Abstract

Protein DNA binding is the core of transcriptional regulation, the process which controls the flow of information stored in an organism’s genome to react to its environment and to maintain its functionality. The initial event of gene expression is the binding of a transcription factor (TF) to its target site. These binding events are integrated over several binding sites and TFs by which a fine tuned regulation can be achieved. The number, combination and strengths of the different binding sites encode the desired gene expression level and the plasticity of the regulated gene. Efforts have been devoted with the goal of identifying the specific DNA sequences bound by different TFs. For more than two decades, it was thought that mutations at each position in this sequence independently contribute to the binding probability of a TF. This binding preference has therefore been described through position weight matrices (PWMs). PWMs describe the binding preference of a TF towards its target sites by assuming that each nucleotide position contributes independently to the total specificity (linearity assumption). However, current research has shown that this simplified view lacks a significant part of the information needed to precisely describe the binding preference of a TF. It was also shown that the most information missing in the PWM is encoded in dinucleotide mutations. Two questions are important in this regard: (1) Which information about TF-DNA interaction are we missing and are currently employed methods able to provide them? and (2) What is a comprehensive description of non-linearity that is based on biophysical properties rather then on abstract probabilities? One important aspect is the three dimensional configuration of the DNA strand (DNA shape) which is known to affect TF binding to a varying degree. Through recent work by the group of Remo Rohs it is possible to predict shape parameters (features) from a DNA sequence and investigate to which degree they influence binding for any given set of measurements. The first aim of this thesis is therefore to determine non-linearity in TF-DNA interaction and investigate the influence of DNA shape on them. Protein-DNA interactions were studied with a variety of methods using structural biology (NMR, crystallography, cryo EM) or quantitative Methods (EMSA, DNA binding arrays, ChIPSeq, B1H, SELEX, MITOMI, Simile-Seq). Most of these quantitative methods to measure TF-DNA interactions, however, are not very sensitive to weak binders due to stringent washing steps or cutoffs they employ. Especially sequences with two positions differing from the consensus can be very weakly bound - therefore a sensitive method is needed to investigate non-linearity. The method called High Performance Fluorescence Anisotropy (HiP-FA, recently developed in our lab) provides the necessary sensitivity. Using HiP-FA, I determined the affinities of 13 TFs from the Drosophila melanogaster segmentation network and found most of them to contain a significant non-linearity in their specificity. The binding energies of the TFs correlated significantly with certain DNA shape features suggesting shape readout by the TFs. These results could be confirmed in existing structural biology data. Besides the influence of information directly encoded in the DNA sequence, the binding of a TF in the genome is most influenced by the DNA accessibility. This property is a result of the genomic DNA being wrapped around histone octamers forming nucleosomes. Since the underlying sequence can also influence the binding of the histone complex to the DNA, a natural question to ask is which features of the DNA sequence are the major determinant of histone-DNA interaction. Attempts to address this question used existing methods which were either MNase based and are therefore prone to the enzymes intrinsic cutting bias or based on dialysis and/or EMSA readout and have in consequence a low throughput and can only be automated to a small degree. This leads to a limited set of measurements which are usually only based on a single measurement point instead of a complete titration curve. The second aim of my thesis is therefore to develop an in vitro assay to determine free energies of nucleosome formation which improves on the limitations of existing methods. Using the sensitive FA-microscopy setup, I developed an automated assay to determine the free energy of nucleosome formation in a competitive titration. In contrast to existing methods, the throughput of the assays allows for full competitor titration curves. By measuring the free binding energies of 42 sequences, I showed that GC-content is the factor most contributing to the free energy. The relationship between these quantities is non-monotonous with an optimal GC-content of 49 percent. The results provided in this thesis give insight into the nature of non-linearity in TF-DNA interactions and highlight the DNA shape readout therein. Methodical advancements developed in this work can be used as a foundation to investigate other kinds of molecular interactions making use of the high sensitivity of FA-based microscopy.