Research concept
The general theme of our research is focused on the question how small molecule ligands can interfere, act upon or modulate the properties
and function of a protein. Such processes involve the inhibition and/or modulation of enzymes, the activation or blocking of receptors or
receptor complexes or the interference with signaling cascades via the perturbance of protein-ligand or protein-protein recognition interfaces.
Fundamental to all these processes is the specific and selective recognition of mutually interacting biomolecules. Over the years, our research
has been reflected by the following topics:
1. Analysis and classification of structural data of protein-ligand complexes
a. Relibase - an integrated database for protein-ligand complexes
b. Waterbase - an analysis tool to study the influence of water on ligand binding
c. Cavbase - a database to compare binding pockets across proteins
d. Secbase - combining secondary structural information and folding patterns with ligand binding data
e. Analysis and interference of protein-protein interfaces
2. Development of experimental and computational tools for the discovery and design of specific modulators of protein function
a. Development of scoring functions for docking, scoring and affinity prediction
b. Hot-Spot analysis of binding sites
c. Strategies for lead discovery by virtual screening
3. Experimental and computational approaches towards a better understanding of binding thermodynamics, selectivity and
interaction kinetics including to role of water in ligand binding
a. Thermodynamic analysis of protein-ligand binding
b. Site specific mutagenesis among members of a protein family to study selectivity determinants
c. 3D-QSAR analysis and clustering of binding cavities to elucidate affinity and selectivity determinants
4. Design projects to modulate the function of proteins relevant for different therapeutic areas such as infectious diseases or used as
engineered biocatalysts
a. De novo crystal structure determination of proteins, involved in infectious and other diseases, structure-based ligand design
b. Inhibitor design of tRNA guanine transglycosylase: A putative target for Shigellosis therapy
c. Design of inhibitors against family members of serine and aspartyl proteases
d. Structural studies on Shigella-specific pathogenicity factors as a basis for structure-based drug design
e. Design of substrate specificity of biocatalysts, hybrid enzymes, catalysts for click chemistry
5. Crystallographic high-throughput screening for fragments directly on protein crystals to obtain starting points for subsequent drug
development
Home
Prof. Klebe
Group Members
DrugDesign
Research
Publications
Textbook
Instrumentation
Funding
Conferences
Job Offers
Download-Area
School leavers
Pharmacy
Pharm. Chemist
Contact
Imprint
Privacy Policy
We have tackled the above-mentioned projects via a multi-disciplinary approach that involves the following techniques:
1. Crystal structure analysis of proteins, ligands and protein-ligand complexes
2. Development of computational tools for data analysis, docking and ligand design
3. Thermodynamic and interaction kinetic characterization of protein-ligand interactions by micro calorimetry and binding assays
4. Expression, purification and mutagenesis of proteins, enzyme kinetics
5. Chemical synthesis of lead structures
In the following, reference is taken to different research papers (in parenthesis, cf. list of publications linked to the corresponding doi, digital
object identifier) of our group, funding agencies, and research cooperations.
Our research was funded by: DFG, EU, BMBF, ERC, CCDC, Synmikro, LOEWE (Hessen) and different pharma companies. We would
like to gratefully acknowledge all the support over many years.
Top of Page
Home
Prof. Klebe
Group Members
DrugDesign
Research
Publications
Textbook
Instrumentation
Funding
Conferences
Job Offers
Download-Area
School leavers
Pharmacy
Pharm. Chemist
Contact
Imprint
Privacy Policy
1. Analysis and Classification of Structural Data of Protein-Ligand Complexes
Top of Page
a. Relibase - an integrated database for protein-ligand complexes
Knowledge discovery from the exponentially growing body of structurally characterized protein-ligand complexes as a source of information
in structure-based drug design is one of the major needs in contemporary drug research. Given the requirement for powerful data retrieval,
integration and analysis tools, Relibase was developed as a database system particularly tailored to handle protein-ligand related problems (119,
133, 134, 135, 141). In many of our implemented projects, Relibase has been used as routine tool for data analysis (133, 141) (collaboration:
Cambridge Crystallographic Data Center (CCDC)).
Top of Page
b. Waterbase - an analysis tool to study the influence of water on ligand binding
Biological transformations are performed in aqueous solution, thus water is present as intimate partner in all ligand-binding processes. In two
third of the known protein-ligand complexes water molecules mediate interactions between protein and ligand. Accordingly, a profound
understanding and an adequate consideration of water molecules is a prerequisite for the prediction of ligand-binding modes, e.g. in docking,
QSAR considerations or affinity and binding thermodynamics predictions. We have equipped Relibase with retrieval and analysis tools to study
the properties of water molecules involved in ligand binding. Descriptors have been developed to predict the probability for the occurrence of
water molecules in the protein-ligand binding interface (135).
Top of Page
c. Cavbase - a database to compare binding pockets across proteins
The function of proteins is almost invariably linked with the specific recognition of substrates and ligands in given binding pockets, thus proteins
of related function should share comparable recognition properties exposed into these pockets. We have developed the module Cavbase for
Relibase that stores protein cavities in terms of simple surface-exposed physicochemical properties. These descriptors allow for fast retrieval
of proteins with functional relationships independent of a particular sequence or fold homology (113, 130, 304). The approach also allows
detecting unexpected cross-reactivity of ligands among unrelated proteins (146). Classification of binding pockets across protein family
members allows elucidating selectivity determinants, e.g. in carbonic anhydrases, kinases, nuclear hormone receptors or proteases (186, 203).
Via spatial graph alignment, consensus binding epitopes are extracted and correlated to detect those physicochemical features that are conserved
across individual members of protein families (164, 249, 178, 319, 337). This novel taxonomy has been applied to cluster the protein space of
enzymes. The obtained clustering shows differences in the grouping based on sequence similarity but agrees with similarities in fold space or
of the ligands hosted in the commonly structured binding pockets (239). Clustering of cofactor binding pockets in cavity (Cavbase) and fold
space (DALI) reveals virtually the same data structuring. Remarkable relationships can be found among the different spaces and show how
conformations are conserved across the host proteins and which distinct local cavity and fold motifs recognize the different portions of the
cofactors. In those cases, where different cofactors are found to be accommodated in a similar fashion to the same fold motifs, only a commonly
shared substructure of the cofactors is used for the recognition process (283). Algorithms are developed for the automated analysis and
decomposition of binding pockets in subpockets. Alternative approaches for the description of binding pockets by a set of surface patches to
classify proteins (181, 339) or new graph matching algorithms to cluster pockets have been developed (267, 278, 290). Representing the pocket
features by sets of distance histograms, the search engine could the accelerated dramatically (321, 324) (funding; CCDC, DFG, collaboration:
Prof. Hüllermeier Univ. Marburg, now Paderborn, Prof. Ultsch, Univ. Marburg, CCDC, Sanofi).
Top of Page
d. Secbase - combining secondary structural information and folding patterns with ligand binding data
The increasing body of structurally determined protein-ligand complexes indicates that particular secondary structural elements and folding
motifs in proteins provoke preferred ligand-binding patterns. To search for such binding motifs possibly formed due to the recursive placement
of physicochemical properties of secondary structural elements in special folding patterns of protein families, Relibase has been equipped with
search facilities to combine information about helices, sheets and turns with ligand binding modes. An exhaustive classification of turn motifs
has been performed and implemented into Relibase. It allows better coverage of turn motifs in fold prediction (240, 241, 246) (funding and
collaboration: CCDC).
Top of Page
e. Analysis and interference of protein-protein interfaces
A large part of the presently known small molecule drugs are either enzyme inhibitors, allosteric effectors or receptor agonists or antagonists.
They replace natural substrates or endogenous ligands mostly in deeply buried and stringent binding pockets. The presently available drug
design tools are all methodologically focused on the competitive replacement of such ligands or substrate portions by appropriate lead structures
that exploit a similar region of the deeply buried binding pocket. However, functional regulation of a biological system can also be achieved
via inference with protein-protein interactions. In many processes, activation of protein function depends on co-activation through the assembly
of several protein components, e.g. through the formation of a multi-domain complex. Many signal transduction cascades operate via the
formation of protein-protein interfaces.
Accordingly, the interference with this recognition process using small molecule drugs would - in principle - allow for the development of an
entirely new class of drugs. At present, we do not yet understand sufficiently the nature and the driving force for the formation of protein-
protein interactions, in particular with respect to their strength and dynamic stability. We have analyzed the structural patterns of exposed
physicochemical properties exhibited by permanent and transient protein-protein interfaces (200). Also the reverse concept might interfere with
information transduction: Stabilization of protein-protein interactions. This can be achieved through accommodation of a small molecule ligand
at the interface, thus enhancing the stability of a protein-protein contact (205) (collaboration: Prof. Hüllermeier, Univ. Marburg).
As an own experimentally example, we studied the tRNA modifying enzyme tRNA-guanine transglycosylase (TGT). It is a putative target for
new selective antibiotics against Shigella bacteria and only active as homodimer (see also Section 4b). By noncovalent mass spectrometry, we
could confirm its dimeric oligomerization state and a 2:1 binding stoichiometry of the complex formed between TGT and its full-length substrate
tRNA. To study whether dimer formation via a large protein-protein interface (>1600 Ã…
2
) is essential for TGT to accomplish its catalytic
function, point mutations were first investigated by a computational alanine scan (322). This approach indicated a cluster of aromatic residues
as most crucial for interface stabilization. Subsequently, the crucial residues have been mutated and enzyme kinetics reveal reduced catalytic
activity of the mutated variants. The achieved destabilization toward monomeric state has been further evidenced by both noncovalent mass
spectrometry and X-ray crystallography (253, 322, 392). Furthermore, the interface has been analyzed by computational methods to elucidate
hot-spot residues that stabilize the interface (314). (collaborations: Prof. F. Diederich, ETH-Zürich, Switzerland, Dr. Sarah Sanglier-Cianférani,
Univ. Strasbourg, France, funding: DFG FO 806).
Top of Page
2. Development of experimental and computational tools for the discovery and design of specific modulators of protein function
Top of Page
a. Development of scoring functions for docking, scoring and affinity prediction
Development of scoring functions for docking and affinity prediction
Our group has been involved in the development of computational tools for conformational analysis (58, MIMUMBA), de novo design (56,
LUDI), comparative molecular field analysis (61, 77, 83, 86, 118, 129, 161, 208, 230, CoMSIA, AFMoC), comparative molecular
superpositioning (60, 73, 82, 84, FlexS) and ligand docking (68, FlexX). Key prerequisite in ligand docking (120, 122, 140) is, beside the
generation of relevant binding modes, the correct estimate of binding affinity on the basis of the produced binding geometry (103, 122, 125,
140). Accordingly, we have developed the knowledge-based scoring function DrugScore (93, 185) with the expansions DSX (284) to rank
different poses generated by docking methods. DrugScore employs statistically derived pair potentials using the distance-dependent occurence
frequencies by which a particular ligand atom type is found in contact with a protein atom type. DrugScore can be tailored to a particular protein
target using information about binding affinities of a small training set of ligands (129, AFMoC). As input, crystal data either from the PDB or
the CSD are evaluated (185). Subsequently, we could enhance DrugScore (DSX) by a novel atom type assignment and altered definitions of
reference states (284). Furthermore, the DrugScore potentials can be used as objective function in ligand docking (131) and to optimize and
refine the geometry of protein-ligand complexes (MiniMuDS, 282). In addition, they can be used for the graphical hot-spot analysis of binding
pockets (96). In addition, the regression-based scoring SFCscore has been developed that uses a broad range of different molecular descriptors
as input. The analysis is performed on the basis of about 1000 protein-ligand complexes for which either affinity and structural data are available
(229) (funding and collaboration: industry consortium, CCDC).
Top of Page
b. Hot-Spot analysis of binding sites
To discover favorable areas in a binding pocket, likely to accommodate a particular type of ligand functional group, we map the binding pocket
in terms of "hot spots" of binding (96, 193). For this analysis, either Peter Goodford's force-field based method GRID can be applied, or
knowledge-based approaches such as SuperStar and DrugScore (96, 99, 193). The latter ones are both based on crystal data as input. They
calculate preferred interaction sites by mapping either knowledge-based pair potentials or composite crystal-field environments onto active site-
exposed residues (56). A regularly spaced grid is embedded into the binding site and for different ligand atom types interaction energies or
contact preferences are calculated by systematically placing probe atoms at the various grid intersections. To allow for an intuitive graphical
interpretation of the "hot spots", the obtained grid values are contoured according to a predefined level above the detected global minimum.
Crucial in the analysis is an appropriate assignment of atom types to optimally represent the physicochemical properties of the "hot spot" under
investigation. A sophisticated atom-type classification has been developed to extract knowledge-based potentials either from the PDB and CSD
(275). From an experimental point of view, methods for cocrystallization and soaking of gases, solvent molecules and small molecular fragments
were developed to perform an experimental active-site mapping (269).
Top of Page
c. Strategies for lead discovery by virtual screening
Virtual screening (VS) is an alternative to high-throughput screening (HTS) in lead discovery. Hit selection is attempted in the computer by
predicting binding properties of putative ligands (97, 162, 201). The screened compounds do not necessarily exist and their testing does not
consume valuable substance material. VS requires as key prerequisite knowledge about the three-dimensional structure of the target and the
criteria responsible for ligand binding. It starts with a detailed analysis of the binding pocket of the target protein (Relibase). Using tools for
hot-spot analysis, a protein-based pharmacophore is derived. (102, 128, 136, 145, 147, 157, 165, 168, 195, 242, 287).
Subsequently, this information is used to define a list of essential functional groups to be present in putative candidate ligands. We applied a
and fast pharmacophore matching. In subsequent steps, molecular similarity with known reference ligands was used to re-rank the hits from the
pharmacophore matching. Finally, the best scored candidates were docked flexibly into the protein binding pocket.
We have applied virtual screening successfully to several protein targets (e.g. carbonic anhydrase II, tRNA guanine transglycosylase, aldose
reductase, peptide deformylase 102, 128, 101, 138, 147, 148, 157, 145, 195, 242, 287). The obtained computer hits gave rise to experimental
testing of 10-15 compounds and nano to micromolar binders were discovered. In several cases a crystal structure with the target protein could
be determined. Also in case of a homology-modelled GPCR receptor, our strategy retrieved an antagonist of submicromolar affinity. The
considered model of the NK1 receptor (165, 168) was based on a newly developed approach which constructs proteins by homology including
bound ligand molecules as restraints, thus resulting in more relevant geometries of protein binding sites (104, 155, 255). Initial homology
models of the target protein are iteratively optimized by including information about bioactive ligands as spatial restraints (funding: DFG,
BMBF).
Top of Page
3. Experimental and computational approaches towards a better understanding of binding thermodynamics, selectivity and interaction
kinetics including to role of water in ligand binding
Top of Page
Once a new drug target is discovered, screening technologies are applied to detect prospective hits. However, which hit should be taken to the
next level of development? This decision is most crucial as it allocates a huge financial commitment of the subsequent drug optimization
strategy. Chemogenomic profiling allows compiling parameters that describe the binding characteristics of drug candidates to optimally
interfere with proteins. Membrane-bound proteins demand different properties than viral ones. Either high isoform selectivity or promiscuous
family-wide binding is needed or efficient resistance tolerance is desirable. This calls for very different ligand binding characteristics, requiring
either enthalpy or entropy-driven binding, rigid shape complementarity or pronounced residual mobility at the binding site. Interaction kinetics
determine on-/off-rates and residence times, a drug spends with its target. Distinct adjustment of these properties is essential for drug efficacy.
Chemogenomic binding parameters are increasingly collected but their correlation with drug binding is still rudimentally understood. We
approached this issue by compiling a knowledge base from congeneric protein-ligand series and correlated their structural, thermodynamic,
interaction-kinetic and dynamic behavior to predict the properties a lead has to meet to optimally address a given target. Our investigations
involved high-resolution crystal structure analysis, isothermal titration calorimetry, molecular dynamics simulations, site-directed mutagenesis
and interaction kinetics studies. They helped to provide a more comprehensive picture of our current understanding of drug-protein binding
(323). Furthermore, they made us aware of many contributions to drug binding, that are still insufficiently considered, for example the role of
water in ligand binding, the modulation of protonation states or the change of intrinsic dynamics during the binding process (funding: ERC
Advanced Investigator Grant).
Combination of ITC, Crystallography,
Molecular Dynamics and Binding Kinetics
a. Thermodynamic Analysis of protein-ligand binding
Top of Page
Binding affinity is an essential entity to predict the potency of a ligand in structure-based drug design. We suggested to start ligand optimization
with hits exhibiting largest enthalpic efficacy (256, 323, 362, 375). Isothermal titration calorimetry (ITC) gives access to the thermodynamic
signature of the overall ligand binding event (336), however, which additional effects are overlaid (375) and what are the net criteria to pick
the ligand with the best enthalpic binding properties, means the one with the largest enthalpic efficiency?
While docking programs have meanwhile achieved a level of reliability that makes them a viable tool for database screening of possible leads
on the computer, the ranking of putative hits according to their expected affinity remains the most crucial step in this procedure (67, 74, 105,
125, 201). Accordingly, there is still an vital interest to a better understand what "binding affinity" really means for the recognition of a drug at
its receptor and how this binding relates to thermodynamics (323) and binding kinetics (334). Using isothermal titration calorimetry, protein
crystallography, molecular dynamics simulations and biophysical methods to determine interaction kinetics, we study the binding of series of
low molecular-weight ligands towards a series of model proteins (linked to PDB-entries):
Top of Page
Trypsin, Thrombin Factor Xa: 112, 206, 223, 244, 248, 259, 262, 266, 293, 338, 341, 378, 382, 385, 397, PDB-Trypsin, PDB-Thrombin
Human Carbonic Anhydrase: 325, 384, 386, PDB-hCAII
Endothiapepsin: 362, PDB-Endothiapepsin
Protein kinase A and Pim-1: 371, 373, 377, 395, PDB-PKA, PDB-Pim-1
tRNA guanine transglycosylase: 318, 320, 340, 365, PDB-TGT
Aldose reductase: 189, 207, 209, 218, 277, 355, PDB-AR
In congeneric series of ligands, surprising changes of protonation states can occur (112, 209, 318, 364, 385, 395). They originate from induced
pK
a
shifts experienced by the ligand and protein functional groups upon complex formation (induced dielectric fit). They depend on the local
environment and oxidation state of bound cofactors and involve additional heat effects that must be corrected before any conclusion on the
binding enthalpy (ΔH) and entropy (ΔS) can be drawn. To complement the experimental evidence, we apply computer simulations to predict
changes of protonation states. These calculations involve free energy calculations and we developed a uniform charge model either for the
ligands and protein residues (199, 214, 220, 225, 318).
Some methodological development for the application of ITC have been performed, e.g. the measurement of racemic mixtures (196),
displacement titrations (326), impurity corrections (317, 336), kinetic and thermodynamic data from one ITC experiment (396).
Particular attention was paid to the impact and the contributions of water on the thermodynamic parameters of protein-ligand binding (250,
264, 288, 293, 296, 297, 315, 342, 351, 357, 363, 371). We applied neutron scattering to characterize protonation states and the residual
dynamics of water molecule in ligand binding (364, 368, 397). Based on the observed experimental data, we embarked into the development
of computational method to predict solvation thermodynamic data. In particular the GIST method has been applied (342, 368, 373, 381, 382,
393).
After correction, trends in both contributions can be interpreted in structural terms with respect to the hydrogen-bond inventory or residual
ligand and protein motions or the change in the local water structure. Even across congeneric series, the factorization in enthalpy and entropy
can change significantly, usually in a way that both contributions mutually compensate each other, leaving the free energy of binding ΔG
virtually unchanged. Explanation for these compensating effects are changes in the local water structure (pick-up or release of water molecules),
differences in the residual mobility of protein residues or portions of the ligands, an unbalanced solvation/desolvation inventory and shifts in
the local charge distribution giving rise to modified charge assisted hydrogen bonds (189, 223, 373). Pronounced cooperativity effects are in
operation that demonstrate the inadequacy of simple additivity models to describe functional group contributions to binding affinity (244, 248,
259, 262, 266, 315, 351) (collaboration: Prof. A. Podjarny, CNRS, Illkirch, France; Prof. T. Steinmetzer, Univ. Marburg; Prof. F. Diederich,
ETH-Zürich, Switzerland; Prof. D. Hangauer, SUNY Buffalo, USA, funding: DFG, ERC).
Top of Page
b. Site specific mutagenesis among members of a protein family to study selectivity determinants
Top of Page
step-wise reconstruction of binding pockets in related proteins being members of a particular protein family. As a first example, the binding
site of human factor Xa has been introduced into the structurally related rat and bovine trypsins by site-directed mutagenesis (126, 127). Major
re-organisation of the binding site takes place to yield a geometry virtually identical to that of factor Xa. However, with respect to binding
affinity, still a significant difference is observed between factor Xa and the trypsin variants. To achieve closer relationship in the binding
properties, also residues located in the second coordination sphere around the binding pocket have to be exchanged (126, 127, 151, 156, 206).
As a second example we selected aldose and aldehyde reductase. Both enzymes possess similar binding pockets, however, aldehyde reductase
exhibits an additional loop, comprising 11 residues, that is responsible for differences in substrate specificity. In this part, aldose reductase
opens a hydrophobic specificity pocket and shows pronounced adaptations upon ligand binding (228, 355). Similar investigations were initiated
to study the selectivity differences of aldose reductase and AKR1B10 (302, 333). Binding of two chemically closely related Aldose Reductase
inhibitors has been studied against a series of single-site mutants of the wild-type protein. Overall, the binding mode of the inhibitors is
conserved; but tiny structural changes are responded by partly strong modulation of the thermodynamic profiles and pronounced
enthalpy/entropy compensations. This provides insights how single-site mutations can alter selectivity of closely related ligands against a target
protein (211, 277, 311). In case of the short chain steroid dehydrogenase 17ß-HSD14, site-directed mutagenesis was performed to assess the
functional role of particular residues (376) (collaborations: Prof. M. Schlitzer, Univ. Marburg, Prof. A. Podjarnyi, IGBMC, CNRS Illkirch,
France, Dr. S. Marchais-Oberwinkler, Univ. Marburg; funding: DFG).
c. 3D-QSAR analysis and clustering of binding cavities to elucidate affinity and selectivity determinants
Top of Page
Computational approaches to analyze selectivity start with the comparison of binding affinities across a set of structurally diverse ligands with
respect to several members of a protein family. By means of 3D-QSAR methods, such as comparative molecular field analyses (CoMFA,
CoMSIA (61), AFMoC (129) molecular properties of ligands can be extracted that determine protein binding. If this evaluation is focused on
affinity differences, the mutual comparison of the derived QSAR models allows one to elucidate the most important selectivity determinants
(86, 161, 208, 230). Complementary to this evaluation of ligand data, binding pockets across different members of a protein family can be
analyzed. Generalized probes are used that explore the properties exposed by the protein towards the binding-site cavity (cf. GRID, SuperStar,
DrugScore or CavBase descriptors and surface patches). They provide a set of functionally similar descriptors and can be correlated via
chemometric and clustering analyses (161, 186, 203, 208, 304). In case of Cavbase, the binding pocket-exposed physicochemical properties are
among binding pockets, not apparent through sequence information. Proteins distant in sequence space can be clustered together due to close
relationship in cavity space. Close similarity in the latter space indicates possible cross-reactivity of putative ligands and suggests where to
expect conflicting selectivity profiles (146, 186, 239, 304) (funding: DFG, CCDC, Novartis).
3D-QSAR analysis and clustering of binding cavities to
elucidate affinity and selectivity determinants
4. Design Projects on Protein Targets relevant for Different Therapeutic Areas such as Infectious Diseases or as Engineered Biocatalysts
Top of Page
a. De novo crystal structure determination of proteins, involved in infectious and other diseases, structure-based ligand design
Several crystal structures of novel proteins as putative targets for drug therapy have been determined, e.g., 1-deoxy-D-xylulose-5-phosphate
reductoisomerase, a crucial enzyme in the non-mevalonate isoprenoid biosynthesis (124), the Plasmodium falciparum glutamate
dehydrogenase, a putative target for novel antimalaria drugs (144), lipoamide dehydrogenase from Trypanosoma cruzi, a target for Chagas
disease therapy, the tRNA-modifying enzyme S-adenosylmethionine:tRNA ribosyl transferase/Isomerase, a possible target for Shigella
Dysentry (210).
Structure-based ligand design has been performed on the transglutaminases TG2 (putative target for Coeliac disease, an autoimmune disorder
of the small intestine that occurs in genetically predisposed people) and FXIIIa (final enzyme in the blood clotting cascade (309, 387, PDB-
FXIIIa) (collaboration: Zedira, Darmstadt, funding: BMBF). From the family of short-chain steroid dehydrogenases, 17ß-HSD14 has been
crystallized and new inhibitors have been developed (349, 354, 367, 376, PDB-17ß-HSD14) (collaboration: Dr. S. Marchais-Oberwinkler,
Univ. Marburg).
Top of Page
b. Inhibitor design of t-RNA guanine transglycosylase: A putative target for Shigellosis therapy
Inhibitor design of t-RNA transglycosylase:
A putative target for Shigellosis therapy
Top of Page
Shigella are the causative agents of dysentery (Shigellosis) and effect nearly 270 million infections each year with several 100 thousand lethal
therapy kills bacteria of the natural intestinal flora, which also produces diarrheal symptoms. Thus, there is urgent need for the development of
new and smart antibiotics. Characterization of chromosomal mutants of S. flexneri has resulted in the identification of a gene that contributes
significantly to pathogenicity and codes for tRNA-guanine transglycosylase (TGT). This enzyme, involved in biosynthesis of the highly
modified nucleoside queuine inserted in the anticodon loop of certain tRNAs, has been recognized as essential in the regulation of bacterial
virulence. Accordingly, it has been selected as a target for structure-based inhibitor design (179). Based on the crystal structure of TGT with a
bound substrate, virtual screening has discovered a series of small molecule hits subsequently found to bind in the micromolar range (101, 123,
138, 147). Several of these initial hits served as first lead for an iterative optimization to compounds with improved binding affinity (157, 170).
This also lead to the lin-benzoguanines as a most promising parent scaffold (204, 212). Cycles of design, synthesis, and crystal structure analysis
revealed pronounced induced-fit adaptations of the enzyme with respect to bound ligands (148, 221, 226, 310, 311). The importance of a
conserved water network, not to be perturbed by ligand binding, could be evidenced (250, 254, 296, 330, 365). Significant shift in binding
affinity could be detected by moving from a normal to charge-assisted hydrogen bond. The lin-benzoguanine parent scaffold exhibits two
guanidine-type portions, both likely candidates for proton uptake. Two adjacent aspartates induce a strong pK
a
shift at the ligand site, resulting
in a protonation-state transition of the group with lower (!) pK
a
-value and an array of two parallel H-bonds avoiding secondary repulsive effects
contributes to the high-affinity (226, 243, 318). Interestingly, a set of nanomolar inhibitors shows pronounced residual mobility of one side
chain of the inhibitors. This entropically favorable contribution to binding seems beneficial for ligand binding (320, 340, 365). Kinetic and
mutational studies have elucidated the substrate selectivity profile of TGT originating from different species. Structural modifications of the
bacterial enzyme toward the human protein could be performed to better understand the differences in substrate specificity between the species
(89, 149, 211, 366).
Remarkably, TGT is only active as a homodimer and the active site is located close to the dimer interface (328). Ligands spiking into the
interface increase monomerization and block enzyme function (303). The stability of the dimer interface is determined by a cluster of aromatic
amino acids. This was studied either computationally or by site-directed mutagenesis (253, 314, 322, 392). Surprisingly, ligand-induced TGT
can be transformed from a functionally active to a twisted catalytically incompetent homodimer. These features could be observed by crystal
structure analysis,
19
F-NMR spectroscopy and PELDOR-EPR measurements (372). Fragment-based approaches made aware about the
existence of a transient binding pocket next to the active site (379) and below a regulatory loop important in the dimer interface formation.
Results from both studies provide a perspective for novel ligands with inhibitory potency (collaboration: Prof. F. Diederich, ETH-Zürich,
Switzerland, Prof. G. Garcia, Univ. Michigan, USA, Dr. Sarah Sanglier-Cianférani, Univ. Strasbourg, France, Prof. M. Sattler, HZ-München,
Prof. O. Schiemann, Univ. Bonn, funding: DFG, NIH).
Top of Page
c. Design of inhibitors against family members of serine and aspartyl proteases
Mining the human genome for possible drug targets has revealed pronounced clustering of proteins into several prominent families. Likely,
these families share family-wide commonalities in terms of enzymatic mechanisms and molecular recognition properties. This, in consequence,
suggests that privileged ligand scaffolds can be designed to address entire families. To achieve specificity and selectivity, proper decoration of
the underlying privileged scaffolds is required. Through combinatorial chemistry, a given molecular scaffold can be easily modified with respect
to the decorating side-chains. An appropriate selection of these side-chains is achieved by the detailed analysis of the recognition properties of
the various subsites that compose the binding pocket accommodating the scaffold (217, 263). Taking the families of serine and aspartyl proteases
as case studies, synthetic and computational tools are developed how to optimize a given privileged ligand scaffold towards different family
members. The side-chain selection step is supported by hot-spot analyses and similarity searches with Cavbase. Once a binding pocket similar
to the query pocket is detected, the accommodated ligand might suggest an alternative molecular portion suitable as decoration of the privileged
scaffold under investigation (202, 263, 235).
HIV- protease, plasmepsin II and IV, pepsin and SAP II have been selected as target proteins for structure-based design. Through several cycles
of iterative design, first leads could be optimized with respect to affinity, selectivity and resistance tolerance profile. A novel scaffold could be
detected for HIV protease that shows, compared to all currently available market products, a deviating binding mode. It exhibits a different
resistance profile and shows higher affinity towards one of the major resistance mutants compared to the wildtype (180, 183, 184, 231, 232,
233, 234, 235, 331, PDB-HIVP) (collaboration: Prof. W. Diederich, Univ. Marburg, funding: DFG).
d. Structural studies on Shigella-specific pathogenicity factors as a basis for structure-based drug design
Top of Page
The bacterial invasion of Shigella, the causative agent of bacillary dysentery, depends on its ability to invade colonic epithelial cells via self-
induced macropinocytosis. Invasion of the host cell involves a number of massive cytoskeletal rearrangements within the colonic epithelial cell
which are mainly triggered by so-called invasines. These effectors secreted by the bacterium cause the recruitment of cytoskeletal proteins as
well as actin polymerization at the site of contact with the bacterium. The cytoskeletal rearrangements during Shigella invasion and intracellular
spread are regarded as a model system for common processes during cell movement and adhesion. As there is only few biochemical and
structural information available about the proteins causing these dynamical changes, we started to isolate and crystallize the involved bacterial
effectors in order to determine their 3D structures. With a view to the understanding of their mechanisms of action at a molecular level, we aim
to establish the crystal structures of these proteins not only in their apo forms but also in complex with known protein interaction partners and
small molecule fragments. Not least, these pathogenicity factors will represent highly interesting targets for the structure-based design of drugs
against bacillary dysentery (collaboration: Prof. K. Reuter, Prof. Andreas Heine, Univ. Marburg, funding: DFG, LOEWE DRUID).
Top of Page
e. Design of substrate specificity of biocatalysts, hybrid enzymes, catalysts for click chemistry
reactions. For example, lipases catalyze the hydrolysis, transesterification or amidation of a broad range of esters and amides with distinct
stereopreference. In consequence of their stability and large scale availability, they have found widespread applications in the enantioselective
synthesis of precursors to pharmaceuticals and in the kinetic resolution of racemic mixtures. We have studied the kinetic resolution of pre-
acylated Candida antarctica lipase by enzyme kinetics and crystallography along with computer simulations. We could evidence that the fast-
reacting enantiomer is enthalpically favored through a virtually perfect active-site complementarity, whereas the slow-reacting substrate
compensates some of the discriminating free enthalpy advantage by a beneficial entropic contribution. This is due to a higher residual mobility
and thus a smaller loss in entropy upon binding. It resides less frequently in an orientation productive for the enzyme reaction (142)
(collaboration and funding: BASF).
By tailored design we modify protein structures in a way to catalyze novel chemical reactions. As an example, the Huisgen reaction, which
produces triazole and tetrazole heterocycles via 1,3-dipolar addition has been selected. This reaction is either Cu
I
- or Zn
II
-catalyzed. We used
the Zn
2+
ion in the catalytic center of carbonic anhydrase. As reaction components, small libraries of substituted alkynes or nitrils, and azides
were used. The azide component could be tethered to the enzyme surface via a disulfide bridge, while the alkyne component was reversibly
coordinated via a sulfonamide anchor to the zinc ion in the catalytic center. The incipient orientation of the reactants in the binding site and of
the formed triazole product were characterized by crystallography and the reaction progression could be monitored by HPLC-MS analysis (268)
(collaboration: Prof. U. Koert, Univ. Marburg, funding: DFG-FO 594).
Design of substrate specificity of biocatalysts, hybrid
enzymes, catalysts for click chemistry therapy
Top of Page
5. Crystallographic high-throughput screening for fragments directly protein crystals to obtain starting points for subsequent drug
development
Available from JenaBioscience
Over the last 20 years, methods have been developed to discover small molecular fragments (MW < 250 Da) as first leads including
crystallographic screening approaches. Once soaked or cocrystallized with a target protein (for differences between soaking/cocrystallization:
358, 394) they provide first ideas about putative starting points for ligand development. As a kind of molecular probe, they explore the binding
pocket and highlight regions most favorable to accommodate small molecular fragments (260). We developed and assembled new fragment
libraries which expands the usual Rule of 3 criteria and reveals high hit rates (279, 388). Frequently biophysical assays are applied first to
discover fragments in libraries and to preselect them for a subsequent crystallographic analysis. However, the results of these biophysical assays
show astonishing little mutual overlap (335, 380, 383) and compared to a screening directly on protein crystals they retrieve lower hit rates
(343, 380, 391). We therefore advocate to screen fragment libraries directly on protein crystals. Therefore, we improved experimental soaking
techniques to achieve a higher success rate in getting small molecular fragments into protein crystals. For this purpose, the group developed
special exposure techniques and used a free mounting device for crystals which allows manipulating crystals freely accessible in a humidity
gas stream (269). Together with the macromolecular crystallography group at HBZ in Berlin, synchrotron Bessy II, a dedicated fragment
beamline has been built-up and many examples of direct fragment screening on protein crystals have been performed (344, 345, 353, 380, 389).
Several methodological improvements and validations of fragment discovery could be accomplished (326, 338, 346, 360, 371, 379, 395, 397).
As an extension of the current fragment approach, we embarked on dynamic combinatorial chemistry to efficiently optimize initial fragment
hits (316, 329, 348, 390). In addition, we docked already synthesized compounds that embed as common substructure the geometry of a
crystallographic fragment hit and keep its experimentally determined binding mode. Via an incremental built-up procedure, the initial starting
fragment is grown to larger candidate ligands using the docking program FlexX (collaboration: Proteros, BioSolveIt, Merck, Boehringer
Ingelheim, HZB Berlin, JenaBioscience, CrystalsFirst funding: BMBF).
Top of Page
Home
Prof. Klebe
Group Members
DrugDesign
Research
Publications
Textbook
Instrumentation
Funding
Conferences
Job Offers
Download-Area
School leavers
Pharmacy
Pharm. Chemist
Contact
Imprint
Privacy Policy
