COMMUNICATION

Unmasking Ligand Binding Motifs: Identification of aChemokine Receptor Motif by NMR Studies ofAntagonist Peptides

Valerie Booth1, Carolyn M. Slupsky1, Ian Clark-Lewis2 andBrian D. Sykes1*

1Department of BiochemistryProtein Engineering Networkof Centres of ExcellenceUniversity of Alberta713 HMRC EdmontonAlta., Canada T5G 2S2

2Biomedical Research Centreand Department ofBiochemistry and MolecularBiology, University of BritishColumbia, VancouverBC, Canada V6T 1Z3

Determining the critical structural features a ligand must possess in orderto bind to its receptor is of key importance to the understanding ofvital biological processes and to the rational design of small moleculetherapeutics to modulate receptor function. We have developed a generalstrategy for determining such ligand binding motifs using low tempera-ture NMR structures of peptides with the desired receptor bindingproperties. This approach has been successfully applied to determine abinding motif for the chemokine receptor CXCR4. The motif identifiedprovides a detailed guide for the design of small molecule antagonistsagainst CXCR4, which are much sought after to aid in the treatment ofa number of conditions including human immunodeficiency virus type 1infection and a variety of cancers.

q 2003 Elsevier Science Ltd. All rights reserved

Keywords: HIV-1; CXCR4; drug development; peptide; NMR*Corresponding author

Membrane protein receptors are a class of pro-teins that are pivotal for numerous biologicalfunctions. Knowledge of the structural basis forligand–receptor binding and activation is criticalto the understanding of the biological functions ofthese receptors and to the design of small moleculemimetics for therapeutic intervention. However,the very sparse structural information availablefor the receptors is of limited use in understandingthe details of ligand–receptor interactions. Thus, itmay be advantageous to focus instead on ascertain-ing the critical structural characteristics of theligands involved in binding the receptor.

In many instances, peptides, either natural orengineered, can bind to the receptor of interest asagonists or antagonists. However, there is a funda-mental difficulty in using such peptides to identifystructural features important for binding receptors.In general, peptides in isolation possess a highlydisordered structure and only fold into a definedstructure upon binding to their targets. Formation

of a defined peptide structure requires the estab-lishment of enough favorable enthalpic contacts,both intra-peptide and peptide–receptor, to over-come the unfavorable change in entropy uponfolding into the receptor bound conformation.However, as illustrated by the Gibb’s equation forfree energy, DG ¼ DH 2 TDS; the effects of entropycan be reduced by lowering the temperature.Thus, we have employed a strategy where nuclearmagnetic resonance (NMR) spectroscopy is usedto determine the structure of peptides at lowtemperature (5–8 8C), a condition that ofteninduces the formation of a defined conformationor set of conformations.1,2 NMR is a method ofhigh sensitivity for detecting and describing foldedconformations, even if these conformations consistof only a small fraction of the total peptide popu-lation. Such nascent structure has been the objectof much study in relation to protein folding andprotein interactions.1,3 – 7 The energetically favorableconformation of the peptide in isolation at lowtemperature is determined by favorable intra-peptide contacts and the low energy conformationof the peptide in complex with the receptor isdetermined by these same intra-peptide contactsplus additional receptor–peptide contacts. There-fore, it is not unreasonable to hypothesize that theconformation of the peptide will be similar in both

0022-2836/03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved

E-mail address of the corresponding author:brian.sykes@ualberta.ca

Abbreviations used: NMR, nuclear magneticresonance; DSS, 2,2-dimethyl-2-silapentane-5-sulfonicacid; TOCSY, total correlation spectroscopy; ROESY,rotating frame Overhauser effect spectroscopy.

doi:10.1016/S0022-2836(03)00094-9 J. Mol. Biol. (2003) 327, 329–334

cases. This has been shown to be the case, forexample with the Hox peptide, for which thelow temperature conformation and target proteinbound conformation of the peptide have bothbeen determined and appear to coincide.8 Hence,by reducing the entropy cost of forming a definedstructure, structures of peptides determined at

low temperatures may “unmask” the physio-logically relevant conformation of a peptide(shown schematically in Figure 1).

We have successfully applied this method topeptides derived from chemokines. Chemokinesare small, secretory proteins that play a key role inimmune and inflammatory responses, via theirability to recruit and activate specific subsets ofleukocytes.9 They function by binding to and acti-vating chemokine G protein-coupled receptors.Although they display no defined structure undernormal conditions for protein structure determi-nation, the N-terminal residues of chemokinesare critical for the receptor interaction.10 Peptidesderived from the N termini of chemokines retainthe ability to bind the receptor, although withlower affinity than the full length chemokine.11

We have previously used NMR to solve, atlow temperature, the structures of two N-terminalpeptides that bind the chemokine receptorCXCR4: SDF-1 (residues 1–9)12 and vMIP-II (resi-dues 1–10).13 Here, we present the 5 8C NMR solu-tion structure of a third chemokine N-terminalpeptide, ITAC residues 1–8, that has similarCXCR4 binding activity. The ITAC peptide struc-ture has allowed us to definitively identify astructural motif common to all three peptides thatdefines the structural motif these peptides use tobind the CXCR4 receptor.

The three peptides were tested for their ability tocompete with binding of the full-length chemokineSDF-1 to its receptor CXCR4 (Figure 2). They wereall able to compete with full-length SDF-1 forCXCR4 binding, although as expected, they bound

Figure 1. Although a peptide may possess no definedstructure at physiological temperatures (A), coolertemperatures may induce the formation of structure (B)that is similar to the conformation of the peptide inits receptor bound state (C).

Figure 2. Competition forbinding to CXCR4 of 125I-labeledSDF-1 was measured forunlabeled SDF-1 (B), ITAC 1-8(V), SDF-1 1-9 (P), and vMIPII1-10 (O); at the indicated con-centrations. The control (A) hadmedium alone. The binding isexpressed as a percentage ofmaximal specific counts bound(control), after subtraction of thebackground. Competition assaysfor binding of 125I-labeled SDF-1was performed with CXCR4-transfected B-300-19 cells(CXCR4-B300-19). SDF-1 waslabeled using the Bolton-Hunterprocedure.14 Briefly, 2 £ 106 cellswere incubated with 3 nM oflabeled chemokine and increas-ing concentrations of unlabeledcompetitor (1029–3 £ 1024 M)

in 200 ml of Hank’s balanced salt solution containing 25 mM Hepes (pH 7.4), 10 mg ml21 bovine serum albumin, 0.1%(w/v) and sodium azide. The incubations were carried out for 30 minutes at 4 8C and cell-associated radioactivitywas separated by centrifuging the cells through a 2:3 mixture of diacetylphthalate/ dibutylphthalate and the 125Icounts measured in a gamma counter. The non-specific counts (background) were determined in the presence of a100-fold excess concentration of unlabeled SDF-1 (0%), and was subtracted from each measurement. The maximalbinding was determined using medium alone (control), and after subtraction of background, this value was designated100%.

330 Unmasking Ligand Binding Motifs

with less affinity than the full length protein. Theseresults show that the CXCR4 binding site used bythe peptides is at least partially overlapped withthe CXCR4 site used by full-length SDF-1. The pep-tides bind CXCR4 with similar affinities to eachother, which is consistent with their sharing thesame binding site on CXCR4.

The NMR structures of all three peptides areshown, in Figure 3, as ensembles of structuresthat fulfill the experimentally derived restraints.Although the variation amongst each peptide’sstructure ensemble indicates a degree of flexibilityin the peptides, the peptide structures are stillsignificantly restricted in the conformational spacethey sample. In all three peptides there is a regionof particularly well-defined structure where thebackbone is bent into a semi-compact confor-mation. We hypothesize that such semi-compactconformations reflect the conformation of the pep-tides in complex with the receptor. This hypothesisis strongly supported by the finding that the con-formation of the side-chains in this structurallywell-defined region is remarkably similar in allthree peptides and thus appears to define anessential ligand-binding motif.

Figure 4(A) shows the alignment of backboneatoms of residues in the structurally well-definedregions of the peptides: residues 2–6 of the ITACpeptide, 5–9 of the SDF-1 peptide and 4–8 of thevMIP-II peptide. When aligned in this manner,the peptides display a striking similarity in theposition of three of the side-chains (Figure 4(B)),namely Met3, Phe4 and Lys5 of ITAC, Leu5, Tyr7,Arg8 of SDF-1 and Ser4, His6, Arg7 of vMIP-II.In two out of the three peptides, the first aminoacid residue in this motif is hydrophobic. In threeout of three peptides, the second amino acidresidue contains an aromatic ring and the thirdcontains a basic side-chain. Although the peptidespossess a certain amount of structural flexibility,in the region of the three aforementioned residues,the ensembles of each peptide compare well toeach other (Figure 4(B)) and structures can bechosen from each ensemble that align very well(heavy side-chain atom, RMSD ,1 A) (Figure4(C)), indicating that the motif side-chains of thethree peptides sample overlapping conformationalspace. That all three peptides possess the samestructure for these side-chains, supplies persuasiveevidence that these three side-chains, in thisparticular conformation, form a CXCR4 receptor-binding motif.

Because the peptides do sample a subset of con-formational space, rather than possessing a singledefined structure, there could be some concernthat the similarity in positioning observed for thethree motif side-chains is fortuitous, rather thanindicating a functionally relevant similarity. Toaddress this issue, ensembles of randomized, butenergetically feasible, structures were calculatedfor each peptide. Figure 5 shows that for therandom structures, only 7% of the pairs alignedwith RMSD values of less than 1.5 A, while for theexperimentally determined structures, 33% of thepairs aligned with RMSD values of less than 1.5 A.Additionally, only 0.5% of the pairs of experi-mentally determined structures had RMSD valuesof greater than 2.5 A, while 27% of the randomstructure pairs had RMSD values of greater than2.5 A. Thus, the conformational space sampled by

Figure 3. Ensembles of structures satisfying the NMRexperimental restraints for ITAC 1-8 (A), SDF-1 1-9 (B)and vMip-II 1-10 (C). The structures shown in heavylines indicate a representative structure for each peptide,with the backbone shown in yellow. Residues 1 and 2 ofSDF-1 and vMip-II are not shown due to their excessivedisorder. The peptides were synthesized by stepwisesolid phase methods using t-butoxycarbonyl protectionchemistry. After hydrogen fluoride deprotection, thepeptides were purified by reverse-phase HPLC. Thesamples were lyophilized, stored at 4 8C and later dis-solved in buffer containing 20 mM deuterated sodiumacetate, 1 mM 2,2-dimethyl-2-silapentane-5-sulfonic acid(DSS), 1 mM sodium azide, 90% H20/10% 2H2O andadjusted to pH 5.0. NMR spectra (2D homonucleartotal correlated spectroscopy (TOCSY) with mixing time60 ms, 2D homonuclear rotating frame Overhauser effectspectroscopy (ROESY) with mixing times 300, 400, 500and 600 ms) were acquired on a 600 MHz Varianspectrometer at 5 8C. Assignment of the backbone andside-chain protons was accomplished using TOCSY andROESY. Fifty-five structures were calculated using simu-lated annealing protocols implemented with CNS,23

with 58 ROE-derived distance restraints (38 sequential,20 medium-range). All of the structures fulfilled theexperimental restraints without violations. ROE build-up curves for the four different mixing times wereexamined to ensure that the observed ROEs were notthe consequences of spin diffusion.

Unmasking Ligand Binding Motifs 331

each peptide is limited and this subset of confor-mational space shows a high degree of overlapbetween the different peptides. It should be notedthat for a peptide population containing bothunfolded and compact conformations, structureensembles calculated based on NMR data will be

enriched in the compact conformations. This,however, is inconsequential in this study, as thepoint of interest is the identification of particularcompact structures, rather than the estimation ofwhat fraction of the peptide is in the compactstructure.

It should be noted that while three contiguousresidues of ITAC form the receptor-binding motif,SDF-1 and vMIP-II have an extra residue insertedbetween the first and second residues of thereceptor-binding motif. As well, the position ofthe motif, with respect to the position of theconserved N-terminal cysteine residues of thefull-length chemokines, is varied among the threechemokines:

This illustrates the great difficulty in identifyingsuch binding motifs from sequence informationalone, as there may be many peptide sequencespossible that lead to the presentation of thesethree motif side-chains in this configuration. Italso has some interesting ramifications in terms ofchemokine function. Assuming the N-terminalregions of the intact chemokines bind the receptorsin the same way the peptides do, the variation inthe location of the motif within the N-terminalregion will lead to a different positioning of themain body of the chemokine with respect to thereceptor and this may bring about variations inhow the three chemokines function.

Although all three peptides are antagonists forthe receptor CXCR4, there are differences in howthe proteins from which they are derived function.vMIP-II is a viral chemokine and both full lengthvMIP-II and the peptide consisting of residues1–10 are antagonists for the receptor CXCR4,13 i.e.they bind to the receptor but do not induce acti-vation. Full length SDF-1 and the SDF-1 1-9 peptidebind CXCR4 as well, but are agonists, i.e. they bindto the receptor and induce activation.11 SDF 1-9 can

Figure 4. (A) Backbone alignmentof ITAC (magenta), SDF-1 (blue)and vMip-II (green) peptide struc-ture families. The N-terminal tworesidues of SDF-1 and one residueof vMip-II are not shown. (B) Struc-tures aligned according to theposition of the three motif side-chains, with the backbone shownin grey. (C) One well-aligning struc-ture from each peptide structurefamily. (D) Alignment of the drugAMD-3100 (orange) with ITAC(side-chains magenta; backbonegrey). Blue spheres indicate thelocation of nitrogen atoms.

Figure 5. Similarity between inter-peptide structurespairs (i.e. ITAC/SDF-1 SDF-1/vMip-II and vMip-II/ITAC) for the set of experimentally derived structuresand a set of randomized, but energetically feasible, struc-tures. The structures were calculated using X-PLOR3.851,24 by starting with an extended peptide with idealgeometry and randomly rotating all the f and c dihedralangles. After refinement, this set of structures possesseda high degree of structural variability, but were similarto the experimentally derived structures in their non-nuclear Overhauser enhancement energy profiles andpercentage of residues in the most favored and addition-ally allowed regions of the Ramachandran plot. Thisrandom set of peptide structures was then analyzed toascertain the number of well-aligning inter-peptidepairs. Pair-wise RMSD values were calculated using thefollowing atoms: ITAC residue 3 Cb Cg, residue 4 Cb Cg

Cd1 Cd2, residue 5 Cb Cg Cd C1; SDF-1 residue 5 Cb Cg, resi-due 7 Cb Cg Cd1 Cd2, residue 8 Cb Cg Cd N1; vMIP-II resi-due 4 Cb Og, residue 6 Cb Cg Nd1 Cd2, residue 7 Cb Cg Cd

N1.

332 Unmasking Ligand Binding Motifs

be converted into an antagonist by mutations inresidues 1 and 2.14 These residues of SDF-1 do notcontribute to the binding motif identified in thisstudy, implying that the motif side-chains identi-fied are involved in binding, but not activation.Full length ITAC binds CXCR3 as an agonist andthe ITAC 1-8 peptide binds to a broad range ofchemokine receptors: to CXCR3 as an agonist andto CCR1, CCR2, CCR3, CCR5 and CXCR4 as anantagonist (I. C.-L., unpublished data). The dif-ference in the positioning of the receptor-bindingmotif within the N terminus of these chemokinesmay partially explain the difference in functioningbetween the three proteins.

The development of small molecule antagonistsagainst chemokine receptors is of great interest inthe treatment of several diseases. The bindingmotif identified here provides a detailed startingpoint for the rational design of such molecules.CXCR4 small molecule antagonists have beenmuch sought after in the treatment of AIDSbecause proteins and peptides that bind CXCR4have been shown to prevent entry into the cell oflate infection strains of HIV-1.15 There is alsopotential for CXCR4 antagonists in the treatmentof cancer, as CXCR4 has been implicated in meta-static spreading of tumors in several types ofcancer.16 Other possible applications for manipu-lation of CXCR4 activation are in improving theoutcome of clinical stem cell transplantation,17

treatment of malaria18 and in attenuating allergiclung inflammation.19

Although a number of small molecule antagon-ists to CXCR4 have been developed, none haveyet been presented that possess the features necess-ary for a therapeutic agent, such as high efficacy,tolerance and bio-availability.20 Some small mol-ecules and peptides that have been developed dopossess at least some of the structural featuresdefined by the peptide motif uncovered in thispaper. For example, the peptide ALX40-4C21

consists of several tyrosine and lysine residues.The CXCR4 antagonist AMD310022 consists of twoazamacrocycles (large, nitrogen-containing cyclicgroups), each containing an excess charge of þ4at physiological pH, connected by a phenyl ring.If the phenyl group of AMD3100 is placed on topof the ring from the peptide receptor-bindingmotif, part of one of the positively charged aza-macrocycle groups follows the path of the basicside-chain of the peptide motif (Figure 4(D)). Themotif described here will help in refining theselead drugs by indicating which structural featuresof the drug can be modified in order to improveits pharmaceutical behavior and which are essen-tial to maintain CXCR4 antagonist activity. Theidentified motif should also aid in the choice ofnew compounds to be tested for CXCR4 antagonistactivity.

In summary, we have shown that low tempera-ture NMR structures of peptides can indicate theimportant structural features responsible for aligand’s ability to bind a receptor. Knowledge of

these structural features is key to understandingbiologically important ligand–receptor interactionsand to rational drug design. This method has beensuccessfully applied to define a structural motifthat binds the chemokine receptor CXCR4, a targetof much pharmaceutical interest. We anticipatethis method will be of utility in many othersystems where peptides that bind a receptor ofinterest are available.

Acknowledgements

This work was funded by the Protein Engineer-ing Network of Centres of Excellence. V.B. is alsosupported by the Canadian Institutes for HealthResearch and the Alberta Heritage Foundation forMedical Research.

References

1. Wright, P. E., Dyson, H. J. & Lerner, R. A. (1988).Conformation of peptide fragments of proteins inaqueous solution: implications for initiation ofprotein folding. Biochemistry, 29, 7167–7175.

2. Brown, J. E. & Klee, W. A. (1971). Helix-coil tran-sition of the isolated amino terminus of ribonuclease.Biochemistry, 10, 470–476.

3. Baldwin, R. L. & Zimm, B. H. (2000). Are denaturedproteins ever random coils? Proc. Natl Acad. Sci.USA, 97, 12391–12392.

4. Baldwin, R. L. & Rose, G. D. (1999). Is protein foldinghierarchic? I. Local structure and peptide folding.Trends Biochem. Sci. 24, 25–33.

5. Dyson, H. J. & Wright, P. E. (2002). Coupling offolding and binding for unstructured proteins. Curr.Opin. Struct. Biol. 12, 54–60.

6. Dobson, C. M. & Karplus, M. (1999). The funda-mentals of protein folding: bringing together theoryand experiment. Curr. Opin. Struct. Biol. 9, 92–101.

7. Ma, B., Shatsy, M., Wolfson, H. J. & Nissinov, R.(2002). Multiple diverse ligands binding at a singleprotein site: a matter of pre-existing populations.Protein Sci. 11, 184–197.

8. Slupsky, C. M., Sykes, D. B., Gay, G. L. & Sykes, B. D.(2001). The HoxB1 hexapeptide is a prefoldeddomain: implications for the Pdx1/Hox interaction.Protein Sci. 10, 1244–1253.

9. Fernandez, E. J. & Lolis, E. (2002). Structure,function, and inhibition of chemokines. Annu. Rev.Pharmacol. Toxicol. 42, 469–499.

10. Loetscher, P. & Clark-Lewis, I. (2001). Agonistic andantagonistic activities of chemokines. J. Leukoc. Biol.69, 881–884.

11. Loetscher, P., Gong, J. H., Dewald, B., Baggiolini, M.& Clark-Lewis, I. (1998). N-terminal peptides ofstromal cell-derived factor-1 with CXC chemokinereceptor 4 agonist and antagonist activities. J. Biol.Chem. 273, 22279–22283.

12. Elisseeva, E. L., Slupsky, C. M., Crump, M. P., Clark--Lewis, I. & Sykes, B. D. (2000). NMR studies of activeN-terminal peptides of stromal cell-derived factor-1.Structural basis for receptor binding. J. Biol. Chem.275, 26799–26805.

Unmasking Ligand Binding Motifs 333

13. Crump, M. P., Elisseeva, E., Gong, J., Clark-Lewis, I.& Sykes, B. D. (2001). Structure/function of humanherpesvirus-8 MIP-II (1-71) and the antagonistN-terminal segment (1-10). FEBS Letters, 489,171–175.

14. Crump, M. P., Gong, J. H., Loetscher, P., Rajarathnam,K., Amara, A., Arenzana-Seisdedos, F. et al. (1997).Solution structure and basis for functional activityof stromal cell-derived factor-1; dissociation ofCXCR4 activation from binding and inhibition ofHIV-1. EMBO J. 16, 6996–7007.

15. Horuk, R. (2001). Chemokine receptors. CytokineGrowth Factor Rev. 12, 313–335.

16. Baggiolini, M. (2001). Chemokines in pathology andmedicine. J. Intern. Med. 250, 91–104.

17. Petit, I., Szyper-Kravitz, M., Nagler, A., Lahav, M.,Peled, A., Habler, L. et al. (2002). G-CSF inducesstem cell mobilization by decreasing bone marrowSDF-1 and up-regulating CXCR4. Nature Immunol. 3,687–694.

18. Garnica, M. R., Souto, J. T., Silva, J. S. & de Andrade,H. F. (2002). Stromal cell derived factor 1 synthesisby spleen cells in rodent malaria, and the effects ofin vivo supplementation of SDF-1alpha and CXCR4receptor blocker. Immunol. Letters, 83, 47–53.

19. Lukacs, N. W., Berlin, A., Schols, D., Skerlj, R. T. &Bridger, G. J. (2002). AMD3100, a CXCR4 antagonist,attenuates allergic lung inflammation and airwayhyperreactivity. Am. J. Pathol. 160, 1353–1360.

20. Condra, J. H., Miller, M. D., Hazuda, D. J. & Emini,E. A. (2002). Potential new therapies for the treat-ment of HIV-1 infection. Annu. Rev. Med. 53, 541–555.

21. O’Brien, W. A., Hartigan, P. M., Martin, D., Esinhart,J., Hill, A., Benoit, S. et al. (1996). Changes in plasmaHIV-1 RNA and CD4 þ lymphocyte counts andthe risk of progression to AIDS. Veterans AffairsCooperative Study Group on AIDS. J. Virol. 70,2825–2831.

22. De Clercq, E., Yamamoto, N., Pauwels, R., Balzarini,J., Witvrouw, M., De Vreese, K. et al. (1994). Highlypotent and selective inhibition of human immuno-deficiency virus by the bicyclam derivative JM3100.Antimicrob. Agents Chemother. 38, 668–674.

23. Brunger, A. T., Adams, P. D., Clore, G. M., DeLano,W. L., Gros, P., Grosse-Kunstleve, R. W. et al. (1998).Crystallography & NMR system: a new softwaresuite for macromolecular structure determination.Acta Crystallog. sect. D, 54, 905–921.

24. Brunger, A. T. (1993). X-PLOR Manual, Yale Univer-sity, New Haven, CT.

Edited by M. F. Summers

(Received 19 September 2002; received in revised form 18 December 2002; accepted 18 December 2002)

334 Unmasking Ligand Binding Motifs