain, liver and placenta of E16.5 embryos sired by ethanol (n=24) compared to sucrose (n=24) treated male mice; (3) quantify DNA methylation changes in candidate genes in the three embryonic tissues. (4) Lastly, previously generated microarray data were reanalysed using bioinformatics tools to generate a top ranked candidate differentially expressed gene list that was used to identify and analyse biological functions or pathways significantly over represented among these genes using PANTHER and DAVID. This study was unable to provide validation for most of the significant differences observed in the sperm DNA methylome in the original study, most likely because of the low sperm DNA concentration. Significant methylation differences were however observed at individual CpG sites in three candidate genes (Igf1r, Odc1, Depdc1b) in specific tissues of embryos sired by ethanol-exposed males relative to embryos sired by sucrose-treated males. There was concordance in the direction of altered gene expression between the cases and controls using the microarray and real-time PCR approaches for two genes in the brain (Grm7 and Zfp317), three genes in the liver (Igf1r, Vwf and Depdc1b) and one gene in the placenta vii (Vwf). However, none of the candidate genes selected for validation showed statistically significant changes. This may be a result of the modest fold changes observed in the microarray experiment that as shown in many cases, often do not replicate. The remainder of the genes showed no changes in expression in the test embryos relative to the control. The functional enrichment analysis revealed biological processes that were over represented in the brain and liver indicating that they may be more vulnerable to the effects of alcohol, compared to the placenta. Overall, the study could not provide a statistically significant correlation between methylation changes in the sperm that were inherited by the offspring which subsequently dysregulated gene expression in the embryo. However, as trends toward significance and significant DNA methylation changes were observed in the embryonic tissues, this study supports the idea that preconception paternal alcohol exposure can induce epigenetic alterations in a locus and organ specific manner within offspring. / MT2016

Genome-wide association studies (GWAS) have been tremendously successful in identifying genetic variants associated with complex diseases, such as rheumatoid arthritis (RA). However, the majority of these associations lie outside traditional protein coding regions and do not necessarily represent the causal effect. Therefore, the challenges post-GWAS are to identify causal variants, link them to target genes and explore the functional mechanisms involved in disease. The aim of the work presented here is to use high level bioinformatics to help address these challenges. There is now an increasing amount of experimental data generated by several large consortia with the aim of characterising the non-coding regions of the human genome, which has the ability to refine and prioritise genetic associations. However, whilst being publicly available, manually mining and utilising it to full effect can be prohibitive. I developed an automated tool, ASSIMILATOR, which quickly and effectively facilitated the mining and rapid interpretation of this data, inferring the likely functional consequence of variants and informing further investigation. This was used in a large extended GWAS in RA which assessed the functional impact of associated variants at the 22q12 locus, showing evidence that they could affect gene regulation. Environmental factors, such as vitamin D, can also affect gene regulation, increasing the risk of disease but are generally not incorporated into most GWAS. Vitamin D deficiency is common in RA and can regulate genes through vitamin D response elements (VDREs). I interrogated a large, publicly available VDRE ChIP-Seq dataset using a permutation testing approach to test for VDRE enrichment in RA loci. This study was the first comprehensive analysis of VDREs and RA associated variants and showed that they are enriched for VDREs, suggesting an involvement of vitamin D in RA.Indeed, evidence suggests that disease associated variants effect gene regulation through enhancer elements. These can act over large distances through physical interactions. A newly developed technique, Capture Hi-C, was used to identify regions of the genome which physically interact with associated variants for four autoimmune diseases. This study showed the complex physical interactions between genetic elements, which could be mediated by regions associated with disease. This work is pivotal in fully characterising genetic associations and determining their effect on disease. Further work has re-defined the 6q23 locus, a region associated with multiple diseases, resulting in a major re-evaluation of the likely causal gene in RA from TNFAIP3 to IL20RA, a druggable target, illustrating the huge potential of this research. Furthermore, it has been used to study the genetic associations unique to multiple sclerosis in the same region, showing chromatin interactions which support previously implicated genes and identify novel candidates. This could help improve our understanding and treatment of the disease. Bioinformatics is fundamental to fully exploit new and existing datasets and has made many positive impacts on our understanding of complex disease. This empowers researchers to fully explore disease aetiology and to further the discovery of new therapies.

New advances in genomic technology make it possible to address some of the most fundamental questions in biology for the first time. They also highlight a need for new approaches to analyze and model massive amounts of complex data. In this thesis, I present six research projects that illustrate the exciting interaction between high-throughput genomic experiments, new machine learning algorithms, and mathematical modeling. This interdisci- plinary approach gives insights into questions ranging from how variations in the epigenome lead to diseases across human populations to how the slime mold finds the shortest path. The algorithms and models developed here are also of interest to the broader machine learning community, and have applications in other domains such as text modeling. / Mathematics

archives@tulane.edu / 1 / Yuntong Bai

Pour enrichir les connaissances scientifiques sur les effets biologiques des radionucléides et risques des contaminations chroniques sur la descendance, une étude multigénérationnelle in vivo d’exposition a été réalisée à doses non toxiques d'uranium. Ce modèle, a permis de suivre les effets biologiques de l’uranium sur trois générations de rats (F0, F1 et F2) par des analyses cliniques et le suivi de marqueurs biologiques. Dans cette étude, des analyses métabolomiques, transcriptomiques et épigénomiques ont été réalisées à partir d’échantillons de sang, d’urine et de rein.Pour la première génération des rats contaminés (F0), des différences dépendant du sexe des animaux sont observables par l’analyse des niveaux d’expression géniques (ARNm et micro-ARN) dans les reins, des profils métabolomiques et biochimiques dans les reins, l’urine et le sang. Aucune modification épigénétique des profils de méthylation de l’ADN rénal n’est à noter. Pour les deux générations suivantes (F1 et F2), un effet multigénérationnel dépendant aussi du sexe des rats est observable au niveau des profils métabolomiques urinaires et rénaux ainsi qu’au niveau des profils épigénétiques de méthylation de l'ADN des reins. Une baisse de poids corporel et des reins a aussi été observée pour la troisième génération de rats chez les mâles (F2).En conclusion, les travaux de cette thèse montrent qu’une contamination chronique à faible dose d'uranium entraine des effets biologiques sur plusieurs générations de rats. Ils sont observables à différents niveaux moléculaires des systèmes de régulation cellulaires et dépendent du sexe des rats. Ces effets, étroitement liés à des systèmes biologiques intégrés, sont utiles à la compréhension des mécanismes biologiques des expositions à l'uranium et à l’évaluation des risques de nocivités à long termes. Dans le domaine de la radioprotection, ces résultats justifient la nécessité de considérer les dimorphismes sexuels des individus et les conséquences des expositions sur les générations à venir. / In order to deepen scientific knowledge regarding biological effects of radionuclides and associated risk to offspring, an in vivo multigenerational study of chronic exposure to a non-toxic dose of uranium was performed by monitoring three generation of rats (F0, F1 and F2). Clinical parameters and biological markers, including metabolomics, transcriptomics and epigenomics high throughput analysis were conducted in blood, urine and kidney samples.For the first generation of contaminated rats (F0) sex-differences to uranium effects were observed in kidney for gene expression (mRNA, miRNA) and in kidney, urine and blood for biochemical parameters and metabolomics profiles. No epigenetic modification of DNA methylation profiles was shown in kidney. For the next two generations (F1, F2), a multigenerational sex-specific effect is observed for both metabolomics and renal DNA methylation profiles of contaminated rats. Moreover, for the last generation of male rats (F2), a decrease of both total body and kidney weight was shown.In conclusion, low-dose chronic contamination of rats to uranium leads to multigenerational effects. Including sex-differences, they can be shown at different molecular levels of the cellular system. Depending of integrated system biology, data of this thesis are useful in the understanding of biological mechanisms of uranium effect and risk of delayed harmful effect. In the field of radiation protection, these results prove the requirement of considering sexual dimorphisms and consequences of such exposures to offspring.

Barley (Hordeum vulgare) is an economically important crop species with a large diploid genome. Around a half of the barley genome and a fifth of the genes are constrained within a low-recombining pericentromeric (LR-PC) region. I explored the LR-PC gene component with a genomic investigation of gene expression, diversity and evolution. Chromatin environments were also explored in the LR and high recombining (HR) regions by surveying the genic and genomic distributions of nine histone modifications. Firstly, regions of HR and LR were identified and compared for gene evolution, expression and diversity. LR regions of the barley genome were found to be restrictive for gene evolution and diversity, but not gene expression. I employed a bioinformatics approach to identify ancient gene pairs in barley to determine the long-term effects of residency in those regions upon gene evolution. Gene pair loss in LR regions was found to be elevated relative to the HR regions. Applying the same method to rice and Brachypodium distachyon revealed the same situation, suggesting a universal process in the grasses for loss of gene pairs in LR regions. The chromosomal distributions of transposable elements (TEs) were also explored and examined for correlations with recombination rate. Secondly, I developed a chromatin immunoprecipitation followed by Next Generation Sequencing (ChIP-seq) protocol for the investigation of histone modifications in barley seedlings. A protocol was optimised for the fixation, extraction and sonication of barley chromatin. The protocol was applied using antibodies against 13 different histone modifications. Following DNA library construction and Illumina sequencing, a bioinformatics pipeline was devised to analyse the sequence data. NGS reads were mapped to a custom assembly of the barley cultivar Morex reference genome sequence before peak calling. Genomic and genic locations were determined for the covalently modified histones. Four modifications were discarded from further study on the basis of low peak numbers or unexpected chromosomal locations. The remaining nine modifications were classified into four groups based on chromosomal distributions. Groupings were closely mirrored by peak sharing relationships between the modifications except histone H3 lysine-27 tri-methylation (H3K27me3). In addition, chromatin states representing local chromatin environments were defined in the barley genome using the peak sharing data. Mapping the states onto the genome revealed a striking chromatin structure of the gene-rich chromosome arms. A telomere-proximal region bearing high levels of H3K27me3-containing states was found adjacent to an interior gene-rich region characterised by active chromatin states lacking H3K27me3. The LTR retroelement-rich interior was found to be associated with repressive chromatin states. The histone modification status of TE classes were also probed revealing unexpected differences relating to the genomic and genic distributions of these elements. Finally, a genome browser was created to host the information publicly.

Cancer progression is driven by cumulative changes that promote and maintain the malignant phenotype. Epigenetic alterations are central to malignant transformation and to the development of therapy resistance. Changes in DNA methylation, histone acetylation and methylation, noncoding RNA expression and higher-order chromatin structures are epigenetic features of cancer, which are independent of changes in the DNA sequence. Despite the knowledge that these epigenetic alterations disrupt essential pathways that protect cells from uncontrolled growth, how these modifications collectively coordinate cancer gene expression programs remains poorly understood. In this dissertation, I utilize molecular and informatic approaches to define and characterize the genome-wide epigenetic patterns of two important human cancer cell models. I further explore the dynamic alterations of chromatin structure and its interplay with gene regulation in response to therapeutic agents. In the first part of this dissertation, pancreatic ductal adenocarcinoma (PDAC) cell models were used to characterize genome-wide patterns of chromatin structure. The effects of histone acetyltransferase (HAT) inhibitors on chromatin structure patterns were investigated to understand how these potential therapeutics influence the epigenome and gene regulation. Accordingly, HAT inhibitors globally target histone modifications and also impacted specific gene pathways and regulatory domains such as super-enhancers. Overall, the results from this study uncover potential roles for specific epigenomic domains in PDAC cells and demonstrate epigenomic plasticity to HAT inhibitors. In the second part of this dissertation, I investigate the dynamic changes of chromatin structure in response to estrogen signaling over a time-course using Estrogen Receptor (ER) positive breast cancer cell models. Accordingly, I generated genome-wide chromatin contact maps, ER, CTCF and regulatory histone modification profiles and compared and integrated these profiles to determine the temporal patterns of regulatory chromatin compartments. The results reveal that the majority of alterations occur in regions that correspond to active chromatin states, and that dynamic chromatin is linked to genes associated with specific cancer growth and metabolic signaling pathways. To distinguish ER-regulated processes in tamoxifen-sensitive and in tamoxifen-resistant (TAMR) cell models, we determined the corresponding chromatin and gene expression profiles using ER-positive TAMR cancer cell derivatives. Comparison of the patterns revealed characteristic features of estrogen responsiveness and show a global reprogramming of chromatin structure in breast cancer cells with acquired tamoxifen resistance. Taken together, this dissertation reveals novel insight into dynamic epigenomic alterations that occur with extrinsic stimuli and provides insight into mechanisms underlying the therapeutic responses in cancer cells.

Le cancer de la prostate est le plus fréquent et représente la troisième cause de mortalité par cancer chez l’homme en France. Outre la théorie de la génétique dans le développement des cancers, l’implication des modifications épigénétiques au cours de la carcinogenèse prostatique ne fait plus aucun doute. L’épigénétique est définie comme l’étude des modifications de l’expression des gènes qui sont transmissibles lors de la mitose et/ou la méiose, mais ne découlent pas de modifications dans la séquence de l’ADN (Berger et al., 2009). Autrement dit, les altérations épigénétiques sont capables de moduler le niveau d’expression des gènes en agissant sur la chromatine. Depuis quelques années, l’attention des chercheurs porte de plus en plus sur l’implication des modifications épigénétiques dans la carcinogenèse prostatique. Une activité anormale d’un ou de plusieurs acteurs épigénétiques entraîne des modifications de profil d’expression des gènes pouvant être associées à la carcinogenèse. Dans le cancer de la prostate, on retrouve ainsi une forte activité des ADN méthyltransférases (DNMT1) et une dérégulation des modificateurs d’histones, notamment de l’histone méthyle transférase EZH2, associée à la répression de la transcription de certains gènes (Gillio-Tos et al., 2012; Gravina et al., 2013; Koh et al., 2011; Yu et al., 2007). En plus de ces facteurs épigénétiques, des facteurs environnementaux et plus particulièrement la nutrition intervient dans les processus de carcinogenèse. La thématique du laboratoire a été longtemps axée sur la nutrition et le développement des cancers. En effet, des études sur les effets préventifs des phyto-oestrogènes du soja dans la carcinogenèse prostatique ont largement été abordées (Adjakly et al., 2011). Les phyto-oestrogènes du soja ont montré leur effet déméthylant sur les promoteurs des gènes suppresseurs de tumeurs. L’hypothèse de ces travaux a été établie du fait de la faible incidence du cancer de la prostate retrouvée dans les pays asiatiques où une forte consommation de soja a été remarquée. Les micronutriments contenus dans le soja auraient ainsi la capacité de moduler les modifications épigénétiques dans le cancer de la prostate (Vardi et al., 2010). Dans le cadre de cette thèse, nous nous sommes focalisés sur les modifications des histones notamment la méthylation des histones qui reste peu étudiée dans le cancer de la prostate. Premièrement, nous avons étudié le rôle de la triméthylation de la lysine 27 de l’histone H3 (H3K27me3) dans l’implication et la progression du cancer de la prostate. La deuxième partie a consisté en l’identification de nouveaux marqueurs pronostiques et épigénétiques liés à la méthylation des histones afin d’établir un profil épigénétiques des tumeurs prostatiques. Nous avons également évoqué l’effet des phyto-oestrogènes du soja sur la modulation des modifications épigénétiques dans le cancer de la prostate et leurs rôles protecteurs dans le développement tumoral. / Prostate cancer is the most common and represents the third leading cause of cancer death in men in France. In addition to the theory of genetics in the development of cancers, the involvement of epigenetic changes during prostate carcinogenesis is no longer in doubt. Epigenetics is defined as the study of changes in the expression of genes that are transmissible during mitosis and / or meiosis, but do not result from modifications in the DNA sequence. In other words, epigenetic alterations are able to modulate the level of gene expression by acting on chromatin. In recent years, the attention of researchers has increasingly focused on the involvement of epigenetic modifications in prostatic carcinogenesis. Abnormal activity of one or more epigenetic actors leads to changes in gene expression patterns that may be associated with carcinogenesis. In prostate cancer, there is thus a strong activity of DNA methyltransferases (DNMT1) and a deregulation of histone modifiers, especially histone methyl transferase EZH2, associated with the repression of the transcription of certain genes (Gillio- Tos et al., 2012, Gravina et al., 2013, Koh et al., 2011, Yu et al., 2007). In addition to these epigenetic factors, environmental factors and more particularly nutrition is involved in the processes of carcinogenesis. The work of the laboratory has long focused on nutrition and cancer development. Indeed, studies on the preventive effects of soy phytoestrogens in prostate carcinogenesis have been widely discussed (Adjakly et al., 2011). The soy phytoestrogens showed their demethylating effect on the promoters of tumor suppressor genes. The hypothesis of this work was established because of the low incidence of prostate cancer found in Asian countries where high consumption of soybeans was noticed. The micronutrients contained in soy thus have the ability to modulate epigenetic modifications in prostate cancer (Vardi et al., 2010). In this thesis, we focused on histone modifications including histone methylation, which remains poorly studied in prostate cancer. First, we investigated the role of trimethylation of histone H3 lysine 27 (H3K27me3) in the involvement and progression of prostate cancer. The second part consisted of the identification of new prognostic and epigenetic markers related to histone methylation in order to establish an epigenetic profile of prostate tumors. We also discussed the effect of soy phytoestrogens on the modulation of epigenetic changes in prostate cancer and their protective roles in tumor development.

Modern healthcare research demands collaboration across disciplines to build preventive measures and innovate predictive capabilities for curing diseases. Along with the emergence of cutting-edge computational and statistical methodologies, data generation and analysis has become cheaper in the last ten years. However, the complexity of big data due to its variety, volume, and velocity creates new challenges for biologists, physicians, bioinformaticians, statisticians, and computer scientists. Combining data from complex multiple profiles is useful to better understand cellular functions and pathways that regulates cell function to provide insights that could not have been obtained using the individual profiles alone. However, current normalization and artifact correction methods are platform and data type specific, and may require both the training and test sets for any application (e.g. biomarker development). This often leads to over-fitting and reduces the reproducibility of genomic findings across studies. In addition, many bias correction and integration approaches require renormalization or reanalysis if additional samples are later introduced. The motivation behind this research was to develop and evaluate strategies for addressing data integration issues across data types and profiling platforms, which should improve healthcare-informatics research and its application in personalized medicine. We have demonstrated a comprehensive and coordinated framework for data standardization across tissue types and profiling platforms. This allows easy integration of data from multiple data generating consortiums. The main goal of this research was to identify regions of genetic-epigenetic co-ordination that are independent of tissue type and consistent across epigenomics profiling data platforms. We developed multi-‘omic’ therapeutic biomarkers for epigenetic drug efficacy by combining our biomarker regions with drug perturbation data generated in our previous studies. We used an adaptive Bayesian factor analysis approach to develop biomarkers for multiple HDACs simultaneously, allowing for predictions of comparative efficacy between the drugs. We showed that this approach leads to different predictions across breast cancer subtypes compared to profiling the drugs separately. We extended this approach on patient samples from multiple public data resources containing epigenetic profiling data from cancer and normal tissues (The Cancer Genome Atlas, TCGA; NIH Roadmap epigenomics data).

Roughly ~50% of the human genome, contains noncoding sequences serving as regulatory elements responsible for the diverse gene expression of the cells in the body. One very well studied category of regulatory elements is the category of enhancers. Enhancers increase the transcriptional output in cells through chromatin remodeling or recruitment of complexes of binding proteins. Identification of enhancer using computational techniques is an interesting area of research and up to now several approaches have been proposed. However, the current state-of-the-art methods face limitations since the function of enhancers is clarified, but their mechanism of function is not well understood. This PhD thesis presents a bioinformatics/computer science study that focuses on the problem of identifying enhancers in different human cells using computational techniques. The dissertation is decomposed into four main tasks that we present in different chapters. First, since many of the enhancer’s functions are not well understood, we study the basic biological models by which enhancers trigger transcriptional functions and we survey comprehensively over 30 bioinformatics approaches for identifying enhancers. Next, we elaborate more on the availability of enhancer data as produced by different enhancer identification methods and experimental procedures. In particular, we analyze advantages and disadvantages of existing solutions and we report obstacles that require further consideration. To mitigate these problems we developed the Database of Integrated Human Enhancers (DENdb), a centralized online repository that archives enhancer data from 16 ENCODE cell-lines. The integrated enhancer data are also combined with many other experimental data that can be used to interpret the enhancers content and generate a novel enhancer annotation that complements the existing integrative annotation proposed by the ENCODE consortium. Next, we propose the first deep-learning computational framework for identifying enhancers. The proposed system called Dragon Ensemble Enhancer Predictor (DEEP) is based on the novel deep learning two-layer ensemble algorithm capable of identifying enhancers characterized by different cellular conditions. Experimental results using data from ENCODE and FANTOM5, demonstrate that DEEP surpasses in terms of recognition performance the major systems for enhancer prediction and shows very good generalization capabilities in unknown cell-lines and tissues. Finally, we take a step further by developing a novel feature selection method suitable for defining a computational framework capable of analyzing the genomic content of enhancers and reporting cell-line specific predictive signatures.