Receptor recognition and specificity of interleukin-8 is determined by residues that cluster near a surface-accessible hydrophobic pocket.

To determine the regions of interleukin-8 (IL-8) that allow high affinity and interleukin-8 receptor type 1 (IL8R1)-specific binding of chemokines, we produced chimeric proteins containing structural domains from IL-8, which binds to both IL8R1 and interleukin-8 receptor type 2 (IL8R2) with high affinity, and from GRO gamma, which does not bind to IL8R1 and binds to IL8R2 with reduced affinity. Receptor binding activity was tested by competition of 125I-IL-8 binding to recombinant IL8R1 and IL8R2 cell lines. Substitution into IL-8 of the GRO gamma sequences corresponding to either the amino-terminal loop (amino acids 1-18) or the first beta-sheet (amino acids 18-32) reduced binding to both IL8R1 and IL8R2. The third beta-sheet of IL-8 (amino acids 46-53) was required for binding to IL8R1 but not IL8R2. Exchanges of the second beta-sheet (amino acids 32-46) or the carboxyl-terminal alpha-helix (amino acids 53-72) had no significant effect. When IL-8 sequences were substituted into GRO gamma, a single domain containing the second beta-sheet of IL-8 (amino acids 18-32) was sufficient to confer high affinity binding for both IL8R1 and IL8R2. The amino-terminal loop (amino acids 1-18) and the third beta-sheet (amino acids 46-53) of IL-8 had little effect when substituted individually but showed increased binding to both receptors when substituted in combination. Individual amino acid substitutions were made at positions where IL-8 and GRO gamma sequences differ within the regions of residues 11-21 and 46-53. IL-8 mutations L49A or L49F selectively inhibited binding to IL8R1. Mutations Y13L and F21N enhanced binding to IL8R1 with little effect on IL8R2. A combined mutation Y13L/S14Q selectively decreased binding to IL8R2. Residues Tyr13, Ser14, Phe21, and Lys49 are clustered in and around a surface-accessible hydrophobic pocket on IL-8 that is physically distant from the previously identified ELR binding sequence. A homology model of GRO gamma, constructed from the known structure of IL-8 by refinement calculations, indicated that access to the hydrophobic pocket was effectively abolished in GRO gamma. These studies suggest that the surface hydrophobic pocket and/or adjacent residues participate in IL-8 receptor recognition for both IL8R1 and IL8R2 and that the hydrophobic pocket itself may be essential for IL8R1 binding. Thus this region contains a second site for IL-8 receptor recognition that, in combination with the Glu4-Leu5-Arg6 region, can modulate receptor binding affinity and IL8R1 specificity.

Interleukin-8 (IL-8) 1 and the related GRO proteins are members of a superfamily of proinflammatory cytokines that stimulate neutrophil activation and chemoattraction (see Refs. 1-3 for review). These proteins contain four conserved cysteine residues with a single intervening amino acid between the first two cysteine residues and are designated the C-X-C chemokines. IL-8 mediates the recruitment and activation of neutrophils during inflammation and has been implicated in multiple pathologic conditions involving chronic and acute inflammation and in neutrophil-mediated injury (4 -11). Two receptors for IL-8 have been identified by molecular cloning and account for the observed effects of IL-8 and other C-X-C chemokines on neutrophils (12,13). Although both are seven-transmembrane, G-protein-coupled receptors, Type 1 and Type 2 IL-8 receptors differ in ligand specificity. The Type 1 receptors (IL8R1) have restricted specificity and bind IL-8 exclusively with high affinity (14,15). In contrast, the Type 2 receptors (IL8R2) bind IL-8 with a similarly high affinity but also recognize several other C-X-C chemokines with varying affinities (14,15).
Structures for both the crystal and solution forms of IL-8 have been determined (16,17). IL-8 comprises five discrete structural domains: an amino-terminal loop, three antiparallel ␤-sheets, and a carboxyl-terminal ␣-helix. The highest regions of amino acid homology among the C-X-C chemokines occur at the conserved cysteine residues and at other key structural residues, suggesting that the basic structural elements of IL-8 are conserved among family members (1)(2)(3). The solution structure for GRO␣ (18) and crystal structure for neutrophil-activating peptide-2 (19) show similar organization.
Several studies have addressed the structure-activity relationships required for C-X-C chemokine binding to neutrophils. A conserved sequence that is essential for binding, Glu 4 -Leu 5 -Arg 6 (ELR), was identified by scanning mutagenesis of IL-8 (20) and by amino-terminal truncated analogs (21). The exclusive binding of IL-8 to IL8R1 and the different affinities among C-X-C chemokines for binding to IL8R2 suggest that a second site on IL-8 determines receptor specificity. Hybrid proteins * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed: Chiron Corp., 4560 Horton St., Emeryville, CA 94608. Tel.: 510-601-2939; Fax: 510-658-0329. § Current address: Gryphon Sciences, 250 E. Grand Ave. #90, South San Francisco, CA 94080. 1 The abbreviations used are: IL-8, interleukin-8; IL8R1, interleukin-8 receptor type 1; IL8R2, interleukin-8 receptor type 2; GRO, growth related protein (note that GRO␣ is also known as melanocyte growth-stimulating activity, GRO/MGSA, and that GRO␤ and GRO␥ are also known as macrophage inflammatory proteins MIP2␣ and MIP2␤, respectively); ELR, Glu 4 -Leu 5 -Arg 6 sequence; PCR, polymerase chain reaction; CHO, Chinese hamster ovary. derived from IL-8 and IP10, a C-X-C chemokine that lacks the ELR sequence and that does not bind neutrophil IL-8 receptors, indicated that the ELR sequence as well as IL-8 sequences for amino acids 10 -22 and 30 -34 are required for neutrophil recognition (22). Chimeric analysis of IL-8 with the murine chemokine N51/KC indicated a site that enhances binding to neutrophils is within amino acids 13-29 (23). Receptor specificity was addressed directly on recombinant IL-8 receptors with chimeric proteins derived from IL-8 and GRO␣ or rabbit IL-8 and demonstrated that IL8R1 specificity element(s) reside between Cys 7 and Cys 50 of IL-8 (24). In this study we have generated chimeric proteins containing domains from IL-8 and GRO␥ and report that the second site for receptor binding and specificity is localized in the region of a unique hydrophobic surface pocket on IL-8. Through amino acid replacement studies we have identified residues in and around the hydrophobic pocket that influence binding of IL-8 to IL8R1 and IL8R2.
Construction of Chimeric Chemokines-Corresponding domains of IL-8 and GRO␥ were identified by protein sequence alignments shown in Fig. 1. DNA encoding chimeric IL-8/GRO␥ were generated by recombinant polymerase chain reaction (PCR) (26) and/or by ligation with synthetic oligonucleotides encoding the desired amino acid as depicted in Fig. 2. A synthetic IL-8 gene, il8syn (27), and a MIP2␤ cDNA clone, MIP540 (28), were used as initial templates for IL-8 and GRO␥ sequences. PCR primers and synthetic oligonucleotide linkers are listed in Table I. For G18I, sequences between the Asp718 and MluI restriction sites were generated with oligonucleotides accX, Xmlu, Xacc, and mluX. For I18G, sequences between Asp718 and BglII were from oligonucleotides accZ, Zbgl, Zacc, and bglZ. Chimeric DNA were used as PCR templates in some cases: G32I and I46G for G32I46G, I32G and G46I for I32G46I, and G46I53G for I18G46I53G. Constructs were ligated, as Asp718/SalI fragments a transfer vector containing the GAP promoter fused to the ␣-factor leader (29).
Construction of Mutant Chemokines-All constructs were generated by PCR according to Shyamala and Ames (30) with the exception that vent polymerase was used in the place of Taq polymerase. The DNA template was either the native IL-8 or il8syn (27). Amino acid substitutions were introduced by overlap PCR using the sense and antisense mutated primers in combination with appropriate end primers as listed in Table II (31). The PCR-amplified fragment was digested with Asp718/XhoI and ligated into the transfer vector.
Protein Expression and Purification-Expression cassettes for yeast secretion were transferred as BamHI restriction fragments into vector pAB24 (29) and introduced into Saccharomyces cerevisiae strain MB2-1 by electroporation. Chimeric and mutant chemokines were purified from 50 -200 ml of yeast culture broth by batch adsorption on S-Sepharose FF (Pharmacia Biotech Inc.) after adjustment to pH 5.5 with 50 mM sodium acetate and eluted in 20 mM HEPES, pH 8.3, 1 M NaCl to a final concentration of 0.2-2 mg/ml. SDS-polyacrylamide gel electrophoresis on 18% Tris/glycine gels (Novex) indicated 80 -95% purity. Protein concentrations were estimated by Coomassie-stained polyacrylamide gels and by BCA (Pierce) protein assays. Amino acid composition and amino-terminal sequencing were performed on selected proteins and agreed with predicted protein sequences.
Binding Assays-Competitive binding assays for the chimeric proteins were performed on CHO-IL8R1 and CHO-IL8R2 cells essentially as described (25). Assays were performed in triplicate and data were analyzed by GraFit (32).
Chemotaxis Assays-Assays were performed in triplicate on freshly isolated human neutrophils as described (25). Chemotaxis to f-Met-Leu-Phe (100 nM) was measured as a positive control for each experiment.
Homology Modeling-A homology-based model of GRO␥ was built based on the NMR solution-derived structure of IL-8 (17). The LOOK software program (Molecular Applications Group, Mountain View, CA) was used to align the GRO␥ and IL-8 sequences (33). Then a threedimensional model of GRO␥ was built by Levitt's automatic segment matching (34) and further refined by restrained energy minimization and molecular dynamics (35).

RESULTS
Chimeric IL-8/GRO␥ Proteins-Chimeric IL-8/GRO␥ proteins were designed to test the contribution of each structural domain of IL-8 for binding to IL8R1 and IL8R2. Four conserved amino acid residues were identified as structural domain boundaries for IL-8: His 18 , Pro 32 , Gly 46 , and Pro 53 (Fig. 1). To maintain the overall C-X-C chemokine structure, sequences for complete structural domains between these boundaries were interchanged between IL-8 and GRO␥ (Fig. 2). The corresponding chimeric chemokines were produced in yeast using the ␣-factor mating pheromone secretion pathway (29) and purified to near homogeneity by a single step enrichment/purification protocol. With the exception of I32G, all proteins were ex-

Determinants of IL-8 Recognition by IL8R1 and IL8R2
pressed and recovered in yields sufficient for testing. Truncated variant proteins were observed for several of the mutants in which the amino-terminal sequences of GRO␥ were fused with carboxyl-terminal portions of IL-8 and varied from 20 to 80% of the total chimera protein. Only the predicted aminoterminal protein sequence was detected in these samples. The carboxyl-terminal sequence of IL-8 reveals a Lys-Arg motif at residues 67 and 68 ( Fig. 1). This sequence is a substrate site for the yeast dibasic protease KEX2 (36), and cleavage at this position can account for the observed reduction of approximately 5000 molecular mass units. Although previous studies have suggested that the ␣-helix region is not essential for IL-8 recognition by neutrophil receptors (21), the utilization of a previously silent KEX2 yeast-processing site suggests an increase in the accessibility of this region to protease attack and may indicate aberrant folding for this series of chimeric proteins.
Receptor Binding Activity of Chimeras-Receptor binding activity of chimeric proteins was measured by competition of 125 I-IL-8 binding to recombinant IL8R1 and IL8R2 cell lines. GRO␥ did not bind to IL8R1 and was 15-fold less potent than IL-8 for binding to IL8R2 (Table III). IL-8 or GRO␥ purified in small scale by the procedure used for chimeric proteins had the same potency as that purified in large scale by multiple chro-matography (data not shown). For IL-8/GRO␥ chimeras, substitution of GRO␥ sequences onto the amino terminus of IL-8 (amino acids 1-18 or more) resulted in a loss of affinity to both IL8R1 and IL8R2 (Table III). Chimeric chemokines G32I, G46I, and G53I were even less potent than GRO␥, which may reflect improper folding since these proteins were prone to truncation as described above. In substitutions at the carboxyl terminus, replacement of the ␣-helix residues 53-72 of IL-8 with the corresponding region of GRO␥ had no significant effect on receptor binding activity (Table III, I53G). In contrast, substitution of residues 46 -72 of IL-8 with the corresponding GRO␥ sequence selectively reduced binding to IL8R1 but not IL8R2 (Table III, I46G), suggesting that residue(s) within the third ␤-sheet (amino acids 46 -53) are involved in specific binding to IL8R1 and not IL8R2.
Chimeric proteins with single domains of GRO␥ substituted into IL-8 were used to test the role of each ␤-sheet region. Replacement of amino acids 18 -32, corresponding to the first ␤-sheet, reduced binding of IL-8 to both IL8R1 and IL8R2 (Fig.  3, I18G32I). Exchange of regions within amino acids 32-46 was not important for binding to either IL8R1 or IL8R2 (Fig. 3,  I32G46I). Substitution for amino acids 46 -53 reduced binding to IL8R1 by 10-fold without affecting binding to IL8R2 (Fig. 3,  I46G53I), confirming that the third ␤-sheet of IL-8 participates in IL8R1 specificity.
In the reciprocal experiment, individual domains of IL-8 were substituted into GRO␥ to examine whether any of these regions could enhance binding to IL-8 receptors. Replacement of residues 18 -32, corresponding to the first ␤-sheet of IL-8, was sufficient to allow nearly IL-8-like activity on both IL8R1 and IL8R2 (Fig. 5, G18I32G). Two domains, corresponding to amino acids 1-18 and 32-46, slightly enhanced binding to IL8R1 (Fig. 3, I18G and G32I46G). Since the amino-terminal loop and the third ␤-sheet were both important for IL8R1 binding of IL-8 but had little effect when substituted into GRO␥ Mutations were generated by overlap PCR using sense and antisense mutated primers in combination with the respective 5Ј sense and 3Ј antisense end primers (31). Only the sense strand mutant primers are shown. The nucleotides generating the mutated amino acids are underlined. In the 5Ј and 3Ј end primers the nucleotides corresponding to the restriction site and nonhomologous to IL-8 are in italics.

Name
Sequence (5Ј to 3Ј) The mutations in combination with D52N were obtained by amplifying respective mutated DNA templates with the primers for D52N mutation. individually, a chimera was generated to test whether simultaneous replacement of these regions of GRO␥ with corresponding IL-8 sequences could confer IL-8-like activity. The combined substitution of amino acids 1-18 and 46 -53 allowed high affinity binding of GRO␥ to both IL8R1 and IL8R2 (Fig. 3,  I18G46I53G).
Point Mutations of IL-8 to GRO␥ Residues within the Third ␤-Sheet-Since the replacement of IL-8 amino acids 46 -53 with GRO␥ reduced binding to IL8R1 without affecting the binding to IL8R2, the role of individual amino acids in this region was examined through point mutations to substitute GRO␥ residues into IL-8. Four amino acid residues differ between IL-8 and GRO␥ in this region: Arg 47 , Glu 48 , Leu 49 , and Asp 52 . The GRO␥ amino acids were introduced into IL-8 to test each residue individually and in combinations. The mutation of IL-8 L49A, alone or in combination, selectively decreased the binding to IL8R1 (Table IV). Neither the E48K replacement nor the conservative substitution D52N affected binding to either receptor, although the double mutation of E48K/D52N decreased binding to IL8R1 by 3-fold. R47K marginally decreased the binding to both IL8R1 and IL8R2 but had no impact on receptor binding in combination mutations. The triple mutation R47K/E48K/D52N also had little effect on binding to either receptor and no selectivity for IL8R1. No protein was expressed for IL-8 mutations R47K/E48K and E48K/L49A/D52N, suggesting abnormal folding and protein degradation. IL-8 R47K/ L49A did not generate yeast transformants, implying intolerance of the mutated protein by this host cell. Consistent with the chimeric substitutions in the 46 -53 region, substituting Leu 49 decreased the binding to IL8R1 with no significant effect on binding to IL8R2. These data indicate that Leu 49 is the primary determinant within the third ␤-sheet for IL8R1 specificity of IL-8. terminal Loop-The chimera data demonstrated that IL-8 amino acids 1-18 of the amino-terminal loop are important for binding of IL-8 to both IL8R1 and IL8R2. Within this region there are several amino acid differences; particularly, the five amino acids of IL-8 at positions 13-17 are replaced by only four different residues in GRO␥. IL-8 amino acids 13-15 were substituted with comparable amino acids of GRO␥. No protein was expressed for mutants containing the K15G mutation alone or in combinations, suggesting that this alteration of the chemokine structure is not tolerated. The mutation Y13L increased the binding to IL8R1 over 3-fold without affecting IL8R2 binding (Table IV). S14Q marginally decreased the binding to IL8R1, but the double mutation of Y13L/S14Q significantly decreased the binding to IL8R2 over 4-fold with little effect on IL8R1 (Table IV). These results are consistent with the amino acid 1-18 chimeric substitutions and demonstrate that Tyr 13 and Ser 14 contribute to the specificity of IL-8 binding to both IL8R1 and IL8R2.
Comparison of IL-8 Structure and GRO␥ Homology Model-Based upon the established structure of IL-8 (16,17), residues Tyr 13 , Ser 14 , and Leu 49 lie on a single face of the IL-8 molecule in a region that is unique and distant from both the conserved ELR residues previously identified as essential for IL-8 binding to IL-8 receptors and the strand of hydrophobic residues that participates in the monomer-monomer interface of the dimer form of IL-8 (Fig. 4). These residues coincide with or are adjacent to a surface-exposed hydrophobic pocket on IL-8 that consists of Tyr 13 , Phe 17 , Ile 22 , Val 41 , Leu 43 , Leu 49 , and Leu 51 (Fig.  4). This slot-like hydrophobic pocket is large enough to accommodate a phenyl ring, which fits into the pocket as a coin in a slot. The entrance to this pocket is flanked by Tyr 13 , Lys 15 , Phe 21 , and Arg 47 .
A homology-based model of GRO␥ was constructed to compare the relative positions of the corresponding residues. The predicted structure is shown in the same relative orientation as the structure of IL-8 (Fig. 4). In IL-8, one side of the hydrophobic pocket is formed by the Tyr 13 -Ile 22 strand. In GRO␥, little of this stretch of residues is conserved, and there is a deletion of one residue. Tyr 13 and Phe 17 are conservatively substituted with isoleucine in GRO␥, and Ile 22 is conserved. Thus, the residues contributing to the hydrophobic core appear to be conserved. However, the deletion corresponding to Lys 15 in IL-8 reduces the length of the Tyr 13 -Ile 22 strand, effectively shrinking the hydrophobic pocket in GRO␥. The substitution of Pro 19 by a leucine in GRO␥ affects the structure of this strand and its contribution to the hydrophobic pocket. According to our homology model, the hydrophobic pocket in GRO␥ is much smaller than that in IL-8 and cannot accommodate a phenyl ring. In addition, none of the gateway residues of IL-8 are conserved in GRO␥.

IL-8 Mutations within the Hydrophobic Pocket
Region-Additional point mutations of IL-8 were designed to test the role of the key hydrophobic pocket and flanking residues in the recognition of IL-8 by IL8R1 and IL8R2. Mutations were introduced at positions Phe 21 , Val 41 , Asp 45 , and Leu 49 , corresponding to residues that are unique in IL-8. F21N increased the binding affinity by 5.5-and 2.4-fold to IL8R1 and IL8R2, respectively (Fig. 5). Mutation D45R had no effect on IL8R1 and IL8R2 binding, indicating that this charged residue is not critical for binding. The mutation V41F, which introduces a more bulky hydrophobic group within the pocket, also had little effect on either receptor. Three additional mutants, F21T, V41R, and V41K, did not express the recombinant protein. At  Leu 49 , insertion of the aromatic residue in L49F selectively decreased binding to IL8R1 by 6.2-fold with no significant effect on IL8R2. In contrast the relatively conservative substitution of L49S had little effect on IL-8 receptor activity. Taken together with the observations for L49A and for Y13L, these mutations reveal that key residues for determining receptorspecific binding of IL-8 are clustered around the surface-accessible hydrophobic pocket and suggest that increased access to the pocket by removal of the aromatic residues at Tyr 13 and Phe 21 can enhance binding to IL8R1. Neutrophil Chemotaxis-The functional activation of IL-8 receptors was assessed in neutrophil chemotaxis assays. GRO␥ was less potent than IL-8 and had lower efficacy at optimal chemotaxis concentrations (Fig. 6). Chimeric IL-8/GRO␥ proteins demonstrated neutrophil chemotactic activity consistent with relative receptor binding activity ( Fig. 6 and additional data not shown). Proteins with characteristic binding properties of GRO␥, i.e. an absence of binding to IL8R1 and a reduced affinity binding to IL8R2, displayed GRO␥-like chemotactic activity. An IL-8-like chemotactic efficacy correlated with recognition by IL8R1, and the relative potency paralleled the affinity for IL8R1 observed in competitive binding assays (Fig. 6).
IL-8 variants with point mutations also stimulated chemotaxis with the same efficacy as IL-8 at optimal doses but displayed variable chemotactic potency. Y13L, which showed increased binding to IL8R1, was 2-3-fold more potent than IL-8 in chemotaxis assays (data not shown). L49F had a reduced chemotactic potency proportionate to reduced activity for IL8R1. However, L49A had only slightly reduced chemotactic activity despite a significantly decreased ability to bind to IL8R1 (data not shown). The Y13L/S14Q mutation decreased binding specifically to IL8R2 but had no effect on chemotaxis, consistent with the major role of IL8R1 in chemotaxis (25).
The functional activity of mutant proteins was further analyzed by intracellular signaling assays. Wu et al. (37) have demonstrated that IL-8 receptors signal via G ␣i subunits. Therefore, the ability of IL8R1 and IL8R2 to reduce cAMP concentrations was determined in CHO cells as described in Shyamala et al. (38). IL-8 inhibited the forskolin-induced elevation of cAMP levels in either CHO-IL8R1 or CHO-IL8R2 cells (data not shown). The 11 mutated IL-8 proteins also inhibited cAMP levels upon interaction with either IL8R1 or IL8R2 to varying degrees (data not shown), confirming that these proteins are functional agonists of IL8R1 and IL8R2.

DISCUSSION
By construction of IL-8/GRO␥ chimeras we have demonstrated that the amino-terminal loop (amino acids 1-18) and the first ␤-sheet (amino acids 18 -32) of IL-8 contain residue(s) that are essential for high affinity, IL-8-like binding to both IL8R1 and IL8R2. These domains encompass the regions identified by IL-8/IP10 hybrids (22) and by IL-8/N51 chimeric proteins (23) as necessary for maximal binding to neutrophils. The third ␤-sheet of IL-8 contains residue(s) that confer IL8R1specific binding and are not required for binding to IL8R2. Complete analysis of this region by point mutations indicated that L49A substitution is responsible for the specific reduction of IL8R1 binding. Heinrich et al. (23) observed that a chimeric IL-8/N51 protein containing this substitution (corresponding to I34N50I) exhibited neutrophil binding properties similar to IL-8. That result may reflect the participation of both IL8R1 and IL8R2 receptors in neutrophil binding.
Schraufstatter et al. (24) have reported that the carboxyl terminus of IL-8/GRO␣ chimeric proteins affects specificity of binding to IL8R2 but not to IL8R1. Chimeric I51G␣ behaved like IL-8 in its binding to IL8R1 but had 5-fold lower affinity for IL8R2 (24). The present study demonstrates that the carboxyl terminus of IL-8 beyond amino acid residue 53 does not define   (39) identified Tyr 13 as well as Lys 15 as determinants of the differential IL8R1 affinity for human and rabbit IL-8. These key residues that affect binding to IL8R1 surround a unique hydrophobic pocket on the surface of IL-8. The increase in binding to IL8R1 by the replacement of Tyr 13 with leucine indicates that the removal of the aromatic ring provides increased access for receptor docking. Y13H also decreased binding to IL8R1 (39), and other mutations at this position altered binding to neutrophils (22). The increased binding to IL8R1 of F21N can also be explained as facilitating the receptor interaction by the replacement of the aromatic group with smaller amino acids. F21L had reduced binding to neutrophils (22). Substitutions at Leu 49 are not as clear-cut. L49F decreases binding to IL8R1, which could be interpreted as the large aromatic ring of phenylalanine obstructing the receptor docking. L49A also decreases R1 binding, which might suggest that this is due to an alteration of the hydrophobic pocket itself. The replacement with an intermediately sized residue, L49S, moderately increased binding. The double mutation Y13L/S14Q decreases binding to IL8R2 selectively, indicating that IL8R2 recognition involves different determinants than IL8R1 but is also localized near the hydrophobic pocket on IL-8. Thus this region contributes to IL8R2 binding, and the pocket itself may be essential for binding to IL8R1.
The surface-accessible hydrophobic pocket on IL-8 is sufficiently large to accommodate an aromatic ring structure and could accept a phenylalanine or tyrosine side chain from IL8R1. The region of the hydrophobic pocket, and particularly of residues 12-18 within the amino-terminal loop, corresponds to the site of greatest structural differences between IL-8 and other C-X-C chemokines (GRO␣ and NAP-2) that bind to IL8R2 but not to IL8R1 (18,19). The solution structure of GRO␣ indicates a hydrophobic pocket smaller than that on IL-8 (18). For GRO␥, homology-based structural modeling predicted that the hydrophobic pocket is much smaller and more restricted in access, consistent with the lack of IL8R1 binding and reduced IL8R2 binding.
Interleukin-8 interacts with two sites on IL-8 receptors; conserved charged residues in extracellular domains 3 and 4 are essential for IL-8 binding to IL8R1 or IL8R2 (40), and the amino-terminal extracellular domains confer the ligand specificity profiles characteristic of each receptor type (41,42). Our data indicate that the hydrophobic pocket and/or surrounding residues contribute to the binding of C-X-C chemokines to IL-8 receptors and serve as the second binding site on IL-8 to modify receptor recognition in conjunction with the ELR sequence. Thus, receptor extracellular domains 3 and 4 provide the primary interaction with ELR (residues 4 -6) of IL-8, and receptor amino-terminal domains interact with a secondary binding site FIG. 6. Neutrophil chemotactic activity of IL-8/GRO␥ chimeras. A, effect of the amino-terminal loop residues 1-18. Chemotaxis assays were performed with IL-8 (E), GRO␥ (q), G18I (Ⅺ), and I18G (f). B, effect of first ␤-sheet residues 18 -32. Chemotaxis assays were performed with IL-8 (E), GRO␥ (q), I18G32I (Ç), and G18I32G (å). C, effect of third ␤-sheet residues 46 -53. Chemotaxis assays were performed with IL-8 (E), GRO␥ (q), I46G (ϫ), I46G53I (É), and G46I53G (ç). Data are the mean Ϯ S.D. of triplicate determinations from a single experiment and are representative of at least three independent experiments for each chimera. localized in or near the surface hydrophobic pocket on IL-8 bounded by residues Tyr 13 , Lys 15 , Phe 21 , and Arg 47 .