Thesis (M.Phil.) - University of Queensland, 2006. / Includes bibliography.

Thesis (M.Ch.E.)--University of Delaware, 2005. / Principal faculty advisor: Abraham M. Lenhoff, Dept. of Chemical Engineering. Includes bibliographical references.

Thesis (M. Phil.)--University of Hong Kong, 2006. / Restricted as requested by author. Restricted access for 1 year 2007-03-31. Also available online.

Thesis (Ph. D.)--Ohio State University, 2006. / Title from first page of PDF file. Includes bibliographical references (p. 140-154).

Jena, Univ., Diss., 2005

Techn. Hochsch., Diss., 2005--Aachen.

Protein-protein interactions modulate cellular processes in health and disease. However, tracing weak or rare associations or dissociations of proteins is not a trivial task. Kinases are often regulated through interaction partners and, at the same time, themselves regulate cellular interaction networks. The use of kinase domains for creating a synthetic sensor device that reads low concentration protein-protein interactions and amplifies them to a higher concentration interaction which is then translated into a FRET (Fluorescence Resonance Energy Transfer) signal is here proposed. To this end, DNA constructs for interaction amplification (split kinases), positive controls (intact kinase domains), scaffolding proteins and phosphopeptide - SH2-domain modules for the reading of kinase activity were assembled and expression protocols for fusion proteins containing Lyn, Src, and Fak kinase domains in bacterial and in cell-free systems were optimized. Also, two non-overlapping methods for measuring the kinase activity of these proteins were stablished and, finally, a protein-fragment complementation assay with the split-kinase constructs was tested. In conclusion, it has been demonstrated that features such as codon optimization, vector design and expression conditions have an impact on the expression yield and activity of kinase-based proteins. Furthermore, it has been found that the defined PURE cell-free system is insufficient for the active expression of catalytic kinase domains. In contrast, the bacterial co-expression with phosphatases produced active kinase fusion proteins for two out of the three tested Tyrosine kinase domains.

Late SV40 factor (LSF) is a CP2 family transcription factor involved in cell cycle regulation. In liver cancer, LSF is an oncogene, in part due to its role in upregulation of osteopontin leading to increase tumor size. As a result, LSF is a potential target for drug discovery. LSF binds the p65 subunit of the transcription factor NFkB and also the transcription factor ying yang 1 (YY1). In this thesis, I show that binding of both YY1 and p65 occurs at the ubiquitin-like domain of LSF in U2OS cell extracts. Interestingly, when phosphatase inhibitors are added during preparation of U2OS cell extracts, the binding of YY1 and p65 to LSF shifts from the ubiquitin-like domain of LSF to the DNA binding domain. The role of a yet unidentified docking protein may be responsible for this shift in binding. In an attempt to map the specific region of the LSF sequence that is involved in these interactions, I have developed a peptide identification assay which utilizes protease digestion, protein mediated peptide capture, and LC ESI-MS. Through the use of this assay, I'm confident that the sequence(s) involved in these LSF protein-protein interactions can be further defined.

Die AAA+ ATPase p97 ist ein essenzielles Protein, das an einer Vielzahl zellulärer Prozesse beteiligt ist und eine Schlüsselrolle in der Protein-Homöostase spielt. Die funktionale Diversität von p97 beruht auf der Interaktion zahlreicher unterschiedlicher Kofaktoren, die vorwiegend an die N-Domäne von p97 binden. Aufgrund seiner Bedeutung in der Regulierung diverser physiologischer und pathologischer Prozesse stellt p97 eine interessante Zielstruktur für die Entwicklung neuer Wirkstoffe dar, die insbesondere in der Krebstherapie von Bedeutung sein könnte. Bekannte p97-Inhibitoren greifen vor allem die ATPase-Funktion des Proteins an. Ein neuer pharmakologischer Ansatz stellt die Inhibierung der Kofaktorbindung an die N-Domäne dar. Ein solcher Protein-Protein-Interaktionsinhibitor wäre nicht nur von therapeutischem Interesse, sondern hätte auch einen besonderen Nutzen für die Entschlüsselung molekularer und zellulärer Funktionen von p97-Kofaktoren. In dieser Arbeit wurde ein fragmentbasierter Ansatz für die Identifizierung von chemischen Startstrukturen für die Entwicklung eines Protein-Protein- Interaktionsinhibitors verfolgt. Als Zielstruktur wurde die SHP-Bindestelle in der N-Domäne gewählt. Die Identifizierung von Liganden erfolgte sowohl durch computergestützte Methoden (insbesondere virtuelles Screening und Molekulardynamik-Simulationen) als auch experimentell durch biophysikalische Techniken (wie Biolayer-Interferometrie, Röntgenstrukturanalyse und ligandbasierte NMR-Techniken). Die Grundlage des computerbasierten Designs stellte eine Analyse der bekannten Kristallstrukturen der p97-Komplexe mit den SHP-Motiven der Kofaktoren UFD1 und Derlin-1 dar. Darüber hinaus dienten Molekulardynamik-Simulationen der Analyse der Wassereigenschaften innerhalb der SHP-Bindestelle. Darauf aufbauend wurden verschiedene Pharmakophormodelle entwickelt, die die Grundlage des im Anschluss durchgeführten virtuellen Screenings und Dockings bildeten. Anhand der Ergebnisse von Molekulardynamik-Simulationen wurden zehn Verbindungen für die experimentelle Validierung ausgewählt. Hiervon konnten zwei Fragmente in STD-NMR- und Biolayer-Interferometrie-Experimenten als Liganden bestätigt werden. In einem parallel durchgeführten biophysikalischen Fragmentscreening mittels Biolayer-Interferometrie wurden unter mehr als 650 Verbindungen 22 identifiziert, die an die N-Domäne binden. 15 dieser Fragmente wurden durch einen orthogonalen STD-NMR-Assay bestätigt. Fünf dieser Verbindungen zeigten Affinitäten mit KD-Werten kleiner 500μMund günstigen Ligandeffizienzen. Des Weiteren konnte die Bindungskinetik und Affinität des in der Literatur als p97-Inhibitor berichteten Naturstoffes Xanthohumol bestimmt und eine Bindung an die N-Domäne bestätigt werden. Zur Identifizierung möglicher Bindestellen dieser fünf Fragmente wurden mixed-solvent Molekulardynamik-Simulationen durchgeführt. Diese ergaben, dass alle Verbindungen die SHP-Bindestelle in der N-Domäne adressieren. Die Regionen fielen mit hot spots der Kofaktorwechselwirkungen zusammen und stellen somit mögliche Ankerpunkte für die Weiterentwicklung dar. Für zwei Fragmente konnten die postulierten Bindestellen mittels Röntgenstrukturanalyse bzw. STD-NMR-Messungen an p97-Alanin-Mutanten bestätigt werden. Die erhaltene Röntgenstruktur ist die erste p97-Struktur, die ein gebundenes Fragment an der N-Domäne zeigt. / The AAA+ATPase p97 is an essential protein involved in numerous cellular pro-cesses and plays a key role in multiple aspects of protein homeostasis. Its functio-nal diversity is mediated through the interaction with a large number of distinctcofactors binding to the N-domain of p97. Due to its significant role in regulatinga variety of physiological responses, p97 has emerged as a potential therapeu-tic target. A small molecule inhibiting the cofactor binding would be importantto dissect the molecular and cellular functions of p97 cofactors, thus helping tounravel their specific role in controlling p97 activity. Such compounds may alsoopen routes to new cancer therapies.In this work, a fragment-based approach was pursued for the identification ofchemical starting points for the development of a protein-protein interaction in-hibitor addressing the SHP binding site. Therefore, computer-assisted methods,such as virtual screenings and molecular dynamics simulations, as well as bio-physical techniques including biolayer interferometry, X-ray crystallography, andligand-based NMR techniques, were applied.The computer-based design started with an analysis of the known p97 crystalstructures in complex with the SHP motifs of cofactors UFD1 and Derlin-1. In ad-dition, molecular dynamics simulations were used to analyze the water proper-ties within the SHP binding site. Based on these results, pharmacophore modelswere developed and utilized in the subsequent virtual screening and dockingprocess. With the help of molecular dynamics simulations, ten compounds wereselected for experimental validation. Two of these were confirmed as ligands inSTD-NMR and biolayer interferometry experiments.In parallel, a biophysical fragment screening of over 650 compounds was perfor-med using the biolayer interferometry method. This led to the identification of22 compounds binding to the N-domain. Fifteen of these fragments were con-firmed in an orthogonal STD-NMR assay. Five compounds showed affinities withKDvalues below 500 μM and favourable ligand efficiencies for further optimiza-tion. Furthermore, the binding kinetics and affinity of xanthohumol, a naturalproduct reported in the literature as a p97 inhibitor, were determined and bin-ding to the N-domain was confirmed. xToidentify possible binding sites of these five fragments, mixed solvent mole-cular dynamics simulations were performed. These revealed that all compoundsaddress the SHP binding site in the N-domain. The regions coincide with hotspots of the cofactor binding and, thus, represent potential anchor points for aprotein-protein interaction inhibitor. For two fragments, the postulated bindingsites were confirmed by X-ray crystallography and STD-NMR measurements onp97 alanine mutants, respectively. The X-ray structure obtained is the first p97structure showing a fragment bound to the N-domain.

The recognition of small organic compounds and metabolites is essential for living systems, enabling the cell to sense environmental stimuli and respond appropriately. Developing quantitative models of living systems which can incorporate these environmental stimuli would accordingly benefit from comprehensive mapping of interactions between proteins and small molecules of interest. While high-throughput experimental methods provide a wealth of interaction data, the scale of chemical space currently precludes comprehensive enumeration of protein-compound interaction space. Computational methods can help to bridge this gap by inferring proteome-scale protein-compound interactomes, elucidating structural features within protein families which mediate specificity of binding to specific small molecules, and inferring the affinity of binding for specific protein-compound interactions. In this thesis, we attempt to use, and in some cases develop, methods to study protein-compound interactions at these three scales. First, we describe recent work in extending our structure-based algorithm for predicting protein-compound interactions throughout the proteome to include a wider array of small molecules. We demonstrate that this method performs comparably to existing methods and describe an online database storing the results of this analysis. We also report several case studies illustrating how this database can be used along with cautionary vignettes indicating areas where the method fails and directions for future improvement. We subsequently analyze druggable pockets occurring within protein-protein interfaces (PPIs) to assess whether they are less structurally conserved than analogous pockets of conventional drug sites. We find that PPI interfacial pockets are associated with fewer expected off-targets than conventional drug sites, however that this finding is specific to individual protein families, rather than a general feature of interfacial PPI pockets. Finally, we use Free Energy Perturbation to predict the binding affinity of an array of small volatile odorants with an olfactory receptor from the jumping bristletail, Machilis hrabei, as well as attempt to further optimize the system in order to study the effects of mutating receptor binding site residues on binding affinity to its active ligands.