Resolving chromatin interaction and transcriptional networks in mammalian nuclei
by Yajie Zhu
Date of Examination:2024-06-20
Date of issue:2024-08-15
Advisor:Prof. Dr. Argyris Papantonis
Referee:Prof. Dr. Argyris Papantonis
Referee:Dr. Johannes Soeding
Persistent Address:
http://dx.doi.org/10.53846/goediss-10618
Files in this item
Name:PhD_thesis_revised_YajieZhu.pdf
Size:7.30Mb
Format:PDF
The following license files are associated with this item:
Abstract
English
The nucleus is a highly organized machine. Chromatin organization, transcription, and RNA processing are working harmoniously in a specialized but communicative way to pass the information from DNA to mature RNA. In the nucleus, chromatin is organized in a hierarchical order to serve gene expression. Chromatin organization can be captured in an interaction map by Hi-C-based technology. The interaction map reveals structural and functional building blocks of the genome including A/B compartments, topologically associating domains (TADs), and chromatin loops. Chromatin loops are point-to-point genome interactions between genome loci. Loops can be divided into different types, including CTCF loops, active loops, and repressive loops, mediated by different proteins and histone modifications. Currently, the annotation of loop types relies on peak calling tools and manual assignment. With deep learning, we can classify loops in a more accurate and efficient way. While chromatin organization regulates the timing and spacing of the birth of transcripts, RNA processing facilitates the maturation of the transcripts. Alternative splicing plays an important role in RNA processing. It expands the transcriptome diversity by generating different RNA isoforms from one single gene. Accurate RNA splicing depends on mainly three cis-elements in pre-RNA: the 5' splice site (the donor site), the 3' splice site (the acceptor site), and the branch point. Recognition of those elements is mainly carried out by the spliceosome machine. One core player in the major splicing pathway is U1 snRNP, which recognizes the 5' splice site. It also plays a role in transcription initiation and mRNA 3ʹ-end processing. There exists a group of U1 snRNA variants (vU1), which are highly expressed at the early stages of development. They may recognize non-canonical 5' splice sites in a special splicing process termed recursive splicing. In the nucleus, the splicing factors are enriched in particular condensates, such as nuclear speckles, but the exact function of nuclear speckles remains unclear. Here, I studied chromatin organization, transcription, and RNA processing in three biological models: stem cells, senescent cells, and RNAPII-degraded cells. In Chapter 1, I explored the functions of human variant U1 snRNA (vU1) in stem cells. By analyzing the transcriptome of vU1-knockout IPS cells, I found the role of two vU1 snRNAs in regulating cell cycle progression. The knockout globally shifted alternative splicing and influenced 3’UTR processing of genes that regulate the cell cycle. Depending on which mutations they contain, the two vU1s showed preference to binding different donor site sequences in recursive splicing. In Chapter 2, I investigated the influence of SICCs (senescence-induced CTCF clusters) and nuclear speckles on RNA splicing in senescent cells. I found that the HMGB2 knockdown, which induced SICCs, showed a strong positive correlation in the splicing changes with ICM (a drug)-induced senescence, and that the SRRM2 knockdown, which disrupted SICCs, showed a strong negative correlation with that senescence. In Chapter 3, I developed LoopBin, a deep learning model to classify loop types, and applied it to the data of RNAPII-degraded cells. The model classified loops into six clusters with distinct biological meanings and captured the difference between the control and the RNAPII degradation group. Finally, I discussed the functions and the potential applications of vU1, especially its role in regulating recursive splicing, the difficulties and solutions in analyzing recursive splicing, the potential ways to rejuvenate senescent cells via SICCs and splicing, and the considerations in model selection.
Keywords: genome organization; deep learning; variant U1 snRNA; stem cell; RNA splicing; CTCF; replicative senescence
