<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>

京都大学 / 新制・論文博士 / 博士(医学) / 乙第13517号 / 論医博第2267号 / 新制||医||1061(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 戸井 雅和, 教授 小川 誠司 / 学位規則第4条第2項該当 / Doctor of Medical Science / Kyoto University / DFAM

No description available.

During development, multipotent stem cells establish tissue-specific programmes of gene expression that underlie a process of differentiation into specialized cell types. It was shown in the study that changes in the nuclear architecture during terminal keratinocyte differentiation show correlation with the dynamics of the transcriptional and metabolic activity. In particular, terminal differentiation is accompanied by the decrease of nuclear volume, elongation of its shape, reduction of the number and fusion of nucleoli, increase in the number of centromeric clusters and a dramatic decrease of the transcriptional activity. Global changes in the nuclear architecture of epidermal keratinocytes are associated with marked remodelling of the higher-order chromatin structure of the epidermal differentiating complex (EDC). EDC is positioned peripherally in the epidermal nuclei at E11.5 when its genes show low expression levels and relocates towards the nuclear interior at E16.5 when EDC genes are markedly upregulated. P63 transcription factor serving as a master regulator of epidermal development is involved in the control of EDC relocation in epidermal progenitor cells. The epidermis of E16.5 p63KO exhibits significantly more peripheral positioning of the EDC loci, compared to wild-type. The genome organizer Satb1 serving as a direct p63 target controls higher order chromatin folding of the central part of EDC and Satb1 knockout mice show alterations of epidermal development and expression of the EDC encoded genes. Thus, this study shows that the programme of epidermal development and terminal differentiation is regulated by p63 and other factors and include marked remodelling of three-dimensional nuclear organization and positioning of tissue specific gene loci. In addition to the direct involvement of p63 in controlling the expression of tissue-specific genes, p63 via regulation of the chromatin remodelling factors such as Satb1 promotes establishing specific conformation of the EDC locus required for efficient expression of terminal differentiation-associated genes.

During tissues development multipotent progenitor cells establish tissue-specific gene expression programmes, leading to differentiation into specialized cell types. It has been previously shown that the transcription factor p63, a master regulator of skin development, controls the expression of adhesion molecules and essential cytoskeleton components. It has also been shown that p63 plays an important role in establishing distinct three-dimensional conformations in the Epidermal Differentiation Complex (EDC) locus (Fessing et al., 2011). Here we show that in p63-null mice about 32% of keratinocytes showed altered nuclear morphology. Alterations in the nuclear shape were accompanied by decreased expression of nuclear lamins (Lamin A/C and Lamin B1), proteins of the LINC complex (Sun-1, nesprin-2/3) and Plectin. Plectin links components of the nuclear envelope (nesprin-3) with cytoskeleton and ChIP-qPCR assay with adult epidermal keratinocytes showed p63 binding to the consensus binding sequences on Plectin 1c, Sun-1 and Nesprin-3 promoters. As a possible consequence of the altered expression of nuclear lamins and nuclear envelope-associated proteins, changes in heterochromatin distribution as well as decrease of the expression of several polycomb proteins (Ezh2, Ring1B, Cbx4) has been observed in p63-null keratinocytes. Moreover, recent data in our lab have showed that p63 directly regulates Cbx4, a component of the polycomb PRC1 complex. Here we show that mice lacking Cbx4 displayed a skin phenotype, which partially resembles the one observed in p63-null mice with reduced epidermal thickness and keratinocyte proliferation. All together these data demonstrate that p63-regulated gene expression program in epidermal keratinocytes includes not only genes encoding adhesion molecules, cytoskeleton proteins (cytokeratins) and chromatin remodelling factors (Satb1, Brg1), but also polycomb proteins and components of the nuclear envelope, suggesting the existence of a functional link between cytoskeleton, nuclear architecture and three dimensional nuclear organization. Other proteins important for proper epidermal development and stratification, are cytokeratins. Here, we show that keratin genes play an essential role in spatial organization of other lineage-specific genes in keratinocytes during epidermal development. In fact, ablation of keratin type II locus from chromosome 15 in epidermal keratinocytes led to changes in the genomic organization with increased distance between the Loricrin gene located on chromosome 3 as well as between Satb1 gene located on chromosome 17 and keratin type II locus, resulting in a more peripheral localization of these genes in the nucleus. As a possible consequence of their peripheral localization, reduced expression of Loricrin and Satb1 has also been observed in keratins type II-deficient mice. These findings together with recent circularized chromosome conformation capture (4C) data, strongly suggest that keratin 5, Loricrin and Satb1 are part of the same interactome, which is required for the proper expression of these genes and proper epidermal development and epidermal barrier formation. Taken together these data suggest that higher order chromatin remodelling and spatial organization of genes in the nucleus are important for the establishment of lineage-specific differentiation programs in epidermal progenitor cells. These data provide an important background for further analyses of nuclear architecture in the alterations of epidermal differentiation, seen in pathological conditions, such as psoriasis and epithelial skin cancers.

The epidermal differentiation complex (EDC) encodes co-ordinately regulated genes critically involved in epidermal differentiation, however knowledge of the molecular mechanisms involved in co-ordinating EDC gene expression is limited. Recent findings indicate p63 dependent changes in the nuclear localisation and higher-order chromatin folding the EDC coincide with the onset of epidermal stratification during embryonic development. Here it is demonstrated that a direct transcription target of p63, the chromatin-remodelling enzyme Brg1, modulates the specific nuclear positioning of the EDC and transcription of differentiation-specific gene encoded at the EDC. In addition, the results of high-resolution 5C-based analyses of the spatial chromatin interactome at a 5.3Mb region spanning the murine EDC in epidermal keratinocytes, and the silenced EDC in thymocytes, are presented. Chromatin interactions at the EDC region in keratinocytes include long-range interactions between multiple proximal and distal candidate gene regulatory regions. Many candidate regulatory elements involved in looping chromatin interactions at the EDC region are enriched for both active (H3K4me1, H3K27ac) and repressive (H3K27me3) chromatin marks and are bound by Sin3a and RBP2 co-repressor complexes. The chromatin interactome at the EDC in epidermal progenitor cells is enriched for bound chromatin architectural proteins Satb1, Satb2, and the cohesin subunit Rad21. Further, a substantial degree of co-localisation is observed between these chromatin architectural proteins, transcription factors and co-factors. Findings presented here suggest that a functional chromatin interactome, mediated by Satb proteins and cohesin, acts in conjunction with transcriptional repressor complexes to facilitate co-ordinated gene expression at the EDC in epidermal progenitor cells upon differentiation. These results provide a foundation for further study of the mechanisms controlling EDC gene expression in health and disease.

Melatonin is an indoleamine hormone with neuromodulatory and neuroprotective properties. It mediates many of its effects by its two G protein-coupled receptors, MT1 and MT2. We have shown that valproic acid (VPA) induces melatonin receptor expression in cultured rat C6 glioma cells, and in the rat hippocampus. VPA is known to affect gene expression through several mechanisms, including the modulation of intracellular kinase pathways and/or transcription factors, as well as the inhibition of histone deacetylase (HDAC) activity. In this study, we show that HDAC inhibitors of distinct chemical classifications, including suberanilohydroxamic acid (SAHA) and 4-(dimethylamino)-n-[7-(hydroxyamino)-7-oxoheptyl] benzamide (M344), parallel the effects of VPA on MT1 induction in vitro. However valpromide, a VPA analogue that lacks the ability to inhibit HDAC activity, does not. The observed increase in MT1 expression by VPA is matched by an increase in global histone H3 acetylation. More importantly, an enrichment of histone H3 acetylation occurs along the rat MT1 promoter following treatment with VPA, indicating that histone acetylation and chromatin remodelling are a primary mechanism underlying this induction. Independent of VPA, the rat MT1 gene may be regulated by a number of intracellular kinase pathways and transcription factors, which are also targeted by VPA. KG501-mediated CREB inhibition did not block MT1 upregulation by VPA. Blockade experiments targeting the PKC, PI3K/AKT, or GSK3β signaling pathways suggest that VPA induces melatonin receptor expression independent of these intracellular signaling cascades as well. The relevance of melatonin receptor upregulation was assessed using in vivo VPA and melatonin combination treatments on neuroprotective gene expression. The results of this study provide evidence that expression of the melatonin receptor is epigenetically induced by VPA by means of promoter histone acetylation. Melatonin receptor upregulation by VPA, or other HDAC inhibitors, may represent a therapeutic strategy for the management of several nervous system disorders. / Dissertation / Doctor of Philosophy (PhD)

Chromatin regulation is a major aspect of cancer development, progression, and treatment. Several small molecule inhibitors of chromatin regulators are currently used for treatment of certain hematological malignancies. However, there is still opportunity for many more patients to benefit from therapeutic approaches that target chromatin regulation, especially in the context of solid tumors. A critical unmet need is the identification of robust biomarkers that can guide the application of epigenetic inhibitors in a precise and personalized manner. In my dissertation, I aim to address this important knowledge gap by studying how perturbation of chromatin can target metabolic and lineage abnormalities in solid tumors for therapeutic benefit. To do this, I have focused on genetic and pharmacological perturbations of chromatin pathways in two cancer models: (1) lung adenocarcinoma (LUAD) with NRF2 activation and (2) neuroendocrine esophageal carcinoma (NEC). In the study on NRF2-active LUAD, we found that histone deacetylase (HDAC) inhibitors can be repurposed to reprogram the epigenomic and metabolic landscape, which leads to specific and potent anti-tumor effects in the context of NRF2 activation. Specifically, we employed a chromatin-focused genetic screen to identify dependencies on chromatin regulators. The screen revealed an NRF2-specific dependency on class I histone deacetylases. Experiments in mouse and human LUAD cell lines in vitro and in vivo indicated an NRF2-specific sensitivity to the class I HDAC inhibitor Romidepsin. Mechanistically, profiling of histone acetylation and gene expression upon Romidepsin treatment revealed a relative loss of histone H4 acetylation at promoters which was associated with reduced gene expression. Many downregulated genes were more essential for the survival of NRF2 hyperactive cancer cells, including genes involved in glutamine and serine metabolism, c-Myc and several of its targets involved in purine and pyrimidine synthesis. These transcriptional changes had corresponding effects on altering the metabolic pathways that NRF2-active cells selectively require for survival. In the study on neuroendocrine esophageal carcinoma (NEC), we identified a crucial role for epigenetic regulation of lineage fate through transcriptional control of the key epidermal transcription factor p63. This project originated from data from my collaborators that indicates a role for p63 in the suppression of basal-to-neuroendocrine identity transition in the developing esophagus. Consistently, I found that p63 is silenced in NEC through a non-genetic mechanism. Reintroducing p63 isoforms in a human NEC cell line showed that ΔNp63α was sufficient to restore squamous marker expression. An epigenetic drug screen assessing p63 gene expression and subsequent validation experiments revealed that inhibition of EZH2, a histone methyltransferase, induced expression of ΔNp63α and genes related to the squamous identity. Analysis of the chromatin state in the TP63 locus showed that EZH2 inhibition led to a loss histone H3 methylation and a gain of histone H3 acetylation and its reader BRD4. These results support the hypothesis that the squamous identity can be reactivated epigenetically in NEC through de-repression of ΔNp63α as a potential therapeutic strategy. Together, these studies contribute to our understanding of the transcriptional response to chromatin perturbation and show that this can be leveraged to modulate cell metabolism and identity, as well as to achieve therapeutic benefit in new contexts of cancer.

No description available.

No description available.