Search

The genetic information in eukaryotic genomes exists in a compacted state allowing for several billions of DNA base pairs to fit into the nucleus of the cells which is only micrometers wide. Nuclear enzymes thus require a highly regulated network of proteins able to decondense chromatin at specific locations, making it accessible to cellular processes such as DNA replication, DNA transcription, DNA recombination and DNA repair. Two types of enzymes exercise this function: chromatin modifying enzymes which create non-coded DNA modifications by the addition of covalent post-translational modifications called epigenetic marks and ATP-dependent chromatin remodellers which use a conserved ATPase motor to alter contacts within the nucleosome core particle, the basic subunit of chromatin. ATP-dependent chromatin remodellers are essential to maintenance of genetic integrity and all biological processes involving nucleic acids. Their dysregulation is observed in a variety of cancers and neurodevelopmental disorders, thus reinforcing their central role. The aim of the project is the analysis of the three ATP-dependent chromatin remodellers of the CHD subfamily of ATPases (CHD4, CHD5 and CHD7) using biochemical, biophysical and structural analysis techniques. The main interest is the resolution of the structures from subdomains to full-length proteins and the understanding of the interaction between each of these proteins and the nucleosome. Integrative structural biology methods have become more prevalent for the study of large proteins and macromolecular complexes and were used during this project. The first chapter of the thesis [Chapter 1 - Introduction] covers a literature review resuming the body of knowledge available about the topic of chromatin remodelling by CHD proteins in eukaryotic cells. The second chapter [Chapter 2 - Material and Methods] lists the different techniques used as well as protocols necessary to reproduce the experiments. The third chapter [Chapter 3 - CHD4] focusses on the study of CHD4 using biochemical techniques as well as negative staining EM and preliminary cryoEM. The fourth chapter [Chapter 4 - CHD5] details the work performed to analyse CHD5 in the context of mammalian and insect cells expression systems. The fifth chapter [Chapter 5 - CHD7] centres around the study of CHD7 and more particularly its tandem chromodomains and their interaction with the nucleosome particle. High throughput tests in insect cells were performed to identify optimal expression conditions of different human CHD7 constructs. Biochemical analysis by gel electrophoresis, Western blots, pulldowns, EMSAS were applied as well as structural biology techniques such as SEC-SAXS, X-ray crystallography and cryoEM. The thesis concludes with a general discussion of the results obtained from the integrative structural analysis approach of these three chromatin remodellers as well as from the use of structural prediction tools. General conclusions. Taken together, the work performed produced an ab initio SAXS model for the tandem chromodomains of human CHD7. In addition, low resolution maps of mouse CHD4 and human CHD7 truncated proteins have been obtained by cryoEM. Biophysical characterisation showed that the tandem chromodomains of CHD7 are not sufficient to retain the binding of the nucleosome core particle, therefore it is likely that the presence of additional DNA and histone binding domains are required to secure this interaction. Finally, bioinformatics and structural prediction analysis also revealed the presence of new domains in the C-terminal region of CHD7, CHD4, CHD5 and CHD3 corresponding to a potential SANT-like domain, a DNA binding domain commonly found in transcription factors and nuclear proteins.

Cilia elongation and homeostasis require the transport of building blocks and signalling molecules between the ciliary base and tip through a process called intraflagellar transport (IFT). IFT is driven by multiple copies of the ciliary motor proteins kinesin-2 and dynein-2 attached to train-like assemblies of IFT-A and IFT-B protein subcomplexes. Kinesin-2 transports IFT-trains along a microtubule-based ciliary cytoskeleton towards the ciliary tip during anterograde IFT. The ultrastructure of in situ anterograde IFT-trains was determined by cryo-electron tomography (cryo-ET) in the Pigino lab, but high-resolution structural information of the interactions within and between the IFT-A and IFT-B subcomplexes remains absent. I set out to purify IFT-complexes from the motile cilia of the protist Tetrahymena for structural analysis using negative-stain and cryo-electron microscopy (cryo-EM). I identified Bmh 14-3-3 as a potential novel component of the purified IFT-B complex and found IFT43, IFT25 and IFT56 are likely lost in Tetrahymena IFT. Negative stain EM averages of IFT-A complexes partially fit in the in situ IFT structure, but the complex disintegrated during cryo-EM sample preparation with all tested conditions. This report of purification and sample-preparation conditions contributes to the future optimisation of the IFT-complexes for cryo-EM. Anterograde IFT-trains reorganize into dynein-2 mediated retrograde trains during IFT-turnaround. IFT-turnaround takes place at the ciliary tip, close to but independent of ciliary microtubule capping structures. The role of the ciliary capping structures remains ambiguous and high-resolution information is lacking. Moreover, high-resolution in situ observations of IFT in mammalian motile cilia and the process of IFT turnaround are missing. I used cryo-ET on motile cilia from porcine trachea to survey IFT and ciliary cap structures in the ciliary tip. I observed ciliary cap structures consisting of two stacked amorphous plates and identified dynein densities in the distal-most plate. This suggests porcine ciliary cap structures consist of a single plate and IFT-turnaround densities, in contrast to five stacked plates previously observed in rodent trachea. I identified retrograde IFT-trains, and putative anterograde IFT-trains with high-contrast membrane-associated densities. A larger dataset of porcine tracheal cilia combined with subtomogram averaging could further the initial observations of novel membrane-associated densities, IFT-turnaround and the ciliary caps. In interphase cells, centriole-pairs are connected by striated fibrils composed of the coiled-coil protein rootletin along with CEP68, CCDC102B and potentially other proteins. Upon ciliogenesis, the centriole-pairs template the ciliary cytoskeleton and the fibrils merge into a striated fibre called the ciliary rootlet. The rootlet is highly conserved and critical for ciliary functioning and maintenance through an unknown mechanism. The rootlet's structural organisation, direct protein-protein interactions and striation composition remain unknown. I set out to purify recombinant and native rootlets for analysis with EM and cryo-ET to analyse the rootlet structure and ultrastructure. Tomograms of purified mouse photoreceptor cell rootlets showed a longitudinal periodicity with an amorphous striation and three confined striations each 81.3 nm. Subtomogram averaging showed the absence of lateral regularity in both the filaments and confined striations. The confined striations consist of globular densities associated to filaments, frequently found as parallel filament pairs, with no visible difference in filament density on either side of a striation. These findings rule out a previously proposed collagen-like filament organisation. Extensive optimisation of subtomogram alignments, classification and averaging resulted in a dimer of two-helix coiled-coils. Lateral oligomerisation of rootletin coiled-coils was not seen before in situ, but it remains unknown if the structure represents parallel or staggered rootletin contacts. Finally, I set out to establish a system for recombinantly expressed rootlets in SF9 insect cells and mammalian cells. Regardless of the presence of CEP68 or CCDC102B during expression, the purified fibres from insect or mammalian cells did not contain any striations. The previously observed striations are either not an intrinsic part of rootletin and may arise from unidentified rootlet interacting proteins, or the striations only form with factors such as chaperones that are missing in the current expression systems.

GABAA receptors (GABAARs) are pentameric ligand-gated ion channels that modulate the activity of neural circuits in response to binding of the neurotransmitter γ-aminobutyric acid (GABA) or small molecule therapeutics, such as benzodiazepines or general anaesthetics. GABAARs assemble from a pool of 19 paralogous genes, each conferring the receptor unique properties, such as ligand binding, gating kinetics or subcellular localization. Yet, the landscape of possible receptor assemblies, their signalling properties and the principles governing assembly remain elusive. In this thesis, electron cryo-microscopy (cryo-EM) was used to structurally characterize synaptic and extrasynaptic receptors isolated from cells expressing various combinations of α1, α4, β3, γ2 and δ subunits. I discovered that, from a pool of two or more subunits, individual cells can assemble more than one receptor subtype. The subtypes differ in relative subunit stoichiometry or arrangement and have unique signaling properties because most ligands bind at inter-subunit interfaces. Structural evidence is provided to show that differential assembly diversifies receptor responses to physiological and synthetic ligands by creating or eliminating ligand-binding pockets. For example, the α4β3δ receptor can simultaneously bind two neurotransmitters, GABA and histamine. Finally, I show that subtype stoichiometry depends on the identity of the α subunit and that, when co-expressed, α1 and α4 segregate into distinct subcomplexes. A computational model of receptor assembly was developed and combined with published single-cell RNA sequencing data to calculate the upper boundary for receptor diversity in recombinant systems and in vivo. These findings suggest that differential assembly of GABAARs may be a pervasive mechanism for regulating cellular responses to physiological and synthetic ligands. The rich diversity discovered in the GABAARs system may have facilitated the evolution of complex developmental programs and intelligence in animals.

Electron cryomicroscopy (cryoEM) is an important imaging technique for determining the atomic structures of biological macromolecules and complexes. Success in determining a structure by cryoEM depends on being able to prepare a thin frozen specimen of the molecules of interest on a small metal support, called a grid. Most high-resolution information loss in cryomicrographs stems from (1) radiation damage, and (2) particle movement in the vitrified specimen during imaging. This movement has been an unexplained problem for the field ever since Jaques Dubochet invented the method, more than three decades ago. In this work, experiments were performed to elucidate the physical causes of this movement and to formulate a theory that describes it. The movement is caused by buckling during freezing and subsequent deformation during imaging of the suspended ice within the holes of the grid. Surprisingly, there exists a critical threshold for this phenomenon, that depends directly on the shape of the frozen water layer. Below this threshold (width:thickness ratio for a given geometry), specimen movement is eliminated. Based on these findings, a simple solution to the movement problem in single-particle cryoEM is proposed: a new all-gold specimen support, HexAuFoil, where the foil hole diameter is tuned to be below the threshold, only 10 times larger than the specimen thickness (in contrast to the typical factor of 100 in commercially available supports). Movement-free imaging increases the information content of cryoEM movies, increases the throughput of current microscopes, and reduces the computational demands on processing them. Movement-free imaging also allows for a new approach to cryoEM reconstruction from 2D images: zero-dose extrapolation. A program that uses a radiation damage model to calculate the 3D structure of the undamaged molecule, before the onset of radiation damage, from movement-free data is developed. In collaboration with colleagues, we used HexAuFoil specimen supports and this approach to determine the structures of three protein complexes: (1) the SARS-CoV-2 RNA-dependent RNA polymerase in presence of RNA and the inhibitor favipiravir, (2) the light-harvesting 2 antenna complex of the purple bacterium Mch. purpuratum, and (3) the double-ring photosystem of G. phototrophica. These examples illustrate how movement-free imaging allowed us to better visualise ligand binding, metal coordination, solvent molecules, alternative side chain conformations, lipid bilayers and other important structural features of the molecules. This work also addresses a major practical bottleneck in modern cryoEM: the availability of high quality grids. Starting from the HexAuFoil design, the first integrated, wafer-scale method for manufacturing specimen supports by the thousand is developed. This method is based on techniques from the semiconductor industry. In contrast to current manufacturing methods, it does not require handling of individual grids, allowing large batches of various types of specimen supports to be produced in a scaleable manner with little labor. Understanding how and why plunge frozen vitreous specimens move unlocks the potential for further improvements in cryoEM. For example, movement-free imaging on a HexAuFoil support at 13 Kelvin is now demonstrated, meaning that we may soon be able to take advantage of the reduced secondary effects of radiation damage in cryoEM at lower temperature.

DNA replication is an essential biochemical process that underlies genetic inheritance. In eukaryotic cells, this process is carried out by a multi-subunit protein complex called replisome. The core of replisome is made up of six-subunit protein ring Mcm2–7, which contains weak DNA helicase activity. In order to fully activate the helicase activity, protein co-activators of Cdc45 and GINS have to be recruited to assemble a replicative DNA helicase called the CMG (Cdc45-Mcm2–7-GINS) complex. The active CMG subsequently unwinds the parental DNA duplex to provide single-stranded DNA templates for replicative DNA polymerases. The precise mechanism for CMG assembly and activation is still poorly understood. Recent studies using model organisms suggested the importance of protein called Sld3/Treslin for CMG assembly. The aim of the project was to understand in detail how human Treslin participates in the assembly of the CMG helicase. The interactions of Treslin with Cdc45, Mcm2–7, and DNA were characterised by a series of biochemical, biophysical, and structural biology experiments. Cdc45–Treslin interaction was shown to be conserved from budding yeast to human, in which it is mediated by a domain called Sld3/Treslin Homology Domain (SHD). However, human Treslin contains an additional Cdc45-binding site—next to the C-terminus of SHD— in an intrinsically disordered region referred to as Treslin ’extension’ (Treslinext). Treslinext was also demonstrated to interact with Mcm2–7 ring and DNA. S-phase Cyclin-Dependent Kinase (S-CDK) appeared to regulate Treslin’s interaction with Cdc45 and Mcm2–7, in which it targets a number of potential S-CDK phosphorylation sites present in Treslinext. As a part of this project, low-resolution Cdc45–Treslin and medium-resolution of MCM single hexamer 3D maps were obtained through single-particle cryo-electron microscopy (cryo-EM) analyses. The results provide molecular basis for Treslin’s role in the assembly of the CMG complex. Treslin acts as a molecular chaperone for Cdc45 recruitment to the DNA-bound Mcm2-7 ring. Treslin’s role during CMG assembly may involve its binding ability to the Mcm2–7 and DNA. Furthermore, the regulatory role of S-CDK in Treslin’s interactions with Cdc45 and Mcm2–7 suggests a link of the CMG assembly control to the cell cycle in human.

Eukaryotic genes contain non-coding introns, removal of which during gene expression is a pre-requisite for gene function. Removal of introns and ligation of coding exons --a process called splicing--is catalysed by a dynamic macromolecular machine called the spliceosome. The spliceosome assembles anew on precursor mRNA (pre-mRNA) from its component small nuclear ribonucleoproteins (snRNPs) and protein factors. A series of conformational and compositional changes produces the catalytically active spliceosome, in which the snRNA components of the snRNPs mediate intron recognition and catalysis of splicing. Despite the importance of splicing to fundamental eukaryotic molecular biology, and the contribution alternative splicing makes to organismal diversity and disease aetiology, a structural view of the spliceosome during catalysis has been lacking. Therefore, it is unclear how the spliceosome's active site is formed and stabilised, how protein factors contribute to the chemistry, fidelity, and dynamics of splicing, and exceptionally how the 3' splice site-i.e. the end of an intron-is recognised. This thesis first describes two atomic structures of the spliceosome from budding yeast, stalled immediately after each chemical reaction of splicing and resolved by electron cryomicroscopy. The structure of the spliceosome immediately after branching shows how the 5' splice site is ligated to the branch point adenosine upstream of the 3' splice site. It reveals the three-dimensional conformation of the RNA-based active site, its stabilisation by tertiary interactions and an extensive protein scaffold, and how the branching factors Cwc25, Yju2, and Isy1 promote docking of the branch point adenosine into the active site. The position of the Prp16 ATPase gave the first structural insight into how the DExD/H-box helicase proteins might remodel ribonucleoprotein complexes. The structure of the spliceosome immediately after exon ligation shows the connected 5' and 3' exons and how the active site is remodelled to allow two different reactions in a single active site. It shows large-scale conformational changes in the spliceosome associated with exon ligation and how these might be stabilised by the exon-ligation factors Prp18, Slu7, and Prp17. Most importantly it shows how the AG dinucleotide defining the 3' splice site is recognised by non-Watson-Crick base pairs to the branch point adenosine and 5' splice site guanosine. This explains the importance of the most conserved intron nucleotides and justifies a branching mechanism for splicing. Finally, it is unclear how the 5' splice site, which is initially recognised by the U1 snRNP, is transferred into the spliceosome's active site. The electron cryomicroscopy structure of a fully assembled pre-catalytic human spliceosome is described, stalled immediately before 5' splice site transfer from U1 to U6 snRNA. The structure shows in high-resolution the human U4/U6.U5 tri-snRNP and how U6 snRNA is primed by RBM42 and SNRNP-27K for receiving the 5' splice site. It shows how the DEAD-box ATPase Prp28 disrupts the 5' splice site/U1 snRNA duplex and how this might be coupled to Brr2-mediated spliceosome activation.

Complex I (NADH:ubiquinone oxidoreductase), located in the mitochondrial inner membrane, is a major electron entry point to the respiratory chain. It couples the energy released from electron transfer (from NADH to ubiquinone) to the concomitant pumping of protons across the membrane, to generate an electrochemical proton motive force. Mammalian complex I is composed of 45 subunits, 14 of which comprise its simpler bacterial homologues. It is encoded by both the mitochondrial and nuclear genomes, and pathological mutations in both sets of subunits result in severe neuromuscular disorders such as Leigh syndrome. Several structures of mammalian complex I from various organisms have been determined, but the limited resolutions of the structures, which typically refer to poorly characterised enzyme states, has hampered detailed analyses of mechanistic features. The first part of this thesis describes development of a method for purifying complex I from the genetically amenable and medically relevant model organism Mus musculus (mouse), in a pure, stable and active state. The enzyme from mouse heart mitochondria was then comprehensively characterised, to ensure the presence of all the expected subunits and co-factors, and to define its kinetic properties. The second part of this thesis describes structural studies by single particle electron cryomicroscopy (cryo-EM) on the purified mouse enzyme in two distinct states, the 'active' and 'de-active' states. The active state was determined to 3.3 Å resolution, the highest resolution structure of a eukaryotic complex I so far. Subsequently, comparison of the two mouse structures, together with previously determined mammalian and bacterial structures, revealed variations in key structural elements in the membrane domain, which may be crucial for the catalytic mechanism. Moreover, in the high-resolution active mouse complex I structure a nucleotide co-factor was observed bound to the nucleoside kinase subunit NDUFA10. Finally, complex I from the Ndufs4 knockout mouse model, which recapitulates the effects of a human mutation that causes Leigh syndrome, was purified and subjected to kinetic and proteomic analyses. Following cross-linking and preliminary structural studies, it was concluded that the detrimental effects of deleting NDUFS4 are due to lack of stability of the mature complex.

The tubulin/FtsZ superfamily contains a large set of proteins that spans through all kingdoms of life, with αβ-tubulins being the eukaryotic representatives and FtsZ being the best studied prokaryotic homologue. It is believed that all tubulin/FtsZ-related proteins have evolved from a common ancestor, however, members from this superfamily have diverged in many aspects. αβ-tubulins polymerise into giant and hollow microtubules in the presence of GTP. Despite the size of around 25 nm wide, microtubules display sophisticated dynamics. In particular, dynamic instability, the stochastic change between fast growth and rapid shrinkage, is a hallmark of microtubules. In contrast to αβ-tubulins, FtsZ lacks the C-terminal domain of tubulins and it probably functions as single homopolymeric protofilaments, possibly through treadmilling dynamics. There is strong divergence of the biological functions in the tubulin/FtsZ superfamily. Microtubules are involved in fundamental processes such as motility, transport and chromosomal segregation, whereas FtsZ is involved in bacterial cytokinesis (bacterial cell division), and the equivalent role of FtsZ is carried out by actin-based and ESCRTIII-based systems in eukaryotes. It seems that there is a big evolutionary gap between αβ-tubulins and FtsZ, and the only properties that are conserved within the tubulin/FtsZ superfamily are fold, protofilament formation and GTPase activity. In 2002, a pair of tubulin-like genes, btuba and btubb were identified in Prosthecobacter bacteria, with higher sequence homology to eukaryotic tubulins than FtsZ or any other bacterial homologues. The crystal structures solved later revealed, again, a closer similarity to αβ-tubulins than to their prokaryotic equivalents. It has been known for a while that BtubAB form filaments in the presence of GTP, however, little knowledge has been available regarding the filament architecture. In this project, I determined the near atomic resolution structure of the in vitro BtubAB filament using cryoEM and cryoET, revealing a hollow tube that consists of four protofilaments. A closer look showed that BtubAB filaments have many conserved microtubule features including: an overall polarity, similar longitudinal contacts, M-loops in lateral interfaces, and the presence of the seam, a structural hallmark of microtubules. My study also shows that BtubC, a protein with a TPR fold, binds to the BtubAB filaments in a stoichiometric manner, similar to some MAPs on microtubules. Based on this work, I concluded that BtubAB from Prosthecobacter form bacterial ‘mini microtubules’, and my work provided interesting insight into the evolution of tubulin/FtsZ-related proteins.

PSD95 is the canonical member of the Membrane Associated Guanylate Kinase class of scaffold proteins. PSD95 is a five-domain major scaffolding protein abundant in the postsynaptic density (PSD) of the neuronal excitatory synapse. Within PSD95 three PDZ domains modulate protein-protein interactions by selectively binding to short peptide motifs of target proteins. Under the direction of the multivalent PDZ domain interactions, the interacting proteins tend to cluster at the PSD, a phenomenon that is critical for synaptic signalling regulation. Earlier studies have shown that the N-terminal PDZ domains of PSD95 are obligatory for the clustering to occur. This thesis focuses on the strong inwardly rectifying potassium channel, Kir2.1 as the PSD95 binding partner. Kir2.1 is known to maintain membrane resting potential and control cell excitability. Previous studies have reported that Kir2.1 clustered into ordered tetrad complexes upon association with PSD95.This study investigates the detailed clustering mechanisms of Kir2.1 by PDZ domains. To achieve this, components that are involved in the formation of a complex namely PSD95 sub-domains comprising single PDZ and the tandem N terminal PDZ double domain (PDZ1-2), and Kir2.1 cytoplasmic domains(Kir2.1NC) are studied in detail via different structural and biophysical approaches; 1) PDZ1-2 is examined in apo- and bound ligand form with a Kir2.1 Cterminal peptide in crystal and solution via X-ray crystallography and small angle X-ray scattering; 2) the tandem and the single PDZ domain interaction with ligand are measured thermodynamically via isothermal calorimetry (ITC); 3) the complex of full length PSD95 with Kir2.1NC is analyzed with electron microscopy (EM). The protein components are produced in high quality by protein expression and multiple-step protein purification techniques. PDZ1-2 crystallographic structures were solved at 2.02A and 2.19A in theapo- and the liganded forms respectively. The solution state analysis showed domain separation and structural extension of the tandem domain when incorporated with the ligand. The ITC experiment revealed PDZ1-2 to have greater affinity towards the peptide ligand relative to the single PDZ domains. These combinatorial outcomes lead to the conclusion that PSD95 clusters Kir2.1 by adopting an enhanced binding interaction which is associated with increased PDZ1-2 inter-domain separation. The preliminary analysis of PSD95-Kir2.1NC complex with cryo-EM showed the establishment of a tetrad and led to a reconstruction at 40A resolution. The work in obtaining a higher resolution complex structure is promising with further data collection required to allow the employment of more sophisticated model reconstruction processes.