NMR spectroscopy in structure-based drug design Gordon cK roberts NMR methods for the study of motion in proteins continue to structure and dynamics of the binding site improve, and a number of studies of protein-ligand complexes Methodological developments in NMR of macromolecules relevant to drug design have been reported over the past year, continue apace [10, 11]. Important recent developments for example, studies of fatty-acid-binding protein and SH2 and che use of residual dipolar couplings and of SH3 domains. These studies have begun to give a picture of anisotropic contributions to relaxation to provide new the structural dynamics of protein-ligand complexes and to kinds of structural information [12, 13]; measurements of relate the changes in dynamics on ligand binding to the origins dipolar couplings in partially oriented proteins are a partic of specificity. NMR is also valuable in locating binding sites, ularly important addition to NMR methods for structure both qualitatively from changes in chemical shift and more determination, as they provide information on the orienta- precisely from distances measured from relaxation effects. The tion of individual bond vectors within the molecule conformation of the bound ligand can provide useful information Another exciting development is the exploitation of rela for drug design, and over the past year improvements in ects to obtain methods have made it easier to obtain quantitative information proteins -up to -150 kDa[14]. The ability of nmr to from transferred nuclear Overhauser effect experiments provide information on dynamic processes has long been recognised, and developments in NMR studies of molecu Addresses lar motion in proteins have been reviewed [15I Road, Leicester LE1 9HN, UK; e-mail: gcr@le ac uk In thinking about structure-based drug design, molecular motion in the binding site is of fundamental importance for Current Opinion in Biotechnology 1999, 10: 42-47 the access of the ligand and the kinetics of association and dissociation, and in many cases for determining the struc tural range of ligands that can bind. In this context, motions C Elsevier Science Ltd ISSN 0958-1665 in the microsecond-millisecond range are probably more important than those in the nanosecond range [16], because E nuclear Overhauser effect this is the ale in which ligand binding and dissociation take place. In structures determined by NMR, there are invariably regions that are 'less well determined than oth ntroduction ers-which differ more than the average between As recently noted in this journal by Amzel [1], it seems members of the ensemble of structures. A recent analysis kely that combinatorial chemistry and structure-based provides evidence that ill-definedregions in the methods will cease to be rival approaches to drug design NMR-derived ensemble correspond to regions une and instead will be combined to maximise the effective- relatively large-amplitude low-frequency motions zL an ness of the design process [1, 2. NMR spectroscopy has in practice the regions of the structure that are poorly deter a Valuable role to play in both approaches, and a number mined often also show relaxation behaviour characteristic of significant recent advances in its application are internal motion, The motion responsible for relaxation reviewed in this issue(see Watts, pp 48-53; Moore, effects may, however, be on a very different,shorter pp 54-58)and elsewhere [3] timescale from that responsible for the apparent'disorder' which leads to the observation of"ill defined regions in the In recent years, NMR has contributed in a major way to structure. Although relaxation experiments can qualitatively protein and nucleic acid structure determination, with demonstrate nanosecond mobility without ambiguity over 100 new structures reported annually. NMR can detailed description of its nature and frequency is more thus, like crystallography, provide the 'raw material' for ficult[15]. Rotating frame relaxation measurements afford structure-based drug design -the structure of the target a direct approach for measuring dynamics on the microsec molecule and of its complexes with candidate drugs. ond-millisecond timescale [15, 18), whereas in favour Recent examples of NMR structures of drug-target com- cases, lineshape analysis allows both rate measurement and plexes include complexes of dihydrofolate reductase [4] characterisation of the nature of the motion. For example, and HIV protease [5, 6, with inhibitors, and complexes experiments of this kind have led to the interesting conclu of antibiotics with oligonucleotides [7, 8"]. The focus sion that an arginine guanidino group and a ligand of this review will, however, be on recent developments in two other aspects of NMR in drug design: first, going and dihydrofolate reductase undergo correlated motion(191 beyond the static (or average) structure to characterise so that the ion-pair is maintained as the structure fluctuates the binding site flexibility; and second, obtaining infor- mation short of a full struct n the location of the Several examples have been reported of proteins in binding site and the conformation of the ligand within it. which parts of the ligand binding site are the
NMR spectroscopy in structure-based drug design Roberts 43 mobile parts of the structure, and become more rigid spectrum having single peaks for the backbone amide of when the ligand binds. Thus, in both cyclophilin A[20] most residues in the protein. Provided that the residues from and in the intestinal fatty-acid binding protein which all these peaks arise have been identified, changes in [21,22,23, local regions of the structure are less well the position of these peaks on addition of ligand can quickly defined than the remainder and appear by a number of give an indication of where the ligand binds. This approach criteria to be mobile. On the addition of ligand, these has been adapted for screening for lead compounds(see regions become structurally better-defined and less Moore, this issue, pp 54-58). It has also been applied at a mobile- although in the intestinal fatty-acid binding later stage in the design process to establish whether new homology(PH)and phosphotyrosine binding(PTB) or derived from chemical libraries)do bind in the same way domains [24, 25, 26], some residues involved in ligand as the parent'compound [36, 37 ]. Additional information of binding are already restricted in the unligated protein, the same kind but including backbone carbonyl as well as while others are mobile and become restricted only NH groups can be obtained from the HNCO experiment when the ligand binds [38], and the TROSY experiment [14"]will allow the exten sion of this kind of experiment to much larger proteins In all these examples, the observed mobility involves a These experiments give information only on the protcin very limited number of residues(-5-20) One function of backbone, rather than on the sidechains that may be more this localised mobility in cyclophilin and the fatty-acid directly involved in binding: an approach binding protein may be to allow access of the ligand to a sidechain interactions is selective 13C-labelling of individual rather deep binding pocket. In cases where the residues residue types(e. g. methionines)[39] mobile are directly involved in ligand bindi degree of active-site mobility is required to allow the pro- Nuclear Overhouser effects(NOEs), which provide infor- rein to bind a range of ligands, Fluctuations, perhaps mation on inthernuclear distances, have the advantage correlated, in both ligand and binding site ar s of peptide for the proximity of an atom on the ligand and one on the [27, 28 In these and other cases, a knowledge of this y Stein, and hence a clear identification of the binding site recognition, for example by SH2 and SH3 domains disadvantage of this technique at the experiments mobility could in principle be exploited in the design of required to identify and assign intermolecular NOEs can new structurally diverse ligands be more time consuming. If the ligand binds relatively weakly(Kd210-7M), the transferred NOE approach(see Although the observation of residues in the binding site below )is very useful, whereas for tightly bound ligands a xhibiting relatively large-amplitude motions in the absence number of 'editing experiments ([40,41] and references of ligand and becoming'immobilisedon ligand binding is a therein)make it possible to specifically identify NOes common one, it is far from universal. Increases in mobility of between a 1.C-and/or 15N-labelled protein and an unla some residues on complex formation have been reported belled ligand. If a sufficient number of intermolecular [28, 29, 30]. The entropy changes resulting from these NOEs can be identified, they can be used to'dock'the lig Canges in local mobility[29-31] are in several cases signifi- and into the site; this has been successfully applied to the cant in relation to the overall free energy of binding, and binding of inhibitors to dihydrofolate reductase [42-44] must be considered in a detailed analysis of the individual and stromelysin [45 and a valuable assessment of the contributions to binding. A report of the changes in mobility effectiveness of this method has been reported [461 f the binding site of SH2 domains on binding of phospho- peptides includes a valuable discussion of the relation If the protein of interest contains a paramagnetic centre, between binding-site mobility and binding energy [28""I. such as a metal ion, relaxation effects of the unpaired elec tron(s)of the metal on the nuclei of the ligand can be In addition to molecular motion per se, other dynamic to determine distances from the metal to atoms of the processes which are significant in the context of drug design bound ligand. These distances can then be used to dock can also influence NMR spectra, notably tautomeric and the ligand into the site without the need for protein reso- ionisation equilibria. These can only be inferred from X-ray nance assignment. This has recently been applied to the crystallography but can often be studied much more direct- study of substrate binding to the drug-metabolising ly by NMR, as illustrated tudies of substrate and cytochromes P450 147, 48", 49, 50,51l, offering the prospect nhibitor binding to dihydrofolate reductase [32, 33, 34 of a structurally-based understanding of the specificity of these enzymes for use in drug design Structures are not yet Location of binding sites and docking available for mammalian cytochromes P450: the NMR nMr has proved to be particularly valuable in providing a information has been used either to compare the positions rapid identification of the location of a ligand binding site. An of a number of substrates relative to the haem 51l, or, by experiment widely used in this connection is 15N-'H het- combining the nmR and homology information together to pronuclear single or multiple quantum correlation(HSOC or construct models of individual enzyme-substrate complex HMOC, respectively), which quickly [35] yield a simple es [47, 48], which simultaneously satisfy the constraints of
44 Analytical biotechnology he NMr-determined distances and of the homology with Conclusion bacterial cvtochrome P450s of known structure The diversity of applications of NMR to structure-based drug design parallels the range of its applications to protein struc- Determination of the conformations of ture in general, and both have seen significant advances over the past year. NMR is particularly useful in characterising In the absence of detailed information on the structure of the motion in proteins; understanding binding site flexibility will drug-target complex, a knowledge of the conformation of be important in designing novel ligands and, at a fundamer the bound drug for the natural ligand one wishes to mimic) tal level, in accounting for entropic contributions to affinity at can provide a valuable constraint on the drug design process d of indivi The transferred NOE complex is not available, the fact that Nmr can be used to in this regard (52-55); although limited to ligands with study a number of compounds relatively rapidly gives it an Kd210-7M, it requires relatively little protein and can be advantage in identifying the binding site for candidate ligands and determining their conformation in the bound describes its use to study the conformation of a macrolide Several recent methodological advances discussed here antibiotic bound to the ribosome 156. Recent advances in promise to improve the precision of the information obtained the quantitative analysis of the experiment have made pre. and to extend the range of applicability of the methods, lead- cise structures of the bound ligand easier to obtain. The two g to an expansion in the use of NMR as a valuable key steps in the quantitative analysis are to account for any component in the armoury of the drug designer contributions from the exchange process and to ensure that the NOEs between ligand protons that yield distance con- References and recommended reading of review straints are direct effects and not mediated by magnetisation Papers of particular interest, published within the annual perid transfer via other protons of the ligand or the protein(spin diffusion ). Identification of spin diffusion can be helped by using the rotating frame NOE experiment [57, 58), in which he indirect NOE contributions have opposite sign in con 1. Amzel LM: Structure-based drug design. Curr Opin Biotechno 1998,9:366-369 junction with the normal NOE experiment. Spin diffusion 2. integration of structure-based drug design and combi through ligand protons and exchange contributions can accounted for by using a combined relaxation and exchange chemistry for efficient drug discovery. Structure 1997,5: 319-324 matrix in the data analysis [58, 591, but spin diffusion through 3. Stockman BJ: NME the protein presents more of a problem. The most general A drug design solution is to remove the problem by per-deuteration of the tic dy -triet syste ms, incudis a nm thein aplin inv protease. r of sre- have been described for quenching spin diffusion by use of pulse to select both thesourceand'target'nuclei and thus obtain distance constraints free of the effects of spin diffu- 5. Wlodawer A, Vondrasek Inhibitors of Hiv-1 protease: a major sion: its application to NAD+ bound to lactate dehydrogenase [63] illustrates its promise in studies of An informative and critical review of the contributions of crystallography and 6. Yamazaki T, Hinck Y, Nicholson LK, Torchia DA, Wingfield Illustrative examples of recent analyses of the conformation of bound ligands by analysis of transferred NOEs include a novel cyclic urea-type inhibito etermined by nuclear magnetic resonance Protein Sci1996 studies of enzyme-substrate/inhibitor 158, 63, 64, 65]. 5:495-506 protein-carbohydrate 166,67,68""] and protein-peptide 7. Chen H, Liu X, Patel DJ: DNA bending and unwinding associated [69,70, 71,72.interactions, Among the interesting points actinomycin D antibiotics bound to partially overlapping sites om this work are evidence in several sy on DNA. J Mol Biol 1996. 258: 457-479 that two-step kinetic mechanisms(E+LAEL 6EL 8. Kumar RA, Ikemoto N, Patel DJ: Solution structure of the esperamicin A1-DNA complex. J Mol Bio/1997, 265: 173-186 where E is enzyme and L is the ligand)can have important See annotation to [g] effects on the experiment, under some circumstances the 9. Kumar RA, Ikemoto N, Patel DJ: Solution structure of conformation obtained being that in the intermediate(el) This and the prentitumg r apei B] d rather than the final complex [58, 59, 65), and evidence for the struc ligands bound in a high-energy conformation [66]. There structures provide explanation ved sequencespecificity of the are also examples where peptides corresponding to key positions the proradical centres inding loops of a protein have a different conformation oproprately for double-strand when bound to a target protein than in the free parent' pro- 0. Clore GM, Gronenborn AM: New methods of structure refinement rein [72,73]; this has obvious implications for the design of peptidomimetic inhibitors by NMR, inciwiew of some recent developments in ture determination ding the application of the methods in [12, 13
NMR spectroscopy in structure-based drug design Roberts 45 Gardner KH: Solution NMR spectroscopy beyond 25 kDa striction of loop motions upon binding inositol trisphosphate pin Struct Bio/ 1997, 7: 722-731 J Mol Biol1998,280:8798s 12. Tjandra N, Garrett DS, Gronenborn AM, Bax A, Clore GM: Defining nination from the ected diffusion anisotropy. Nat Struct Bio/ 1997, 4: 443-449 be obtained from simple (T1 T2 and NOE)relaxation measurements. Tjandra N, Bax A: Direct measurement of distances and angles in nolecules by Nmr in a dilute liquid crystalline medium. 6. Olejniczak ET, zhou M-M, Fesik SW: Changes in the NMR-derived 997278:1111-1114 is an important addition to the available methods for obtaining structural are motionally restricted in the free protein; others become more restricted information from Nmr on peptide binding -including some residues that do not contact the ligand 14. Peruvshkin K, Riek R, Wider G, Wuthrich K: Att 27. Wittekind M, Mapelli Lee V, Goldfarb V, Friedrichs MS, Meyers CA, The in shtinsmemt gans g2 267-g38o5ings and H and t-C chemical netic field leads to differing linewidths for the components of The NmR structure of this peptide-SH3 domain complex shows that the to extend substantially the upper size limit for studying proteins by nm ity of fluctuating hydrogen bonding interactions is discussed in which a protein can interact with alternative acceptors on the Palmer AG Il; Probing molecular motion by NMR. Curr Opin Struct 28. Kay LE, Muhandir Wolf G, Shoelson SE, Forman- Kay JD studying internal motion in proteins. 8:io 1998. at sH2 domain 16. Jardetzky O: Protein d allosteric proteins. Prog Brophys Mol Bio/ 1996, 65: 171-218 L, Hilbers Cw, Nilges M: Esser le discussion of the rela- y NMR structure ensembl tionships between dynamics, binding thermodynamics and specificity S Whitman CP: 15N NMR of the structural ensembles obtained by nmr amics and entropy changes of an kke M, Bruschweiler R, Palmer AG ll: NMR order parameters and criterion for inclusion of structures in the NMI Ca2+ binding by calbindin D9k. J Am Chem Soc 1993 ensemble to ensure that amplitudes of motion are better represented vector fluctuations measured from NMR- derived order ibronectin type lll domain. Nat Struct Bio/ 1998, 5: 55 59 protein folding. J Mol Bio/ 1996 Nieto PM, Birdsall B, Morgan WD, Frenkiel TA, Gargaro AR, Feeney 3: 32. Feeney J: NMR studies of interactions of ligands with dihydrofolate reductase, Biochem Pharmaco/ 0:141-152 and carboxylate groups in protein ligand complexes. FEBS Le 997,405:1620 eductase and their Design. Edited by Craik D. New York mation of unligated human cyclophilin A. J Mol Bio/ 1997 2726481 34. Birdsall B, Casarotto MG, Cheung AHT, Basran J, Roberts GCK 21. Hodson ME Discrete backbone disorder in the nuclear feeley j: The influence of aspartate 26 on the tautomeric forms of tobacillus casei dihydrofolate reductase. FEBS the mechanism of ligand entry. An illustration of the value of NMR in identifying 97.36:1450-1460 of folate and NADP+ with th See annotation to 22 22. Hodson ME Cistola, DF Iters the backbone conformations g protein as monitored by of the complex, and selectively [3C-Aspl-labelled enzyme was used to study veste aspartIc 35. Ross A, Slazmann M, Senn H: Fast-HMQc using Ernst angle acid binding protein and the way in proteins. J Biomol NMR 389396. required for access of the fatty acid 36. Tilley JW, Chen L, Fry DC on SD, Powers GD, Biondi D. tification of a 3. Zhang F, Lucke C, Baier LJ, Sacchettini JC, Hamil which binds to IL-2. J Am Chem Soc 1997,119: 7589-7590 See annotation to[22·」 bor TM, Feng S, Shirai F, Schre ased selection 266:173-194. 38. Farmer B Goldfarb V, Friedrichs MS, Wittekind M 25. Gryk MR, Abseher R, Simon B, Nilges M, Oschkinat H e MurB enzyme by nmR. Nat Struct Bio! 1996 Heteronuclear relaxation study of the Ph domain of B-spectrin 3:995997
ytical biotechnology 39. Kleerekoper Q, Liu w, Choi D, Putkey JA: Identification of binding 56. Bertho G, Ghart es for bepridil and trifluoperazine on cardiac troponin C. J Biol Chem1998,27381538160. 二邮 cor tion between antibiotic 40. Otting G, wuthrich nuclear filters in two-dimensional ations. Bioorg Med Chem 1998 6209221 for studies 1990233996 entification of pr C, Legault P, Vincent SJF, Greenblatt J, Konrat R, Kay LE: eson B1995,106:195198 ex by transferred ROESY. J Magn NMR spectroscopy: application to a bacteriopha B RNA complex. J Am Chem Soc 1997. 58. Barsukov IL, Lian L-Y, Ellis J, Sze K-H. Shaw WV. Roberts GCK: The enzyme a bound to chloramphenic rase determined by transferred NOE experiments. 2. Martorell G, Gradwell M, Birdsall B, Bauer C, Frenkiel TA, J Mol Bio11996,262:543558 T, Palmer AG lll, Mc Dermott AE: Conformation dihydrofolate reductase. Biochemistry 1994, 33: 12416-12426. ar Overhauser studies, Biochemistry 1997 36:14202 ucture of a brodimoprim analog in its 60. Massefski W. Redfield us casei dihydrofolate reductase. Bio mination of multiple-step spin 34:1169011702 Johnson」 EM, Wnight JE, Pardo J, Rosowsky A, Wagner G 61. Zonai z, Juranic N, Markley JL, Macura S: Magnetization exchange NADPH ture of the antitumor compound PT523 and network editing- mathematical principles and experimental demonstration. Chem Phys 191 reductase msty199736:43994411 45. Gonnella NC, Li YC, Zhang XL, Paris CG: Bioactive conformation of "assition of spin diffusion and multidimensional NMR spectroscopy. Bioorg Med Chem 1997 46. Gradwell M, Feeney J: Validation of the use of intermolecular NOE docked structures of protein ligand at/ Acad Sci USA 1997. 9 omplexes, J Biomol NMR 1996. 7: 48-58 This paper describes the application of the QUIET-NOESY experiment to 47. Modi S, Paine ML, Sutcliffe M, Lian L-Y, Primrose wU. Wolf CR an cytochrome P450 2D6 based on homology modelling and NMR studies of substrate binding. Biochemistry 1996, 35: 4540-4550. 48. Modi S, Gilham DE, Sutcliffe M, Lian L-Y, Primrose WU. Wolf CR nyl-1, 2, 3, 6-tetrahydropyridi glycosidic bond was not, and the coexistence of two co ations about this bond in the complex is suggested 2D the presence of pound to e dis. 64. Cox Ra,ycosiasu EH: Biologically important conformations of er effect spectroscopy Ba by transferred nuclear 36:23532359 ty1997 of the paramagnetic relaxation approach in studying the specificity65 M, Paci M, Sette M, Oakley Aj, Parker MW. Lo Bello M Primrose WU, Lian L-Y, Roberts GCK o human glutathione transferase P m of cytochrome p450 BM3 involves a 6A ments with the crystal st 1996,3:41 bound substrate on reduction Nat Struct Bio/ Transferred NOE experiments lead to on of bound glutathione tion but sidechain conformations 0. Oliver CF, Modi S, Sutcliffe M, Primrose WU, Lian L-Y. Roberts GCK The origins of this difference ar A single-residue mutation of docking calculations and kinetic experiments; it is con- cluded that the NMR structure may correspond to that of a'pre-complex' 36:1567-1572 51. Pol-Scaife S, Attias R, Dansette PM, Mansuy D: The substrate mM0门hmR nergy conformation of C-lactose, a non hydrolyzable substrat nalogue. NMR and m te(e. g. the location of a putative positive charge)from the of Chem Soc1998,120:1309-1318 52. Albrand JP, Birdsall B, Feeney J, Roberts CK, Burgen ASV: The use us repla indicate that C-lactose (in which the glyco- transferred nuclear Overhauser effects in the study of th tosidase adopts a conformation which is only slightly populated (-5%)in the free ligand all molecules bound to proteins. int JBi 4 67. Montero M, Vallmitjana M, Perez-Pons JA, Querol confor cellobiose bound to a B-glucosidase from Streptomyces sp. FEBS Le1998,421:243248 982,48:402417 54. Campbell AP, Sykes BD: The two-dimensional transferred nuclear L-selectin. J Am Chem Soc 1997.119: 1727-1736 Biomol Struct 1993, 22: 99-122 55. NiF: Recent advances in transferred NOE methods. Prog NMR ed NOE experiments Spectrosc 1994 informers of the tetrasaccharide are recognised by E- and P-selectin on
NMR spectroscopy in structure-based drug design Roberts 47 69. LaPlante SR iizachew D. Moffett DB, Bus Wester WM. Dratz EA e conformation of the CD4 36-59 Transferred NOE experiments, including identification of spin diffusion mali F Do a-T. Doan B-T A, Palmas P. Sk te MR and MD study Proteins crystal structures of CD4, suggesting a change in conformation on binding Two peptide inhibitors of factor Xa, obtained by library screening, were stud. 73. Benitez BAS. ith EGF-like domain of thrombomodulin an EGF-like domai isolation and in the binding site (see also [58]), systematica d ng possible orientation of the 273:913-926. otained, suggesting that the inhibitors bind differently from the substrate, The structure of a loop in this domain which forms part of its thrombi made it possible to explain the specificity of the inhibitors. 71 but clearty not identical. This and the on from the Nmr structure of preceding paper [72] show the value of transferred NOE experiments in band 3 peptide inhibitor bound to glyceraldehyde-3phosphate g access to the peptide conformation in the bound state, as a me dehydrogenase. Biochemistry 1998, 37: 867-877 tic drugs