Receptor tyrosine kinases (RTKs) are a major class of cell surface receptors, important in cell signalling events associated with a variety of functions. High-throughput (HTP), coarse-grained molecular dynamics (CG-MD) simulations have been used to investigate the dimerization of the transmembrane (TM) domain of selected RTKs, including epidermal growth factor receptor (EGFR) and muscle-specific kinase (MuSK). EGFR activation requires not only a specific TM dimer interface, but also a proper orientation of its juxtamembrane (JM) domain. Phosphatidylinositol 4,5-bisphosphate (PIP₂) is known to abolish EGFR phosphorylation through interaction with basic residues within the JM domain. Here, a multiscale approach was used to investigate anionic lipid clustering around the TM-JM junction and how such clustering is modulated by the mutation of basic residues. The simulations demonstrated that PIP₂ may help stabilize the JM-A antiparallel dimer, which may in turn help stabilize TM domain helix packing of the N-terminal dimerization motif. A proximal TM domain residue has been implicated in the inhibition of ganglioside GM3 in phase-separated membranes. Here, CG simulations were used to explore the dynamic behaviour of the EGFR TM domain dimer in GM3-containing and GM3-depleted bilayers designed to resemble lipid-disordered (Ld) and phase-separated (Ld/Lo) membranes. The simulations suggest that the presence of GM3 in Ld/Lo bilayers can disrupt and destabilize the TM dimer, which helps to explain why GM3 may favour monomeric EGFR in vivo. To gain insights into the dynamic nature of the intact EGFR, a nearly complete EGFR dimer was modelled using available structural data and embedded in an asymmetric compositional complex bilayer, which resembles the mammalian plasma membrane. The results demonstrated the dynamic nature of the EGFR ectodomain and its predicted interactions with lipids in the local bilayer. Strong protein-lipid interactions, as well as lipid-lipid interactions, affect the local clustering of lipids and the diffusion of lipids in the vicinity of the protein on both leaflets.

Protein aggregation is a major challenge in the development of antibody-based therapeutics. Therapeutic antibodies are produced and stored in high concentrations and under fluctuating conditions unfavourable for their stability. Aggregation of these proteins in solution leads to serious consequences for patients, with the initiation of immune reactions, which have the potential to be fatal, and in the loss of clinical potency. The types of aggregates formed by antibodies, and the processes that lead to their propagation are poorly understood. By studying these molecules via computational approaches, we are able to simulate and probe their tendency to aggregate on experimentally comparable timescales. By performing small numbers of coarse grained simulations of immunoglobulin frag- ments it is shown that specific regions of proteins are involved in self-self interactions, and these regions are targets for reducing the self-association of experimental molecules. Techniques developed here are integrated within a high throughput approach that is able to generate information on aggregation for a large number of candidate antibody structures. The methodology was refined via development of a novel technique for coarse grained simulations of oligosaccharides. This method was initially tested on glycolipids, and then extended to glycoproteins. The primary outcome is a coarse grained model for a glyco- sylated antibody Fc fragment. The glycosylated Fc was then simulated, and compared to experimental data. Coarse grained simulations support the hypothesis that the protein be- comes more flexible in the absence of glycosylation.

The physical properties of viral capsids are major determinants of vaccine efficacy for several picornaviruses which impact on human and animal health. Current picornavirus vaccines are frequently produced from inactivated virus. Inactivation often reduces the stability of the virus capsid, causing a problem for Foot and Mouth Disease Virus (FMDV) where certain serotypes fall apart into pentameric assemblies below pH 6.5 or at temperatures slightly above 37°C, destroying their effectiveness in eliciting a protective immune response. As a result, vaccines require a cold chain for storage and animals need to be frequently immunised. FMDV is a member of the Aphthovirus genus of the Picornaviridae. Globally there are seven FMDV serotypes: O, A, Asia1, C and SAT-1, -2 and -3, contributing to a dynamic pool of antigenic variation. As part of collaboration between the Division of Structural Biology, Oxford University, The Pirbright Institute, Reading University and ARC, Ondespoort, South Africa we sought to rationally engineer thermo-stable FMDV capsids either as infectious copy virus or recombinant empty capsids with improved thermo-stability for improved vaccines. In this project, in silico molecular dynamics (MD) simulations, molecular modelling, free energy calculations, X-ray crystallography, electron microscopy and various biochemical/biophysical techniques were used to design and help characterise the capsids. For the most unstable FMDV serotypes (O and SAT2), panels of stabilising mutants were characterised by MD. Promising candidates were then engineered and shown to confer increased thermo- and pH-stability. Thus, in silico predictions translate into marked stabilisation of both infectious and recombinant empty viral capsids. A novel in situ method was used to determine crystal structures for quality assessment and to verify that no unanticipated structural changes have occurred as a consequence of the modifications made. The structures of the wildtype and two of the stabilised mutants were solved and the antigenic surfaces shown to be unchanged. Animal trials showed stabilised particles can generate a similar or improved neutralising antibody response compared to the traditional vaccines and may therefore lead to a new generation of stable and safe vaccines.

Membrane proteins perform a variety of functions essential for the viability of the cell, including transport and signalling across the membrane. Most membrane proteins are formed from bundles of transmembrane helices. In this thesis molecular dynamics simulations have been used to investigate helix insertion into bilayers and helix association within bilayers. The potentials of mean force for the insertion of helices derived from the cystic fibrosis transmembrane conductance regulator into lipid bilayers were calculated using coarse-grained molecular dynamics simulations. The results showed that the insertion free energy increased with helix length and bilayer hydrophobic width. The insertion free energies obtained were significantly larger than comparable quantities obtained from translocon- mediated insertion experiments, consistent with a variety of previous studies. The implications of this observation for the interpretation of in vivo translocon-mediated insertion experiments, and the function of the translocon, are discussed. Coarse-grained and atomistic molecular dynamics simulations of the transmembrane region of the receptor tyrosine kinase EphA1 suggested that the transmembrane helix dimer was most stable when interacting via the glycine zipper motif, in agreement with a structure obtained by NMR spectroscopy. Coarse-grained simulations of the transmembrane region of EphA2 suggested that the dimer has two stable orientations, interacting via a glycine zipper or a heptad motif. Both structures showed right-handed dimers, although an NMR structure of the transmembrane region of EphA2 shows a left-handed dimer interacting via the heptad motif. Both structures obtained from coarse-grained simulations proved unstable when simulated at an atomistic level of detail. The potentials of mean force for dissociating the EphA1 and EphA2 dimers were calcu- lated using coarse-grained molecular dynamics calculations. Convergence of the detailed structure of the profiles was not conclusively shown, although association free energies cal- culated from the profiles were consistent over a variety of simulation times. The association free energies were slightly larger than experimental values obtained for comparable sys- tems, but consistent with similar computational calculations previously reported. However, direct comparisons are difficult owing to the influence of environmental factors on reported association free energies. The potential of mean force profiles showed that the interaction via the glycine zipper motif for EphA1 was significantly more stable than any other confor- mation. For EphA2 the potential of mean force profiles suggested that interaction via the glycine zipper and heptad motifs both provided stable or metastable conformations, with the interaction via the glycine zipper motif probably at least as stable as that via the heptad motif.

Transcription lies at the centre of gene expression. In eukaryotes, transcription occurs not only at genes but also across the non-coding portion of the genome, an apparently pervasive process that gives rise to a wide array of different transcripts. In recent years, it has emerged that genes themselves are frequently subject to non-coding transcription of their antisense strand. This antisense transcription is evident in eukaryotes from yeast to mammals; however its general genome-wide role, if indeed it has one, remains elusive. Here, the nature of antisense transcription in the budding yeast Saccharomyces cerevisiae is explored on a genome-wide scale. Antisense transcription is ubiquitous and often abundant, and appears to be driven by a promoter architecture at the 3’ end of genes, one which shows evidence of regulation, and which mirrors that found at the 5’ end. Furthermore, antisense transcription shows evidence of changing gene behaviour. It is associated with a drastically altered chromatin environment at the 5’ promoter and across the gene body; however it is not associated with a change in the level of gene transcription itself. Rather, these chromatin changes appear to enforce a change in the mode of gene transcription, promoting rapid bursts of transcription re-initiation that result in noisier gene expression – a hitherto unknown role of antisense transcription. It is proposed that antisense transcription represents a fundamental layer of gene regulation, and that it should be considered a canonical feature of eukaryotic genes.

Membrane active peptides with the ability to cross the plasma membrane represent a promising class of therapeutic compounds. However, translocation efficacy and membrane toxicity of these peptides appear correlated and a better understanding of their mechanisms of action is needed to achieve the desired effect. Here, a range of coarse grain molecular dynamics simulations have been performed to systematically investigate the interactions of such cell-penetrating peptides (CPPs) with biologically relevant membranes. Challenges associated to the development of a suitable asymmetric mammalian membrane model demonstrated the importance of lipid species distribution on the bilayer mechanical properties, as well as the effect of coarse graining on its electrostatic properties. However, simulations successfully discriminated between two CPPs, penetratin and transportan, and were consistent with the experimental data available for these. The results obtained suggest that the ability of transportan peptides to aggregate into flexible, dynamic, transmembrane bundles is responsible for their relative membrane toxicity. The stability and structure of these aggregates, as well as the extent of the bilayer perturbations they induced, were shown to depend on the membrane composition and asymmetry, thus providing a molecular basis to explain how the toxicity of CPPs is modulated by membranes. In particular, bilayer destabilisation was enhanced by the presence of anionic lipids and hampered by that of cholesterol. Transportan aggregates were also observed to trigger lipid flip-flops above a certain size and a new pathway for such events, not relying on the formation of water defects, was characterised.

Specific recognition of pathogenic molecules by the immune system is mediated by proteins known as antigen receptors. One such component is the antibody. Binding properties of natural and engineered antibodies can be understood by studying the structure of their variable domains, VH and VL. In this thesis we investigate how the two variable domains orientate with respect to one another and therefore influence the geometry of the antigen binding site which is formed between them. We describe a method which fully characterises the VH-VL orientation in a consistent and absolute sense using five angles and a distance. The ABangle method is used to investigate variable domain orientation in structures collected by our database SAbDab. Using the ABangle method we compare VH-VL orientation to the corresponding property in a different component of the immune system, the T-cell receptor (TCR). Despite having similar individual domain structures the variable domain orientations of antibodies and TCRs are found to be distinct. This is found to affect an antibody’s ability to mimic TCR specificity. ABangle's characterisation is used to find determinants of the VH-VL orientation. We identify sequence and structural properties that influence the variable domain pose. A feature based method for predicting VH-VL orientation is presented and assessed. Future directions of this research and its application to the development of antibody therapeutics are described.

Integrins are large heterodimeric (αβ) cell surface receptors that play a key role in the formation of focal adhesion complexes and are involved in various signal transduction pathways. They are ‘activated’ to a high affinity state by the formation of an intracellular complex between the membrane, the integrin β-subunit tail and talin, a process known as ‘inside-out activation’. The head domain of talin, a FERM domain homologue, plays a vital role in the formation of this complex. Recent studies also suggest that kindlins act in synergy with talin to induce integrin activation. Despite much available structural and functional data, details of how talin activates integrins remain elusive. In this thesis a multiscale simulation approach (using a combination of coarse-grained and atomistic molecular dynamics simulations) together with NMR experiments were employed to study talin-mediated integrin inside-out activation. A number of novel insights emerged from these studies including: (i) the crucial role of negatively charged lipids in talin/membrane association; (ii) a novel V-shape conformation of the talin head domain which optimizes its interactions with negatively charged lipids; (iii) that interactions of talin with negatively charged moieties in the membrane orient talin to optimize interactions with the β cytoplasmic tail; (iv) that binding of talin to the β cytoplasmic tail promotes rearrangement of the integrin TM helices and weakens the integrin α/β association; and (v) that an increase in the tilt angle of the β integrin TM helix initiates a scissoring movement of the two integrin TM helices. These results, combined with experimental data, provide new insights into the mechanism of integrin inside-out activation. The same multiscale approach was used to demonstrate the crucial role of the Gx3G motif in the packing of the integrin transmembrane helices. Using recent structural data the integrin/talin complex was modelled and inserted in bilayers which resemble the biological plasma membrane. The results demonstrate the dynamic nature of the integrin receptor and suggest that the integrin/talin complex alters the lipid organization and motion in the outer and inner bilayer leaflets in an asymmetric way and that diffusion of lipids in the vicinity of the protein is slowed down. The work in this thesis demonstrates that multiscale simulations have considerable strengths when applied to investigations of structure/function relationships in membrane proteins.

At the time of writing this thesis, the complete genomes of more than 180 organisms have been sequenced and more than 80000 biological macromolecular structures are available in the Protein Data Bank (PDB). While the number of sequenced genomes and solved three-dimensional structures are rapidly increasing, the functional annotation of protein sequences and structures is a much slower process, mostly because the experimental de-termination of protein function is expensive and time-consuming. A major class of in silico methods used for protein function prediction aim to transfer annotations between proteins based on sequence or structural similarities. These approaches rely on the assumption that homologous proteins of similar primary sequences and three-dimensional structures also have similar functions. While in most cases this assumption appears to be valid, an increasing number of examples show that proteins of highly similar sequences and/or structures can have different biochemical functions. Thus the relationship between the divergence of protein sequence, structure and function is more complex than previously anticipated. On the other hand, there is mounting evidence suggesting that minor changes of the sequences and structures of proteins can cause large differences in their conformational dynamics. As the intrinsic fluctuations of many proteins are key to their biochemical functions, the fact that very similar (almost identical) sequences or structures can have entirely different dynamics might be important for understanding the link between sequence, structure and function. In other words, the dynamic similarity of proteins could often serve as a better indicator of functional similarity than the similarity of their sequences or structures alone. Currently, little is known about how proteins are distributed in the 'dynamics space' and how protein motions depend on structure and sequence. These problems are relevant in the field of protein design, studying protein evolution and to better understand the functional differences of proteins. To address these questions, one needs a precise definition of dynamic similarity, which is not trivial given the complexity of protein motions. This thesis is intended to explore the possibilities of describing the similarity of proteins in the 'dynamics space'. To this end, novel methods of characterizing and comparing protein motions based on molecular dynamics simulation data were introduced. The generally applicable approach was tested on the family of PDZ domains; these small protein-protein interaction domains play key roles in many signalling pathways. The methodology was successfully used to characterize the dynamic dissimilarities of PDZ domains and helped to explain differences of their functional properties (e.g. binding promiscuity) also relevant for drug design studies. The software tools developed to implement the analysis are also introduced in the thesis. Finally, a network analysis study is presented to reveal dynamics-mediated intramolecular signalling pathways in an allosteric PDZ domain.

Électroporation (EP) est une technique utilisée pour affecter l’intégrité des membranes cellulaires de plasma et/ou organites internes, conséquence de l’application d’un champ électrique d’énergie suffisante, dépendant de son intensité et sa durée. Il a été montré in- directement par de nombreuses études expérimentales et in-silico que ce phénomène résulte de la perméabilisation de la membrane par la formation pores aqueux. L’EP permet ainsi la vectorisation de molécules normalement non perméantes. Les applications de l’EP vont de l’électrochimiothérapie, à la vaccination à ADN. Les impulsions électriques utilisées dans l’EP sont classées en deux familles: Les msPEF dont la longueur des impulsions est de l’ordre de la microseconde et l’amplitude de l’ordre de quelques kV/cm. Ils affectent principalement la membrane cellulaire plasmique. Les nsPEFs d’intensité de MV/m de durée de l’ordre de la nanoseconde, ceux eux sont capables de perméabiliser organites internes ainsi que la membrane plasmique et présentent l’avantage d’éviter les effets thermiques indésirables. Les simulations de dynamique moléculaire (DM) qui permettent la description atomique, de la structure de la membrane et de son interaction avec la solution environnante, constituent un appui précieux aux résultats expérimentaux. Plusieurs études utilisant la DM été consacrées à décrire certains des aspects de l’EP (par exemple la formation de pores, leur évolution, le rôle de l’eau et des groupes de tête lipidiques, ...) néanmoins des questions en suspens restent inexplorées : • Comment la composition de la membrane affecte le seuil d’EP ? • Quelles sont la morphologie, la taille et la conductance des pores formés ? • Quels sont le mécanisme et l’échelle de temps de translocation de petites molécules à travers ces électropores ? • Y-a-t-il une différence notoire entre les effets des msPEFs et des nsPEFs ? Dans le cadre de ce travail, en utilisant des simulations de DM nous avons abordé ces questions pertinentes. Nous avons quantifié le seuil d’EP de bicouches lipidiques contenant des concentrations croissantes de cholestérol utilisant des protocoles qui miment les deux modes types de pulses nsPEFs et msPEFs. Les résultats obtenus indiquent que dans les deux cas les modèles de membranes à concentration en cholestérol croissante, nécessitent un voltage transmembranaire plus élevé pour perméabiliser la bicouche lipidique. Nous avons développé une procédure, mimant l’effet des msPEFs en adéquation avec les expériences, qui permet de stabiliser les voltages appliqués à la membrane suffisamment longtemps pour déterminer la dimension des pores, leur conductance et sélectivité ionique. Nous avons utilisé le même protocole pour étudier le transport de petites molécules chargées, utilisés dans l’administration de médicaments, et comparé nos résultats avec des études similaires menées dans des conditions nsPEFs. Nous avons montré que le transport assisté par EP a lieu dans la même échelle de temps (ns) que sous nsPEFs. Bien que les nsPEF ont l’avantage d’affecter les membranes cellulaires et celles des organites internes, la possibilité d’exploiter de telles impulsions pour la vectorisation de médicaments est encore en cours d’étude, car la capacité à fournir de manière fiable à des échantillons «biologiques» ces impulsions intenses ultra-courtes n’est pas trivial. Une attention particulière doit être accordée à la conception de micro-chambres afin de réaliser un dispositif à large bande passante afin de transmettre sans atténuation et distorsion les pulses ns, qui sont caractérisés par une grandes composante spectrale, jusqu’à GHz. Une partie importante de cette thèse mené en cotutelle, a été consacrée à la conception théorique (utilisant la Méthode des éléments Finis) d’un dispositif d’exposition, basé sur des systèmes de propagations de micro-ondes, capable de délivrer des impulsions aussi courtes que la ns avec des temps de monté et de chute de 0,5 ns / Electroporation (EP) is a technique used to affect the integrity of plasma cell membranes and/or internal organelles, consequence of the application of an external pulsed electric field of sufficient energy content, tuned by its strength and duration. It is proven by extensive indirect experimental and in silico evidences that this phenomenon results in the permeabilization of membrane structures by aqueous pores, allowing the transport of poorly- or non-permeant molecules, e.g. salts, ions, genetic material, and any other small solutes present. Applications of the techniques range from electrochemoterapy DNA vaccination and gene regulation. The electric pulses used in EP are categorized in two main families: msPEF, the length of the pulses is in the µs- ms scale and the amplitude in the order of kV/cm, their effect takes place mainly at the plasma cell membrane of cells; nsPEFs, higher magnitude (MV/m) over ns time scale, they act are able to permeabilize internal organelles as well as the plasma cell membrane, presenting the advantage of avoiding undesired thermal effects. Molecular dynamics simulations allow the microscopic description, with atomic resolution, of the membrane structure and its interaction with the surrounding solution, providing a substantial support to experimental findings. A considerable amount of work have been devoted to describe some of the aspects of EP using MD, (e.g. the pore formation, its evolution and reseal, the role of water and of lipid headgroups, …) nevertheless outstanding questions remain unexplored: • How does the composition of the bilayer affect the EP threshold? • What are the morphology, size and conductance of pores formed? • What are the mechanisms and time scales of translocation of small molecules through the electropores? • Is there any difference when modeling nsPEFs and msPEFs? As part of the present work, using MD simulations and comparing our results to other findings from our group, we addressed some relevant questions. We quantified the EP threshold of libid bilayes for the increasing concentration of cholesterol (0, 20, 30, 50 mol %) when the two protocol to model nsPEFs and msPEFs are exploited. The results obtained applying the two approaches indicate that in both cases an increase in cholesterol concentration requires a higher transmembrane voltage to porate the membrane bilayer. We developed a procedure, mimicking msPEFs, to stabilize electropores under different transmembrane voltages in mechanical condition similar to experiments for a time long enough to determine the pore dimension, its conductance and selectivity to ion species. We employed the same method to investigate the transport of small charged molecules, used in drug delivery, comparing our findings with similar studies conducted under nsPEFs conditions with the attempt to rationalize the molecular uptake. Interestingly we found that that the dynamic of the transport process takes place in the same time scale (nanosecond) that for nsPEFs. Despite the fact that nsPEFs have the advantage to affect both cell membranes and internal organelles and to further reduce thermal effects, the possibility to exploit nsPEFs for drug delivery is an ongoing research since the ability to reliably deliver to biological loads these ultra-short intense pulses is not trivial. Particular attention must be paid in the design of microchambers to realize a broadband devices to transmit without attenuation and distortion nsPEF, which exhibit large spectral components, i.e. spanning from MHz up to GHz. An important part of the current work has been devoted to the design (with Finite Element Method) of an exposure device, based on microwave propagating systems, able to deliver pulses down to 1 ns with rise and fall time of 0.5 ns