This thesis focuses on searching for CpG islands of DNA sequences based on analysis of DNA spectrograms. The first part is theoretical and deals with the significance CpG island, and a description of the algorithms that are used or have been proposed for their search. The theoretical basis were implemented two algorithms based on the analysis of DNA spectrogram. One is based on the assumption that the region CpG islands has a higher content of guanine and cytosine than the region outside the CpG island and the other on the assumption of a higher frequency of occurrence of CG dinucleotides in the CpG island. The algorithms are implemented through MATLAB programming interface. For evaluation usefulness and effectiveness of solutions, results achieved on the selected DNA sequences implemented algorithms are compared with the results achieved by search engines CpG islands, which are freely available on the internet.

In higher eukaryotes, the DNA base Cytosine can exist in a variety of modified forms when in the dinucleotide CpG. Although a methylated form tends to dominate within the genome, approximately 1% of all CpG dinucleotides are found unmodified at high densities spanning around 1Kb and tend to co-localise to the 5’ ends of around 60% of annotated gene promoters. These unique DNA sequences are known as CpG islands (CGIs) and their role within the genome to date is largely unknown. Methylation of CGIs in cancers however has been linked to silencing of associated genes implying a role in gene regulation. Furthermore these sites are also interesting as they remain specifically nonmodified within a genome rich in methylated CpG. We set out to better understand the roles for CGIs through the characterisation of any specific CGI binding proteins. Digestion of nuclei with methyl sensitive restriction enzymes facilitates the purification of CGI fragments. Subsequent immunohistochemistry on the CGI chromatin fragments along with ChIP-PCR over several CGIs revealed an enrichment of the “active” histone modifications including H3K4me3, a depletion of the “silencing” marks such as H3K27me3, as well as a group of CGI specific binding factors. These latter proteins contained a domain previously shown to bind to non-methylated CpG dinucleotides (the CXXC domain) and as such were ideal candidates for CGI specific factors, in particular a protein called Cfp1. Genome wide sequencing revealed a striking correlation between Cfp1 and H3K4me3 which were both seen at around 80% of islands. Furthermore, the presence of Cfp1/H3K4me3 at islands tended to have a negative correlation with the presence of chromatin rich in the silencing histone modification H3K27me3. Closer investigation of the Cfp1 protein reveals it to be a true non-methyl CGI binding factor in vivo and shRNA reduction of Cfp1 levels to around 10% of wild type resulted in a precipitous drop in H3K4me3 levels over CGIs without a dramatic reduction in global H3K4me3 levels. As Cfp1 has been shown to be part of the Set1 histone H3K4 methyltransferase complex responsible for this modification, this CXXC protein may be attracting this histone modifying complex and as such represents a method whereby the underlying DNA sequence (CpG) can drive the overlying epigenetic state. This study may go some way to understanding the functional significance of CGIs within the genome.

Abstract As the defects in DNA repair and cell cycle control are known to promote tumorigenesis, a proportion of inherited breast cancers might be attributable to mutations in the genes involved in these functions. In the present study, three such genes, TP53, CHK2 and ATM, which are also associated with known cancer syndromes, were screened for germline mutations in Finnish breast cancer patients. In combination with our previous results, three TP53 germline mutations, Tyr220Cys, Asn235Ser and Arg248Gln, were detected in 2.6% (3/108) of the breast cancer families. The only observed CHK2 alteration with a putative effect on cancer susceptibility, Ile157Thr, segregated ambiguously with the disease, and was also present in cancer-free controls. The available functional data, however, suggests that the altered CHK2 in some way promote tumorigenesis. Furthermore, compared to the other studied populations, Ile157Thr seems to be markedly enriched in Finland. Thus, the clinical significance of Ile157Thr requires further investigation among Finnish cancer patients. ATM germline mutations appear to contribute to a small proportion of the hereditary breast cancer risk, as two distinct ATM mutations, Ala2524Pro and 6903insA, were found among three families (1.9%, 3/162) displaying breast cancer. They all originated from the same geographical region as the AT families with the corresponding mutations, possibly referring to a founder effect concerning the distribution of these mutations in the Finnish population. The genes important for tumorigenesis in sporadic disease might also contribute to familial breast cancer. Therefore, four putative LOH targets genes in chromosome 11q21-q24 were screened for intragenic mutations, and five were analyzed for epigenetic inactivation in sporadic breast tumors. The lack of somatic intragenic mutations in MRE11A, PPP2R1B, CHK1 and TSLC1 led us next to investigate promoter region hypermethylation as a mechanism capable of silencing these genes, as well as the ATM gene. Only TSLC1 demonstrated involvement of CpG island methylation, which was especially prominent in three tumors. This suggests that together with LOH, methylation could result in biallelic inactivation of the TSLC1 gene in breast cancer.

This thesis focuses on searching for CpG islands of DNA sequences based on analysis of DNA spectrograms. The first part is theoretical and deals with the significance CpG island, and a description of the algorithms that are used or have been proposed for their search. The theoretical basis were implemented two algorithms based on the analysis of DNA spectrogram. One is based on the assumption that the region CpG islands has a higher content of guanine and cytosine than the region outside the CpG island and the other on the assumption of a higher frequency of occurrence of CG dinucleotides in the CpG island. The algorithms are implemented through MATLAB programming interface. For evaluation usefulness and effectiveness of solutions, results achieved on the selected DNA sequences implemented algorithms are compared with the results achieved by search engines CpG islands, which are freely available on the internet.

http://www.impactjournals.com/oncotarget/index.php?journal=oncotarget&page=article&op=view&path[]=8620&pubmed-linkout=1 / 京都大学 / 0048 / 新制・課程博士 / 博士(医学) / 甲第19927号 / 医博第4147号 / 新制||医||1017(附属図書館) / 33013 / 京都大学大学院医学研究科医学専攻 / (主査)教授 山田 泰広, 教授 妹尾 浩, 教授 武田 俊一 / 学位規則第4条第1項該当 / Doctor of Medical Science / Kyoto University / DFAM

With the advent of high-throughput genomics, biological big data brings challenges to scientists in handling, analyzing, processing and mining this massive data. In this new interdisciplinary field, diverse theories, methods, tools and knowledge are utilized to solve a wide variety of problems. As an exploration, this dissertation project is designed to combine concepts and principles in multiple areas, including signal processing, information-coding theory, artificial intelligence and cloud computing, in order to solve the following problems in computational biology: (1) comparative gene structure detection, (2) DNA sequence annotation, (3) investigation of CpG islands (CGIs) for epigenetic studies. Briefly, in problem #1, sequences are transformed into signal series or binary codes. Similar to the speech/voice recognition, similarity is calculated between two signal series and subsequently signals are stitched/matched into a temporal sequence. In the nature of binary operation, all calculations/steps can be performed in an efficient and accurate way. Improving performance in terms of accuracy and specificity is the key for a comparative method. In problem #2, DNA sequences are encoded and transformed into numeric representations for deep learning methods. Encoding schemes greatly influence the performance of deep learning algorithms. Finding the best encoding scheme for a particular application of deep learning is significant. Three applications (detection of protein-coding splicing sites, detection of lincRNA splicing sites and improvement of comparative gene structure identification) are used to show the computing power of deep neural networks. In problem #3, CpG sites are assigned certain energy and a Gaussian filter is applied to detection of CpG islands. By using the CpG box and Markov model, we investigate the properties of CGIs and redefine the CGIs using the emerging epigenetic data. In summary, these three problems and their solutions are not isolated; they are linked to modern techniques in such diverse areas as signal processing, information-coding theory, artificial intelligence and cloud computing. These novel methods are expected to improve the efficiency and accuracy of computational tools and bridge the gap between biology and scientific computing.

A principle challenge of modern biology is to understand how the human genome is organised and regulated within a nucleus. The field of chromatin biology has made significant progress in characterising how protein and DNA modifications reflect transcription and replication state. Recently our lab has shown that the human genome is organised into large domains of altered DNA helical twist, called DNA supercoiling domains, similar to the regulatory domains observed in prokaryotes. In my PhD I have analysed how the maintenance and distribution of DNA supercoiling relates to biological function in human cells. DNA supercoiling domains are set up and maintained by the balanced activity of RNA transcription and topoisomerase enzymes. RNA polymerase twists the DNA, over-winding in front of the polymerase and under-winding behind. In contrast topoisomerases relieve supercoiling from the genome by introducing transient nicks (topoisomerase I) or double strand breaks (topoisomerase II) into the double helix. Topoisomerase activity is critical for cell viability, but the distribution of topoisomerase I, IIα and IIβ in the human genome is not known. Using a chromatin immunoprecipitation (ChIP) approach I have shown that topoisomerases are enriched in large chromosomal domains, with distinct topoisomerase I and topoisomerase II domains. Topoisomerase I is correlated with RNA polymerase II, genes and underwound DNA, whereas topoisomerase IIα and IIβ are associated with each other and over-wound DNA. This indicates that different topoisomerase proteins operate in distinct regions of the genome and can be independently regulated depending on the genomic environment. Transcriptional regulation by DNA supercoiling is believed to occur through changes in gene promoter structure. To investigate DNA supercoiling my lab has developed biotinylated trimethylpsoralen (bTMP) as a DNA structure probe, which preferentially intercalates into under-wound DNA. Using bTMP in conjunction with microarrays my lab identified a transcription and topoisomerase dependent peak of under-wound DNA in a meta-analysis of several hundred genes (Naughton et al. (2013)). In a similar analysis, Kouzine et al. (2013) identified an under-wound promoter structure and proposed a model of topoisomerase distribution for the regulation of promoter DNA supercoiling. To better understand the role of supercoiling and topoisomerases at gene promoters, a much larger-scale analysis of these factors was required. I have analysed the distribution of bTMP at promoters genome wide, confirming a transcription and expression dependent distribution of DNA supercoils. DNA supercoiling is distinct at CpG island and non-CpG island promoters, and I present a model in which over-wound DNA limits transcription from both CpG island promoters and repressed genes. In addition, I have mapped by ChIP topoisomerase I and IIβ at gene promoters on chromosome 11 and identified a different distribution to that proposed by Kouzine et al. (2013), with topoisomerase I maintaining DNA supercoiling at highly expressed genes. This study provides the first comprehensive analysis of DNA supercoiling at promoters and identifies the relationship between supercoiling, topoisomerase distribution and gene expression. In addition to regulating transcription, DNA supercoiling and topoisomerases are important for genome stability. Several studies have suggested a link between DNA supercoiling and instability at common fragile sites (CFSs), which are normal structures in the genome that frequently break under replication stress and cancer. bTMP was used to measure DNA supercoiling across FRA3B and FRA16D CFSs, identifying a transition to a more over-wound DNA structure under conditions that induce chromosome fragility at these regions. Furthermore, topoisomerase I, IIα and IIβ showed a pronounced depletion in the vicinity of the FRA3B and FRA16D CFSs. This provides the first experimental evidence of a role for DNA supercoiling in fragile site formation.

Mammalian genomes are characterised by unique regions of non-methylated DNA known as CpG islands (CGIs). These genomic elements are characterised by a high density of CpGs and an elevated GC content compared to the surrounding, bulk of the genome. CGIs are prevalently associated with the 5’ end of genes and represent key nucleation sites where specific transcription factors and chromatin modifiers are recruited to impact on gene function. This thesis is focused at understanding the biochemical properties of the recently discovered H3K36-specific histone demethylase, KDM2A. This enzyme is specifically recruited to CGIs but how it interfaces with local chromatin in vivo remains unknown. Using defined chromatin templates in vitro, this study demonstrates that KDM2A binding to DNA relies on a zinc finger CXXC domain that preferentially recognizes non-methylated CpGs. In particular, nucleosomes represent a major barrier to KDM2A binding and chromatin substrates are interpreted by the CXXC domain through specific interaction with CpGs within linker DNAs. Moreover, the adjacent PHD domain does not contribute to KDM2A binding to chromatin. Together these observations suggest that sequence, methylation status and accessibility of DNA define how CGI chromatin is interpreted by CXXC domain proteins. In particular, the precise targeting of KDM2A to CGIs contributes to the creation of a unique chromatin architecture that highlights gene regulatory regions within large and complex mammalian genomes.

Understanding the mechanism of recombination has important implications for genome evolution and genomic variability. The work presented in this thesis studies the properties of recombination by investigating the effects it has on genome evolution in humans and dogs. Using alignments of human genes with chimpanzee and macaque orthologues we studied substitution patterns along the human lineage and scanned for evidence of positive selection. The properties mirror the situation in human non-coding sequences with the fixation bias ‘GC-biased gene conversion’ (gBGC) as a driving force in the most rapidly evolving regions. By assigning candidate genes to distinct classes of evolutionary forces we quantified the extent of those genes affected by gBGC to 20%. This suggests that human-specific characters can be prompted by the fixation bias of gBGC, which can be mistaken for selection. The gene PRDM9 controls recombination in most mammals, but is lacking in dogs. Using whole-genome alignments of dog with related species we examined the effects of PRDM9 inactivation. Additionally, we analyzed genomic variation in the genomes of several dog breeds. We identified that non-allelic homologous recombination (NAHR) via sequence identity, often GC-rich, creates structural variants of genomic regions. We show that these regions, which are also found in dog recombination hotspots, are a subset of unmethylated CpG-islands (CGIs). We inferred that CGIs have experienced a drastic increase in biased substitution rates, concurrent with a shift of recombination to target these regions. This enables recurrent episodes of gBGC to shape their distribution. The work presented in this thesis demonstrates the importance of meiotic recombination on patterns of molecular evolution and genomic variability in humans and dogs. Bioinformatic analyses identified mechanisms that regulate genome composition. gBGC is presented as an alternative to positive selection and is revealed as a major factor affecting allele configuration and the emergence of accelerated evolution on the human lineage. Characterization of recombination-induced sequence patterns highlights the potential of non-methylation and establishes unmethylated CGIs as targets of meiotic recombination in dogs. These observations describe recombination as an interesting process in genome evolution and provide further insights into the mechanisms of genomic variability.

No description available.