Download Transcriptional repressor CopR Structure model-based localization of the deoxyribonucleic ...
PROTEINS: Structure, Function, and Genetics 38:361–367 (2000)
Receptor-Binding Conformation of the “ELR” Motif of IL-8: X-Ray Structure of the L5C/H33C Variant at 2.35 Å Resolution Nancy Gerber,1 Henry Lowman,1 Dean R. Artis,2 and Charles Eigenbrot1* 1 Department of Protein Engineering Genentech, Inc., South San Francisco, California 2 Department of Bioorganic Chemistry, Genentech, Inc., South San Francisco, California
ABSTRACT The “ELR” (Glu-Leu-Arg) tripeptide sequence near the N-terminus of interleukin-8 (IL-8) contributes a large part of the receptor binding free energy. Prior X-ray and nuclear magnetic resonance (NMR) structures of IL-8 have shown this region of the molecule to be highly mobile. We reasoned that a hydrophobic interaction between the leucine and the neighboring ␤-turn might exist in the receptor binding conformation of the Nterminus. To test this hypothesis, we mutated two residues to cysteine and connected the N-terminus to the ␤-turn. The mutant retains receptor binding affinity reasonably close to wild type and allows the characterization of a high-affinity conformation that may be useful in the design of small IL-8 mimics. The L5C/H33C mutant is refined to R-values of R ⴝ 20.6% and Rfree ⴝ 27.7% at 2.35 Å resolution. Other receptor binding determinants reside in the “N-loop” found after “ELR” and preceding the first ␤-strand. All available structures of IL-8 have been found with one of two distinct N-loop conformations. One of these is relevant for receptor binding, based on NMR results with receptor peptides. The other conformation obscures the receptor-peptide binding surface and may have an undetermined but necessarily different function. Proteins 2000;38:361–367. ©
2000 Wiley-Liss, Inc.
Key words: minimization; inhibitor; proline; rational; design; intermolecular INTRODUCTION Interleukin-8 (IL-8) is a well characterized proinflammatory member of the family of chemoattractant proteins called chemokines.1 Biological effects of IL-8 result from it binding to extracellular regions of seven-transmembrane, G protein-linked receptors IL8-RA and IL8-RB with nanomolar affinity. The various structurally characterized members of the family differ in their extent of oligomerization (monomers, dimers, and tetramers) and/or the surface used for the oligomerization interaction.2– 4 All the monomers, however, have a three-stranded ␤-sheet, followed by an ␣-helix and preceded by an extended N-terminal segment of approximately 20 amino acids. Four subclasses are currently discriminated on the basis of their pattern of cysteines near their N-terminus. IL-8 is a “CXC” chemokine, and its wild-type disulfides link Cys7 to Cys34 and ©
2000 WILEY-LISS, INC.
Cys9 to Cys50. The X-ray and nuclear magnetic resonance (NMR) structures determined for IL-8 both have a widely variable N-terminus up to the first cysteine.5,6 Receptor binding determinants on IL-8 are restricted to two distinct parts of the N-terminal polypeptide preceding the first ␤-strand, referred to as “ELR” (Glu4-Leu5-Arg6) and the N-loop. The N-loop receptor binding determinants include most or all of the residues from about Ser14 to Lys20.7–9 The ELR site shows the largest effects of single Alanine substitutions,10 –13 but the N-loop epitope clearly offers additional binding and specificity interactions. Despite the predominance of ELR side chains in receptor interactions, simple peptide mimics of this motif have negligible affinity for the IL-8 receptors, suggesting a special three-dimensional context in the intact protein. The possibility that the disulfides play a distinctly functional role in addition to their important structural role has recently been tested.14 The structural variability of the extreme N-terminal region of IL-8 has hampered efforts to characterize and exploit the conformation required for high-affinity receptor binding. In an attempt to circumvent this limitation and provide a structural template for design of small mimics of IL-8, we assigned a key structural role, but not a receptor interaction, to the side chain of Leu5. We have made the L5C/H33C mutant of IL-8 that adds a significant structural constraint to the ELR epitope in the context of the intact protein and report on its receptor affinity and complete structural characterization using X-ray crystallography. MATERIALS AND METHODS Design of the Restricted N-Terminus IL-8 Mutant L5C/H33C We hypothesized that the receptor-binding conformation of the IL-8 N-terminal tripeptide Glu4-Leu5-Arg6 (ELR) might involve a hydrophobic interaction between Leu5 and the adjacent ␤-turn containing His33 reminiscent of that found for the related chemokine MGSA.15 According to the hypothesis, Leu5’s putative interaction
Grant sponsor: National Science Foundation; Grant number: DMR9311772; Grant sponsor: National Institutes of Health; Grant number: RR-01646. *Correspondence to: Charles Eigenbrot, Department of Protein Engineering, Genentech, Inc., One DNA Way, South San Francisco, CA 94080. E-mail: [email protected]
Received 26 July 1999; Accepted 21 October 1999
N. GERBER ET AL.
Fig. 2. Refined isotropic temperature factors plotted as the average of main-chain (solid) and side-chain (dotted) atoms.
TABLE I. Refinement Statistics for L5C/H33C IL-8
Fig. 1. Electron density maps of L5C/H33C IL-8 contoured at 1.0 rmsd. a: The initial 2Fo-Fc map at 3.0 Å resolution after rigid body refinement of the molecular replacement solution. The search model had residue 33 as alanine and did not include N-terminal residues before Cys7. This map shows that a change to Cysteine at residue 33 and N-terminal additions are warranted. b: Final 2Fo-Fc map and model in the same region.
with the ␤-turn serves to orient the Glu4 and Arg6 side chains but does not itself contact receptor. His33 seems not very important to receptor interactions, because mutation
Refinement resolution (Å) No. of reflections (F/F ⱖ 2) R-value (F/F ⱖ 2) (%) R-free (F/F ⱖ 2) (%) R-value (I ⬎ 0) (all) (%) R-value with no waters (all) (%) rmsd bond distances (Å) rmsd bond angles (°) rmsd (°) rmsd dihedrals (all) (°) rmsd improper dihedrals (°) No. of atoms No. of atoms occ. ⫽ 0 No. of residues No. of waters No. of sulfate ions rmsd main chain B-factors (Å2) (bond) (angle) rmsd side-chain B-factors (Å2) (bond) (angle)
20.0–2.35 11770 20.6 27.7 21.1 29.8 0.006 1.00 1.3 27.6 0.61 2510 82 278 231 3 2.3 3.1 4.9 6.4
RECEPTOR-BINDING ELR CONFORMATION OF IL-8
Fig. 3. ELR from wild-type (red) and L5C/H33C (monomer B) (blue) after superpositioning using C␣ atoms from residues 7–12 and 30 –36 (rmsd ⫽ 0.27 Å). Wild-type’s X-ray structure was resolved starting at residue Leu5. Shown are main-chain atoms of residues 5–10, 49 –51, and 29 –38 and selected side chains or parts of side chains.
Fig. 4. Two distinct conformations for N-loop residues among IL-8 variants. Main chain ribbons of seven crystallographically determined IL-8 monomers after superpositioning based on ␤-strand and ␣-helical C␣ atoms. The “closed” conformation (blue) is exhibited by wild-type, monomers “B” and “C” of the L5C/H33C variant, and monomer “B” of the E38C/C50A variant. The “open” conformation (red) is exhibited by monomers “A” and “D” of L5C/H33C and by monomer “A” of E38C/C50A IL-8. The distance between C␣ atoms of Phe17 from the two conformations is about 3.7 Å.
to alanine has little effect.12 A robust connection between the ELR and ␤-turn was designed by simultaneous mutation of Leu5 and His33 to cysteines and their linkage in a disulfide bond. Production and Purification of L5C/H33C IL-8
Fig. 5. The ELR region of the IL-8 variant L5C/H33C. Monomers “B” (magenta) and “D” (blue), which have well-resolved N-terminal segments starting at Ala2, are depicted after superpositioning using ␤-strand and ␣-helix C␣ atoms. Residues preceding Glu4, some main-chain carbonyl oxygen atoms, and some side chains are omitted for clarity. Glu4 and Arg6 are found in conformations energetically close to those used in binding IL8-receptors. They may, in fact, also be close structurally.
Site-directed mutagenesis was performed by the method of Kunkel,16 with synthetic oligodeoxynucleotides prepared on an Applied Biosystems DNA synthesizer. The desired mutations were confirmed by sequencing. Fermentation broth was centrifuged and the supernatant was filtered (0.4 mm), loaded on an S-sepharose FF column in phosphate at pH 7.3, and eluted in an NaCl gradient between 0.0 and 0.6 M. An S-sepharose column one tenth as large was used next to concentrate the protein, in the presence of 1 g/mL ␣2-macroglobulin. A phenylsuperose reverse phase column was used next, the
N. GERBER ET AL.
Fig. 6. Intermolecular (crystal packing) interactions characteristic of the two-state N-loop behavior in L5C/ H33C IL-8. a: The interplay between an “open” monomer (A) (dark gray) and a “closed” monomer (B⬘) (light gray), shown by using main-chain worms and selected side chains. b: A close-up of specific interactions from monomer B⬘ where it is inserted between the ␣-helix and N-loop of monomer A. c: A close-up of the intact Pro16⬘/Trp57⬘ interaction in monomer B⬘ and Pro16 from monomer A inserted in the hydrophobic pocket formed by Phe17⬘, Phe21⬘, and Leu43⬘.
protein eluting at 30 – 40% ammonium sulfate. The target fractions were exchanged into phosphate-buffered saline, and ␣2-macroglobulin was replaced at 1 g/mL. Heparinagarose chromatography with elution in a 0 –1.0 M NaCl gradient yielded pure protein, which was concentrated, desalted, and exchanged into 20 mM MES pH 6.2/150 mM NaCl on a PD10 column. Binding Assay Binding of L5C/H33C to IL-8 receptor A (CXCR1) and B (CXCR2) displayed on stably transfected 293 cells was performed as described previously.17 Crystallization Small crystal blocks (approximately 0.05 mm on each side) grew at 20°C via vapor diffusion from hanging drops
of 4.2 mg/mL protein in 100 mM NaCl, 4 mM MES pH 6.2, 60 mM ammonium sulfate, and 18% (wt/vol) PEG 8000 suspended over a solution of 100 mM ammonium sulfate and 30% (wt/vol) PEG 8000. Crystals were assayed for free cysteine using 5,5⬘-dithio-bis(2-nitrobenzoic acid) (DTNB), and none was found. Data Collection, Structure Solution, and Refinement Crystals were rinsed in reservoir and frozen by immersion in liquid nitrogen. Data extending beyond 1.8 Å resolution were collected at CHESS beamline A1 by using a Princeton 1K CCD detector.18 Data extending to 2.35 Å resolution were available via data reduction and scaling with DENZO/SCALEPACK19 in space group P21 with cell parameters a ⫽ 37.54 Å, b ⫽ 71.77 Å, c ⫽ 57.86 Å, and ␤ ⫽
RECEPTOR-BINDING ELR CONFORMATION OF IL-8
Fig. 7. Solvent accessible surface representation of the differences in the “open” and “closed” N-loop conformations of IL-8. In the “closed” state, there are hydrophobic interactions between the Pro16 and Trp57 side chains, and a hydrophobic pocket formed by Phe17, Phe21, and Leu43 is apparent. In the “open” state, the Pro16/Trp57 interaction is absent, and the hydrophobic pocket is diminished.
90.46°. Completeness overall was 95% (outer shell 86%). The overall signal-to-noise ratio (I/I) was 14.7 (outer shell 5.2). The average redundancy was 4.3. A Vm of 2.44 Å3/Da and solvent content of 43% result from assuming two IL-8 dimers per asymmetric unit. The structure was solved by molecular replacement (X-PLOR20) using a modified wild-type IL-8 as search model5 (Fig. 1). The modifications were designed to provide confirmation of putative molecular replacement solutions. Iterative adjustments to the model and refinement, including the use of simulated-annealing A-weighted omit maps, noncrystallographic symmetry restraints on well-conserved main chain segments, simulated annealing refinement, and anisotropic temperature factor and bulksolvent corrections to the data, were performed by using X-PLOR and Xsight (MSI, San Diego, CA). The principal changes required from the starting model were additions to the N-termini in two of the four independent monomers and extensive refitting of residues Lys11 to Phe21. Solvent waters were added when inspection of the top (Fo-Fc) peaks revealed suitable H-bond geometry. The highest water B-factor is 43 Å2. A graph of average isotropic temperature factors appears in Figure 2. Main chain (, ) torsion angles are tightly bunched within lowest energy regions of Ramachandran space (94.3%) or nearby (5.3%).21 The single outlier is residue Ala2 in monomer “D,” the only monomer in which Ser1 has been placed (in weak electron density). The final model includes all or part of 278 amino acid residues, 231 waters, and three sulfate ions. Final statistics appear in Table I. Coordinates have been deposited and are available on publication (code 1QE6). RESULTS AND DISCUSSION Description of the Structure The structure contains two independent dimers of the IL-8 mutant: one comprised of monomers designated “A” and “B” and the other of monomers “C” and “D.” They are all similar, in most respects, to each other and to wild-type and to an additional IL-8 variant (E38C/C50A) we have reported previously.17 The rms deviations of C␣ atoms
Fig. 8. Proline binding site is used by a receptor peptide. The hydrophobic pocket between the N-loop and the 40s loop (gray) is displayed together with the receptor peptide (red) from the NMR structure of its complex with IL-826 (PDB code 1ILQ) and the proline-containing N-loop (green) from a crystal contact found commonly among IL-8 crystal structures, after superposition of the corresponding proline binding sites. Note the close correspondence is restricted to the Proline side chains.
from wild type and L5C/H33C after superpositioning using C␣ atoms from the ␤-sheet and ␣-helix are between 0.26 and 0.44 Å (Fig. 3). The relationships between monomers within a dimer are also like the wild-type X-ray structure.5 Significant differences among the monomers are restricted to the N-termini and N-loops. Our intention that the engineered disulfide link between residues 5 and 33 would provide a well-ordered N-terminus for IL-8 has proven only partially successful. For two of the four independent monomers, the 5/33 disulfide is well resolved, and the N-terminus can be discerned starting at Ser1 (monomer “D”) or Ala2 (monomer “B”). However, the other two monomers have only scattered electron density at the 5/33 position. Because in these monomers the ␤-turn containing Cys33 is reasonably well resolved, we conclude that this variant undergoes a low-energy disulfide torsion angle transition that alters N-terminal residues up to
N. GERBER ET AL.
about Arg6 and the S␥ atom of Cys33, but not other parts of the ␤-turn containing Cys33. The well-resolved Ntermini of two monomers are the results of serendipitous intermolecular interactions. The fact that a well-ordered N-terminus is observed in L5C/H33C but not in wild-type or other IL-8 mutants could be pure chance. It remains possible, however, that the added disulfide link has predisposed L5C/H33C toward such a structure. In the latter case, we are justified in pursuing its characterization. There are two distinct conformations of the N-loop segment between Ser14 and His18 that differ by about 3.7 Å at the main chain of Phe17 (Fig. 4). One is like that seen in the wild-type X-ray structure, with the N-loop close to the ␣-helix (closed), whereas the other conformation is farther away from the ␣-helix (open). These two conformations are the same as were seen in the E38C/C50A IL-8 variant. This difference is associated with specific intermolecular interactions. Receptor Binding ELR The absolute Kd values for our measurements of wild type are 2.0 and 1.3 nM for CXCR1 and CXCR2, respectively, in agreement with prior determinations,22 and for L5C/H33C are 95 and 61 nM for CXCR1 and CXCR2, respectively. This means that the L5C/H33C variant has a Kd for CXCR1 and CXCR2 about 50-fold worse than wild type (48-fold and 46-fold, respectively). For comparison, the L5A variant has been reported to be about 100-fold worse than wild type,10 and H33A is essentially unchanged from wild type.12 This relatively small binding effect of the significant restraint imposed by the engineered Cys5-Cys33 has encouraged us to pursue the characterization of a template for IL-8 minimization. Monomers “B” and “D” offer similar views of the receptor binding conformation of the ELR epitope (Fig. 5). Ile10 has been shown by others10,11 and us17 to be important for receptor interactions. Because it is proximal to ELR, we include it in our discussion of the ELR epitope. The small loss in binding affinity suffered in creating the 5/33 disulfide link supports our hypothesis that the Leu5 side chain may not actually contact receptor. The link at residue 5 means that the side chains of Glu4 and Arg6 are within low-energy torsion angle transitions of the optimal receptor binding arrangement. The lack of significant change at Ile10 means all the important ELR binding elements are displayed close to (energetically) as they are when bound to IL-8 receptor(s). If we assume they are also structurally close, we are in a position to design small mimics of IL-8 that reproduce these elements in an equivalent combination. The challenge offered by this result is to design a peptide or other small molecule that faithfully recreates the threedimensional array of important receptor contacts. The simplest cyclic peptide reductions are complicated by the parallel nature of the relationship between the Glu4 to Ile10 segment and the Cys33 to Cys34 segment. Cyclic peptides small enough to be conformationally constrained require Ile10 to be exocyclic, raising a new requirement that its display be directed appropriately by some addi-
tional means. On the other hand, nonpeptide small molecules suffer from generally greater synthetic challenges, but allow greater freedom of design. Such an approach would at some stage provide information on the role, if any, that the main chain of ELR plays in receptor interactions. Two Conformations for the N-Loop There is now ample evidence that IL-8 has a characteristic two-state flexibility in the N-loop region. Crystallographic results offer a total of seven monomers that can be placed in one of two classes based on their N-loop conformation (one from the wild type,5 two from E38C/C50A,17 and four from the present work). Segregation into these classes is not caused by the mutations among the X-ray structures, because for each mutation, both classes are found. This view is supported by solution studies via NMR.23,24 Either the N-loop is close to the ␣-helix (closed) or near to Arg47 (open) (Fig. 4). The main chain atoms of the two conformations are about 3.7 Å apart at Phe17. Each N-loop conformation is associated with a specific collection of intermolecular contacts (Fig. 6). This collection of interactions has been observed almost without change in all the crystallographic structures, and they are more specific than most intermolecular packing contacts. In the closed state, a hydrophobic pocket or groove between the N-loop and Arg47 is exposed25 (Fig. 7). This site is occupied by a proline from an intermolecular contact in all crystal structures where the closed form is found: wild-type, two monomers in the present structure, and one of the two monomers in E38C/C50A. In wild-type IL-8, the proline is Pro32, and in the present structure and E38C/ C50A it is Pro16 (Fig. 6). This “proline binding site” is also where an IL8-RA peptide (containing prolines) has been shown to bind26,27 (Fig. 8). In addition to this intermolecular contact, the closed form has a hydrophobic interaction between Pro16 and the Trp57 side chains and an H-bond from the Lys20 amide nitrogen to the His18 side chain. In the open state, the proline binding site is obscured, and a groove opens up between the N-loop and the ␣-helix (Fig. 7), as part of which the Pro16/Trp57 interaction is lost (Fig. 6). In its place is found an intimate arrangement of interactions between Trp57, the main chain of Phe17, and His18 from the open monomer, and the 40s loop from a neighboring molecule (Fig. 6). The side chain of Arg47⬘ forms a hydrophobic contact with Trp57, the Asp45⬘ side chain (apparently not ionized) H-bonds to the Phe17 main chain carbonyl oxygen, and the Ser44⬘ side chain H-bonds to the His18 side chain. This same collection of interactions is found for all instances of the open-form N-loop: two monomers in the present structure and one of the two independent monomers in the E38C/C50 variant. If we consider the NMR analysis that in solution there are two distinct states for the N-loop,24 the consistent intermolecular arrangements we have observed crystallographically could be viewed as an adventitious use of a characteristic flexibility in IL-8. However, the consistency with which these specific interactions have been observed suggests they mimic relevant biological interactions. In
RECEPTOR-BINDING ELR CONFORMATION OF IL-8
light of evidence that the proline binding site is part of the IL8-RA binding site, it is interesting to speculate that other specific and biologically relevant interactions may pertain to the alternate, open state in which this site is absent. If such interactions were maintained even in the presence of IL-8 receptors, they would prevent binding via the proline binding site and may act as inhibitors of the known biological effects of IL-8.
ACKNOWLEDGMENTS This work is based on research conducted at the Cornell High Energy Synchrotron Source (CHESS), which is supported by the National Science Foundation under award DMR-9311772, using the Macromolecular Diffraction at CHESS (MacCHESS) facility, which is supported by award RR-01646 from the National Institutes of Health.
REFERENCES 1. Baggiolini M, Loetscher P, Moser B. Interleukin-8 and the chemokine family. Int J Immunopharmacol 1995;17:103–108. 2. Fairbrother WJ, Skelton NJ. Three-dimensional structures of the chemokine family. In: Horuk R, editor. Chemoattractant ligands and their receptors. Cleveland Ohio: CRC Press; 1996. p 55– 86. 3. Sticht H, Escher SE, Schweimer K, Forssmann WG, Rosch P, Adermann K. Solution structure of the human CC chemokine 2: a monomeric representative of the CC chemokine subtype. Biochemistry 1999;38:5995– 6002. 4. Crump MP, Rajarathnam K, Kim KS, Clark-Lewis I, Sykes BD. Solution structure of eotaxin, a chemokine that selectively recruits eosinophils in allergic inflammation. J Biol Chem 1998;273:22471– 22479. 5. Baldwin ET, Weber IT, St. Charles R, et al. Crystal structure of interleukin 8: symbiosis of NMR and crystallography. Proc Natl Acad Sci USA 1991;88:502–506. 6. Clore GM, Appella E, Yamada M, Matsushima K, Gronenborn AM. Three-dimensional structure of interleukin 8 in solution. Biochemistry 1990;29:1689 –1696. 7. Lowman HB, Slagle PH, DeForge LE, et al. Exchanging IL-8 and melanoma growth-stimulating activity receptor binding specificities. J Biol Chem 1996;271:14344 –14352. 8. Williams G, Borkakoti N, Bottomley G, et al. Mutagenesis studies of interleukin-8: identification of a second epitope involved in receptor binding. J Biol Chem 1996;271:9579 –9586. 9. Schraufsta¨tter IU, Ma M, Oades ZG, Barritt DS, Cochrane CG. The role of Tyr13 and Lys15 of interleukin-8 in the high affinity interaction with the interleukin-8 receptor type A. J Biol Chem 1995;270:10428 –10431. 10. He´bert CA, Vitangcol RV, Baker JB. Scanning mutagenesis of interleukin-8 identifies a cluster of residues important for receptor binding. J Biol Chem 1991;266:18989 –18994. 11. Clark-Lewis I, Schumacher C, Baggiolini M, Moser B. Structure-
20. 21. 22.
activity relationships of interleukin-8 determined using chemically synthesized analogs: critical role of NH2-terminal residues and evidence for uncoupling of neutrophil chemotaxis, exocytosis, and receptor binding activities. J Biol Chem 1991;266:23128 – 23134. Clark-Lewis I, Dewald B, Loetscher M, Moser B, Baggiolini M. Structural requirements for interleukin-8 function identified by design of analogs and CXC chemokine hybrids. J Biol Chem 1994;269:16075–16081. Rajarathnam K, Clark-Lewis I, Sykes BD. 1H NMR studies of interleukin 8 analogs: characterization of the domains essential for function. Biochemistry 1994;33:6623– 6630. Rajarathnam K, Sykes BD, Dewald B, Baggiolini M, Clark-Lewis I. Disulfide bridges in interleukin-8 probed using non-natural disulfide analogues: dissociation of roles in structure and function. Biochemistry 1999;38:7653–7658. Fairbrother WJ, Reilly D, Colby TJ, Hesselgesser J, Horuk R. The solution structure of melanoma growth stimulating activity. J Mol Biol 1994;242:252–270. Kunkel TA, Roberts JD, Zakour RA. Rapid and efficient sitespecific mutagenesis without phenotypic selection. Methods Enzymol 1987;154:367–382. Eigenbrot C, Lowman HB, Chee L, Artis DR. Structural change and receptor binding in a chemokine mutant with a re-arranged disulfide: x-ray structure of E38C/C50A IL-8 at 2 Å resolution. Proteins 1997;27:556 –566. Gruner SM, Ealick SE. Charge coupled device X-ray detectors for macromolecular crystallography. Structure 1995;3:13–15. Otwinowski Z, Minor W. Processing of x-ray diffraction data collected in oscillation mode. Methods Enzymol 1996;276:307– 326. Bru¨nger AT. X-PLOR: a system for x-ray crystallography and NMR. New Haven CT: Yale University Press; 1992. CCP4. The CCP4 suite: programs for protein crystallography. Acta Crystallogr 1994;D50:760 –763. He´bert CA, Lowman HB. Structure-function relationships of IL-8 and its two neutrophil receptors: IL-8-RA and IL-8-RB. In: Horuk R, editor. Chemoattractant ligands and their receptors. Boca Raton, Florida: CRC Press; 1996. p 29 –53. Grasberger BL, Gronenborn AM, Clore GM. Analysis of the backbone dynamics of interleukin-8 by 15N relaxation experiments. J Mol Biol 1993;230:364 –372. Bonvin AM, Bru¨nger AT. Conformational variability of solution magnetic resonance structures. J Mol Biol 1995;250:80 –93. Hammond M, Shyamala V, Siani M, et al. Receptor recognition and specificity of interleukin-8 is determined by residues that cluster near a surface-accessible hydrophobic pocket. J Biol Chem 1996;271:8228 – 8235. Skelton NJ, Quan C, Reilly D, Lowman H. Structure of a CXC chemokine-receptor fragment in complex with interleukin-8. Structure 1999;7:157–168. Clubb RT, Omichinski JG, Clore GM, Gronenborn AM. Mapping the binding surface of interleukin-8 complexes with an N-terminal fragment of the type 1 human interleukin-8 receptor. FEBS Lett 1994;338:93–97.