The work described in this thesis is an experimental study into the application of Electron Paramagnetic Resonance (EPR) Spectroscopy for the study of biological systems. Using a variety of methods of site-directed spin-labelling (SDSL), this thesis aims to explore long range structure in an assortment of recombinant and native proteins, and complexes thereof. The work described in this thesis covers all aspects of the work, from experimental design, molecular biology and cloning, protein expression and purification, as well as functional characterisation, and finally EPR distance measurements, data analysis and interpretation. Challenges and pitfalls will also be addressed. Chapters 1 and 2 introduce EPR spectroscopy, and its application in the study of long range structure in biological systems. The experimental techniques employed throughout this thesis are also introduced. Chapter 3 details an investigation into the complement C3b:factor H complex. This chapter addresses the challenges associated with the SDSL of cysteine rich proteins. Utilising hidden cysteine residues in native proteins for spin-labelling purposes will also be addressed. Chapter 4 looks at the interactions of the human myosin regulatory light chain (RLC) with cardiac myosin binding protein C (cMyBP-C). Optimisation of expression and purification protocols will be the focus, as well as addressing issues with protein solubility and spin labelling efficiencies. Chapter 5 explores the development of new methods of SDSL, for the specific labelling of cysteine rich proteins. The ability of Escherichia coli to read through the amber stop codon will be exploited for the incorporation of unnatural amino acids for labelling purposes, and novel spin labels, specific for labelling cysteine pairs tested in several model systems. Furthermore, native paramagnetic centres in recombinant proteins will be explored as potential labelling sites.

The fidelity of biological processes and reactions, inspite of the widespread diversity, is programmed by highly specific physico-chemical principles. This underlines our basic understanding of different interesting phenomena of biological relevance, ranging from enzyme specificity to allosteric communication, from selection of fold to structural organization / states of oligomerization, from half-sites-reactivity to reshuffling of the conformational free energy landscape, encompassing the dogma of sequence-structure dynamics-function of macromolecules. The role of striking an optimal balance between rigidity and flexibility in macromolecular 3D structural organisation is yet another concept that needs attention from the functional perspective. Needless to say that the variety of protein structures and conformations naturally leads to the diversity of their function and consequently many other biological functions in general. Classical models of allostery like the ‘MWC model’ or the ‘KNF model’ and the more recently proposed ‘population shift model’ have advanced our understanding of the underlying principles of long range signal transfer in macromolecules. Extensive studies have also reported the importance of the fold selection and 3D structural organisation in the context of macromolecular function. Also ligand induced conformational changes in macromolecules, both subtle and drastic, forms the basis for controlling several biological processes in an ordered manner by re-organizing the free energy landscape. The above mentioned biological phenomena have been observed from several different biochemical and biophysical approaches. Although these processes may often seem independent of each other and are associated with regulation of specialized functions in macromolecules, it is worthwhile to investigate if they share any commonality or interdependence at the detailed atomic level of the 3D structural organisation. So the nagging question is, do these diverse biological processes have a unifying theme, when probed at a level that takes into account even subtle re-orchestrations of the interactions and energetics at the protein/nucleic acid side-chain level. This is a complex problem to address and here we have made attempts to examine this problem using computational tools. Two methods have been extensively applied: Molecular Dynamics (MD) simulations and network theory and related parameters. Network theory has been extensively used in the past in several studies, ranging from analysis of social networks to systems level networks in biology (e.g., metabolic networks) and have also found applications in the varied fields of physics, economics, cartography and psychology. More recently, this concept has been applied to study the intricate details of the structural organisation in proteins, providing a local view of molecular interactions from a global perspective. On the other hand, MD simulations capture the dynamics of interactions and the conformational space associated with a given state (e.g., different ligand-bound states) of the macromolecule. The unison of these two methods enables the detection and investigation of the energetic and geometric re-arrangements of the 3D structural organisation of macromolecule/macromolecular complexes from a dynamical or ensemble perspective and this has been one of the thrust areas of the current study. So we not only correlate structure and functions in terms of subtle changes in interactions but also bring in conformational dynamics into the picture by studying such changes along the MD ensemble. The focus was to identify the subtle rearrangements of interactions between non-covalently interacting partners in proteins and the interacting nucleic acids. We propose that these rearrangements in interactions between residues (amino acids in proteins, nucleic acids in RNA/DNA) form the common basis for different biological phenomena which regulates several apparently unrelated processes in biology. Broadly, the major goal of this work is to elucidate the physico-chemical principles underlying some of the important biological phenomena, such as allosteric communication, ligand induced modulation of rigidity/flexibility, half-sites-reactivity and so on, in molecular details. We have investigated several proteins, protein-RNA/DNA complexes to formulate general methodologies to address these questions from a molecular perspective. In the process we have also specifically illuminated upon the mechanistic aspects of the aminoacylation reaction by aminoacyl-tRNA synthetases like tryptophanyl and pyrrolysyl tRNA synthetase, structural details related to an enzyme catalyzed reaction that influences the process of quorum sensing in bacteria. Further, we have also examined the ‘dynamic allosterism’ that manipulates the activity of MutS, a prominent component of the DNA bp ‘mismatch repair’ machinery. Additionally, our protein structure network (PSN) based studies on a dataset of Rossmann fold containing proteins have provided insights into the structural signatures that drive the adoption of a fold from a repertoire of diverse sequences. Ligand induced percolations distant from the active sites, which may be of functional relevance have also been probed, in the context of the S1A family of serine proteases. In the course of our investigation, we have borrowed several concepts of network parameters from social network analysis and have developed new concepts. The Introduction (Chapter-1) summarizes the relevant literature and lays down a suitable background for the subsequent chapters in the thesis. The major questions addressed and the main goal of this thesis are described to set an appropriate stage for the detailed discussions. The methodologies involved are discussed in Chapter-2. Chapter-3 deals with a protein, LuxS that is involved in the bacterial quorum sensing; the first part of the chapter describes the application of network analysis on the static structures of several LuxS proteins from different organisms and the second part of this chapter describes the application of a dynamic network approach to analyze the MD trajectories of H.pylori LuxS. Chapter-4 focuses on the investigation of human tryptophanyl-tRNA synthetase (hTrpRS), with an emphasis to identify ligand induced subtle conformational changes in terms of the alternation of rigidity/flexibility at different sites and the re-organisation of the free energy landscape. Chapter-5 presents a novel application of a quantum clustering (QC) technique, popular in the fields of pattern recognition, to objectively cluster the conformations, sampled by molecular dynamics simulations performed on different ligand bound structures of the protein. The protein structure network (PSN) in the earlier studies were constituted on the basis of geometric interactions. In Chapters 6 and 7, we describe the networks (proteins+nucleic acids) using interaction energy as edges, thus incorporating the detailed chemistry in terms of an energy-weighted complex network. Chapter-6 describes an application of the energy weighted network formalism to probe allosteric communication in D.hafniense pyrrolysyl-tRNA synthetase. The methodology developed for in-depth study of ligand induced changes in DhPylRS has been adopted to the protein MutS, the first ‘check-point protein’ for DNA base pair (bp) mismatch repair. In Chapter-7, we describe the network analysis and the biological insights derived from this study (the work is done in collaboration with Prof. David Beveridge and Dr. Susan Pieniazek). Chapter-8 describes the application of a network approach to capture the ligand-induced subtle global changes in protein structures, using a dataset of high resolution structures from the S1A family of serine proteases. Chapter-9 deals with probing the structural rationale behind diverse sequences adopting the same fold with the NAD(P)-binding Rossmann fold as a case study. Future directions are discussed in the final chapter of the thesis (Chapter-10).

The work in this thesis focuses on the study of three main aspects of proteins, viz, Protein structure, stability, and dynamics. Chapter 1 is a general introduction to the topics studied in this thesis. Chapter 2 deals with the first aspect, i.e., protein structure in which we describe an approach to use saturation mutagenesis phenotypes to guide protein structure prediction. Chapters 3 and 4 discuss how to increase protein stability using surface electrostatics, and Chapter 5 details a method to predict whether a proline substitution in a given protein would be stabilizing or destabilizing. Hence, Chapters 3-5 can be associated with the second aspect, i.e., protein stability. The third aspect, namely protein dynamics, is dealt with in Chapters 6 and 7 which study conformational dynamics of adenylate kinase. Protein structure prediction is a difficult problem with two major bottlenecks, namely, generation of accurate models and the selection of the most appropriate models from a large pool of decoys. In Chapter 2, the problem of model discrimination is addressed using mutant phenotype information derived from saturation mutagenesis library. A library of ~1500 single-site mutants of the E. coli toxin CcdB (Controller of Cell Division or Death B) has been previously constructed in our lab. The pooled library was characterized in terms of individual mutant phenotypes at various expression levels which were derived from the relative populations of mutants at each expression level. The relative populations of mutants were estimated using deep sequencing. Mutational tolerances were derived from the phenotypic data and were used to define an empirical parameter which correlated with a structural parameter, residue depth. We further studied how this new parameter can be used for model discrimination. Increasing protein stability in a rational way is a challenging problem and has been addressed by various approaches. One of the most commonly used approaches is optimization of protein core residues. Recently, optimization of protein surface electrostatics has been shown to be a useful approach for increasing stability of proteins. In Chapter 3, from analyses of a dataset of ~1750 non-homologues proteins, we show that proteins having a pI away from physiological pH, possess a significant fraction of unfavorably placed charged amino acids on their surface. One way to increase protein stability in such cases might be to alter these surface charges. This hypothesis was validated experimentally by making charge reversal mutations at putative unfavorable positions on the surface of maltose binding protein, MBP. The observed stabilization can potentially be increased by combining multiple individually stabilizing mutations. Different combinations of such mutations were made and tested in Chapter 4 to decide which mutants can be combined to achieve net stabilization. Ideas were tested through systematic experimentation which involved generation of two-site, three-site, and four-site mutations. A maximum increase in melting temperature (Tm) of 3-4 °C over wild-type protein was achieved upon combination of individually stabilizing mutants. Proline (Pro) has two special stereo-chemical properties when it is a part of a polypeptide chain. First the φ value of Pro has a very constrained distribution and second, Pro lacks an amide hydrogen. Due to these properties, introduction of Pro might perturb stability/activity of the protein. In Chapter 5 we describe a procedure to accurately predict the effects of Pro introduction on protein stability. Pro scanning mutagenesis was carried out on the model protein CcdB and the in vivo activity of the individual mutants was also examined. A decision tree was constructed, using the special stereo-chemical properties of Pro to maximize correlation of predicted phenotype with the in vivo activity. Binary classification as perturbing or non-perturbing of every Pro substitution was possible using the decision tree. The performance of the decision tree was assessed on various test systems, and the average accuracy was found to be ~75%. The role of conformational dynamics in enzyme catalysis has been explored in great detail in the literature. In Chapter 6, with the help of very long (350 ns), fully atomistic, explicit solvent molecular dynamics simulations, we studied conformational dynamics of adenylate kinase. We found the existence of a relatively stable state which lies intermediate between the open and closed conformations of the enzyme. The finding was further confirmed by computing a two dimensional configurational free energy surface when motions along each of the two movable domains (LID and NMP) are considered as reaction coordinates. We also discussed possible roles of the intermediate state during enzyme catalysis. The role of water in stabilization of intermediate states was also discussed. In Chapter 7, we studied dynamical coupling between LID and NMP domains of adenylate kinase during domain opening. Our observation suggests that the LID domain should start opening prior to the NMP domain. On the domain opening trajectory, the free energy surface of LID domain was found to be very rugged. We discuss a possible role of water in the ruggedness of the domain motions. The Appendix contains 3 supplementary parts of the thesis. Appendix I is a mutant dataset obtained from 454 sequencing analysis. It includes the normalized number of reads per mutation at each expression level along with mutational sensitivity score. Appendix II is parameters used for one of the electrostatic calculations. Appendix III contains a list of PDB ids used for database analysis in surface electrostatics work discussed in Chapter 3.

Missense single-nucleotide polymorphisms (mSNPs) in titin are emerging as a main causative factor of heart failure. However, distinguishing between benign and disease-causing mSNPs is a substantial challenge. Here, we research the question of whether a single mSNP in a generic domain of titin can affect heart function as a whole and, if so, how. For this, we studied the mSNP T2850I, seemingly linked to arrhythmogenic right ventricular cardiomyopathy (ARVC). We used structural biology, computational simulations and transgenic muscle in vivo methods to track the effect of the mutation from the molecular to the organismal level. The data show that the T2850I exchange is compatible with the domain three-dimensional fold, but that it strongly destabilizes it. Further, it induces a change in the conformational dynamics of the titin chain that alters its reactivity, causing the formation of aberrant interactions in the sarcomere. Echocardiography of knock-in mice indicated a mild diastolic dysfunction arising from increased myocardial stiffness. In conclusion, our data provide evidence that single mSNPs in titin's I-band can alter overall muscle behaviour. Our suggested mechanisms of disease are the development of non-native sarcomeric interactions and titin instability leading to a reduced I-band compliance. However, understanding the T2850I-induced ARVC pathology mechanistically remains a complex problem and will require a deeper understanding of the sarcomeric context of the titin region affected.

Malaria is the cause of over one million deaths per year, primarily in African children. The parasite responsible for the most virulent form of malaria, is Plasmodium falciparum. Protein structure plays a pivotal role in elucidating mechanisms of parasite functioning and resistance to anti-malarial drugs. Protein structure furthermore aids the determination of protein function, which can together with the structure be used to identify novel drug targets in the parasite. However, various structural features in P. falciparum proteins complicate the experimental determination of protein three dimensional structures. Furthermore, the presence of parasite-specific inserts results in reduced similarity of these proteins to orthologous proteins with experimentally determined structures. The lack of solved structures in the malaria parasite, together with limited similarities to proteins in the Protein Data Bank, necessitate genome-scale structural annotation of P. falciparum proteins. Additionally, the annotation of a range of structural features facilitates the identification of suitable targets for structural studies. An integrated structural annotation system was constructed and applied to all the predicted proteins in P. falciparum, Plasmodium vivax and Plasmodium yoelii. Similarity searches against the PDB, Pfam, Superfamily, PROSITE and PRINTS were included. In addition, the following predictions were made for the P. falciparum proteins: secondary structure, transmembrane helices, protein disorder, low complexity, coiled-coils and small molecule interactions. P. falciparum protein-protein interactions and proteins exported to the RBC were annotated from literature. Finally, a selection of proteins were threaded through a library of SCOP folds. All the results are stored in a relational PostgreSQL database and can be viewed through a web interface (http://deepthought.bi.up.ac.za:8080/Annotation). In order to select groups of proteins which fulfill certain criteria with regard to structural and functional features, a query tool was constructed. Using this tool, criteria regarding the presence or absence of all the predicted features can be specified. Analysis of the results obtained revealed that P. falciparum protein-interacting proteins contain a higher percentage of predicted disordered residues than non-interacting proteins. Proteins interacting with 10 or more proteins have a disordered content concentrated in the range of 60-100%, while the disorder distribution for proteins having only one interacting partner, was more evenly spread. Comparisons of structural and sequence features between the three species, revealed that P. falciparum proteins tend to be longer and vary more in length than the other two species. P. falciparum proteins also contained more predicted low complexity and disorder content than proteins from P. yoelii and P. vivax. P. falciparumprotein targets for experimental structure determination, comparative modeling and in silico docking studies were putatively identified based on structural features. For experimental structure determination, 178 targets were identi_ed. These targets contain limited contents of predicted transmembrane helix, disorder, coiled-coils, low complexity and signal peptide, as these features may complicate steps in the experimental structure determination procedure. In addition, the targets display low similarity to proteins in the PDB. Comparisons of the targets to proteins with crystal structures, revealed that the structures and predicted targets had similar sequence properties and predicted structural features. A group of 373 proteins which displayed high levels of similarity to proteins in the PDB, were identified as targets for comparitive modeling studies. Finally, 197 targets for in silico docking were identified based on predicted small molecule interactions and the availability of a 3D structure. / Dissertation (MSc)--University of Pretoria, 2008. / Biochemistry / unrestricted

Besides helping us advance the understanding of the physicochemical principles governing the three-dimensional folding of proteins and their mechanisms of action, the ability to build, evaluate, and optimize reliable 3D protein models has provided valuable tools for the development of different applications in the fields of biotechnology, medicine, and synthetic biology. The development of automated algorithms has made many of the current methodologies for protein modelling and visualization available to researchers from all backgrounds, without the need to be familiarized with the inner workings of their statistical and biophysical principles. However, there is still a lack in some areas where the learning curves are too steep for the methods to be widely used by the average non-programmer molecular biologist, or the implementation of the methods lacks key features to improve the interpretability and impact of their results. Throughout this work, I will focus on two different applications in the field of structural biology where computational methods provide useful tools to aid in synthetic biology or medical research. The first application is the implementation of a pipeline to build models of protein complexes by joining structured domains with disordered linkers, in individual or multiple chains, and with the possibility of building symmetric structures. Its capabilities and performance for the generation of complex constructs are evaluated, and possible areas of improvement described. The second application, but not less important, involves the structural analysis of patient-derived protein mutants using protein modelling techniques and visualization tools, to elucidate the potential molecular basis for the patient’s phenotype. The methodology for these analyses is described, along with the results and observations from 22 such cases in 13 different proteins. Finally, the need for a dedicated pipeline for the structure-based prediction of the effect of different types of mutations on the stability and function of proteins, complementary to available sequence-based approaches, is highlighted.

<div> <div> <div> <p>Proteins are drivers of almost all biological processes in the cell. The functions of a protein are dependent on their three-dimensional structure and elucidating the structure and function of proteins is key to understanding how a biological system operates. In this research, we developed computational methods using machine learning techniques to predicts the structure and function of proteins. Protein 3D structure prediction has advanced significantly in recent years, largely due to deep learning approaches that predict inter-residue contacts and, more recently, distances using multiple sequence alignments (MSAs). The performance of these models depends on the number of similar protein sequences to the query protein, wherein some cases similar sequences are few but dissimilar sequences with local similarities are more and can be helpful. We have developed a novel deep learning-based approach AttentiveDist which further improves over the previous state of art. We added an attention mechanism where dis-similar sequences are also used (increasing number of sequences) and the model itself determines which information from such sequences it should attend to. We showed that the improvement of distance predictions was successfully transferred to achieve better protein tertiary structure modeling. We also show that structure prediction from a predicted distance map can be further enhanced by using predicted inter-residue sidechain center distances and main-chain hydrogen-bonds. Protein function prediction is another avenue we explored where we want to predict the function that a protein will perform. The crux of the approach is to predict the function of protein based on the function of similar sequences. Here, we developed a method where we use dissimilar sequences to extract additional information and improve performance over the previous approaches. We used phylogenetic analysis to determine if a dissimilar sequence can be close to the query sequence and thus can provide functional information. Our method was ranked highly in worldwide protein function prediction competition CAFA3 (2016-2019). Further, we expanded the method with a neural network to predict protein toxicity that can be used as a safety check for human-designed protein sequences.</p></div></div></div>

This work is focused on the interactome study of 14-3-3ζ protein, a regulatory protein found in all eucaryotic cells. An important 14-3-3 protein feature is the ability to bind a number of structurally and functionally distinct protein ligands. This link is usually implemented through phosphorylated serine and threonine motifs. The first aim of this work is the preparation of sufficient amount of recombinant 14-3-3ζ protein with incorporated photoactivatable analogue of methionine (foto-Met, L-2-amino-5,5- azihexan acid). The four different conditions of recombinant expression in auxotrophic E. coli B834 (DE3) strain were tested to obtain a protein with a maximal rate of photoactivatable methionine analogue incorporation into the sequence 14-3-3 protein. The second aim is to study the methionine 121, 160 and 218 participation in the 14-3-3ζ protein binding groove and finding of potential covalent bond with the phosphorylated peptide 251-266 of Raf-1 kinase (phosphorylation on Ser259). The photo-initiated cross-linking method was used (photolysis), to form a reactive biradical of methionine analogue capable to attack any amino acid residues in close vicinity (till 5Å). Finally, the products of photo-initiated cross-linking were analyzed by cross-linking reactions using MALDI-TOF MS, LC-MS and...

Nuclear magnetic resonance (NMR) is one of the main approaches for protein struc- ture determination. The biggest advantage of this approach is that it can determine the three-dimensional structure of the protein in the solution phase. Thus, the natural dynamics of the protein can be studied. However, NMR protein structure determina- tion is an expertise intensive and time-consuming process. If the structure determi- nation process can be accelerated or even automated by computational methods, that will significantly advance the structural biology field. Our goal in this dissertation is to propose highly efficient and error tolerant methods that can work well on real and noisy data sets of NMR. Our first contribution in this dissertation is the development of a novel peak pick- ing method (WaVPeak). First, WaVPeak denoises the NMR spectra using wavelet smoothing. A brute force method is then used to identify all the candidate peaks. Af- ter that, the volume of each candidate peak is estimated. Finally, the peaks are sorted according to their volumes. WaVPeak is tested on the same benchmark data set that was used to test the state-of-the-art method, PICKY. WaVPeak shows significantly better performance than PICKY in terms of recall and precision. Our second contribution is to propose an automatic method to select peaks pro- duced by peak picking methods. This automatic method is used to overcome the limitations of fixed number-based methods. Our method is based on the Benjamini- Hochberg (B-H) algorithm. The method is used with both WaVPeak and PICKY to automatically select the number of peaks to return from out of hundreds of candidate peaks. The volume (in WaVPeak) and the intensity (in PICKY) are converted into p-values. Peaks that have p-values below some certain threshold are selected. Ex- perimental results show that the new method is better than the fixed number-based method in terms of recall. To improve precision, we tried to eliminate false peaks using consensus of the B-H selected peaks from both PICKY and WaVPeak. On average, the consensus method is able to identify more than 88% of the expected true peaks, whereas less than 17% of the selected peaks are false ones. Our third contribution is to propose for the first time, the 3D extension of the Median-Modified-Wiener-Filter (MMWF), and its novel variation named MMWF*. These spatial filters have only one parameter to tune: the window-size. Unlike wavelet denoising, the higher dimensional extension of the newly proposed filters is relatively easy. Thus, they can be applied to denoise multi-dimensional NMR-spectra. We tested the proposed filters and the Wiener-filter, an adaptive variant of the mean-filter, on a benchmark set that contains 16 two-dimensional and three-dimensional NMR- spectra extracted from eight proteins. Our results demonstrate that the adaptive spatial filters significantly outperform their non-adaptive versions. The performance of the new MMWF* on 2D/3D-spectra is even better than wavelet-denoising. Finally, we propose a novel framework that simultaneously conducts slice picking and spin system forming, an essential step in resonance assignment. Our framework then employs a genetic algorithm, directed by both connectivity information and amino acid typing information from the spin systems to assign the spin systems to residues. The inputs to our framework can be as few as two commonly used spectra, i.e., CBCA(CO)NH and HNCACB. Different from existing peak picking and resonance assignment methods that treat peaks as the units, our method is based on slices, which are one-dimensional vectors in three-dimensional spectra that correspond to certain (N, H) values. Experimental results on both benchmark simulated data sets and four real protein data sets demonstrate that our method significantly outperforms the state-of-the-art methods especially on the more challenging real protein data sets, while using a less number of spectra than those methods. Furthermore, we show that using the chemical shift assignments predicted by our method for the four real proteins can lead to accurate calculation of their final three-dimensional structures by using CS-ROSETTA server.

English Abstract Shortly after the identification of HIV as a causative agent of AIDS, an aspartic protease was identified in the viral genetic information. The very same time protease has become one of the dominant therapeutical targets in AIDS therapy. The introduction of protease inhibitors into the antiretroviral therapy has led to a significant improvement in the quality and length of life of HIV patients. However, the virus is still able to effectively prevent the impact of an inhibitor via generating inhibitor-resistant mutated protease variants. Thus, there is a constant need for novel types of inhibitors that would be capable of effectively blocking these resistant variants and simultaneously not supporting the development of novel resistant viral strains. One way to identify such inhibitors could be searching for compounds interacting with the enzyme at different sites than the active cavity, via the mechanisms of noncompetitive or uncompetitive inhibition. The group of compounds called metallacarboranes - inorganic compounds consisting of carbon, boron, hydrogen and metall ion - were shown to exhibit such an activity against HIV-1 protease. However, for further optimization of these inhibitors, detailed biophysical investigation of the enzyme-inhibitor complex is needed. This work focuses on the...