Drug Discovery Today: Technologies

Editors-in-Chief Kelvin Lam Pzer, Inc., USA Henk Timmerman Vrije Universiteit, The Netherlands
DRUG DISCOVERY

Vol. 3, No. 3 2006

TODAY

TECHNOLOGIES

Medicinal chemistry

Structure-based drug design: NMR-based approach for ligandprotein interactions

Xu Zhang, Huiru Tang*, Chaohui Ye, Maili Liu*
State Key Laboratory of Magnetic Resonance and Atomic and Molecular Physics, Wuhan Centre for Magnetic Resonance, Wuhan Institute of Physics and Mathematics, Chinese Academy of Sciences, Wuhan, 430071, PR China

The realization of the powerfulness in analyzing ligand protein interactions at the atomic resolution has made NMR techniques increasingly attractive in drug discovery and development. With some signicant new method developments during the past few years, NMR-based approaches will undoubtedly be helpful in high throughput screening and in providing structural and interaction information benecial for new drug developments. Here in this review, instead of providing an exhaustive account of applications of NMR, we will seek to highlight some key points related to the solution state NMR methods for extracting information from both ligands and proteins.

Section Editors: Li-he Zhang School of Pharmaceutical Science, Peking University, Beijing, China Kaixian Chen Shanghai Institute of Materia Medica, Chinese Academy of Sciences, Shanghai, China

achieve better understandings for the structure of ligand protein complex and dynamics of the interaction. A number of reviews have already been published [4,710] on using NMR approach in drug discovery and screening. The aims of this review will seek to provide a description of the technical aspects of solution state NMR-based methods applicable to ligandprotein interactions and to address some of the key issues including the current NMR strategies aimed to extract interaction information from both proteins and ligands in free and bound states.

Introduction
As an important drug discovery tool, nuclear magnetic resonance (NMR) can provide valuable information on the molecular structure, dynamics and interactions between drugs and their targets. Ever since the well-known NMR method, socalled SAR by NMR, was introduced in 1996 [1] to reveal the structure-activity relationships (SAR) in discovering highafnity ligands for proteins, many new NMR techniques for measuring proteinligand interactions have been developed during the past few years [26], although such development will be predictably accelerated even further to
*Corresponding authors: : H. Tang (Huiru.tang@wipm.ac.cn), M. Liu (ml.liu@wipm.ac.cn) 1740-6749/$ 2006 Elsevier Ltd. All rights reserved. DOI: 10.1016/j.ddtec.2006.09.002

Protein-based experiments
The binding of ligands to a protein often results in changes in the proteins structure and dynamics, in addition to the local environmental changes on the binding surface. Properties of the ligand are also affected by the interactions. These changes will be in turn manifested in the NMR properties of both ligands and the protein, such as chemical shifts, relaxation rates and line-shapes of NMR peaks. Therefore, these parameters can be used as indicators for the interactions, the relative orientation and motions of domains within proteins, which are key factors affecting the multivalent recognition.
241

Drug Discovery Today: Technologies | Medicinal chemistry

Vol. 3, No. 3 2006

Chemical shift perturbation

The strong binding of a ligand to a protein often results in changes of protein chemical shifts. This has made the chemical shift perturbation experiment one of the most widely used methods for probing ligandprotein interactions [11 13] by measuring the chemical-shift changes of the protein during titration using a ligand. The amino acid residues that experienced progressive chemical changes were generally located on the binding surface or in the domain of structural changes; the larger the chemical-shift change is, the closer the residues or spins are to the binding site. Using the fast exchanging model, the dissociation constant (kd) can be derived from chemical-shift changes as function of the ligand concentration [14].

cules that always exhibit slow relaxation (T1, T2) and fast diffusion. When binding to a large receptor, nuclear spin relaxation time and molecular self-diffusion coefcient of the ligand will be dramatically reduced. These changes enable the ligandprotein interaction to be accessed by directly observing the ligand NMR signals.

Relaxation-edited methods
Relaxation-based methods are more sensitive than diffusionbased methods because of increment of apparent molecular weight upon binding [21]. In particular, the paramagnetic induced relaxation approach, such as so-called SLAPSTIC (the spin labeled attached to protein side chains as a tool to identify interacting compounds), offers a sensitive method for interaction mapping [22]. This is because that the magnetic moment of the electron is thousands times larger than that of the proton, and the amount of the biological targets required will be reduced by 1 to 2 orders of magnitude, alleviating the necessity of high ligand solubility. In addition, because relaxation is a spin property, a spin or nucleus that closed to the target protein or binding site will experience larger relaxation-rate change (similar to chemical shift). Therefore, the relaxation rate measurement can also provide information of the ligand orientation at the binding site.

Protein dynamics
Dynamics are vital to protein functions and alterations in its three dimensional structure due to specic molecular interactions will lead to changes to its dynamics. The dynamics and their alteration can often be readily measured in autorelaxation that is one of the most extensively explored methods for accessing the dynamics of the protein backbone. The spin properties of 15N, 13C or 2H sites in proteins, including both longitudinal and transverse relaxation rates (R1 and R2), are sensitive to the dynamics of the proteins in particular the local exibility [15]. Upon binding, the changes will occur in the relaxation rates of the nuclear spins at the interactive interface. In addition, the transferred cross-correlated relaxation [16] and transferred residual dipolar coupling are also important parameters for probing the ligandproteins interactions [17] in terms of the binding geometries. Furthermore, a number of novel experiments have been introduced to probe the side chain dynamics [18] during past few years, because the side chains are always related to the protein surface properties and their motions are essential for ligand binding. For example, methyl groups of the human intestinal fatty acid binding proteins (IFABP) are often found in the hydrophobic cores, the binding of oleic acid results in decreased mobility of most of its methyl groups in the binding region [19]. The protein-targeted experiment can provide information about interaction interface and structure. However, extra care has to be taken to distinguish the structural and dynamic changes in the binding surface and in more extended regions of the protein molecule. It is also worth noting that the chemical shift perturbation may also result from the variation of pH and/or salt concentration although it is an excellent indicator of allosteric processes [20].

Diffusion-edited experiment
Measuring self-diffusion coefcient is a convenient way to study the ligand binding without prior separation of the components [23]. Because diffusivity is a principle property of the whole molecule, all NMR signals of a ligand are decayed identically in the experiment with pulsed-eld-gradients (PFG). In solution containing a target protein and ligands, the bound ligands will have small diffusion coefcients and their NMR signal intensities will be less attenuated in the PFG-NMR diffusion experiment. By selecting a suitable diffusion-attenuation factor such as the gradient strength, the NMR signals from nonbinding ligands will be ltered out because of their fast diffusion, and thus, the bound ligands will be selectively detected from the mixture. The efciency for PFG-NMR depends on the differences in the observed diffusion coefcients between binding and nonbinding ligands. Because free and bound ligands are generally in fast exchanging on the diffusion time scale, the measured diffusion coefcients are the weighted average of the free and bound forms. This will narrow the gap of diffusivities between binding and nonbinding ligands, leaving a small window to identify active components in the mixture [24,25]. In using PFG-NMR, one often has to disentangle the transferred nuclear Overhauser effect (trNOE) from the diffusion. Nevertheless, the decay of trNOE can also be used as an effective way to gain insight into ligandprotein interaction during the diffusion time [2427].

Ligand-based experiments
The ligand-based methods are established on the basis of the fact that upon binding to a target protein, the molecular tumbling rates of ligands in solution are altered. Generally, most of the NMR screening libraries consist of small mole242
www.drugdiscoverytoday.com

Vol. 3, No. 3 2006

Drug Discovery Today: Technologies | Medicinal chemistry

NOE pumping and reverse NOE pumping experiment

The nuclear Overhauser effect (NOE) results from dipole dipole interactions of atomic spins (through space); its strength is inversely related to the interspin distances (r6). In the NOE and reverse NOE pumping (RNP) experiments, the details in proteinligand interactions are obtained through the dipoledipole interactions to transfer the magnetization from protein to bound ligands or from binding ligands to protein. Therefore, a state where only the ligand signals or the protein signals are preserved has to be prepared by the diffusion [24] or relaxation [25] weighted experiments. Because of fast exchange between free (slow relaxation) and bound (fast relaxation) ligands, the magnetization transferred from protein is accumulated in the free ligands, resulting in a pumping effect for the NOE. The method can give reliable results for an interacting pair without the necessity of chemical separation or isotope labeling. However, The NOE/RNP is generally limited to the systems with large differences in transverse relaxation time or diffusion coefcients between the receptor and ligands, and requires the binding reaction to be in the fast exchange regime between the free and bound states. In the case of slow exchange systems, isotope ltering/editing technique can be used to prepare nonequivalent states without the T2 relaxation limitation.

one can get a spectrum contained the resonances of a protein and its binding ligands. NMR peaks of the nonbinding ligands in the sample are not affected by the saturation and, thus, canceled by the difference. Therefore, the STD experiment is useful in identifying all potential ligands that interact with protein in a mixture. For weak interactions, the STD experiment is more sensitive than the NOE-based approach and, therefore, low concentration of protein can be used [31,32]. WaterLOGSY experiment (WaterLigand Observation via Gradient SpectroscopY) is another intermolecular magnetization-transfer difference experiment based on saturation of water resonance [33] instead of protein resonance used in the STD approach. Because water plays an important role in protein surface, saturation of water signals denitely saturated the protein and the bound compounds. Because of the fast exchange between bound water and bulk water, the large quantity of bulk water magnetization increases the sensitivity of the WaterLOGSY experiment.

NOE/trNOE experiment
Nuclear Overhauser effect provides another mechanism for both inter and intramolecular magnetization transfers. Because magnitude of NOE enhancement between two nuclei spins are exponentially related to the internuclear distance (r6), NOE related experiments have been widely used for determining three dimensional structures of protein and ligandprotein complex, orientations of the binding domain in protein and ligand inside or toward the binding site, as well as for deriving dynamic information for the ligandprotein interactions. It is worth noting that NOE experiments work well for the high afnity interactions, where stable ligand protein structure is formed. In the case of low afnity binding or weak interactions, transferred NOE (trNOE) spectroscopy [3436] can be the method of choices. Transferred NOE provides information about orientation of the ligand at the protein binding site. In practice, trNOE as detected in NOESY experiments generally reaches a maximum within a few hundred milliseconds after which differential trNOE for individual ligand functionalities is equalized by spin diffusion of the NOE throughout the ligand proton spin system. Therefore, when long mixing times are used, NOE may diffuse throughout a proton spin system, affecting remote nuclei not directly involved in dipolar coupling [37].

Isotope ltering/editing NMR techniques

Isotope ltering/editing NMR techniques make use of isotopic (13C, 15N, etc) labeling to selectively detect the spectrum of the labeled protein or ligand [28], and the spectral signals of the other components are ltered out. By combining ltering/editing approach with conventional two dimensional or three dimensional experiments, the structure of the bounded proteins or ligands can be directly obtained. To maintain high sensitivity, the target component has to be completely labeled with NMR active isotopes followed with higher performance of the isotope ltering technique to suppress the unlabeled component. This kind of experiment can also be carried with direct detection of the labeling nuclear spin, 13C or 15N NMR. Because of low gyromagnetic ratio (g), 13C and 15N NMR is less sensitive than proton (1H) NMR.

Saturation transfer difference (STD) experiment

Saturation transfer difference usually starts from saturating one of the resonances (peaks) of protein [29,30] and until a stead state of saturation is achieved. There are magnetizationtransfer mechanisms, spin diffusion and cross-relaxation, happened during the saturation. Because of spin diffusion, all signals of the protein will be saturated. Most importantly, cross-relaxation carries the saturation effect to ligands that interacts with the target protein. By taking difference of two spectra obtained with and without saturation, respectively,

Future perspective
It is obvious that study of ligandprotein interaction in biological systems at atomic resolution is an ideal way for understanding the true pharmacological mechanism happening in living systems. From a technical point of view, such a goal can only be achieved by a combination of methods that are used for characterizing the target protein or ligand, solving the spectral overlapping and the other problems associated with the complexity of the system.
www.drugdiscoverytoday.com

243

Drug Discovery Today: Technologies | Medicinal chemistry

Vol. 3, No. 3 2006

One of the examples is the study on interactions between ibuprofen, a potential drug for inhibiting low-density lipoprotein oxidation and lipoproteins in intact blood plasma [38]. Using the diffusion-edited experiment, chemical shift up-eld drifts were observed for certain resonances of the lipoproteins after ibuprofen was introduced into the blood plasma. From two-dimensional (2D) correlation spectroscopy (COSY) experiment, chemical shift and line-shape analysis, it was possible to assign the resonances that experienced the chemical shift perturbation to the protons of N+(CH3)3 moieties of phosphatidylcholine (PC) and sphingomyelin (SM), olenic chains (CH CH, CH CHCH2CH CH, (CH2)nCH2CH ), and (CH2)n and CH3 groups from unsaturated lipids in lipoprotein particles. This revealed that the negatively charged carboxyl group of ibuprofen may interact with the positively charged trimethyl head group of PC and SM, whereas the phenyl and isobutyl groups of the ibuprofen may take part in the hydrophobic interaction with the olenic chains, (CH2)n and CH3 groups of the unsaturated lipids [38]. The results were further improved by observation of intermolecular NOE, relaxation rate and diffusion coefcient measurements. These results may elucidate the mechanism by which the drug protects against oxidative modication of lipoproteins. The recent development of the in-cell NMR technique [39] makes it possible to study proteinprotein interactions in physiological conditions [40] without the time-consuming purication procedures and the use of buffer. In solution NMR, buffering is critical in drug-protein interactions because the solubility and stability of the investigated macromolecules are seriously affected by the solvent conditions. It is conceivable that studies of ligandprotein interaction in intact biological system will attract more and more attentions. When the proteins are not stable in solution or have poor solubility, solid state NMR may be a method of choice. This is exemplied by the 13P-19F REDOR measurements [41] as the rst example of the distance measurements in an RNA-peptide complex.

Acknowledgements
We acknowledge the nancial supports from the National Natural Science Foundation of China (20610104, 20575074, 20475061), Chinese Academy of Sciences (Hundred Talent Program, 2005 [35]) and the Ministry of Science and Technology of China (973 project, 2002CB713806).

References
1 Shuker, S.B. et al. (1996) Discovering high-afnity ligands for proteins: SAR by NMR. Science 274, 15311534 2 Maurer, T. (2005) NMR studies of ligandprotein interactions. Methods Mol. Biol. 305, 197214 3 Wishart, D. (2005) NMR spectroscopy and protein structure determination: applications to drug discovery and development. Curr. Pharm. Biotechnol. 6, 105120

4 Lepre, C.A. et al. (2004) Theory and applications of NMR-based screening in pharmaceutical research. Chem. Rev. 104, 36413676 5 Lepre, C.A. et al. (2002) Applications of SHAPES screening in drug discovery. Combinat. Chem. High Throughput Screen. 5, 583590 6 Clarkson, J. et al. (2003) Studies of proteinligand interactions by NMR. Biochem. Soc. Trans. 31, 10061009 7 Watts, A. (2005) Solid-state NMR in drug design and discovery for membrane-embedded targets. Nat. Rev. Drug Discov. 4, 555568 8 Rees, D.C. et al. (2004) Fragment-based lead discovery. Nat. Rev. Drug Discov. 3, 660672 9 Rudin, M. and Weissleder, R. (2003) Molecular imaging in drug discovery and development. Nat. Rev. Drug Discov. 2, 123131 10 Pellecchia, M. et al. (2002) NMR in drug discovery. Nat. Rev. Drug Discov. 1, 211219 11 Peng, C. et al. (2004) Automated evaluation of chemical shift perturbation spectra: New approaches to quantitative analysis of receptor-ligand interaction NMR spectra. J. Biomol. NMR 29, 491504 12 Ross, A. et al. (2000) Automation of NMR measurements and data evaluation for systematically screening interactions of small molecules with target proteins. J. Biomol. NMR 16, 139146 13 Chen, Y. et al. (1993) Mapping of the binding interfaces of the proteins of the bacterial phosphotransferase system. HPr and IIAglc. Biochem. 32, 3237 14 Cui, Y.F. et al. (2003) Interaction between calcium-free calmodulin and IQ motif of neurogranin studied by nuclear magnetic resonance spectroscopy. Anal. Biochem. 315, 175182 15 Jarymowycz, V. and Stone, M.J. (2006) Fast time scale dynamics of protein backbone: NMR relaxation methods applicaiton, and functional consequences. Chem. Rev. 106, 16241671 16 Carlomagno, T. et al. (1999) Transferred cross-correlated relaxation: application to the determination of Sgar Puker in an aminoacylated tRNAMimetic weakly bound to EF-Tu. J. Am. Chem. Soc. 121, 19451948 17 Bolon, P.J. et al. (1999) Residual dipolar coupling derived orientational constrains on ligand geometry in a 53Kda ligandprotein complex. J. Mol. Biol. 293, 107115 18 Igumenva, T.I. et al. (2006) Characterization of the fast dynamics of protein amino acid side chains using NMR relaxation in solution. Chem. Rev. 106, 16721699 19 Zhang, X. et al. (2006) Probing methyl dynamics from 13C autocorrelated and cross-correlated relaxation. J. Am. Chem. Soc. 128, 50735081 20 van Nuland, N.A.J. et al. (1993) The NMR determination of the IIAmtl binding site on HPr of the Escherichia coli phosphoenol pyruvatedependent phosphotransferase system. FEBS Lett. 315, 1115 21 Hajduk, P.J. et al. (1997) One-dimensional relaxation- and diffusion-edited NMR methods for screening compounds that bind to macromolecules. J. Am. Chem. Soc. 119, 1225712261 22 Jahnke, W. et al. (2001) Sping Label enhanced NMR screening. J. Am. Chem. Soc. 123, 31493150 23 Lin, M. et al. (1997) Diffusion-edited NMR-afnity NMR for direct observation of molecular interactions. J. Am. Chem. Soc. 119, 52495250 24 Chen, A. and Shapiro, M.J. (1998) NOE pumping: a novel NMR technique for Identication of compounds with binding afnity to macromolecules. J. Am. Chem. Soc. 120, 025810259 25 Chen, A. and Shapiro, M.J. (2000) NOE pumping. 2. a high-throughput method to determine compounds with binding afnity to macromolecules by NMR. J. Am. Chem. Soc. 122, 414415 26 Lucas, L.H. et al. (2003) Transferred nuclear Overhauser effect in nuclear magnetic resonance Diffusion measurements of ligandprotein binding. Anal. Chem. 75, 627634 27 Yan, J. et al. (2002) Epitope mapping of ligand-receptor interactions by diffusion NMR. J. Am. Chem. Soc. 124, 99849985 28 Breeze, A.L. (2000) Isotope-ltered NMR methods for the study of biomolecular structure and interactions. Prog. NMR Spectrosc. 36, 323372 29 Mayer, M. and Mayer, B. (2001) Group epitope mapping by saturation transfer difference NMR to idensify segments of a ligand in direct contact with a protein receptor. J. Am. Chem. Soc. 123, 61086117 30 Megy, S. et al. (2005) STD and TRNOESY NMR studies on the conformation of the oncogenic protein beta-catenin containing the phosphorylated

244

www.drugdiscoverytoday.com

Vol. 3, No. 3 2006

Drug Discovery Today: Technologies | Medicinal chemistry

31

32

33

34 35

motif DpSGXXpS bound to the beta-TrCP protein. J Biol Chem 280, 29107 29116 Vogtherr, M. and Peters, T. (2000) Application of NMR based binding assays to identify key hydroxy groups for intermolecular recognition. J. Am. Chem. Soc. 122, 60936099 Klein, J. et al. (1999) Detecting binding afnity to immobilized receptor proteins in compound libraries by HR-MAS STD NMR. J. Am. Chem. Soc. 121, 53365337 Dalvit, C. et al. (2001) WaterLOGSY as a method for primary NMR screening: Practical aspects and range of applicability. J. Biomol. NMR 21, 349359 Feng, N. (1994) Recent developments in transferred NOE methods. Prog. NMR Spectrosc. 26, 517606 Vincent, S.J.F. et al. (1997) The conformation of NAD1 bound to lactate dehydrogenase determined by nuclear magnetic resonance with suppression of spin diffusion. Proc. Natl. Acad. Sci. 94, 43834388

36

37

38

39 40 41

Massefski, W., Jr and Alfred, R. (1988) Elimination of multiple-step spin diffusion effects in two-dimensional NOE spectroscopy of nucleic acids. J. Magn. Reson. 78, 150155 Lucas, L.H. et al. (2004) Epitope mapping and competitive binding of HSA drug site II ligands by NMR diffusion measurements. J. Am. Chem. Soc. 126, 1425814266 Yang, Y.H. et al. (2004) H-1 NMR spectroscopic evidence of interaction between ibuprofen and lipoproteins in human blood plasma. Anal. Biochem. 324, 292297 Serber, Z. and Dotsch, V. (2001) In-cell NMR spectroscopy. Biochem. 40, 1431714323 Burz, D.S. et al. (2006) mapping structural interactions using in-cell NMR spectroscopy (STINT NMR). Nat. methods 3, 9193 Olsen, G.L. et al. (2005) Monitoring tat peptide binding to TAR RNA by solid-state 31P-19F REDOR NMR. Nucleic Acids Res. 33, 34473454

www.drugdiscoverytoday.com

245