Mihiret T. Sisay

Identification of the first low-molecular-weight inhibitors of matriptase-2.

As recently discovered, matriptase-2, a type II transmembrane serine protease, plays a crucial role in body iron homeostasis by down-regulating hepcidin expression, which results in increased iron levels. Thus, matriptase-2 represents a novel target for the development of enzyme inhibitors potentially useful for the treatment of systemic iron overload (hemochromatosis). A comparative three-dimensional model of the catalytic domain of matriptase-2 was generated and utilized for structure-based virtual screening in combination with similarity searching and knowledge-based compound design. Two N-protected dipeptide amides containing a 4-amidinobenzylamide as P1 residue (compounds 1 and 3) were identified as the first small molecule inhibitors of matriptase-2 with K(i) values of 170 and 460 nM, respectively. An inhibitor of the closely related protease matriptase (compound 2, K(i) = 220 nM), with more than 50-fold selectivity over matriptase-2, was also identified.

Integrating structure- and ligand-based virtual screening: comparison of individual, parallel, and fused molecular docking and similarity search calculations on multiple targets.

Similarity searching is often used to preselect compounds for docking, thereby decreasing the size of screening databases. However, integrated structure‐ and ligand‐based screening schemes are rare at present. Docking and similarity search calculations using 2D fingerprints were carried out in a comparative manner on nine target enzymes, for which significant numbers of diverse inhibitors could be obtained. In the absence of knowledge‐based docking constraints and target‐directed parameter optimisation, fingerprint searching displayed a clear preference over docking calculations. Alternative combinations of docking and similarity search results were investigated and found to further increase compound recall of individual methods in a number of instances. When the results of similarity searching and docking were combined, parallel selection of candidate compounds from individual rankings was generally superior to rank fusion. We suggest that complementary results from docking and similarity searching can be captured by integrated compound selection schemes.

Structural interpretation of activity cliffs revealed by systematic analysis of structure-activity relationships in analog series.

Discontinuity in structure-activity relationships (SARs) is caused by so-called activity cliffs and represents one of the major caveats in SAR modeling and lead optimization. At activity cliffs, small structural modifications of compounds lead to substantial differences in potency that are essentially unpredictable using quantitative structure-activity relationship (QSAR) methods. In order to better understand SAR discontinuity at the molecular level of detail, we have analyzed different compound series in combinatorial analog graphs and determined substitution patterns that introduce activity cliffs of varying magnitude. So identified SAR determinants were then analyzed on the basis of complex crystal structures to enable a structural interpretation of SAR discontinuity and underlying activity cliffs. In some instances, SAR discontinuity detected within analog series could be well rationalized on the basis of structural data, whereas in others a structural explanation was not possible. This reflects the intrinsic complexity of small molecule SARs and suggests that the analysis of short-range receptor-ligand interactions seen in X-ray structures is insufficient to comprehensively account for SAR discontinuity. However, in other cases, SAR information extracted from ligands was incomplete but could be deduced taking X-ray data into account. Thus, taken together, these findings illustrate the complementarity of ligand-based SAR analysis and structural information.

Allosteric modulation of caspases

Caspases are proteolytic enzymes mainly involved in the induction and execution phases of apoptosis. This type of programmed cell death is an essential regulatory process required to maintain the integrity and homeostasis of multicellular organisms. Inappropriate apoptosis is attributed a key role in many human diseases, including neurodegenerative disorders, ischemic damage, autoimmune diseases and cancer. Allosteric modulation of the function of a protein occurs when the regulatory trigger, such as the binding of a small effector or inhibitor molecule, takes place some distance from the proteins active site. In recent years, several caspases have been identified that possess allosteric sites and binding of small molecule to these sites resulted in the modulation of enzyme activities. Regulation of caspase activity by small molecule allosteric modulators is believed to be of great therapeutic importance. In this review we give brief highlights on recent developments in identifying and characterizing natural and synthetic allosteric inhibitors as well as activators of caspases and discuss their potential in drug discovery and protein engineering.

Inhibition of human leukocyte elastase by brunsvicamides a-C: cyanobacterial cyclic peptides.

Human leukocyte elastase (HLE) belongs to the chymotrypsin family of serine proteases. It has a primary specificity for small aliphatic residues in the P1 position of the substrate. HLE is able to catalyze the cleavage of fibrous elastin, an important extracellular matrix protein with the unique property of elastic recoil. Under normal conditions, the activity of HLE is regulated by endogenous inhibitors but excessive, uncontrolled HLE activity may result in several pathological states, including emphysema, chronic obstructive pulmonary disease, cystic fibrosis and rheumatoid arthritis. Due to its involvement in such pathophysiological processes, HLE has become an important pharmaceutical target. Potent and selective HLE inhibitors can be used to reduce or treat HLE-mediated inflammatory disorders. In recent years, several cyanobacterial secondary metabolites have been identified belonging to structurally novel cyclic peptides and depsipeptides. These natural products possess an attractive molecular architecture with a constrained conformation. They have an increased metabolic stability and display a variety of biological effects. Many of them inhibit enzymes, for example, peptides of the microcystin class and oscillamides B and C were found to inhibit protein phosphatases, while the cyanopeptolins, scyptolin A and B, insulapeptolides A–D, and the anabaenopeptins B and F are inhibitors of elastases. Anabaenopeptins G, H, I, J, and T 10] inhibit carboxypeptiACHTUNGTRENNUNGdase A. Trypsin and chymotrypsin inhibitory activities were reported for the cyanopeptolins A90720A and symplocamide A, respectively. 12] The brunsvicamides A, B and C were isolated from the cyanobacterium Tychonema sp. 14] and are related to the sponge-derived mozamides. They are characterized by six amino acids, five of which form a 19-membered ring structure, closed by an amide bond between the carboxylic group of the C-terminal Phe and the e-amino group of the N-terminal d-Lys. The sixth amino acid is attached to the a-amino group of the d-Lys via a urea moiety. The brunsvicamides, anabaenopeptins, oscillamides, 16] and the nodulapeptins are structurally related; they have a d-Lys-urea motif attached to a terminal amino acid and an N-methylated peptide bond in common, but differ in their amino acid sequence. The remarkable inhibitory properties of brunsvicamides against the tyrosine phosphatase B of Mycobacterium tuberculosis have been reported, however, their protease inhibiting potential remains unexploited. In this study, based on previous findings that several cyclic cyanopeptides inhibit elastases, the brunsvicamides A–C were evaluated as inhibitors of HLE. The concentration-dependent inhibition by brunsvicamide B is presented in Figure 1. The progress curves of the HLE-catalyzed

Ligand-target interaction-based weighting of substructures for virtual screening.

A methodology is introduced to assign energy-based scores to two-dimensional (2D) structural features based on three-dimensional (3D) ligand-target interaction information and utilize interaction-annotated features in virtual screening. Database molecules containing such fragments are assigned cumulative scores that serve as a measure of similarity to active reference compounds. The Interaction Annotated Structural Features (IASF) method is applied to mine five high-throughput screening (HTS) data sets and often identifies more hits than conventional fragment-based similarity searching or ligand-protein docking.

Insights into matriptase-2 substrate binding and inhibition mechanisms by analyzing active-site-mutated variants.

Matriptase-2 (transmembrane protease, serine 6, TMPRSS6; MT2), predominantly expressed in the liver, was initially identified in human and mouse. 2] MT2 belongs to the family of type II transmembrane serine proteases, representing an emerging class of cell surface proteolytic enzymes. Recent findings have revealed a link between mutations in MT2 and iron-refractory iron deficiency anemia. 5] MT2 suppresses the expression of hepcidin, 6] the main regulator of systemic iron homeostasis, through cleavage of the bone morphogenetic protein co-receptor hemojuvelin. Due to the important role of MT2 in iron homeostasis, the enzyme represents a novel target for the development of inhibitors, potentially useful in the treatment of hemochromatosis or beneficial as pharmacological tools. In a previous work, we identified two dipeptide amides as the first low-molecular weight inhibitors of MT2. MT2 shares high sequence similarity with matriptase (membrane-type serine protease 1, MT-SP1, suppressor of tumorigenicity 14), particularly in the catalytic domain which is about 45 % identical. Matriptase is expressed in epithelial cells and involved in tumor pathogenesis. Owing to its potential as a therapeutic target, synthetic inhibitors for matriptase have been described, such as mono/bisbenzamidines and derivatives of the sunflower trypsin inhibitor. In this study, we have investigated the interaction of MT2 with prototype low-molecular weight ligands using site-directed mutagenesis, kinetic analysis and molecular modeling. A peptidic substrate and inhibitors from our previous study were chosen to perform an active site scan of MT2. Our data reveal insights into substrate/inhibitor–enzyme interactions and structural differences in the active sites of MT2 and matriptase. Furthermore, amino acids that enhanced (Phe 665) or reduced (Tyr 712, Asp 785) the affinity of peptide ligands were identified. For the design of potent and low-molecular weight inhibitors, an X-ray crystal structure of the target enzyme is beneficial. Although MT2 was confirmed as a key regulator of iron homeostasis, no crystal structure is currently available, in contrast to the already well-characterized matriptase. As such, the previously generated model of the MT2 active site, based on the crystal structure of the closely related matriptase, was used in the present study. The catalytic domain of MT2 features the common fold of chymotrypsin-like serine proteases. In the active site, MT2 and matriptase vary in relevant amino acids participating in the formation of the S1 and S2 pockets and the S3/S4 binding region. To obtain insights into these structural differences between MT2 and matriptase and the resulting enzyme–ligand interactions, four important MT2 amino acids (His 665, Glu 712, Ala 757, Leu 785; corresponding to His 99, Glu 146, Ala 190, Leu 217 according to the chymotrypsinogen numbering) were mutated to the equivalent residues in matriptase (Figure 1).

Hit expansion through computational selectivity searching.

Finding small molecules that selectively interact with individual target proteins within target families is a major task in medicinal chemistry and chemical biology. Currently, the identification of suitable small molecules relies primarily on the screening of diverse or specialized compound libraries. In contrast, computational methods have thus far contributed only very little to the identification of molecules that are target selective or that have a selective tendency. We have previously adapted ligand-based computational screening methods for selectivity searching and we now apply these methodologies in the search for cathepsin-K-selective inhibitors. Herein we report the identification of two nonpeptidic cathepsin K inhibitors that are weakly selective over cathepsins L and S. These inhibitors were identified by assaying only 16 candidate molecules taken from ~3.7 million virtually formatted database compounds. One inhibitor lacks an electrophilic “warhead” that is usually a prerequisite for the inhibition of cathepsins. In virtual screening, computational methods are applied to search large databases for compounds that have a desired biological activity using ligand and/or target structure information as input. However, target selectivity has thus far not been explicitly considered in computational compound screening. We recently evaluated computational approaches for their ability to recognize target-selective compounds. The results of these studies suggest that computational screening models may also have the potential to recognize target-selective compounds, and this has triggered our current investigation. Cathepsins K, S, and L are cysteine proteases belonging to the papain superfamily. These closely related enzymes are involved in important physiological processes such as antigen presentation, bone remodeling, and apoptosis. 11] Accordingly, cathepsins have become relevant drug targets for the potential treatment of various diseases including cancer, osteoporosis, rheumatoid arthritis, and autoimmune disorders. Cathepsin inhibitors are typically substrate analogues with an electrophilic warhead. First-generation inhibitors contain a strongly reactive group that covalently modifies the active site cysteine. More recently, second-generation reversible covalent inhibitors have also been introduced with a less-reactive functional group, typically a nitrile, which renders these inhibitors reversible and more desirable for therapeutic applications. Among these cysteine proteases, cathepsin K has attracted particular interest. It is predominantly expressed in osteoclasts that mediate bone resorption and is capable of cleaving native type I collagen, the major component of bone matrix, and other components of bone matrix such as osteopontin and osteonectin. Because bone matrix degradation is necessary for osteoclastic bone resorption, cathepsin K constitutes a major therapeutic target for the treatment of osteoporosis and is also implicated in rheumatoid arthritis and osteoarthritis. 11] There have been considerable efforts to develop selective cathepsin K inhibitors because simultaneous inhibition of cathepsins S or L is thought to be associated with unwanted side effects. 14] However, the design proved to be challenging owing to the high degree of structural and mechanistic similarity between the cathepsins. Only recently, the first inhibitors of human cathepsin K with selectivity over cathepsins S and L, balicatib and odanacatib, have proceeded to clinical evaluation. Given their therapeutic potential and the difficulties experienced in the design of selective cathepsin K inhibitors, we considered the search for cathepsin K inhibitors that are selective over both cathepsins S and L a challenging and relevant test case for selectivity searching. For our analysis, we implemented a search protocol that involves two conceptually different in silico methods developed in our laboratory, a compound mapping algorithm termed DynaMAD and a specialized type of molecular fingerprint consisting of compound class characteristic substructures, ACCS-FP. These methods were practically applied here because they have been benchmarked for selectivity searching in computational studies. 9] Other virtual screening methods could also be applied, but have thus far not been evaluated for selectivity searching. Briefly, DynaMAD is designed to map database compounds to activity-specific consensus positions in chemical space representations of stepwise increasing dimensionality, and ACCS-FP is used for search calculations in which fingerprints of reference and database molecules are compared and fingerprint overlap is quantified as a measure of molecular similarity. Further methodological details including ACCS-FP generation and details of the search calculations are provided in the Supporting Information. The computational screening methods applied here extrapolate from known ligand information in order to identify structurally diverse compounds with desired properties. Thus, the design of compound reference sets is generally very important for the outcome of the search calculations. For the assembly of a reference set for selectivity searching, we used differential [a] D. Stumpfe, Dr. J. Batista, Dr. I. Vogt, Prof Dr. J. Bajorath Department of Life Science Informatics Bonn–Aachen International Center for Information Technology Rheinische Friedrich-Wilhelms-Universit t Bonn Dahlmannstr. 2, 53113 Bonn (Germany) Fax: (+ 49) 228-2699-341 E-mail : bajorath@bit.uni-bonn.de [b] M. Frizler, M. T. Sisay, Prof Dr. M. G tschow Pharmazeutisches Institut, Pharmazeutische Chemie I Rheinische Friedrich-Wilhelms-Universit t Bonn An der Immenburg 4, 53121 Bonn (Germany) [] These authors contributed equally to this work. Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/cmdc.200800304.

Inhibitors of cathepsins K and S identified using the DynaMAD virtual screening algorithm.

In ligand-based virtual screening (LBVS), computational methods are applied to extrapolate from known active compounds and to identify structurally diverse small molecules with similar biological activity, an objective often referred to as scaffold or lead hopping. Herein we report the results of a virtual screening project designed to identify inhibitors of cathepsins K and S. Many of the cysteine proteases of the papain-like family are considered attractive drug targets. Among these are cathepsins K and S, which are involved in a variety of physiological processes such as bone remodeling and antigen presentation. 5] Cathepsin K plays a critical role in the osteoclast-mediated degradation of collagen and is predominantly expressed in osteoclasts that mediate bone resorption. It is capable of cleaving native type I collagen and other components of the bone matrix such as osteopontin and osteonectin. Accordingly, cathepsin K has become an attractive target for the development of drugs to treat osteoporosis and other disorders characterized by increased bone resorption. Cathepsin S is expressed by adipocytes and antigen-presenting cells such as macrophages and B cells, and is involved in the control of antigen presentation by the major histocompatibility complex class II (MHC-II). The enzyme selectively degrades the MHC-IIassociated invariant chain, which is a prerequisite for peptide antigen loading and presentation by the MHC-II complex. 11] Therefore, among other therapeutic applications, the inhibition of cathepsin S is thought to provide a promising route to the treatment of autoimmune diseases. A number of efforts have been made to identify inhibitors of cathepsins K and S, including computational screening. Most inhibitors discovered early on contain an electrophilic warhead that reacts with the catalytic cysteine residue, resulting in reversible or irreversible inhibition. 16] In more recent years, a number of potent inhibitors of cathepsins K and S have been identified that contain less reactive electrophilic functionalities and inhibit via reversible, covalent interaction. 5, 14, 18, 19] Among the prominent group of nitrile-based cathepsin inhibitors, recent investigations have revealed different electrophilic reactivity depending on the chemical environment of the cyano group. 21] Moreover, several inhibitors have also been reported that lack an electrophilic group and inhibit cathepsins noncovalently. 22, 23] In a previous study carried out by us, we targeted cathepsin K by combining in-house developed computational screening methods, including mini-fingerprint searching, and compound mapping using the “Dynamic Mapping to Activityclass-specific Descriptor value ranges” (DynaMAD) algorithm. In this study, which was based on a reference set focused on selective cathepsin K inhibitors, two inhibitors with a selective tendency for cathepsin K over cathepsin S were identified. One of these inhibitors contains a nitrile group, whereas the other lacks an electrophilic moiety. The study reported herein is a continuation of virtual screening efforts directed at cathepsin inhibitors, albeit with different focal points. For this project, we are interested in: a) identifying inhibitors of both cathepsins K and S with previously unobserved scaffolds (as starting points for chemical exploration), and b) further evaluating the DynaMAD algorithm. DynaMAD is an unconventional virtual screening method, for which only limited experience in practical applications is available as of yet. The characteristic feature of DynaMAD that sets it apart from other LBVS tools is that the method is designed to navigate molecular descriptor spaces of increasing dimensionality (whereas most compound classification techniques use low-dimensional reference space representations). Figure 1 summarizes the DynaMAD approach. The first step is the assignment of molecular descriptors to so-called dimension extension levels (DEL). The underlying idea is that descriptors that are most responsive to a biological activity represented by a reference set of known active compounds should only adopt very narrow value ranges. Descriptors are scored to account for the reference set specificity of their value ranges (scores range from 100 to 0). A score layer interval of 5 is then applied so that a total of 20 DEL (0–20) are obtained. On the basis of their score, different numbers of descriptors are assigned to each layer, and hence the number of descriptors and the dimensionality of the descriptor reference space increase in a stepwise manner from layer to layer. In the next step, descriptor values are calculated for individual database compounds, and they are mapped to the descriptor value ranges of the reference set at each layer. Only compounds whose values fall into the value range of each descriptor are retained for the next dimension [a] Dr. D. Stumpfe, Dr. I. Vogt, Prof. Dr. J. Bajorath Department of Life Science Informatics, B-IT, LIMES Program Unit Medicinal Chemistry and Chemical Biology Rheinische Friedrich-Wilhelms-Universit t Bonn Dahlmannstr. 2, 53113 Bonn (Germany) Fax: (+ 49) 228-2699-341 E-mail : bajorath@bit.uni-bonn.de [b] M. T. Sisay, M. Frizler, Prof. Dr. M. G tschow Pharmazeutisches Institut, Pharmazeutische Chemie I Rheinische Friedrich-Wilhelms-Universit t Bonn An der Immenburg 4, 53121 Bonn (Germany). [c] Dr. I. Vogt Current address : Chemogenomics Laboratory Research Unit on Biomedical Informatics Municipal Institute of Medical Research, and University Pompeu Fabra Doctor Aiguader 80, 08003 Barcelona (Spain) [] These authors contributed equally to this work. Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/cmdc.200900457.