The extensive and sometimes incorrect and noncompliant use of various types of antimicrobial agents has accelerated the development of antimicrobial resistance (AMR). In fact, AMR has become one of the greatest global threat to human health in this era. The broad-spectrum antibiotics aminoglycosides (AGs) display excellent potency against most Gram-negative bacteria, mycobacteria, and some Gram-positive bacteria, such as Staphylococcus aureus. The AG antibiotics amikacin, gentamicin, kanamycin, and tobramycin are still commonly prescribed in the U.S.A. for the treatment of serious infections. Unfortunately, bacteria evolve to acquire resistance to AGs via four different mechanisms: i) changing in membrane permeability to resist drugs from entering, ii) upregulating efflux pumps for active removal of intracellular AGs, iii) modifying the antimicrobial target(s) to prevent drugs binding to their targets, and iv) acquiring resistance enzymes to chemically inactivate the compounds. Amongst all, the acquisition of resistance enzymes, AG-modifying enzymes (AMEs), is the most common resistance mechanism identified. Depending on the chemistry each enzyme catalyzes, AMEs can be further divided into AG N-acetyltransferases (AACs), AG O-phosphotransferases (APHs), and AG O-nucleotidyltransferases. To overcome AME-related resistance, we need to better understand these resistance enzymes and further seek ways to either escape or inhibit their actions. In this dissertation, I summarized my efforts to characterize the AAC(6') domain and its mutant enzymes from a bifunctional AME, AAC(6')-Ie/APH(2")-Ia as well as another common AME, APH(3')-IIa. I also explained my attempt to inhibit the action of various AAC enzymes using metal salts. In an effort to explore the current resistance epidemic, I evaluated the resistance against carbapenem and AG antibiotics and the correlation between the resistance profiles and the AME genes in a collection of 122 Pseudomonas aeruginosa clinical isolates obtained from the University of Kentucky Hospital System. Besides tackling the resistance mechanisms in bacteria, I have also attempted to explore a new antifungal option by repurposing an existing antipsychotic drug, bromperidol, and a panel of its derivatives into a combination therapy with the azole antifungals against a variety of pathogenic yeasts and filamentous fungi.

Les déshydrogénases sont faiblement actives en présence de fortes concentrations (> 50 % (v/v)) de liquides ioniques (LIs) miscible à l’eau et les raisons précises de cette inactivation ne sont pas connues. La structure de la formiate déshydrogénase de Candida boidinii (FDH) en présence de ces LIs miscibles à l’eau a été étudiée par atténuation de fluorescence par de l’iode ou de l’acrylamide. Une concentration critique, la CILc ("Critical Ionic Liquid concentration"), au dessus de laquelle la fluorescence n’est pas exploitable, est déterminée. Elle se situe entre 30 et 40 % (v/v) des LIs de cette étude. Les LIs s’avèrent être de forts agents dénaturants : les constantes d’atténuation de fluorescence en présence de LI sont de 1,4 à 2 fois plus importantes qu’en présence d’urée. La FDH a été modifiée chimiquement par des cations analogues aux cations des LIs afin de la préserver de l’interaction avec les LIs. Les enzymes modifiées par des cations de plus petite taille présentent une importante activité résiduelle (30-45%) en présence de 70% (v/v) de LI tandis que l’enzyme sauvage est inactive. En présence de 30% (v/v) de LI, l’efficacité catalytique (kcat/KM) des enzymes modifiées augmente de 1,3 à 3,6 fois et la constante de Michaelis (KM) diminue de 1,7 à 4,6 fois suivant le cation utilisé. Selon la taille du cation greffé, le temps de demi-vie des enzymes modifiées augmente de 3 à 5 fois en solution tamponnée. Enfin, la structure des enzymes modifiées est préservée en présence de 40% (v/v) de LI tandis que l’enzyme native commence à se dénaturer. La technique d’évolution dirigée a également été utilisée. A ce jour, 987 mutants ont été criblés. Un mutant (M60) présente une activité résiduelle de 35% à 70% (v/v) de LI tandis que l’enzyme sauvage est inactive. Ce dernier porte deux mutations, N187S et T321S, situées à la surface de la structure protéique. En présence de 30 % (v/v) de LI, le mutant 60 fixe les substrats avec des valeurs de de KMNAD et de KDN3 (N3 est l’azide, un inhibiteur compétitif du formiate) 2 fois plus faibles que celles de l’enzyme sauvage. / The dehydrogenases are weakly active in the presence of high concentration (> 50% (v/v)) of water-miscible ionic liquids (ILs) and the mechanism of enzyme inactivation in ILs is not fully understood. The structure of the formate dehydrogenase Candida boidinii (FDH) in ILs has been studied by quenching of fluorescence experiments in the presence of iodide or acrylamide. A critical concentration, the CILc (“Critical Ionic liquid concentration"), is defined as the concentration of IL above which the fluorescence is not relevant. In this work, the CILc is comprised between 30 and 40 % (v/v) of the ILs. The ILs have revealed to be denaturing agents which increase the quenching of fluorescence efficiency by 1.4 to 2 times more than in urea. The FDH is chemically modified by analogous cations found in ILs in order to preserve the biocatalyst from the direct interaction with ILs. The enzymes modified by the smaller cations shown 30 45% residual activity at 70% (v/v) of ILs while the native enzyme is fully inactive. In the presence of 30% (v/v) of ILs, the kinetic efficiency (kcat/KM) of the grafted enzymes is improved by a 1.3-3.6 fold factor and the constant of Michaelis (KM) is reduced by a 1.7-4.6 fold factor depending on the grafted cations. In aqueous solution, the half-lives of modified enzymes are 3 to 5 fold higher than the native FDH depending on the size of the cation grafted. Finally, the structure of the grafted enzymes is somehow maintained in the presence of 40% (v/v) of ILs while the native FDH begins to unfold. We also used directed evolution to improve the FDH activity in the presence of ILs. At this stage, 987 mutants have been screened and one mutant (M60) shows 35% of residual activity at 70% (v/v) of ILs. In the presence of 30% (v/v) of ILs, the mutant binds the substrates in a greater extent than the native FDH, the values of KMNAD and the KDN3 (N3 is the azide, a competitive inhibitor of formate) are reduced by a 2 fold factor by comparison to the native enzyme.

This thesis presents the application of different protein engineering methods on enzymes and non-catalytic proteins that act upon xyloglucans. Xyloglucans are polysaccharides found as storage polymers in seeds and tubers, and as cross-linking glucans in the cell wall of plants. Their structure is complex with intricate branching patterns, which contribute to the physical properties of the polysaccharide including its binding to and interaction with other glucans such as cellulose. One important group of xyloglucan-active enzymes is encoded by the GH16 XTH gene family in plants, including xyloglucan endo-transglycosylases (XET) and xyloglucan endo-hydrolases (XEH). The molecular determinants behind the different catalytic routes of these homologous enzymes are still not fully understood. By combining structural data and molecular dynamics (MD) simulations, interesting facts were revealed about enzyme-substrate interaction. Furthermore, a pilot study was performed using structure-guided recombination to generate a restricted library of XET/XEH chimeras. Glycosynthases are hydrolytically inactive mutant glycoside hydrolases (GH) that catalyse the formation of glycosidic linkages between glycosyl fluoride donors and glycoside acceptors. Different enzymes with xyloglucan hydrolase activity were engineered into glycosynthases, and characterised as tools for the synthesis of well-defined homogenous xyloglucan oligo- and polysaccharides with regular substitution patterns. Carbohydrate-binding modules (CBM) are non-catalytic protein domains that bind to polysaccharidic substrates. An important technical application involves their use as molecular probes to detect and localise specific carbohydrates in vivo. The three-dimensional structure of an evolved xyloglucan binding module (XGBM) was solved by X-ray diffraction. Affinity-guided directed evolution of this first generation XGBM resulted in highly specific probes that were used to localise non-fucosylated xyloglucans in plant tissue sections. / QC 20100902

The number of monoclonal antibodies approved for therapeutic use has increased rapidlyover the last decade. As a consequence, precise and robust kinetic characterization techniquesare crucial in order to select the best suitable candidates. A kinetic characterization methodwas developed in Gyrolab with automated sample transfers. The characterization wasperformed in solution in a mixing CD, containing an integrated nanoliter mixing chamberwith affinity binding columns. Association rate constants were determined for four anti-TSHantibodies with values ranging from 3x105 M-1s-1 to 10x105 M-1s-1. The antibodies wereranked according to kass. Reproducibility

Mit Aminotransferasen können chirale Amine auf biotechnologischem Weg hergestellt werden. Diese besitzen große Bedeutung als Bausteine für weitere Synthesen in der pharmazeutischen und agrochemischen Industrie. Da natürlich vorkommende Enzyme oft nicht die gewünschte Substratspezifität für bestimmte industrielle Anwendungen besitzen, ist eine Optimierung durch Mutagenese notwendig. Solche Entwicklungen sind jedoch oft mit hohem Zeit- und Kostenaufwand verbunden. Die Optimierung kann entweder ungezielt durch empirische Methoden oder gezielt unter Einbeziehung von Informationen über das Enzym erfolgen. Die notwendigen Daten können als Vorbereitung zu konkreten Produktentwicklungen durch Untersuchungen an potentiell geeigneten Enzymen gewonnen werden. Um einen solchen rationalen Ansatz bei der einer speziellen Amin-Pyruvat-Aminotransferase zu ermöglichen, war es Ziel der vorliegenden Arbeit die Grundlagen für die Veränderung der Substratspezifität dieses Enzyms zu erarbeiten. Zunächst wurden strukturelle Informationen durch ein Homologie-Modell gewonnen und später durch eine experimentell bestimmte Struktur ergänzt. Mit dieser Struktur wurden die Substrat-bindenden Reste identifiziert und zunächst der Einfluss auf die Substratbindung durch ortsgerichtete Mutagenese überprüft. Es konnte gezeigt werden, dass alle acht ausgewählten Aminosäurereste an der Substratbindung beteiligt sind. Zudem wurde unter diesen Positionen nach Mutanten gesucht, die neue Substrate umsetzen können. Eine Reihe von Mutanten wurde identifiziert, die verschiedene neue Substrate umsetzen. Für zwei Positionen konnten eine Reihe von Mutanten identifiziert werden, die neue Substrate akzeptieren. Durch die Art der Seitenketten, die Position der Aminosäuren und der chemischen Struktur der akzeptierten Substrate konnten eine Reihe von Aussagen über den Mechanismus der Substratbindung für diese Amin-Pyruvat-Aminotransferase gemacht werden. Außerdem wurde die Zweckmäßigkeit der eingesetzten theoretischen und experimentellen Methoden für die Anwendung bei Entwicklungen mit Enzymen dieser Klasse gezeigt.

Millions of years of evolution have resulted in an immense number of different proteins, which participate in virtually every process within cells and thus are of utmost importance for allknown forms of life. In addition, there are several examples of natural proteins which have found use in applications outside their natural environment, such as the use of enzymes infood industry and washing powders or the use of antibodies in diagnostic, bioseparation or therapeutic applications. To improve the performance of proteins in such applications, anumber of techniques, all collectively referred to as ‘protein engineering’, are performed in thelaboratory.Traditionally, methods involving ‘rational design’, where a few alterations are introduced atspecific protein locations to hopefully result in expected improvements have been applied.However, the use of more recent techniques involving a simultaneous construction of a large number of candidate variants (protein libraries) by various diversification principles, fromwhich rare clones showing enhanced properties can be isolated have contributed greatly to thefield of protein engineering.In the present thesis, different protein traits of biotechnological importance have beenaddressed for improvements by the use of such methods, in which there is a crucial need tomaintain a clonal link between the genotype and the phenotype to allow an identification of protein library members isolated by virtue of their functional properties. In all protein library investigations included in this thesis this coupling has been obtained by Escherichia coli bacterialcell-membrane compartmental confinement.In a first study, a combination of error prone PCR and gene-shuffling was applied to the Tobacco Etch Virus (TEV)-protease gene in order to produce collections from which genesencoding variants showing an enhanced soluble expression of the enzyme frequently used inbiotechnology to cleave fusion proteins were identified. Using Green Fluorescence Protein(GFP)-based cell fluorescence analysis, a clone with a five-fold increase in the yield of solubly produced protein was successfully isolated. In a second study, a novel and different GFPbased selection system, in addition also involving targeted in vivo protein degradation principles,was employed for investigations of the substrate sequence space of the same protease. In two additional studies, a selection system denoted Protein Fragment Complementation Assay(PCA), based on the affinity driven structural complementation of a genetically split β-lactamase enzyme was used to identify variants having desired target protein binding abilities,including both specificity and affinity. Using Darwinian principles concerning clonal growth advantages, affibody binding proteins showing sub-nanomolar dissociation constants to thehuman cytokine TNF-α were isolated. Taken together, these studies have shown that the bacterial format is very well suited for use in various aspects of protein library selection. / QC 20100729

Abstract Enzymes are potentially superior as catalysts for many industrial chemical processes because of their high specificity, selectivity, minimum energy requirement and environmental friendliness. However, many challenges remain in order to exploit fully the potential of industrial enzymes. The qualities which are needed are catalytic proficiency, availability in high quantities, low price, low product inhibition, and high activity and stability under process conditions. Directed evolution and rational design are the most common strategies to produce enzymes with the desired properties. The TIM barrel is the most frequent and most versatile fold among naturally occurring enzymes. In all known TIM barrel enzymes, the catalytically active residues are located at one end of the barrel structure, while residues maintaining the stability of the fold are found on the opposite end of the barrel. This special architecture of the TIM barrel proteins makes it possible to change catalytic activity of the protein without compromising its stability, which is a perfect start for protein engineering studies. In this research project, a monomeric triosephosphate isomerase (TIM) variant with an engineered binding groove (A-TIM) was created by using a rational design approach. The major aims of this work were (i) to find novel binders and (ii) characterize the new, bigger binding groove using X-ray crystallographic methods. These studies have discovered that monomeric A-TIM can bind compounds completely different from the natural substrate. Studies on three different classes of binder molecules are reported: (i) true substrate analogues of wild type TIM, (ii) substrate analogues that have an extended hydrophobic tail, and (iii) more extended, phosphate containing substrate analogues. In addition to this, the A-TIM active site was shown to be competent. In general these studies illustrate the importance of protein crystallography for characterizing the binding properties of enzyme variants being studied in enzyme discovery projects. / Tiivistelmä Entsyymit voivat toimia ylivoimaisina katalyytteinä monissa kemianteollisuuden prosesseissa johtuen niiden hyvästä spesifisyydestä, valikoimiskyvystä, alhaisesta energiantarpeesta ja ympäristöystävällisyydestä. Näistä ominaisuuksista huolimatta entsyymien kaikkien mahdollisuuksien hyödyntämisen esteenä on monia haasteita. Tarvittavia ominaisuuksia ovat katalyyttinen tehokkuus, saatavuus suurina määrinä, alhainen hinta, alhainen tuoteinhibitio sekä korkea aktiivisuus ja stabiilisuus prosessiolosuhteissa. TIM-tynnyrirakenne on yleisin ja monipuolisin proteiinien laskostumisrakenne luonnossa esiintyvissä entsyymeissä. Tässä rakenteessa katalyyttisesti aktiiviset aminohappotähteet ovat sijoittuneet tynnyrirakenteen toiselle puolelle, kun taas stabiilisuuden kannalta tärkeät aminohappotähteet ovat sijoittuneet kokonaan toiselle puolelle. Tämä erityinen rakenne antaa mahdollisuuden muokata proteiinin katalyyttistä aktiivisuutta vaikuttamatta haitallisesti sen stabiilisuuteen. Tämä on täydellinen lähtökohta proteiininmuokkaukselle. Tässä tutkimusprojektissa käytettiin ns. järkiperäistä suunnittelua monomeerisen trioosifosfaatti-isomeraasivariantin (A-TIM) luomisessa. Tämän tutkimustyön pääasialliset tavoitteet olivat (i) uusien sitoutujien löytäminen ja (ii) uuden, suuremman sitoutumistaskun ominaisuuksien määrittäminen röntgenkristallografisilla menetelmillä. Tässä tutkimuksessa havaittiin, että A-TIM kykenee sitomaan yhdisteitä, jotka ovat täysin erilaisia luonnolliseen substraattiin verrattuna. Tässä tutkimuksessa kuvaillaan kolmenlaisia sitoutujia: (i) todelliset villityypin entsyymin substraattianalogit, (ii) substraattianalogit, joihin on liitetty hydrofobinen hiilivetyketju ja (iii) villityypin substraattia suuremmat sokerifosfaatit. Tämän lisäksi A-TIM:n aktiivisen keskuksen todistettiin olevan toimintakykyinen. Yleisellä tasolla tämä tutkimus osoittaa röntgenkristallografisten menetelmien tärkeyden entsyymienmuokkausprojekteissa, joissa entsyymivarianttien ominaisuuksien määritys on tärkeää.

Comprendre comment les mutations impactent l’activité d’une protéine reste un défi dans le domaine des sciences protéiques. Les méthodes biochimiques traditionnellement utilisées pour résoudre ce type de questionnement sont très puissantes mais sont laborieuses à mettre en œuvre. Des approches bioinformatiques ont été développées à cet égard pour surmonter ces contraintes. Dans cette thèse, nous explorons l'utilisation d'approches bioinformatiques pour comprendre le lien entre mutations et changements d'activité. Notre modèle d'étude est une enzyme bactérienne, la sucrose phosphorylase de Bifidobacterium adolescentis (BaSP). Cette glycosyl-hydrolase de la famille 13 (GH13) suscite l’intérêt de l'industrie en raison de sa capacité à synthétiser des disaccharides et des glycoconjugués originaux. Son activité consiste à transférer un glucose d'un donneur, le saccharose, à un accepteur qui peut être un monosaccharide ou un aglycone hydroxylé. La réaction enzymatique se déroule selon un mécanisme dit « double déplacement avec rétention de configuration », ce qui nécessite la formation d'un intermédiaire covalent dit glucosyl-enzyme. Cependant, la possibilité de contrôler la régiosélectivité de ce transfert pour qu'il soit applicable au niveau industriel est un enjeumajeur. Cette thèse vise d’une part, à fournir une explication rationnelle quant aux modifications de la régiosélectivité de BaSP apportées par des mutations et d’autre part à proposer un canevas pour le contrôle de la régiosélectivité de couplage en vue de la synthèse de disaccharides pré-biotiques rares comme le kojibiose et le nigerose. Dans notre approche, nous avons émis l'hypothèse que les orientations préférées de l'accepteur dans le site catalytique après formation du glycosyl-enzyme déterminent la régiosélectivité de l'enzyme. Nous avons utilisé des approches computationnelles pour étudier l'impact des mutations sur la liaison de l'accepteur à l'intermédiaire covalent, le glucosylenzyme. À cette fin, nous avons construit des modèles à l’échelle atomique du glucosyl-enzyme pour un ensemble de variants de la BaSP pour lesquels des données expérimentales étaient disponibles. Pour y parvenir, nous avons paramétré le glucosyl-aspartyle en tant que nouveau résidu et les avons intégré dans des outils de modélisation tels que Modeller et Gromacs. Nous avons évalué la pertinence de ces paramètres et les avons ensuite appliqués à la vérification de notre hypothèse de travail par le biais d’expériences d'ancrage moléculaire. La méthodologie utilisée dans ce travail ouvre la perspective de l'utilisation d'approches bioinformatiques pour l'ingénierie de la régiosélectivité de la sucrose phosphorylase et plus généralement des glycosylhydrolases possédant un mécanisme similaire. À cet égard, un pipeline de modélisation moléculaire et d'amarrage de molécules accepteurs sur des intermédiaires covalents des enzymes de cette famille (ENZO pour Optimisation d’ENZyme) a été développé au cours de cette thèse. Son application à l’ingénierie d’autres variants de BaSP est en cours. / In this thesis, we explore the usage of computational approaches for understanding the link between mutations and changes in protein activity. Our study model is a bacterial sucrose phosphorylase enzyme from Bifidobacterium adolescentis (BaSP). This glycosyl hydrolase from family 13 (GH13) has been a focus in the industry due to its ability to synthesize original disaccharides and glycoconjugates. In fact, its activity is to transfer a glucose moiety from a donor sucrose to an acceptor which can be a monosaccharide or a hydroxylated aglycone. The enzymatic reaction proceeds by a double displacement with retention of configuration mechanism whereby a covalent glucosyl-enzyme intermediate is formed. However, it is at stake to control the regioselectivity of this transfer for it to be applicable at industrial level. This thesis aimed at providing a rational explanation for the observed impact of mutations on the regioselectivity of BaSP in view of controlling the synthesis of rare pre-biotic disaccharides like kojibiose and nigerose. We hypothesized that the preferred orientations of the acceptor determines the regioselectivity of the enzyme. In that respect, we used computational approaches to investigate the impact of mutations on the binding of the acceptor to the glucosyl-enzyme intermediate. The methodology used in this work opens the perspective of using computational approaches for engineering the regioselectivity of of glycosyl hydrolases with similar mechanism.

Glutathione transferase (GST) A2-2 is the human enzyme most efficient in catalyzing azathioprine activation. Structure-function relationships were sought explaining the higher catalytic efficiency compared to other alpha class GSTs. By screening a DNA shuffling library, five recombined segments were identified that were conserved among the most active mutants. Mutational analysis confirmed the importance of these short segments as their insertion into low-active GSTs introduced higher azathioprine activity. Besides, H-site mutagenesis led to decreased azathioprine activity when the targeted positions belonged to these conserved segments and mainly enhanced activity when other positions were targeted. Hydrophobic residues were preferred in positions 208 and 213. The prodrug azathioprine is today primarily used for maintaining remission in inflammatory bowel disease. Therapy leads to adverse effects for 30 % of the patients and genotyping of the metabolic genes involved can explain some of these incidences. Five genotypes of human A2-2 were characterized and variant A2*E had 3–4-fold higher catalytic efficiency with azathioprine, due to a proline mutated close to the H-site. Faster activation might lead to different metabolite distributions and possibly more adverse effects. Genotyping of GSTs is recommended for further studies. Molecular docking of azathioprine into a modeled structure of A2*E suggested three positions for mutagenesis. The most active mutants had small or polar residues in the mutated positions. Mutant L107G/L108D/F222H displayed a 70-fold improved catalytic efficiency with azathioprine. Determination of its structure by X-ray crystallography showed a widened H-site, suggesting that the transition state could be accommodated in a mode better suited for catalysis. The mutational analysis increased our understanding of the azathioprine activation in alpha class GSTs and highlighted A2*E as one factor possibly behind the adverse drug-effects. A successfully redesigned GST, with 200-fold enhanced catalytic efficiency towards azathioprine compared to the starting point A2*C, might find use in targeted enzyme-prodrug therapies.

Protein tertiary structure is more conserved than amino acid sequence, leading to a diverse range of functions observed in the same fold. Despite < 20 % overall sequence identity, cytochromes P450 all have the same fold. Bacterial Class I P450s receive electrons from a highly specific, often unidentified, ferredoxin, in which case the hemoprotein is termed “orphaned”. CYP199A2, a Class I P450, accepts electrons from ferredoxins Pux and HaPux. Five orientation-dependent and one orientation-independent DEER measurements on paramagnetic HaPux and spin-labelled CYP199A2 yielded vector restraints, which were applied to building a model of the CYP199A2:HaPux complex in silico. A different binding mode was observed compared to P450cam:Pdx and P450scc:Adx, both recently elucidated by X-ray crystallography. This protocol was also applied to the CYP101D1:Arx complex. The first three measurements indicate that this heterodimer does not have a similar orientation to CYP199A2:HaPux, P450cam:Pdx, or P450scc:Adx. P450cam was fused to putidatredoxin reductase (PdR) to explore the kinetic effects with a view to improving electron transfer to orphan P450s. Heme incorporation of this enzyme depends on linker length. In whole cells, the fusion was more active after longer incubations. In vitro kinetics of the fusion exhibited some co-operativity and enhanced kinetics over the unfused system under steady-state conditions. The putative iron-sulfur biosynthesis ferredoxin PuxB had been engineered by rational mutagenesis to support catalysis by CYP199A2. It was confirmed this arose from improved protein-protein recognition. Engineering of E. coli ferredoxin based on these findings was carried out, resulting in electron-transfer to CYP199A4 from a novel engineered alien ferredoxin.