The rationalisation of drug potency using three-dimensional structures of protein-ligand complexes is a central paradigm in medicinal research. For over two decades, a major goal has been to find the rules that accurately relate the structure of any protein-ligand complex to its affinity. Addressing this problem is of great concern to the pharmaceutical industry, which uses virtual screens to computationally assay up to many millions of compounds against a protein target. A fast and trustworthy affinity estimator could potentially streamline the drug discovery process, reducing reliance on expensive wet lab experiments, speeding up the discovery of new hits and aiding lead optimization. Water plays a critical role in drug-protein interactions. To address the often ambiguous nature of water in binding sites, a water placement method was developed and found to be in good agreement with X-ray crystallography, neutron diffraction data and molecular dynamics simulations. The method is fast and has facilitated a large scale study of the statistics of water in ligand binding sites, as well as the creation of models pertaining to water binding free energies and displacement propensities, which are of particular interest to medicinal chemistry. Structure-based scoring functions employing the explicit water models were developed. Surprisingly, these attempts were no more accurate than the current state of the art, and the models suffered from the same inadequacies which have plagued all previous scoring functions. This suggests a unifying cause behind scoring function inaccuracy. Accordingly, mathematical analyses on the fundamental uncertainties in structure-based modelling were conducted. Using statistical learning theory and information theory, the existence of inherent errors in empirical scoring functions was proven. Among other results, it was found that even the very best generalised structure-based model is significantly limited in its accuracy, and protein-specific models are always likely to be better. The theoretical framework developed herein hints at modelling strategies that operate at the leading edge of achievable accuracy.

Macromolecular transitions such as conformational changes and protein-protein association underlie many biological processes. Conformational changes in the N-terminal domain of the transmembrane protein DsbD (nDsbD) were studied by NMR and molecular dynamics (MD) simulations. nDsbD supplies reductant to biosynthetic pathways in the oxidising periplasm of Gram-negative bacteria after receiving reductant from the C-terminal domain of DsbD (cDsbD). Reductant transfer in the DsbD pathway happens via protein-protein association and subsequent thiol-disulphide exchange reactions. The cap loop shields the active-site cysteines in nDsbD from non-cognate oxidation, but needs to open when nDsbD bind its interaction partners. The loop was rigid in MD simulations of reduced nDsbD. More complicated dynamics were observed for oxidised nDsbD, as the disulphide bond introduces frustration which led to loop opening in some trajectories. The simulations of oxidised and reduced nDsbD agreed well with previous NMR spin-relaxation and residual dipolar coupling measurements as well as chemical shift-based torsion angle predictions. NMR relaxation dispersion experiments revealed that the cap loop of oxidised nDsbD exchanges between a major and a minor conformation. The differences in their conformational dynamics may explain why oxidised nDsbD binds its physiological partner cDsbD much tighter than reduced nDsbD. The redox-state dependent interaction between cDsbD and nDsbD is thought to enhance turnover. NMR relaxation dispersion experiments gave insight into the kinetics of the redox-state dependent interaction. MD simulations identified dynamic encounter complexes in the association of nDsbD with cDsbD. The mechanism of the conformational changes in the transport cycle of LacY were also investigated. LacY switches between periplasmic open and cytoplasmic open conformations to transport sugars across the cell membrane. Two mechanisms have been proposed for the conformational change, a rocker-switch mechanism based on rigid body motions and an “airlock” like mechanism in which the transporter would switch conformation via a fully occluded structure. In MD simulations using the novel dynamics importance sampling approach such a fully occluded structure was found. The simulations argued against a strict “rocker-switch” mechanism.

Successful T cell activation depends on the recognition of antigenic peptides in the context of a Major Histocompatibility Complex molecule (pMHC) by the T cell antigen receptor (TCR), together with additional signals from co-stimulatory receptors such as CD28. Despite their importance, a thorough understanding of how TCR-pMHC binding properties relate to T cell functional responses remains unclear. In addition, there are no consensuses to the exact mechanism leading to CD28 receptor triggering. Activation of CD28 is dependent on the phosphorylation of key tyrosine residues within its cytoplasmic domain by Src family kinases. Just like the TCRs, CD28 receptors are susceptible to perturbations of the local kinase: phosphatase ratio. The K-S model postulates that upon ligand engagement, large RPTPs such as CD45 are segregated from the area of close contact, resulting in increased relative kinase concentration and CD28 receptor triggering. This hypothesis was tested in chapter 3, where elongated forms of CD80 were examined for their ability to costimulate primary T cells. CD28 costimulation was indeed diminished and there was reduced CD45 segregation from the elongated CD80 molecules. Additionally, CD28 habouring key Y170F tyrosine mutations were less susceptible to CD28 signal abrogation by elongated CD80 molecules. Interestingly, elongated CD80 molecules remained much less effective in mediating costimulation even when pMHC molecules were also elongated, suggesting that TCR-pMHC and CD28-CD80 size matching is not critical for costimulation. Despite the well-documented MHC-restriction requirement for TCR recognition, the relative energetic contributions of peptide versus MHC in TCR-pMHC interactions remain elusive. To address this question, the energetic footprints of four TCRs (1G4, JM22, A6 and G10) to HLA-A2 were determined via systematic alanine scanning mutagenesis on the HLA-A2 heavy chain in chapter 4. By targeting exclusive TCR contacting residues on the MHC, we conservatively estimate the contribution of MHCs for the four TCRs to range from 15% to over 70%. Several models have been formulated in an attempt to relate TCR-pMHC binding properties to T cell activation. Validity of the models was tested in chapter 5 using a supra-physiological TCR. By mutating key residues within the cognate pMHC, we generated a panel of TCR-pMHC with affinities that varies up to 105-fold. These reagents were used to stimulate Jurkat and primary T cells transduced with the supra-physiological TCR. Results in the Jurkat T cell system demonstrated the presence of an optimal off-rate (k_off) for TCR-pMHC interaction at low concentrations of pMHC concentration. The results argue against affinity models and the basic kinetic proofreading model for T cell activation.

Human immunodeficiency virus (HIV) Rev and Hepatitis B virus (HBV) e-antigen are both viral proteins that have key functions in their respective viral replication cycles. Both have evaded crystallization for decades due to their tendency to aggregate and/or form higher-order species. In this thesis the structure determination of HIV Rev and HBV e-antigen is presented—achieved via complexing with monoclonal antibody Fab fragments—and their structures are analysed. HIV Rev is a small regulatory protein that mediates the nuclear export of viral mRNAs, an essential step in the HIV replication cycle. In this process, Rev cooperatively oligomerises onto a highly structured RNA motif, the Rev response element. The structure of Rev (complexed with Fab), determined to 2.3 Å resolution, reveals a molecular dimer where the ordered portion of each subunit (N-terminal domain; NTD; residues 9-65) contains two coplanar a-helices arranged in hairpin fashion. Rev subunits dimerise via interaction of identical hydrophobic patches that overlap to form a V-shaped assembly. Mating of hydrophobic patches on the outer surface of the dimer promotes higher order interactions. Cryo-electron microscopy and helical image reconstruction of in vitro assembled Rev filaments were performed to better understand higher-order Rev oligomerisation. Reconstructions of Rev filaments were determined to ~13 Å resolution, permitting docking of the Rev NTD structure. Conformational variability of the Rev dimer subunits and use of a third ligomerisation interface engender filaments that can expand and contract. Both characteristics were also observed in the crystal structures of Rev. Surface features of the Rev filaments are altered in different expansion states, which may have implications for the assembled forms that Rev adopts during nuclear export of RNA and subsequent re-import into the nucleus. Various models for Rev oligomerisation onto the viral RNA are proposed. Chronic Hepatitis B virus (HBV) infection afflicts millions worldwide with cirrhosis and liver cancer. HBV e-antigen (HBeAg), a clinical marker for disease severity, is a soluble variant of the protein (core antigen, HBcAg) that forms the building-blocks of capsids. HBeAg is not required for virion production, but is implicated in establishing immune tolerance and chronic infection. The crystal structure of HBeAg clarifies how the short N-terminal propeptide of HBeAg induces a radically altered mode of dimerisation relative to HBcAg (~140 rotation), which is locked into place through formation of intramolecular disulfide bridges. This structural switch precludes capsid assembly and engenders a distinct antigenic repertoire, explaining why the two antigens are cross-reactive at the T-cell level (through sequence identity) but not at the B-cell level (through conformation). The structure offers insight into how HBeAg may establish immune tolerance for HBcAg while evading its robust immunogenicity.

Bacteriorhodopsin (bR) is a 7TM membrane protein expressed in Halobacterium salinarum. Due to its stability and high expression levels, bR serves as a model for other 7TM membrane proteins. Neurotensin receptor 1 (NTS1) is a member of pharmacologically relevant G protein-coupled receptor superfamily, and is the high affinity receptor for neurotensin, a 13mer peptide that can be found in the brain, gut, and central nervous system. NTS1 is a target for Parkinson’s, Schizophrenia, and drug addiction. This thesis aims to develop pulsed magnetic resonance techniques and sample preparation forms for high resolution structural studies on 7TM proteins. In this thesis, pulsed dipolar distance electron paramagnetic resonance (EPR) methods for the study of proteins in their native membrane are established. bR is spin-labeled, and a wellresolved distance distribution is measured in excellent agreement with other structural data. Preliminary distance data for a photoexcited state of bR suggests quaternary rearrangements in the native membrane that are agreement with published AFM results. A fitting method is developed to enable measurements of systems with rapid signal decay, a common feature in reconstituted systems studied by pulsed EPR methods. A physical chemical characterization of nanosized-bilayer discs termed Lipodisqs®, and the successful incorporation of bR is presented. Lipodisqs® are formed from DMPC and a polymer that is able to solubilize DMPC vesicles into discrete particles. Lipodisqs® possess a broad phase transition with increased lipid ordering compared to a DMPC dispersion. The SMA polymer interacts with the lipid tails, but does not perturb the headgroup. BR is incorporated in the monomeric form, and EPR dynamic and distance measurements confirm that Lipodisqs® preserve the native structure of bR, whilst detergent solubilisation increases the overall mobility compared to bR in its native membrane, suggesting that Lipodisqs® serve as an excellent medium for EPR studies on 7TM membrane proteins. A cysteine-depleted mutant of active, ligand binding NTS1 is constructed. Cysteines are reintroduced at positions that may be able to monitor agonist and inverse-agonist induced conformational and dynamic changes. A spin-labeling protocol is developed, and preliminary EPR measurements are discussed. Dynamic nuclear polarization (DNP) results are presented with uniformly-¹³C-labelled bR in the PM, resulting in a DNP enhancement of 16 using the biradical nitroxide polarizing agent, TOTAPOL. DNP-enhanced solid state NMR (ssNMR) is typically carried out at cryogenic temperatures, resulting in poor spectral resolution compared to ambient temperatures. Two different forms of samples are prepared that could potentially lead to better-resolved DNP spectra. BR is reverse labelled by adding natural abundance amino acids to isotopically labelled growth medium, resulting in the partial depletion of resonance signals that may obscure and crowd the NMR spectra. A crystalline sample of bR is prepared using the LCP method for crystallization, which is to date the most successful method for the crystallization of GPCRs. In summary, the first pulsed dipolar measurements of a protein in its native membrane are shown, Lipodisqs® are characterized and found to be a suitable medium for structural and functional studies of 7 TM membrane proteins, the first preliminary EPR studies on a ligand binding GPCR are presented, and novel sample preparation techniques are developed for the nitroxide-based DNP enhancement of ssNMR data. This thesis opens up several avenues for future research into 7TM membrane proteins.

A range of simulations have been conducted to investigate the behaviour of a diverse set of complex biological membrane systems. The processes of interest have required simulations over extended time and length scales, but without sacrifice of molecular detail. For this reason, the primary technique used has been coarse-grained molecular dynamics (CG MD) simulations, in which small groups of atoms are combined into lower-resolution CG particles. The increased computational efficiency of this technique has allowed simulations with time scales of microseconds, and length scales of hundreds of nm. The membrane-permeabilizing action of the antimicrobial peptide maculatin 1.1 was investigated. This short α-helical peptide is thought to kill bacteria by permeabilizing the plasma membrane, but the exact mechanism has not been confirmed. Multiscale (CG and atomistic) simulations show that maculatin can insert into membranes to form disordered, water-permeable aggregates, while CG simulations of large numbers of peptides resulted in substantial deformation of lipid vesicles. The simulations imply that both pore-forming and lytic mechanisms are available to maculatin 1.1, and that the predominance of either depends on conditions such as peptide concentration and membrane composition. A generalized study of membrane protein aggregation was conducted via CG simulations of lipid bilayers containing multiple copies of model transmembrane proteins: either α-helical bundles or β-barrels. By varying the lipid tail length and the membrane type (planar bilayer or spherical vesicle), the simulations display protein aggregation ranging from negligible to extensive; they show how this biologically important process is modulated by hydrophobic mismatch, membrane curvature, and the structural class or orientation of the protein. The association of influenza hemagglutinin (HA) with putative lipid rafts was investigated by simulating aggregates of HA in a domain-forming membrane. The CG MD study addressed an important limitation of model membrane experiments by investigating the influence of high local protein concentration on membrane phase behaviour. The simulations showed attenuated diffusion of unsaturated lipids within HA aggregates, leading to spontaneous accumulation of raft-type lipids (saturated lipids and cholesterol). A CG model of the entire influenza viral envelope was constructed in realistic dimensions, comprising the three types of viral envelope protein (HA, neuraminidase and M2) inserted into a large lipid vesicle. The study represents one of the largest near-atomistic simulations of a biological membrane to date. It shows how the high concentration of proteins found in the viral envelope can attenuate formation of lipid domains, which may help to explain why lipid rafts do not form on large scales in vivo.

Integral membrane proteins of the DP1 and reticulon families are responsible for maintaining the high membrane curvature required for both smooth ER tubules and the edges of ER sheets. Mutations in these proteins lead to motor neurone diseases such as hereditary spastic paraplegia. Reticulon/DP1 proteins contain Reticulon Homology Domains (RHD) that have unusually long (≈30 aa) hydrophobic segments and are proposed to adopt intramembrane helical hairpins that stabilise membrane curvature. I have uncovered the secondary structure and dynamics of the DP1 protein Yop1p and identified a C-terminal conserved amphipathic helix that on its own interacts strongly with negatively charged membranes and is necessary for membrane tubule formation. Analyses of DP1 and reticulon family members indicate that most, if not all, contain C-terminal sequences capable of forming amphipathic helices. Together, these results indicate that amphipathic helices play a previously unrecognised role in RHD membrane curvature stabilisation. This work paves the way towards full structure determination of Yop1p by solution state NMR and marks the first high structural resolution study on an RHD protein.

Shigella flexneri attacks cells of the intestinal tract, causing over 1 million deaths annually from bacterial dysentery. A type III secretion system (T3SS) initiates the host-pathogen interaction and transports virulence factors directly into host cells via a needle complex (NC) comprising an extracellular needle and membrane-spanning basal body. Rings formed by the single-pass membrane proteins MxiG and MxiJ are arranged concentrically within the inner membrane ring (IMR) of the NC. The Neterminal domain of MxiG (MxiG-N) is the predominant IMR cytoplasmic structure, however it was structurally and functionally uncharacterised. Determination of the solution structure of MxiG-N in this study revealed it to be a forkhead associated (FHA) domain, although subsequent analyses of conserved residues suggested it does not have the canonical role in cell-signalling via phospho-threonine recognition. Subsequent positioning of the structure in the electron microscopy (EM) density for the S. flexneri NC supported models with 24-fold symmetry in the IMR. Both MxiG and MxiJ also have significant periplasmic domains, which were purified to homogeneity in this study, facilitating preliminary characterisation of their structures and intermolecular interactions. In addition, the entire IMR within the context of intact basal bodies was isolated and visualised in vitro by EM. The essential function of MxiG-N could be to localise the putative cytoplasmic ring (Cering) at the base of the T3SS. Although absolutely required for secretion, the Csring component, Spa33, was structurally uncharacterised. The crystal structure of the Cvterminal domain of Spa33 (Spa33-C) was determined in this study, showing an intertwined dimer that aligned with homologous structures and exhibited a novel interaction with the N-terminus of the ATPase regulator, MxiN. Subsequently, Spa33-C was identified as an altemative translation product of spa33 that formed a 2: 1 complex with Spa33 in vitro. This complex oligomerised further, demonstrating for the first time that Spa33 has the propensity to form the ordered, high molecular weight assemblies that would be required for C-ring formation in S. flexneri.

The proton dependent oligopeptide transporters (POT family) are members of the Major Facilitator Superfamily of secondary active transporter proteins. They use the transmembrane proton gradient to drive the uptake of di- and tripeptides into the cytoplasm. Members of the family are highly conserved in pro- and eukaryotic genomes, and in humans they are responsible for the oral absorption of many drug families, including -lactam antibiotics. Recently, the crystal structures of PepTSo and PepTSt, two prokaryotic homologues of the human proteins PepT1 and PepT2, captured the proteins in two distinct conformations, providing insight into the structural aspects of the transport mechanism. A protocol was designed for functional liposome reconstitution of POT proteins, and transport assays were conducted to characterise their substrate specificity, pH dependence and kinetic properties. Using site-directed mutagenesis, we identified binding site residues involved in peptide recognition and proton translocation, and distinguished between the two roles by comparing protein activity in proton- and peptide-driven conditions. We also investigated the roles of key residues in the conformational transitions that accompany the transport cycle, using data from biochemical assays, molecular dynamics simulations and modeling, as well as electron paramagnetic resonance measurements. In addition, several bacterial POT members were screened for crystallisation, in order to assess their stability and crystal diffraction quality in different detergents. Further work was performed with bacterial POT homologues YdgR and GkPOT, including binding studies using NMR spectroscopy and assaying drug transport in vivo and in vitro. Together, the data establish bacterial POTs as model systems for studying the mammalian oligopeptide transporters, and a mechanistic model for peptide transport is proposed.

Correct functioning of the brain relies upon the precise connectivity between the billions of neurons that make up this crucial organ. Aberrations in the formation of these elaborate neural networks lead to neurodegenerative and neuropsychiatric disorders. A synapse-spanning molecular triad, involving members of the Neurexin, Cbln and ionotropic glutamate receptor delta families of proteins, is crucial for the accurate formation and proper function of synapses in the cerebellum. This trans-synaptic complex has been implicated in the molecular mechanisms behind motor control and motor learning, and furthermore individual members have been linked to diseases such as Alzheimer’s, autism spectrum disorders and schizophrenia. The major findings presented in this thesis include: crystal structures of the amino-terminal domains (ATD) of the two members of the ionotropic glutamate receptor delta (iGluR-Delta) family, functional characterisation of the effects of disrupting the ATD interface in one member of the iGluR-Delta family, a crystal structure of the C1q domain of Cbln1, biophysical analysis of the molecular interactions within the Neurexin-Cbln1-GluD2 trans-synaptic complex, as well as evidence for the domain arrangement of the ecto-domain of the iGluR-Delta proteins. Together, these data enhance our knowledge of the molecular details of this macro-molecular complex and provide evidence to support models for the mechanisms of their involvement in synapse formation and function, thereby making a contribution to the vast and medically relevant field of molecular neurobiology.