<p>Understanding the mechanisms whereby the sequence of the human genome is interpreted into diverse cellular phenotypes is a critical endeavor in modern biology. A major determinant of cellular phenotype is the spatial and temporal pattern gene expression, which is regulated in part by epigenomic properties such as histone post-translational modifications, DNA methylation, chromatin accessibility and the 3-dimensional architecture of the genome within the nucleus. These properties regulate the dynamic assembly of transcription factors and their co-regulatory proteins upon chromatin. To properly understand the interplay between the epigenomic framework of a cell and transcription factors, integrated analysis of transcription factor-DNA binding, chromatin status, and transcription is required. This work integrates information about chromatin accessibility, as measured by DNaseI hypersensitivity, transcription factor binding, as measured by chromatin immunoprecipitation, and transcription, as measured by microarray or transcriptome sequencing, to further understand the functional role of two important transcription factors, the androgen receptor (AR) and CTCF, in cancer cell line models. Data gathered from a prostate cancer cell line model demonstrate that the AR does not exclusively bind accessible chromatin upon ligand-activation, and induces significant changes in chromatin accessibility upon binding. Regions of quantitative change in chromatin accessibility contain motifs corresponding to potential collaborators for AR function, and are also significantly associated with AR-regulated transcriptional changes. Furthermore, base pair resolution of the DNaseI cleavage profile revealed three distinct patterns of AR-DNA interaction, suggesting multiple modes of AR interacting with the genome. A novel role for the nuclear receptor REV-ERB&#945; in AR-mediated transcription was explored within the same model system. Though preliminary, results thus far indicate that REV-ERB&#945; is required for AR-induced increases in target gene transcription in a manner that is likely dependent on HDAC3. Genetic knockdown of REV-ERB&#945; resulted in notable changes in chromatin accessibility around AR-target genes both before and after AR activation. The function of CTCF was interrogated using stable knockdown in a breast cancer cell line model. CTCF knockdown led to widespread changes in chromatin accessibility that were dependent on DNA sequence. Further analysis suggested that AP-1 and FOXA1 are involved in CTCF function. Together, the work presented in this dissertation offers novel insight into the behavior of two critical transcription factors in cancer cell lines, and describe a framework of analysis that can be extended and applied to any transcription factor within any desired cellular context.</p> / Dissertation

In eukaryotic cells, transcription factors and polymerases must access DNA in the context of nucleosomes and chromatin. The accessibility of DNA sequences to such trans-acting factors is an important feature of gene regulatory elements, including promoters. In vertebrates, the majority of gene promoters coincide with CpG islands (CGIs), which remain free from DNA methylation and exhibit elevated CpG densities. This hypomethylated and CpG-rich state at CGI promoters is associated not only with transcriptional activity, but also with high levels of chromatin accessibility. However, the causes and consequences of such chromatin accessibility remain unclear. To address this, I have profiled chromatin accessibility in mouse embryonic stem cells (ESCs). In addition to confirming that CGI accessibility is independent of transcriptional activity, I was able to demonstrate that the loss of DNA methylation in ESCs resulted in increased chromatin accessibility at a subset of CpG-rich repetitive elements, suggesting that non-methylated CpG-rich sequences may, at least partially, facilitate open chromatin states. This was supported by preliminary work targeting bacterial CpG-rich sequences into the mouse genome, where they were sufficient to establish novel regions of chromatin accessibility. To examine potential mechanisms by which hypomethylated DNA could serve to promote chromatin accessibility, I profiled chromatin accessibility in mouse ESCs lacking various chromatin-modifying proteins which are normally enriched at CGIs, with the histone demethylases KDM2A/B linked to maintaining open chromatin at CGIs. As an alternative approach to understanding the causes of chromatin accessibility in mouse ESCs, I examined the mechanism by which the pioneer transcription factor OCT4 is able to access previously inaccessible chromatin, and reveal that it requires the chromatin remodeller BRG1 to remodel chromatin and facilitate transcription factor binding at distal regulatory elements. Ultimately, this work provides an insight into some of the molecular determinants of chromatin accessibility in mouse ESCs, although many of the consequences of such chromatin states remain unclear.

DNA methylation is an epigenetic mark that is deposited throughout the genome of mammals and plays an important role in the maintenance of transcriptionally repressive states across cell divisions. There are two major mechanisms by which DNA methylation has been proposed to act: one involves the recognition of the mark by protein complexes containing histone deacetylases (HDACs) that can remodel the local chromatin. Alternatively, methylation has been suggested to directly affect the interaction between transcription factors and their cognate binding sequence. The aim of this research was to determine the contributions of these two mechanisms in cells. The importance of HDAC activity in mediating DNA methylation-dependent transcriptional repression was assessed by comparing the genes and retrotransposons that are upregulated in response to DNA methylation loss or the disruption of HDAC activity. To this purpose, we performed whole-genome transcriptional analysis in wild type and DNA methylation-deficient mouse embryonic stem cells (DNMT.TKO mESCs) in the presence and absence of the HDAC inhibitor trichostatin A. Our data suggests that there are few genes whose repression is solely dependent on the recruitment of HDACs by DNA methylation in mESCs. Rather it appears that DNA methylation and HDAC-mediated silencing represent two independent layers of repression that converge at certain transcriptional elements. To investigate the contribution of DNA methylation on the genome-wide occupancy of transcription factors, we compared the global chromatin accessibility landscape and the binding profile of candidate transcription factors in the absence or presence of DNA methylation. We found that loss of DNA methylation associates with localised gains in accessibility, some of which can be linked to the novel binding of transcription factors such as GABPA, MAX, NRF1 and YY1. Altogether, our results present new insights into the interplay between DNA methylation and histone deacetylation and their impact on the localisation of transcription factors from different families.

京都大学 / 0048 / 新制・課程博士 / 博士(医学) / 甲第22829号 / 医博第4668号 / 新制||医||1047(附属図書館) / 京都大学大学院医学研究科医学専攻 / (主査)教授 遊佐 宏介, 教授 斎藤 通紀, 教授 高橋 淳 / 学位規則第4条第1項該当 / Doctor of Medical Science / Kyoto University / DFAM

京都大学 / 0048 / 新制・課程博士 / 博士(医科学) / 甲第19405号 / 医科博第65号 / 新制||医科||5(附属図書館) / 32430 / 京都大学大学院医学研究科医科学専攻 / (主査)教授 河本 宏, 教授 斎藤 通紀, 教授 竹内 理 / 学位規則第4条第1項該当 / Doctor of Medical Science / Kyoto University / DGAM

京都大学 / 0048 / 新制・課程博士 / 博士(医学) / 甲第20788号 / 医博第4288号 / 新制||医||1025(附属図書館) / 京都大学大学院医学研究科医学専攻 / (主査)教授 髙橋 良輔, 教授 高橋 淳, 教授 山下 潤 / 学位規則第4条第1項該当 / Doctor of Medical Science / Kyoto University / DFAM

No description available.

<p>The prediction of chromatin accessibility for cancer and neuron cell lines using genomic sequences is quite challenging. Advances in machine learning and deep learning techniques allow such challenges to be addressed. This thesis investigates the use of both the transfer learning and the multi-task learning techniques. In particular, this research demonstrates the potential of transfer learning and multi-task learning in improving the prediction accu?racy for twenty-three cancer types in human and neuron cell lines. Three different network architectures are used: the Basset network, the network, and the DeepSEA network. In addition, two transfer learning techniques are also used. In the first technique data relevant to the desired prediction task is not used during the pre-training stage while the second technique includes limited data about the desired prediction task in the pre-training phase. The preferred performance evaluation metric used to evaluate the performance of the models was the AUPRC due to the numerous negative samples. Our results demonstrate an average improvement of 4% of the DeepSEA network in predicting all twenty-three cancer cell line types when using the first technique, a decrease of 0.42% when using the second technique, and an increase of 0.40% when using multi-task learning. Also, it had an average improvement of 3.09% when using the first technique, 1.16% when using the second technique and 4.60% for the multi-task learning when predicting chromatin accessibility for the 14 neuron cell line types. The DanQ network had an average improvement of 1.18% using the first transfer learning technique, the second transfer learning technique showed an average decrease of 1.93% and also, a decrease of 0.90% for the multi-task learning technique when predicting for the different cancer cell line types. When predicting for the different neuron cell line types the DanQ had an average improvement of 1.56% using the first technique, 3.21% when using the second technique, and 5.35% for the multi-task learning techniques. The Basset network showed an average improvement of 2.93% using the first transfer learning technique and an average decrease of 0.02%, and 0.63% when using the second technique and multi-task learning technique respectively. Using the Basset network for prediction of chromatin accessibility in the different neuron types showed an average increase of 2.47%, 9 3.80% and 5.50% for the first transfer learning technique, second transfer learning technique and the multi-task learning technique respectively. The results show that the best technique for the cancer cell lines prediction is the first transfer learning model as it showed an improvement for all three network types, while the best technique for predicting chromatin accessibility in the neuron cell lines is the multi-task learning technique which showed the highest average improvement among all networks. The DeepSEA network showed the greatest improvement in performance among all techniques when predicting the different cancer cell line types. Also, it showed the greatest improvement when using the first transfer learning technique for predicting chromatin accessibility for neuron cell lines in the brain. The basset network showed the greatest improvement for the multi-task learning technique and the second transfer learning technique when predicting the accessibility for neuron cell lines. </p>

Les pucerons sont des hémiptères ravageurs des cultures agronomiques particulièrement adaptés à leur environnement. Acyrthosiphon pisum (le puceron du pois) présente un cycle de vie basé sur l’alternance d’une reproduction sexuée ou asexuée en réponse à la photopériode. Ils présentent ainsi un polyphénisme de reproduction aboutissant à la formation de trois phénotypes distincts : femelles asexuées, femelles sexuées, et mâles. Ces derniers étant obtenus par élimination d’un chromosome X, A. pisum est une espèce hétérogamétique mâle présentant un système chromosomique XX chez les femelles et X0 chez les mâles. Le déséquilibre du nombre de chromosome X entre mâles et femelles engendré par cette hétérogamétie nécessite chez certains organismes d’être corrigé par des mécanismes de compensation de dose. Les polyphénismes et compensation de dose impliquent chez d’autres organismes des régulations transcriptionnelles notamment régulées par l’accessibilité de la chromatine.Ma thèse vise ainsi à étudier le polyphénisme de reproduction et la compensation de dose des pucerons sous l’angle d’analyses bio-informatiques de données d’expression des gènes (RNA-seq) et d’accessibilité de la chromatine (FAIRE-seq) dans le but de caractériser l’impact des mécanismes épigénétiques dans ces deux processus biologiques fondamentaux du cycle de vie des pucerons. Les résultats développés dans ma thèse ont permis de montrer d’une part la présence d’une compensation de dose chez le puceron du pois au niveau transcriptomique, supportée par une accessibilité accrue de la chromatine de l’unique X des / Aphids are hemipterous crops pests that are particularly adapted to their environment. Acyrthosiphon pisum (pea aphid) displays a life cycle based on the alternation of sexual or asexual reproduction in response to photoperiod. They thus exhibit a reproductive polyphenism resulting in the formation of three distinct phenotypes: asexual females, sexual females, and males. The latter being obtained by elimination of an X chromosome, A. pisum is a male heterogametic species with a XX chromosomal system in females and X0 in males. The X chromosome number between males and females caused by this heterogamy requires in some organisms to be corrected by dosage compensation mechanisms. Polyphenisms and dosage compensation both involve in other organisms transcriptional regulations that are notably regulated by the chromatin accessibility regulations. My thesis aims to study the reproductive polyphenism and dosage compensation in aphids in the context of bioinformatic analyzes of gene expressioThe results developed in my thesis have shown, on one hand, the presence of dose compensation in pea aphid at the transcriptomic level, which is supported by increased chromatin accessibility of the males’ single X in somatic cells. On the other hand, specific sites of chromatin opening between sexual and asexual embryos seem to participate in the definition of their reproduction mode by modulating the expression of certain genes and by allowing the fixation of transcription factors. Their analysis shows the involvement of ecdysone as a new hormonal pathway that may trigger sexual reproducti

Genetic associations have been discovered for many human complex traits, and yet for most associated loci the causal variants and molecular mechanisms remain unknown. Studies mapping quantitative trait loci (QTLs) for molecular phenotypes, such as gene expression, RNA splicing, and chromatin accessibility, provide rich data that can link variant effects in specific cell types with complex traits. These genetic effects can also now be modeled in vitro by differentiating human induced pluripotent stem cells (iPSCs) into specific cell types, including inaccessible cell types such as those of the brain. In this thesis, I explore a range of approaches for using QTLs to identify causal variants and to link these with molecular functions and complex traits. In Chapter 2, I describe QTL mapping in 123 sensory neuronal cell lines differentiated from human iPSCs. I observed that gene expression was highly variable across iPSC-derived neuronal cultures in specific gene categories, and that a portion of this variability was explained by commonly used iPSC culture conditions, which influenced differentiation efficiency. A number of QTLs overlapped with common disease associations; however, using simulations I showed that identifying causal regulatory variants with a recall-by- genotype approach in iPSC-derived neurons is likely to require large sample sizes, even for variants with moderately large effect sizes. In Chapter 3, I developed a computational model that uses publicly available gene expression QTL data, along with molecular annotations, to generate cell type-specific probability of regulatory function (PRF) scores for each variant. I found that predictive power was improved when the model was modified to use the quantitative value of annotations. PRF scores outperformed other genome-wide scores, including CADD and GWAVA, in identifying likely causal eQTL variants. In Chapter 4, I used PRF scores to identify relevant cell types and to fine map potential causal variants using summary association statistics in six complex traits. By examining individual loci in detail, I showed how the enrichments contributing to a high PRF score are transparent, which can help to distinguish plausible causal variant predictions from model misspecification.