Ulrike Obst

Molecular recognition at the thrombin active site: structure-based design and synthesis of potent and selective thrombin inhibitors and the X-ray crystal structures of two thrombin-inhibitor complexes

BACKGROUND The serine protease thrombin is central in the processes of hemostasis and thrombosis. To be useful, thrombin inhibitors should combine potency towards thrombin with selectivity towards other related enzymes such as trypsin. We previously reported the structure-based design of thrombin inhibitors with rigid, bicyclic core structures. These compounds were highly active towards thrombin, but showed only modest selectivity. RESULTS Here, we describe the rational design of selective thrombin inhibitors starting from the X-ray crystal structure of the complex between the previously generated lead molecule and thrombin. The lead molecule bound with a Ki value of 90nM and a selectivity of 7.8 for thrombin over trypsin. Our design led to inhibitors with improved activity and greatly enhanced selectivity. The binding mode for two of the new inhibitors was determined by X-ray crystallography of their complexes with thrombin. The results confirmed the structures predicted by molecular modeling and, together with the binding assays, provided profound insight into molecular recognition phenomena at the thrombin active site. CONCLUSIONS A novel class of nonpeptidic, selective thrombin inhibitors has resulted from structure-based design and subsequent improvement of the initial lead molecule. These compounds, which are preorganized for binding to thrombin through a rigid, bicyclic or tricyclic central core, could aid in the development of new antithrombotic drugs. Correlative binding and X-ray structural studies within a series of related, highly preorganized inhibitors, which all prefer similar modes of association to thrombin, generate detailed information on the strength of individual intermolecular bonding interactions and their contribution to the overall free energy of complexation.

Synthesis of Novel Nonpeptidic Thrombin Inhibitors

A novel class of nonpeptidic, active, and selective thrombin inhibitors has resulted from X-ray-structure-based design and subsequent improvement of the initial lead molecules. These inhibitors possess a bi- or tricyclic central core structure with attached side chains to reach the three binding pockets (selectivity S1 pocket, distal D pocket, and proximal P pocket) present in the active site of the enzyme. The key step in the preparation of these compounds is the 1,3-dipolar cycloaddition between an azomethine ylide, prepared in situ by the decarboxylative method from an aromatic aldehyde and an α-amino acid, with an N-substituted maleimide (e.g., see Schemes 1 and 2). All potent inhibitors contain an amidinium residue in the side chain for incorporation into the S1 pocket, which was introduced in the last step of the synthesis by a Pinner reaction. The compounds were tested in biological assays for activity against thrombin and the related serine protease trypsin. The first-generation lead compounds (±)-11 and (±)-19 (Scheme 1) with a bicyclic central scaffold showed Ki values for thrombin inhibition of 18 μM and 0.67 μM, respectively. Conformationally more restricted second-generation analogs (Scheme 2) were more active ((±)-22i: Ki=90 nM (Table 1)); yet the selectivity for thrombin over trypsin remained weak. In the third-generation compounds, a small lipophilic side chain for incorporation into the hydrophobic P pocket was introduced (Schemes 7 and 8). Since this pocket is present in thrombin but not in trypsin, an increase in binding affinity was accompanied by an increase in selectivity for thrombin over trypsin. The most selective inhibitor (Ki=13 nM, 760-fold selectivity for thrombin over trypsin; Table 2) was (±)-1 with an i-Pr group for incorporation into the P pocket. Optical resolution of (±)-1 (Scheme 9) provided (+)-1 with a Ki value of 7 nM and a 740-fold selectivity, whereas (−)-1 was 800-fold less active (Ki=5.6 μM, 21-fold selectivity). The absolute configuration of the stronger-binding enantiomer was assigned based on the X-ray crystal structure of the complex formed between thrombin and this inhibitor. Compound (+)-1 mimics the natural thrombin substrate, fibrinogen, which binds to the enzyme with the Ph group of a phenylalanine (piperonyl in (+)-1) in the distal D pocket, with the i-Pr group of a valine (i-Pr in (+)-1) in the proximal P pocket, and with a guanidinium side chain of an arginine residue (phenylamidinium group in (+)-1) in the selectivity S1 pocket of thrombin. A series of analogs of (±)-1 with the phenylamidinium group replaced by aromatic and aliphatic rings bearing OH or NH2 groups (Schemes 10 – 14) were not effectively bound by thrombin. A number of X-ray crystal-structure analyses of free inhibitors confirmed the high degree of preorganization of these compounds in the unbound state. Since all inhibitors prefer similar modes of association with thrombin, detailed information on the strength of individual intermolecular bonding interactions and their incremental contribution to the overall free energy of complexation was generated in correlative binding and X-ray studies. The present study demonstrates that defined mutations in highly preorganized inhibitors provide an attractive alternative to site-directed mutagenesis in exploring molecular-recognition phenomena at enzyme active sites.

6-Alkoxy-5-aryl-3-pyridinecarboxamides, a New Series of Bioavailable Cannabinoid Receptor Type 1 (CB1) Antagonists Including Peripherally Selective Compounds

We identified 6-alkoxy-5-aryl-3-pyridinecarboxamides as potent CB1 receptor antagonists with high selectivity over CB2 receptors. The series was optimized to reduce lipophilicity compared to rimonabant to achieve peripherally active molecules with minimal central effects. Several compounds that showed high plasma exposures in rats were evaluated in vivo to probe the contribution of central vs peripheral CB1 agonism to metabolic improvement. Both rimonabant and 14g, a potent brain penetrant CB1 receptor antagonist, significantly reduced the rate of body weight gain. However, 14h, a molecule with markedly reduced brain exposure, had no significant effect on body weight. PK studies confirmed similarly high exposure of both 14h and 14g in the periphery but 10-fold lower exposure in the brain for 14h. On the basis of these data, which are consistent with reported effects in tissue-specific CB1 receptor KO mice, we conclude that the metabolic benefits of CB1 receptor antagonists are primarily centrally mediated as originally believed.