PhD Defence: Lucas van Duin

Title: Deciphering regulatory mechanisms underlying gene co-activity

Supervisor: Assoc. Prof. Robin Andersson

Committee:
Prof. Albin Sandelin (chair), University of Copenhagen
Assoc. Prof. Pelin Sahlén, KTH Royal Institute of Technology
Assoc. Prof. Kedar Netarajan, Technical University of Denmark

Abstract:
A tight regulation of gene expression in time and space is crucial for organismal development and homeostasis, and disruptions of this regulation is a major cause of disease. It is becoming increasingly clear that mechanisms of chromatin organization and compaction play a crucial role in gene expression regulation. These mechanisms have the potential to affect multiple proximal genes at the same time. The extent to which these mechanisms affect gene expression, relative to the portion of gene expression affected by mechanisms independent of chromatin organization and compaction, is not entirely clear.

To elucidate the relationship between three-dimensional genome organization and transcription, we introduce the Transcriptional Decomposition method. This method separates the portion of gene expression attributable to the genomic context of a gene (Positionally Dependent; PD) from effects independent of positional effects (Positionally Independent; PI). Because the PD component reflects a shared activity between neighboring genes, it can also be termed co-activity.

We demonstrated that the TD components PD and PI were informative of respectively chromatin compartments and repressive histone marks, and effects associated with local gene activation. Furthermore, they could be used to infer TAD boundary locations and enhancer-promoter interactions. Applying TD to 76 human cell types revealed that the PD component and domain boundaries were shared across cell types, while the PI component and interactions were cell type specific. Finally, we showed that TD could aid in the interpretation of disease SNPs, identifying for example significant enrichments of Crohn’s disease associated SNPs in the various components for immune cell types.

Having established the TD method, we next sought to elucidate how co-activity changes from individual to individual, and what mechanisms underlie these changes. We adapted and applied the TD method to expression data from 343 individuals, and we identified domains of co-activity, which were largely invariant across individuals. These domains were reflective of active compartments, and their boundaries were enriched for TAD boundaries. Their co-regulation was reflected by an increased sharing of eQTLs between contained genes. We identified an additional set of domains that showed increased variability, contained within co-activity domains. These variable co-activity domains showed a larger concordance with histone PTM levels, more variable gene interactions, an increased and more widespread effect of genotypic variation, and an enrichment for binding of variably expressed transcription factors. Quantifying the relative effects of these mechanisms through careful construction of linear models revealed that TF expression is a larger determinant of co-activity in variable regions than interactions or cis genotype.