Open Access

A necessary requirement for a pharmacological effect is that a drug molecule tightly interacts with its disease relevant target molecule in the patient. Kinases are regulatory, signal transmitting enzymes and are a large protein family that belongs to the most frequent targets of pharmaceutical industry, as deregulation of kinases has been associated with the development of a variety of diseases, including cancer. In drug discovery, equilibrium binding metrics such as the affinity (Ki, KD) or potency (IC50, EC50) are usually applied for the systematic profiling for potent and selective drug candidates. In recent years, dynamic binding parameters, the drugs association (kon) and dissociation (koff) rates for desired primary-targets and undesired off-targets, were discussed to be better predictors than steady-state affinity per se (KD = koff / kon) for the onset and duration of the drug-target complex in the open in vivo environment and thereby for the therapeutic effect and safety of the drug. It is yet unclear whether and when the binding kinetics parameters can influence drug action in the complex context of pharmacokinetics and pharmacodynamics and how the kinetic rate constants can be optimized rationally. One major obstacle for providing proof for the hypothesis that drug binding kinetics is of importance for drug action is the generation of large and comparable binding kinetic datasets. The aim of this thesis was the comprehensive analysis of the binding kinetic and affinity parameters of a diverse spectrum of 270 small-molecule kinase inhibitors against a panel of pharmacologically relevant kinases to study the role played by binding kinetics for drug discovery: The generated dataset was utilized to assess the effect of chemical properties on drug binding kinetics, and to evaluate the impact of kinetic rate constants on the success of compounds in the drug discovery pipeline. Large scale profiling was made possible by a recently developed “kinetic Probe Competition Assay” (kPCA), whose evaluation is based on Motulsky’s and Mahan’s “kinetics of competitive binding” theory. Monte Carlo analyses performed in this dissertation widened the theoretical knowledge of this theory, provided new insights into its limitations and allowed to derive recommendations about how to best design assays. It was demonstrated that kPCA is indeed high-throughput compatible and that it is comparable to other biochemical and biophysical assay formats in terms of precision and accuracy. Multivariable linear regression for the description of the determined kinase inhibitors’ target binding characteristics (kon or koff or KD) using molecular properties and/or particular kinase-inhibitor interactions as descriptors supported the assumption that molecular properties of compounds might affect binding kinetics, generated new hypothesis about molecular determinants influencing binding kinetic parameters and provided a rational basis for following structure-kinetic relationship studies. Remarkably, the binding kinetic rate constants were better described by the established models than binding affinities. Interestingly, the systematic, quantitative analysis of kinase inhibitors’ target binding kinetics indicated that a slow dissociation rate for the main target is a feature which is more frequently observed in inhibitors that reached approval or late stage clinical testing than in earlier phases of clinical development. In addition, it was demonstrated that binding kinetics of kinase inhibitors is a better predictor for the time course of target engagement in cells as compared to affinity per se. Furthermore, in some study cases simulations using a standard pharmacokinetics model and a modified model considering the inhibitors binding kinetics lead to different in vivo kinase occupancy time profiles. It was illustrated by simulations how the concept of kinetic selectivity can be applied to turn an unselective compound in equilibrium conditions into a more selective compound in the open in vivo situation, where the thermodynamic equilibrium of drug-target binding is not necessarily reached. Thus the generated data and models provide evidence for the importance of binding kinetics in drug discovery and represent a valuable resource for future studies in this field.

The p38α mitogen-activated protein kinase (MAPK) is activated through stress stimuli such as heat shock or hypoxia. In the nucleus, p38α modulates the activity of other kinases and transcription factors, a process that regulates the expression of specific target genes, most importantly pro-inflammatory cytokines. Dysregulation of p38α therefore plays a major role in the development of inflammatory diseases such as rheumatoid arthritis. Despite many years of intensive research, no p38 small-molecule inhibitors have been approved yet. Several inhibitor design strategies have been reported, leading to >100-fold selective compounds for α/β over the γ and δ isoforms. Achieving such a selectivity among the two structurally most related α and β isoforms, however, remains a challenging task. Targeting an inactive DFG-out conformation offers another strategy for the development of potent kinase inhibitors (type-II), exemplified by the BCR/ABL-inhibitor Imatinib. Achieving selectivity with type-II binders is challenging, because many kinases can adopt an inactive DFG-out conformation. This is exemplified by the p38 type-II inhibitor BIRB-796, which exhibits picomolar on-target affinity but only a poor kinome-wide selectivity. A potent and selective type-II chemical probe for p38α/β was still lacking at the start of this thesis. The promising hit VPC-00628, was chosen for a combinatorial synthetic approach to develop a type-II chemical probe. The studies covered the optimization of the hinge-binding head group, the hydrophobic region I and the DFG-out deep pocket of the lead compound VPC-00628. Selectivity for the p38α and p38β isoforms was monitored during the optimization process, which identified several inhibitors with favorable isoform selectivity, providing valuable insights into the potential of isoform-selective inhibitor design for p38. A potent and highly selective p38 MAPK probe (SR-318) was discovered, which showed IC50 values in the low nanomolar range in HEK293T cells. An unusual P-loop conformation induced upon binding of SR-318 to p38α contributed most likely to the impressive selectivity profile within the kinome that surpassed both the parent compound and BIRB-796. A negative control compound, SR-321, was developed, to distinguish between on-target effects and non-specific effects due to cross-reactivity with other cellular proteins. Studies of the metabolic stability in human liver microsomes revealed a high stability of the compounds, with only a small amount of metabolites formed over several hours. Compound SR-318 also exhibited a good in vitro efficacy, quantitatively reducing the LPS-stimulated TNF-α release in whole blood. Taken together, SR-318 is a highly potent and selective type-II p38α/β chemical probe, which will help to gain a better understanding of the catalytic and non-catalytic functions of these key signaling kinases in physiology and pathology. The next studies focused on the exploration of the highly dynamic allosteric back pocket of p38 MAPK, and allosteric BIRB-796 derived compounds for targeting the αC- and DFG-out pockets were synthesized. Kinase activities of allosteric pyrazole-urea fragments were analyzed against a comprehensive set of 47 diverse kinases by differential scanning fluorimetry (DSF), revealing that BIRB-796 off-targets remain a problem when targeting this back-pocket binding motif. Revisiting the recently published compound MCP-081, which combines the allosteric part of BIRB-796 with the active-site directed part of VPC-00628, showed that it displays a clean selectivity profile in our kinase panel. Because the potency of MCP-081 was slightly reduced compared with VPC-00628 and the allosteric tert-butyl pyrazole moiety seemed suboptimal, a set of VPC-00628 derivatives for targeting the αC-out pocket region was synthesized. Through structure-guided extension of the terminal amide of VPC-00628 toward this allosteric site, the potent and selective compound SR-43 was developed, which showed excellent cellular activity on p38 MAPK in NanoBRETTM assays (IC50 [p38α/β] = 14.0 ± 0.1/ 16.8 ± 0.1 nM). SR-43 showed a dose-dependent inhibition of activating phosphorylation of p38 in HCT-15 cells as well as inhibition of phosphorylation of p38 downstream substrates MK2 and Hsp27. In addition, SR-43 induced an anti-inflammatory response by blocking TNF-α release in whole blood and displayed a high metabolic stability. Selectivity profiling of SR-43 revealed a narrow selectivity for additional targets such as the discoidin domain receptor kinases (DDR1/2). DDR kinases play a central role in fibrotic disorders, such as renal and pulmonale fibrosis, atherosclerosis and different forms of cancer. Since selective and potent inhibitors for these important therapeutic targets are largely lacking and the existing inhibitors are of low scaffold diversity, the next study focused on the optimization of SR-43 toward DDR1/2 kinase inhibition. The synthetic work covered the optimization of the hinge-binding head group and the allosteric part of SR-43 toward DDR1/2 kinase inhibition. These studies provided novel insights into the P-loop folding process of p38 MAPK and how targeting of non-conserved amino acids affects inhibitor selectivity. Importantly, they led to the development of a selective dual DDR/p38 inhibitor probe, SR-302, with picomolar affinity for DDR2. SR-302 was efficient in vitro and showed a destabilizing effect on the surface adhesion protein E-cadherin in epithelial cells. In summary, SR-302 and its negative control SR-301 provide a valuable tool set for studying the phenotypic effects of DDR1/2 signaling, e.g., in cancer cell lines.

Guanosine triphosphate (GTP) cyclohydrolase I (GCH1) catalyzes the conversion of GTP to dihydroneopterin triphosphate (H2NTP), the initiating step in the biosynthesis of tetrahydrobiopterin (BH4). Besides other roles, BH4 functions as cofactor in neurotransmitter biosynthesis. The BH4 biosynthetic pathway and GCH1 have been identified as promising targets to treat pain disorders in patients. The function of mammalian GCH1s is regulated by a metabolic sensing mechanism involving a regulator protein, GCH1 feedback regulatory protein (GFRP). GFRP binds to GCH1 to form inhibited or activated complexes dependent on availability of cofactor ligands, BH4 and phenylalanine, respectively. We determined high-resolution structures of human GCH1−GFRP complexes by cryoelectron microscopy (cryo-EM). Cryo-EM revealed structural flexibility of specific and relevant surface lining loops, which previously was not detected by X-ray crystallography due to crystal packing effects. Further, we studied allosteric regulation of isolated GCH1 by X-ray crystallography. Using the combined structural information, we are able to obtain a comprehensive picture of the mechanism of allosteric regulation. Local rearrangements in the allosteric pocket upon BH4 binding result in drastic changes in the quaternary structure of the enzyme, leading to a more compact, tense form of the inhibited protein, and translocate to the active site, leading to an open, more flexible structure of its surroundings. Inhibition of the enzymatic activity is not a result of hindrance of substrate binding, but rather a consequence of accelerated substrate binding kinetics as shown by saturation transfer difference NMR (STD-NMR) and site-directed mutagenesis. We propose a dissociation rate controlled mechanism of allosteric, noncompetitive inhibition.

The aim of this work was to establish a new way of predicting novel dual active compounds by combining classical fingerprint representation with state-of-the-art machine learning algorithms. Advantages and disadvantages of the applied 2D- and 3D-fingerprints were investigated. Further, the impact of various machine learning algorithms was analyzed. The new method developed in this work was used to predict compounds, which inhibit two different targets (LTA4H and sEH) involved in the same disease pattern (inflammation). The development of multitarget drugs has become more important in recent years. Many widespread diseases like metabolic syndrome, or cancer are of a multifactorial nature, which makes them hard to be treated effectively with a single drug. The new in silico method presented in this work can help to accelerate the design and development of multitarget drugs, saving time and efforts. The nowadays readily available access to a large number of 3D-structures of biological targets and published activity data of millions of synthesized compounds enabled this study and was used as a starting point for this work. Four different data sets were compiled (crystalized ligands from the PDB, active and inactive compounds from ChEMBL23, newly designed compounds using a combinatorial library). Those data sets were collected and processed using an automated KNIME workflow. This automation has the advantage of allowing easy change and update of compound sources and adapted processing ways. In a next step, the compounds from the compiled data sets were represented using a variety of well-established 2D- and 3D-fingerprints (PLIF, AtomPair, Morgan, FeatMorgan, MACCS). All those fingerprints share the same underlying bit string scheme but vary in the way they describe the molecular structure. Especially the difference between 2D- and 3D-fingerprints was investigated. 2D-fingerprints are solely based on ligand information. 3D-fingerprints, on the other hand, are based on X-ray structure information of protein-ligand complexes. One major difference between 2D- and 3D-fingerprints usage is the need for a 3D-conformation (pose) of the compound in the targets of interest when using 3D-fingerprints. This additional step is time-consuming and brings further uncertainties to the method. Based on the calculated fingerprints state-of-the-art machine learning algorithms (SVC, RF, XGB and ADA) were used to predict novel dual active compounds. The models were evaluated by 10-fold cross validation and accuracy as the primary measure of model performance was maximized. Second, individual parameters of the four machine learning algorithms were optimized in a grid search to achieve maximal accuracy using the optimized partitioning scheme. Overall accuracies, regardless of fingerprint and machine learning algorithm, are slightly better for LTA4H than for sEH. The goal to predict dual active compounds was realized by comparing the set of predicted to be active compounds for LTA4H and sEH. For the 3D-fingerprint PLIF the machine learning algorithm Random Forest was chosen, from which compounds for synthesis and testing were selected. Of 115 predicted to be active compounds, six compounds were cherry picked. Two compounds showed very good/moderate dual inhibitory activity. Of the 2D-fingerprints, the AtomPair fingerprint in combination with the machine learning algorithm Random Forest was chosen from which compounds were selected for synthesis and testing. 116 compounds were predicted to be dual active against LTA4H and sEH. One of those compounds showed good dual inhibitory activity. In this work it was possible to show advantages and disadvantages of using 2D- and 3D-fingerprints in combination with machine learning algorithms. Both strategies (2D: ligand-based, 3D: structure-based) lead to the prediction of novel dual active compounds with moderate to very good inhibitory activity. The method developed in this work is able to predict dual active compounds with very good inhibitory activity and novel (previously unknown) scaffolds inhibiting the targets LTA4H and sEH. This contribution to in silico drug design is promising and can be used for the prediction of novel dual active compounds. Those compounds can further be optimized regarding binding affinity, solubility and further pharmacological and physicochemical properties.

In der hier vorgestellten Arbeit wurde ein strukturbasierter Ansatz zur gezielten Leitstrukturoptimierung entwickelt. Die Grundlage dafür bildeten die sogenannten Matched Molecular Pairs (MMPs). Dabei handelt es sich um Paare von Molekülen, welche sich lediglich in einer wohldefinierten Modifikation (Transformation) unterscheiden und sich in einer Datenbank mit gemessenen Moleküleigenschaften befinden. Diese Transformationen wurden im Kontext ihrer Targetumgebung untersucht und eine mathematische Beziehung zwischen Transformation und dem Effekt auf die Bindungsaffinität (Transformationseffekt) hergestellt. Auf Basis der generierten Datengrundlage wurde anschließend ein Webserver zur gezielten Leitstrukturoptimierung implementiert und zur freien Nutzung zur Verfügung gestellt.