Open Access

This study focuses on structural features of a particular GPCR type, the family C GPCRs. Structure- and ligand-based approaches were adopted for prediction of novel mGluR5 binding ligand and their binding modes. The objectives of this study were: 1. An analysis of function and structural implication of amino acids in the TM region of family C GPCRs. 2. The prediction of the TM domain structure of mGluR5. 3. The discovery of novel selective allosteric modulators of mGluR5 by virtual screening. 4. The prediction of a ligand binding mode for the allosteric binding site in mGluR5. GPCRs are a super-family of structurally related proteins although their primary amino acid sequence can be diverse. Using sequence information a conservation analysis of family C GPCRs should be applied to reveal characteristic differences and similarities with respect function, folding and ligand binding. Using experimental data and conservation analysis the allosteric binding site of mGluR5 should be characterized regarding NAM and PAM and selective ligand binding. For further evaluation experimental knowledge about family A GPCRs as well as conservation between vertebrate rhodopsins was planned to be compared to results obtained for family C GPCRs (Section 4.1 Conservation analysis of family C GPCRs). Since no receptor structure is available for any family C GPCR, discussion of conserved sequence positions between family A and C GPCRs requires the prediction of a receptor structure for mGluR5 using a family A receptor as template. In order to predict the mGluR5 structure a sequence alignment to a GPCR template protein will have to be proposed and GPCR specific features considered in structure calculation (Section 4.1.4 Structure prediction of mGluR5). The obtained structure was intended to be involved in ligand binding mode prediction of newly discovered active molecules. For discovery of novel selective mGluR modulators several ligand-based virtual screening protocols were adapted and evaluated. Prediction models were derived for selection of possibly active molecules using a diverse collection of known mGluR binding ligands. For that purpose a data collection of known mGluR binding ligands should be established and this reference collection analyzed with respect to different ligand activity classes, NAM or PAM and selective modulators. The prediction of novel NAMs and PAMs using several combinations of 2D-, 3D-, pharmacophore or molecule shape encoding methods with machine learning techniques and similarity determining methods should be tested in a prospective manner (Section 4.2 Virtual screening for novel mGluR modulators). In collaboration with Merz Pharmaceuticals (Merz GmbH & Co. KGaA, Frankfurt am Main, Germany) the modulating effect of a few hundred molecules should be approved in a functional cell-based assay. With the objective to predict a binding mode of the discovered active molecules, molecule docking should be applied using the allosteric binding site of the modeled mGluR5 structure (Section 4.2.4 Modeling of binding modes). Predicted ligand binding modes are to be correlated to conservation profiles that had resulted from the sequence-based entropy analysis and information from mutation experiments, and shall be compared to known ligand binding poses from crystal structures of family A GPCRs.

Die Identifizierung neuartiger Verbindungsklassen für ein pharmakologisches Zielsystem ist eine fordernde Aufgabe für die frühe präklinische Forschung, insbesondere wenn bereits vorherige umfangreiche Studien durchgeführt und viele Leitstrukturserien gefunden wurden. In dieser Arbeit konnte gezeigt werden, dass Scaffold Hopping durch Methoden des Virtual Screenings auch für Systeme möglich ist, für die bereits eine Vielzahl von Referenzsubstanzen beschrieben ist und somit wenig freier chemischer Raum für Innovation zur Verfügung steht. Als Beispielsystem wurde die GlycinB-Bindungsstelle der NR1-Untereinheit des NMDA-Rezeptors betrachtet. Verschiedene zwei- und dreidimensionale Techniken des Virtual Screenings wurden einer umfangreichen retrospektiven Validierung unterworfen. Zur Durchführung der prospektiven Virtual-Screening-Studie wurde eine automatisierte in silico Plattform entwickelt, die 8,9 Millionen käufliche Substanzen aus 46 Substanzkatalogen von 33 verschiedenen Anbietern sammelte, um etwa 5 Millionen unterschiedliche Moleküle in zweidimensionaler Darstellung aufzuarbeiten. Diese Menge an Substanzen stellt den größten Teil der zurzeit kommerziell verfügbaren chemischen Verbindungen, also den „verfügbaren chemischen Raum“ dar. Anhand der retrospektiv validierten Virtual Screening Techniken konnten in einer prospektiven Suche 21 GlycinB-Antagonisten mit neuartigen, d.h. für GlycinB noch unbeschriebenen Scaffolds gefunden werden. Ausgehend von drei dieser Virtual Screening Hits wurden 53 weitere Verbindungen mit insgesamt fünf unterschiedlichen neuartigen Scaffolds und einem gemeinsamen Azo-Motiv identifiziert. Die Struktur-Wirkungsbeziehungen dieser fünf chemischen Serien wurden charakterisiert. Das Ergebnis dieser Arbeit zeigt eindeutig, dass es lohnend ist, alle vorhandenen Methoden auszuschöpfen, da sich die validierten Methoden komplementär zueinander verhielten und kein Virtual Screening Hit von mehr als einer Technik gefunden wurde. Die Flexibilität von Proteinen als Antwort auf die Bindung unterschiedlicher Liganden stellt ein bislang ungelöstes chemieinformatisches Problem dar, welches auch grundlegende pharmakologische Bedeutung hat. So verursachen z.B. bei NMDA/GlycinB agonistische Liganden eine Konformationsänderung des Rezeptors. Diese ruft dann eine direkte funktionale Antwort in Form der Öffnung des Ionenkanals hervor. Auch der Bindungsmodus der Antagonisten von GlycinB ist trotz Vorhandenseins von zwei Kristallstrukturen und mehreren Hundert zum Teil hochaffiner Referenzstrukturen zum großen Teil ungeklärt. Im zweiten Teil dieser Arbeit wurde ein auf Moleküldynamiksimulationen basierendes Verfahren entwickelt, welches flexible Aminosäurereste im Rezeptor und damit induzierbare Bewegungen des Proteinrückgrates bestimmt. Die so identifizierten Reste wurden dann in einem erweiterten Verfahren des Induced-Fit-Dockings als explizit flexibel betrachtet. Hierdurch war die Berechnung verschiedener Bindungsmodi von Antagonisten möglich, die aufgrund ihrer Form und Größe nicht in die verfügbaren Kristallstrukturen von GlycinB passten. Diese benötigten somit einen Induced-Fit-Effekt des Rezeptors, um eine Bindung einzugehen. Für die im ersten Teil dieser Arbeit identifizierten Azo-Liganden wurde auf Basis dieser Methode ein gemeinsamer Bindungsmodus vorgeschlagen. Ebenso konnte anhand der Methodik eine Aussage über die funktionale Auswirkung der Proteinflexibilität beim Übergang vom antagonistischen zum agonistischen Rezeptorzustand von GlycinB getroffen werden. Ein großes Problem aktueller Dockingverfahren ist die mangelnde Verfügbarkeit von Scoringfunktionen, welche die tatsächliche biologische Bindungsaffinität eines Liganden berechnen. Hier wurde ein Verfahren für das Zielsystem GlycinB gezeigt, welches aufgrund der Berechnung des thermodynamischen Entropie- und Enthalpiegewinns durch Verdrängung von hydrophob eingeschlossenen Wasser aus der Bindungsstelle durch den Liganden eine Aussage über dessen zu erwartende Bindungsaffinität trifft. Dieses neuartige Scoringsystem wurde auf die im Virtual Screening identifizierten Serie von Azo-Liganden angewandt und verfügte über eine im Vergleich zu klassischen Scoringfunktionen des Molecular Dockings verbesserte Vorhersagekraft der biologischen Bindungsaffinität.

The goal of this thesis was to gain further insight into the binding behavior of ligands in the heptahelical domain (HD) of group I metabotropic glutamate receptors (mGluRs). This was realized by the establishment of strategies for the detection and optimization of molecules acting as non-competitive antagonists of group I mGluRs (mGluR1/5). These strategies should guarantee high diversity in the retrieved chemotypes of the detected compounds not resembling original reference molecules (“scaffold-hopping”). The detection of new scaffolds, in turn, was divided into two approaches: First the development of pharmacological assays to screen compounds at a certain target for bioactivity (here: affinity towards the allosteric recognition site of mGluR1 and mGluR5), and second the evaluation of computer assisted methods for the identification of virtual hits to be screened afterwards on the pharmacological assays established before. Promising molecules should be optimized with respect to activity/affinity and selectivity, their binding mode investigated and, finally, compared to existing lead compounds. Initially, membrane based binding assays for the HD of mGlu1 and mGlu5 receptors with enhanced throughput (shifting from 24-well plates to 96-well plates) were set up. For the mGluR1 assay the potent antagonist EMQMCM exhibited high affinity towards the binding site (Ki ~3nM), which is in accordance with published data from Mabire et al. (functional IC50 3nM). For mGluR5 the reference antagonist MPEP binds with high affinity to the receptor (binding IC50 13.8nM), which confirmed earlier findings from Anderson et al. (binding IC50 15nM). In another series of experiments the properties of rat cerebellar (mGluR1) and corticalmembranes (mGluR5) as well as of radiotracers were investigated by means of binding saturation studies and kinetic experiments. Furthermore, the influence of the solvent DMSO, necessary for compound screening of lipophilic substances, on positive and negative controls was evaluated. As the precise architecture of the HD of mGluR1 is still not known our efforts in identifying new ligands for this receptor focused on the ligand-based approach. All computer assisted methods that were applied to virtually screen large compound collections and to retrieve potential hits (“activity-enriched subsets”) acting at the heptahelical domain of mGluR1 relied on the existence of a valid dataset of reference molecules. This was realized by an initial compilation of a mGluR reference data collection comprising in total 357 entries predominantly negative but also some positive allosteric modulators for mGluR1 and mGluR5. In the next step a pharmacophore model for non-competitive mGluR1 antagonists was constructed. It was based upon six selective, potent and structurally diverse ligands. Prospective virtual screening was performed using the CATS atom-pair descriptor. The Asinex Gold-Collection was screened for each seed compound and some of the most similar compounds (according to the CATS descriptor) were ordered and tested forbinding affinity and functional activity at mGluR1. A high hit rate of approximately 26% (IC50 < 15 micro M) was yielded confirming the applicability of this method. One compound exerted functional activity below one micro molar (IC50-value of C-07:362nM ± 0.03). Moreover, non-linear principal component analysis was employed. Again the Asinex vendor database served as test database and was filtered by the pharmacophore model for mGluR1 established before. Test molecules that were adjacently located with mGluR1 antagonist references were selected. 15 compounds were tested on mGluR1 in binding and functional assays and three of them exhibited functional activity (IC50) below 15 micro M. The most potent molecule P-06 revealed an IC50-value of 1.11 micro M (± 0.41). The COBRA database comprising 5,376 structurally diverse bioactive molecules affecting various targets was encoded with the CATS descriptor and used for training two selforganizing maps (SOM). The encoded mGluR reference data collection was projected onto this map according to the SOM algorithm. This projection allowed to clearly distinguish between antagonists of mGluR1 and mGluR5 subtype. 28 compounds were ordered and tested on activity and affinity for mGluR1. They exhibited functional activity down to the sub-micro molar range (IC50-value of S-08: 744nM ± 0.29) yielding a final hit rate of 46% (<15 micro M). Then, the Asinex collection was screened using the SOM approach. For a predicted target panel including the muscarinic mACh (M1) receptor, the histamine H1-receptor and the dopamine D2/D3 receptors, the tested mGluR ligands exhibited the calculated binding pattern. This virtual screening concept might provide a basis for early recognition of potential sideeffects in lead discovery. We superimposed a set of 39 quinoline derivatives as non-competitive mGluR1 antagonists that were recently published by Mabire and co-workers. A CoMFA model (QSAR) was established and the influence of several side chains on functional activity was investigated. The coumarine derivative C-07 was obtained as a result of similarity searching. Starting from this compound a series of chemical derivatives was synthesized. This led to the discovery of potent (B-28, IC50: 58nM ± 0.008; Ki: 293nM ± 0.022) and selective (rmGluR5 IC50: 28.6 micro M) mGluR1 antagonists. From a homology model of mGluR1 we derived a potential binding mode for coumarines within the allosteric transmembrane region. Potential interacting patterns with amino acids were proposed considering the difference of the binding pockets between rat and human receptors. The proposed binding modes for quinolines (here:EMQMCM) and coumarines (here:B-04) were compared and discussed considering in particular the influence on activity of several side chains of quinolines obtained from the QSAR studies. The present studies demonstrated the applicability of ligand-based virtual screening for non-competitive antagonists of a G-protein coupled receptor, resulting in novel, potent and selective agents.

The aim of the thesis was to identify structure activity relationships (SAR) in the primary screening data of high-throughput screening (HTS) assays. The strategy was to perform a hierarchical clustering of the molecules, assign the primary screening data to the created clusters and derive models from the clusters. The models should serve to identify singletons, clusters enriched with actives, not confirmed hits and false-negatives. Two hierarchical clustering algorithms, NIPALSTREE and hierarchical k-means have been developed and adapted for this purpose, respectively. A graphical user interface (GUI) has been implemented to extract SAR from the clustering results. Retrospective and prospective applications of the clustering approach were performed. SAR models were created by combining the clustering results with different chemoinformatic methods. NIPALSTREE projects a data set onto one dimension using principle component analysis. The data set is sorted according to the scoring vector and split at the median position into two subsets. The algorithm is applied recursively onto the subsets. The hierarchical k-means recursively separates a data set into two clusters using the k-means algorithm. Both algorithms are capable of clustering large data sets with more than a million data points. They were validated and compared to each other on the basis of different structural classes. NIPALSTREE provided with the loading vectors first insights into SAR whereas the hierarchical k-means yielded superior results. A GUI was developed allowing the display of and the navigation in the clustering results. Functionalities were integrated to analyse the clusters in the dendrogram, molecules in a cluster, and physicochemical properties of a molecule. Measures were developed to identify clusters enriched with actives, to characterize singletons and to analyse selectivity and specificity. Different protease inhibitors of the COBRA database were examined using the hierarchical k-means algorithm. Supported by similarity searches and nearest neighbour analyses thrombin inhibitor singletons were quickly isolated and displayed in the dendrogram. By scaling enrichment factors to the logarithm of the dendrogram level, clusters enriched with different structural classes of factor Xa inhibitors were simultaneously identified. The observed co-clustering of other protease inhibitors provided a deeper insight into selectivity and specificity and shows the utility of the approach for constructing focussed screening libraries. Specificity was analyzed by extracting and clustering relative frequencies of the protease inhibitors from the clusters of dendrogram level 7. A unique ligand based point of view on the pocketome of the protease enzymes was obtained. To identify not confirmed hits and false-negatives in the primary screening data of HTS assays, three assays were retrospectively analysed with the hierarchical k-means algorithm. A rule catalogue was developed judging hits in terminal clusters based on the cluster size, the percent control values of the entries in a cluster, the overall hit rate, the hit rate in the cluster and the environment of a cluster in the dendrogram. It resulted in the identification of a high proportion of not confirmed hits and provided for each hit a rating in context of related non-hits. This allows prioritizing compounds for follow-up studies. Non-hits and hits were retrieved from terminal clusters containing hits. Molecules bearing false-negative scaffolds were co-extracted and enriched. To minimize the number of false-positives in the extracted lists, Bayesian regularized artificial neutral network classification models were trained with the data. Applying the models marked improvement of enrichment factors for the false-negatives was obtained. It proofs the scaffold-hopping potential of the approach. NIPALSTREE, the hierarchical k-means algorithm and self-organising maps were prospectively applied to identify novel lead candidates for dopamine D3 receptors. Compounds with novel scaffolds and low nanomolar binding affinity (65 nM, compound 42) were identified. To provide a deeper insight into the SAR of these molecules, different alternative computational methods were employed. Support vector-based regression and partial least squares were examined. Predictive models for dopamine D2 and D3 receptor binding affinity values were obtained. Important features explaining SAR were extracted from the models. The prospective application of the models to the diverse and novel virtual screening data was of limited success only. Docking studies were performed using a homology model of the dopamine D3 receptor. The visual inspection of the binding modes resulted in the hypothesis of two alternative binding pockets for the aryl moiety of dopamine D3 receptor antagonists. A pharmacophore model was created simultaneously requiring both aryl moieties. Virtual screening with the model identified a nanomolar hit (65 nM, compound 59) corroborating the hypothesis of the two binding pockets and providing a new lead structure for dopamine D3 receptors. The presented data shows that the combined approach of hierarchically clustering a data set in combination with the subsequent usage of the clusters for model generation is suited to extract SAR from screening data. The models are successful in identifying singletons, clusters enriched with actives, not confirmed hits and false-negative scaffolds.

Das Ziel des adaptiven Entwurfs von Substanzbibliotheken ist es, die vollständige biologische Testung einer molekularen Screeningbibliothek zu vermeiden. Stattdessen erfolgt, geleitet durch Optimierungsalgorithmen, eine "intelligente" Navigation durch den chemischen Raum, um so bevorzugt Substanzen mit gewünschten Eigenschaften auszuwählen. In einer retrospektiven Studie wurden die Optimierungsalgorithmen "Zufallssuche", "Simulated Annealing", "Evolutionsstrategie" und "Partikelschwarmoptimierung" im Hinblick auf den Entwurf von Bibliotheken von Serinproteaseinhibitoren systematischen verglichen. Die Gesamtzahl verfügbarer Substanztestungen wurde auf 300 beschränkt, um Laborbedingungen zu simulieren. Als Ergebnis zeigten sich besonders die Evolutionsstrategien für einen Einsatz in einer Niedrigdurchsatzscreening-Kampagne geeignet, da diese effizient mit großen Populationen und wenigen Iterationen arbeiteten. Der zweite Teil dieser Arbeit beschreibt den erfolgreichen Entwurf einer fokussierten Bibliothek von RNA-Liganden. In einer hybriden, prospektiven Optimierungsstudie wurden nach dem Vorbild einer iterativen Niedrigdurchsatzscreening-Kampagne vom Computer vorgeschlagene Moleküle im Labor getestet. Die Substanzen wurden auf Inhibition einer spezifischen molekularen Wechselwirkung im Replikationszyklus von HIV getestet (Tat-TAR-Interaktion). In vier Generationen wurden 9 von 170 untersuchten Verbindungen positiv auf Inhibition der Tat-TAR-Interaktion getestet (Trefferquote: 5,3%), wobei lediglich 0,089% der Verbindungen der Screeningbibliothek untersucht wurden. Die zwei potentesten Kandidaten wiesen einen IC50 von 51 uM bzw. 116 uM auf.