Discrete steps in binding and signaling of interleukin-8 with its receptor.

The mechanisms by which chemokines bind and signal through their receptors are complex and poorly understood. In the present study, we sought to dissect these processes and to map important functional domains of the two CXC chemokine (interleukin-8) receptors, CXCR1 (formally IL-8RA) and CXCR2 (formally IL-8RB), using blocking monoclonal antibodies (mAbs) to the receptors and a series of chimeras between CXCR1 and CXCR2. A panel of specific mAbs against CXCR1 or CXCR2, generated by immunizing mice with transfectants expressing either receptor, were shown to effectively block IL-8- and/or growth-related oncogene α (GROα) -mediated ligand binding, chemotaxis, elastase release, and VCAM-1 binding in CXCR1 and CXCR2 transfectants and/or human neutrophils. Of particular interest was an anti-CXCR1 mAb, 7D9, that inhibited chemotaxis, elastase release, and VCAM-1 binding but had no detectable effects on ligand binding. The epitopes of these blocking mAbs were mapped by using a series of CXCR1/2 chimera transfectants and synthetic peptides. Most of the anti-CXCR1 antibodies, except 7D9, mapped to the amino acid sequence WDFDDL (CXCR1 residues 10-15), and all the anti-CXCR2 antibodies mapped to the amino acid sequence FEDFW (CXCR2 residues 6-10). The epitope of mAb 7D9 mainly involved a region within the first 45 residues of CXCR1, and it appeared to be conformation-sensitive. These results support a model in which the binding and signaling of IL-8 with its receptor occur in at least two discrete steps involving distinct domains of the receptor. This model is consistent with the notion that discrete conformational changes of the receptor secondary to ligand binding are required to trigger various biological responses. Moreover, the ligand binding and chemotaxis properties of each CXCR1/2 chimeric receptor to IL-8 and GROα were determined. It was found that each is distinct in its ability to confer ligand binding and chemotactic response to IL-8 and GROα, and two conclusions could be made. 1) The N-terminal segment of CXCR1 is a dominant determinant of receptor subtype selectivity, consistent with previous studies using rabbit/human CXCR1/2 chimeras; and 2) the specificity determinant for GROα binding in CXCR2 involves sequences in the N terminus, distal to the first 15 residues, as well as other parts of the receptor.

Chemokines (chemoattractant cytokines) play an important role in leukocyte trafficking to inflammatory sites. For example, interleukin-8 (IL-8) 1 is one of the most potent chemoattractants for neutrophils and is produced by many cell types in response to inflammatory stimuli. It induces angiogenesis, mediates cytokine-induced transendothelial neutrophil migration, and triggers a variety of other effects associated with the inflammatory response (1). There is strong evidence indicating that in vivo IL-8 can initiate neutrophil infiltration to the site of inflammation, causing tissue injury. In animal models of acute inflammation such as adult respiratory distress syndrome (ARDS) and acid induced lung injury, anti-IL-8 antibodies have produced beneficial effects in reducing inflammatory tissue damage and improving survival (2)(3)(4). All these data support the notion that if the IL-8 activities could be specifically controlled in humans, we may be able to prevent injuries caused by infiltrating neutrophils and other inflammatory cells. For such purposes, a better understanding of how IL-8 binds and signals with its receptors will certainly help us develop potent receptor antagonists.
Two distinct types of IL-8 receptors, CXCR1 (formally IL-8RA) and CXCR2 (formally IL-8RB), have been cloned and sequenced (5,6). These two subtypes share about 74% amino acid identity. Their transmembrane regions and intracellular loops are extremely similar while the N-and C-terminal regions are very different. The IL-8 receptors, and other chemokine receptors, belong to a family of seven-transmembrane or "serpentine" receptors that interact with GTP-binding proteins to elicit various biological responses (7,8). These two IL-8 receptors display overlapping but distinct ligand specificities. CXCR1 binds IL-8 with high affinity, but not other chemokines, while CXCR2 binds with high affinity to other CXC chemokines such as GRO␣, NAP-2, as well as IL-8. The functional significance of such selectivity is not clear since on peripheral blood cells, where most of the IL-8 activities have been observed, these two receptors are expressed at similar levels on neutrophils and subsets of lymphocytes.
The interaction between IL-8 and its receptors seems to be complex. A number of functional domains on the IL-8 molecule have been identified as essential for this interaction. Results obtained from mutagenesis and synthetic analogs of IL-8 demonstrated that residues Glu 4-Leu 5-Arg 6 of IL-8 were necessary, but not sufficient, for high affinity binding and activities (9 -12). Recent studies identified a second region, comprising residues that cluster near a surface-accessible hydrophobic pocket, that is important for IL-8 function (13,14). This region contains a second site for IL-8 receptor recognition that, in combination with the Glu 4-Leu 5-Arg 6 region, can modulate receptor binding affinity and specificity.
The IL-8 receptors may also provide multiple functional domains contributing to ligand binding and subsequent signaling events. Genetic substitution of each of the aa residues in the extracellular face of CXCR1 with alanine revealed that several residues in the N terminus and extracellular loops of CXCR1 are important for high affinity binding of IL-8 to CXCR1 and IL-8-mediated signal transduction (15,16). Studies with human/rabbit chimeric IL-8 receptors indicated that the N-terminal regions could confer the specificity selection for GRO␣ and NAP-2 between the two receptor subtypes in a cross-species setting (17,18). A more recent report suggested that regions other than the N terminus of the receptors might be important for the difference between human CXCR1 and CXCR2 ligand specificity (19). In the latter study, the authors also showed that IL-8, GRO␣, and NAP-2 could elicit Ca 2ϩ signaling via a CXCR2 chimeric receptor with the entire N terminus being replaced with that of human CC chemokine receptor 1 (CCR1), even though no obvious binding of IL-8 was detected. Although these studies provide some insight as to the structural requirements for IL-8 binding and signaling, they were not aimed to dissect these two processes.
Blocking antibodies against chemokine receptors could provide valuable tools in studying the structure-function relationship of the receptor. Several monoclonal antibodies (mAbs) or polyclonal antibodies against IL-8 receptor have been developed and among them, some block the ligand binding of 125 Ilabeled IL-8 to transfectant cells (20 -22). We have recently described the generation of mAbs against IL-8 receptors and the use of these mAbs to detect receptor expression on various cell populations (23). In the present study, we further characterized these mAbs. The main finding from this work is that the IL-8 binding and signaling with its receptor appear to occur in discrete steps, potentially involving distinct determinants of receptor structure or conformation, as suggested by an anti-CXCR1 mAb that can effectively block IL-8-mediated functional responses without detectable effects on ligand binding.

Construction of Wild-type and Chimeric Human CXCR1 and CXCR2
Expression Constructs-All expression constructs were made in pcDNA3 (Invitrogen, CA). The name, composition, and construction of each chimera receptor are as follows (also refer to Fig. 3; note that for the nomenclature of the chimeras, the letter A or B depicts the receptor subtype of CXCR1 and CXCR2 respectively, and the number preceding the letter indicates the number of aa residues derived from that corresponding receptor subtype).
Chimera 49B, which consists of CXCR2 (aa 1-49) and CXCR1 (aa 46 -350), was constructed by replacing the MscI-MscI-MscI fragments of the wt CXCR2/pcDNA3 construct by that of the CXCR1 construct. Chimera 45A, CXCR1 (aa 1-45) and CXCR2 (aa 50 -355), was the reciprocal chimera to 49B. To make this construct, the first ϳ500 bp of CXCR1 gene was amplified by PCR from CXCR1 using 5Ј-and 3Јprimers. The 3Ј-primer contains a sequence of the CXCR1 gene around 500 bp, followed by a sequence corresponding to the CXCR2 gene around the NdeI site. This PCR fragment was then used to replace the corresponding HindIII-NdeI fragment of CXCR2/pcDNA3 to create 45A. Chimera 15A, CXCR1 (aa 1-15) and CXCR2 (aa 16 -355), was constructed by first generating a PCR fragment using a 5Ј-primer corresponding to the BglII site of the pcDNA3 (upstream from the inserted gene) and a 3Ј-primer corresponding to the aa residue 10 -13 of CXCR1 with a flanking BglII site. This PCR fragment was then subcloned into the BglII-BglII site of the CXCR2-wt construct to replace the corresponding region in CXCR2. Chimera 10A, CXCR1 (aa 1-10) and CXCR2 (aa 11-355), and chimera 5A, CXCR1 (aa 1-5) and CXCR2 (aa 6 -355), were made by PCR using a primer containing the CXCR1 aa 1-9 (for 10A) or aa 1-5 (for 5A) sequence, with a 3Ј-primer of the CXCR2. All constructs made by PCR were sequenced to ascertain the sequence fidelity.
Creation of Cell Lines Stably or Transiently Expressing Wt and Chimeric CXCR1/2-For stable transfectants, the murine pre-B lymphoma cell line (L1-2) was used and transfected as described (24,25). Clones expressing a high level of receptors were selected by staining with appropriate anti-CXCR1 or CXCR2 antibodies, and further enriched either by limiting dilution and rescreening or FACS sorting. For chimera 10A, since none of the anti-CXCR1 or -CXCR2 antibodies recognize it, ligand binding to IL-8 was used as a means of selection for high expressors. Some of the transfected cell lines were treated with 5 mM n-butyric acid for overnight prior to the experiments in order to enhance the receptor expression. For transient transfection, CHO-P cells were transfected in the presence of LipofectAMINE according to the manufacture's instruction (Life Technologies, Inc.).
Antibodies and Flow Cytometry, Isolation of Neutrophils-Anti-CXCR1 and -CXCR2 antibodies were generated as described (23). In addition, two previously reported (21) anti-CXCR2 mAbs, 10H2.12.1 (ATCC HB-11494) and 4D1.5.7 (ATCC HB-11495) were used for the purpose of comparison. Protein G-Sepharose purified mAbs were stored at Ϫ20°C in phosphate-buffered saline until use. Immunofluorescent staining was performed as described and analyzed using a FACScan (Becton Dickinson, Mountain View, CA) (23,25). Human neutrophils were isolated from heparinized venous blood by sedimentation in 1% dextran and centrifugation through Percoll density gradient centrifugation (d ϭ 1.077) at room temperature. Biotinylation of mAbs were carried out using NHS-LC-Biotin (Pierce) following manufacturer protocols. To determine relative affinities of the mAbs, 2 g/ml biotinylated mAb was incubated with cells in the presence of increasing concentrations of unlabeled mAbs. Biotinylated mAb was detected with Streptavidin-fluorescein isothiocyanate, and the mean fluorescence was determined by FACS. Concentration of unlabeled mAbs that gave rise to 50% reduction of mean fluorescence (IC 50 ) was calculated using Kaleida-Graph software.
Ligand Binding Analysis-125 I-labeled human IL-8 and GRO␣ were purchased from Amersham (Arlington Heights, IL) or DuPont NEN (Boston, MA). Recombinant human IL-8 was expressed in E. coli using the pET system (pET3a plasmid and BL21 cells) from Novagen. Recombinant human IL-8 was purified to N-terminal sequence homogeneity from cell supernatants by heparin affinity chromatography (hi-trap, 5 ml, Pharmacia Biotech, Inc.) followed by size exclusion chromatography (Superdex 75, Pharmacia) and C18 reverse-phase high pressure liquid chromatography (Ultrasphere ODS, 1 ϫ 25 cm, Beckman). GRO␣ was purchased from Peprotech (Rocky Hill, NJ). For each binding reaction, 60-l cells, washed and resuspended in Hanks' balanced salt solution containing 0.5% bovine serum albumin at 5 ϫ 10 6 /ml, was mixed with 60 l of 2 ϫ reaction mix and incubated at 37°C for 30 min. For most of the experiments, the final concentration of 125 I-chemokines in the reaction is 0.1 nM. Nonspecific binding was determined in the presence of 100 -250 nM cold chemokines. The binding reaction was stopped by transferring 100 l of the mix to tubes containing 200 l of dibutyl phthalate/bis(2-ethylhexyl) phthalate (1:1 mix) and spun at 12,000 rpm for 3 min. The supernatant was frozen in dry ice, and the cell pellet in the bottom of the tubes was cut off and subjected to counting in a ␥-counter. For Scatchard analysis, 0.05 nM of 125 I-chemokine was added with increasing concentrations of unlabeled chemokine. The data was curve fitted and K d and B max was calculated using the computer program Ligand (26). All experiments were carried out using duplicates and repeated at least twice.
Assay for IL-8-induced Chemotaxis, Adhesion to VCAM-1, and for Elastase Release-Biocoat trans-well tissue culture inserts (Collaborative Biomedical Products, MA) were used for the chemotaxis assays as described (23,25,27). The IL-8-induced adhesion of CXCR1 or CXCR2 transfectants to VCAM-1 were carried out similarly to what was described (24). Cells were labeled with BCECF-AM (Molecular Probes, Eugene, OR) and preincubated with various concentrations of mAbs for 5-10 min. Cells were then left to settle on VCAM-1-coated plates for 5 min before IL-8 (100 nM) was added. The assay was carried out at room temperature for 2 min and stopped by washing the plates, and the fluorescence intensity was measured by a fluorescence concentration analyzer (Pandex). For the elastase release assay, neutrophils treated with cytochalasin B were incubated with mAbs or buffer only for five minutes. Elastase substrate (N-methoxysuccinyl-Ala-Ala-Pro-Val-AMC, Bachem) was added to a final concentration of 50 M before IL-8 or GRO␣ was added. The fluorescence intensity of each well was measured by the fluorescence concentration analyzer immediately and after 10 min of stimulation. Percent inhibition was calculated using readings of relative fluorescence units.
Peptide Competition to the Antibody Staining-Peptides were synthesized by Genemed Biotechnology (San Francisco, CA) and solubi-lized in phosphate-buffered saline. For the competition experiments, the antibody (at subsaturating concentration, 1-2 g/ml) was preincubated with various amounts of peptides at 37°C for 1 h and then mixed with 3 ϫ 10 5 cells that were preresuspended in phosphate-buffered saline containing 5% fetal calf serum. After incubation for 20 -30 min on ice, cells were washed and subjected to staining of fluorescein isothiocyanate-conjugated secondary antibody and FACS analysis as described above. The mean fluorescence was recorded and plotted, and the curve was fitted and IC 50 calculated by KaleidaGraph (Synergy Software, Reading, PA).

Generation of Anti-CXCR1 and -CXCR2 mAbs That Can Block the IL-8/GRO␣-mediated Ligand Binding and/or Other Signal Transduction Responses-Fifteen mAbs against CXCR1
and six mAbs against CXCR2 were generated by immunizing mice with murine L1-2 cells expressing human CXCR1 or CXCR2. No mAbs recognized both CXCR1 and CXCR2. These mAbs stained human neutrophils, monocytes, and subsets of lymphocytes (23).
The ability of these antibodies to inhibit ligand binding, chemotaxis, and other functions mediated through CXCR1 or CXCR2 were examined. In ligand binding assays using 125 Ilabeled IL-8 or GRO␣, the majority of anti-CXCR1 mAbs efficiently blocked the binding of IL-8 to CXCR1 transfectants, and all the anti-CXCR2 mAbs efficiently blocked the binding of IL-8 and GRO␣ to CXCR2 transfectants. For reasons of simplicity, only one anti-CXCR1 mAb, clone 5A12, and one anti-CXCR2 mAb, clone 6C6, are presented here, but most other mAbs gave similar results. Additionally, an anti-CXCR1 mAb, 7D9, is also described as it shows some unusual properties. As shown in Fig. 1, A and B, both 5A12 and 6C6 are very potent in blocking binding of 125 I-labeled IL-8 or GRO␣ to the transfectant cells, and the IC 50 for this is ϳ50 ng/ml. In contrast, mAb 7D9 failed to exhibit any significant inhibition in binding of 125 I-labeled IL-8 to CXCR1 transfectant cells. Furthermore, several irrelevant mAbs did not show any inhibition in IL-8 or GRO␣ binding to the receptor transfectants (not shown).
We next tested the ability of these mAbs to block IL-8 or GRO␣mediated chemotaxis in CXCR1 and CXCR2 transfectants. In a transmembrane chemotaxis assay, these transfectants responded to IL-8 (CXCR1 and CXCR2) and GRO␣ (CXCR2) in a dose-dependent, bell-shaped profile. Incubation of CXCR1-expressing cells with anti-CXCR1 mAbs greatly reduced the number of cells migrating toward IL-8. As shown in Fig. 1C, at an antibody concentration of 20 g/ml, the maximal chemotaxis of these cells were reduced by 97% for 5A12 and 81% for 7D9, respectively. Similarly, anti-CXCR2 mAb 6C6, and other anti-CXCR2 mAbs, inhibited chemotaxis of CXCR2 transfectants to either IL-8 or GRO␣ (Fig. 1D).
The efficiency of these mAbs to inhibit IL-8 activities on neutrophils was not as marked as on receptor transfectants when the anti-CXCR1 or -CXCR2 mAbs were used separately. This is not surprising as neutrophils express both IL-8 receptor subtypes. When anti-CXCR1 and -CXCR2 mAbs were used together to block neutrophil chemotaxis, the potency of inhibition depends on the concentration of IL-8 used. At lower IL-8 concentrations, the inhibition could be as high as 95% ( Fig. 2A), while at higher IL-8 concentrations, the inhibition was only 40% (Fig. 2B). On the other hand, GRO␣-induced neutrophil chemotaxis, which is known to be mediated through CXCR2, could be completely blocked by anti-CXCR2 mAbs while anti-CXCR1 mAbs had little effect (Fig. 2C).
The release of neutrophil granule enzymes such as elastase can be elicited through CXCR1 and CXCR2 (28,29). Here we examined the ability of anti-CXCR1 and -CXCR2 mAbs to inhibit elastase release using human neutrophils. Although single mAbs as high as 200 g/ml only inhibited IL-8-stimulated elastase release by about 20%, a combination of anti-CXCR1 and -CXCR2 mAbs was more effective. 5A12 and 6C6 inhibited elastase release by 68% while 7D9 and 6C6 inhibited elastase release by 52%. GRO␣-stimulated elastase release was inhibited by 6C6 by 75%.
It has been shown that the stimulation of IL-8 receptor transfectants by IL-8 causes transient integrin activation (24). It was found that 5A12, 7D9, and 6C6 were able to inhibit IL-8-induced VCAM-1 binding of the CXCR1 or CXCR2 transfectants by 35-50% (data not shown).
Combining the above data, we noticed that although the two anti-CXCR1 mAbs, 5A12 and 7D9, inhibited functions of receptor transfectants similarly, their ability to block ligand binding was very different as 5A12, but not 7D9, could effectively inhibit IL-8 binding to CXCR1 transfectants. This differential effect does not appear to be due to the different binding ability of the two mAbs to the receptor since they had similar binding affinity to CXCR1 transfectants (IC 50 ϭ 2.7 nM for 5A12 and 4.6 nM for 7D9). To ascertain that the inhibition of 7D9 on CXCR1 transfectants and neutrophils was mediated directly through CXCR1, the effect of this mAb on other cells was examined. It was found that 7D9 had no inhibition on the chemotaxis of transfectants expressing CXCR2 or CXCR2/1 chimera 49B (which was not stained by 7D9, see below). Moreover, 7D9 did not inhibit formyl-methionyl-leucyl-phenylalanine-mediated chemotaxis of neutrophils, whereas it inhibited IL-8-mediated chemotaxis by 40 -50% in the same experiment.
All Anti-CXCR1 and -CXCR2 Antibodies Map to Very Small Segments in the N-terminal Extracellular Region of the Receptor-The unique properties of these mAbs prompted us to identify their binding epitopes. We first took the chimera approach. The most divergent region between CXCR1 and CXCR2 resides in the N-terminal extracellular domain. Since these antibodies are receptor subtype-specific, we first made two CXCR1/2 chimeras by exchanging the N termini. In the CXCR1/2 chimera 45A, the N-terminal extracellular region of CXCR2 was replaced by that of CXCR1. In the reciprocal CXCR2/1 chimera 49B, that region of CXCR1 was replaced by that of CXCR2 (Fig.  3). For these and all the other chimeras, both stable and transient transfectants were generated in L1-2 and CHO-P cells respectively. It was found that all the transient transfectants behaved similarly to each corresponding stable transfectant line. Thus for the purpose of simplicity, only the data with the stable lines will be shown. All 15 anti-CXCR1 antibodies stained wt CXCR1 and 45A but not 49B, and all 6 anti-CXCR2 antibodies stained wt CXCR2 and 49B but not 45A. None of the antibodies showed any staining of untransfected cells. The results of the stable L1-2 transfectants using a representative antibody (5A12 for anti-CXCR1 and 6C6 for anti-CXCR2) are shown in Fig. 4A, and all the other antibodies exhibited similar staining patterns (not shown). These results suggest that each of the mAbs recognizes epitopes that map to the N-terminal extracellular region of the receptors.
To further localize the antibody binding epitope, several more chimeras were generated (Fig. 3), and the properties of each in antibody staining were determined (Fig. 4A). Cells expressing chimera 15A, which contains the first 15 aa residues of CXCR1, were stained very well by all the anti-CXCR1 antibodies except 7D9. These cells were not stained by any of the anti-CXCR2 antibodies. When the N-terminal sequence of CXCR1 was decreased to 10 aa (in chimera 10A), none of the anti-CXCR1 or anti-CXCR2 antibodies showed any staining. The lack of staining of this chimera receptor by any antibodies is not due to the lack of cell surface expression because it exhibited ligand binding and chemotaxis at a level that was comparable with that of wt and other chimera cell lines (see below). When the amount of the CXCR1 sequence is reduced to only the first 5 aa residues (chimera 5A), as expected, this chimera could not be recognized by any of the anti-CXCR1 antibodies. Interestingly, however, this chimera could be stained very brightly by all the anti-CXCR2 mAbs.
These combined results indicate that the aa residues 11-15 of CXCR1 are essential for the anti-CXCR1 antibodies to bind. Since residue 10 (W) is conserved between CXCR1 and CXCR2, the sequence of residues 10 -15 (WDFDDL) is tentatively assigned as the binding epitope for all the anti-CXCR1 antibodies, except 7D9 (see below). Based on the staining properties of chimeras 10A and 5A, the aa residues 6 -10 of CXCR2 (FEDFW) appear to be the binding epitope for the anti-CXCR2 antibodies. Additionally, two anti-CXCR2 antibodies, 10H2 and 4D1, generated by another group (21) were also tested on these chimeras and found to behave similarly to the anti-CXCR2 antibodies we generated (data not shown). Thus their epitopes are also assigned to "FEDFW." The anti-CXCR1 antibody, 7D9, behaved differently from the other anti-CXCR1 antibodies. Although it stained 45A as well as the other antibodies (not shown), unlike other antibodies, it did not stain or stained very weakly chimera 15A (Fig. 4B). These results suggest that the sequence of WDFDDL is, at best, only partially involved in 7D9 binding.

The Synthetic Peptides Derived from the Putative Epitope Region Can Specifically Inhibit the Corresponding Antibody
Binding-To further confirm the epitopes mapped by chimera mutagenesis above, several peptides were synthesized, and their ability to compete for mAb binding to CXCR1 or CXCR2 transfectants were examined (Table I). Peptide A10 -15 (WD-FDDL), the shortest peptide corresponding to the putative CXCR1 antibody epitope, could effectively block the binding of 5A12 (IC 50 ϭ 0.5 M) to CXCR1 transfectants. As expected, this peptide did not block the binding of anti-CXCR2 mAb 6C6 to CXCR2 transfectants, even at a peptide concentration as high as 200 M. Conversely, peptide B6 -10 (FEDFW), the shortest peptide corresponding to the putative CXCR2 antibody epitope, could effectively inhibit 6C6 binding (IC 50 ϭ 0.5 M) but not 5A12 binding. Furthermore, peptide A6 -10 (DPQMW) or B10 -15 (WKGEDL) had no inhibitory effect on 5A12 or 6C6 binding.
The peptide competition properties of 7D9 is different from those of 5A12. Its binding to CXCR1/L1-2 cells could not be effectively inhibited by peptide A10 -15. When this peptide was used at a very high concentration (Ͼ100 M), a slight inhibition (30 -40%) could be observed. Another peptide, A10 -19 (WD-FDDLNFTG), had the similar inhibitory effect (not shown). Similar to 5A12, however, the peptides that did not inhibit 5A12 staining (peptides B6 -10, A6 -10, or B10 -15) did not have any inhibitory effect on 7D9 staining either.
These combined results suggest that the aa residues 10 -16 of CXCR1 (WDFDDL) and 6 -10 of CXCR2 (FEDFW) are the minimal binding sequences required for 5A12 and 6C6 mAb to bind, respectively. On the other hand, for 7D9 mAb, the sequence of CXCR1 residues 10 -15 (WDFDDL) is not its primary binding site. Rather, it may recognize a conformation in which WDFDDL may or may not be a part (under "Discussion").
Despite the fact that these mAbs could effectively block ligand binding and/or chemotaxis, none of the above peptides, shown to contain the mAb binding epitopes, had significant inhibitory effect on 125 I-labeled IL-8 binding to CXCR1 or CXCR2 transfectants at peptide concentrations up to ϳ200 M. Furthermore, these peptides did not inhibit IL-8-mediated che-motaxis in CXCR1 or CXCR2 transfectants or human neutrophils (data not shown).

CXCR1/2 Chimeras Have Distinct Ligand Binding Properties to IL-8 and GRO␣-
The ligand binding properties of wt and each chimera transfectant cell line to IL-8 and GRO␣ were determined ( Fig. 5 and Table II). As expected, cells expressing either wt CXCR1 or CXCR2 showed high affinity ligand binding to IL-8, with a K d of 0.58 nM and 0.54 nM, respectively. Cells expressing CXCR2 also exhibited high affinity ligand binding to GRO␣ (K d ϭ 0.97 nM), and cells expressing CXCR1 did not. These results confirmed the previous reports that CXCR1 is more strictly an IL-8 receptor, and CXCR2 is a more promiscuous receptor for other ␣-chemokines.
The CXCR1/2 chimera 45A consistently showed higher affinity binding to 125 I-labeled IL-8 than wt CXCR1 or CXCR2 (compare K d ϭ 0.1 nM for 45A with K d ϭ 0.5-0.6 nM for wt receptors). This binding could not be competed by cold GRO␣, neither could any specific binding be detected when 125 I-labeled GRO␣, at a concentration up to 5 nM, was used directly.
All the other CXCR1/2 chimeras, 15A, 10A, and 5A, showed high affinity binding to IL-8. The K d values ranged from 1.3 to 3.3 nM (Table II), which are slightly higher than those of wt CXCR1 and CXCR2 (K d ϭ 0.5-0.6 nM). Furthermore, these chimeras also exhibited high affinity binding to GRO␣ with K d

TABLE I Peptide competition of mAb binding to IL-8R transfectants
The mAb (at 1-2 g/ml) was preincubated with various amounts of peptides at 37°C for 1 h and then mixed with 3 ϫ 10 5 CXCR1/L1-2 or CXCR2/L1-2 cells. Flow cytometry staining was carried out as described under "Experimental Procedures," and mean fluorescence was recorded. A dose curve for each peptide was carried out, and the curve was fitted and IC 50 calculated by KaleidaGraph. The IC 50 values shown are the average of at least three experiments. "Ϫ," no inhibition at a peptide concentration up to 200 M.

. Binding of 125 I-labeled IL-8 and 125 I-labeled GRO␣ to L1-2 cells expressing wt CXCR1, CXCR2, and chimeric CXCR1/2.
In most experiments, cells were incubated in the presence of 0.05 nM 125 I-labeled IL-8 (å) or 125 I-labeled GRO␣ (f) with increasing concentrations of cold IL-8 or GRO␣, respectively. Nonspecific binding was determined by adding 250 nM unlabeled chemokines. The cell lines (L1-2 for IL-8 and GRO␣, CXCR1 for GRO␣, 45A for GRO␣, and 49B for IL-8 and GRO␣) that didn't show any specific binding in these conditions were further tested in saturation experiments with labeled chemokines where increasing concentrations of labeled chemokines were used. In those cases, no specific binding was obtained when up to 5 nM 125 I-labeled chemokines were used. Each experiment was done at least twice, and a representative one is shown in the figure. The data were also analyzed using Ligand program, and the K d was shown in Table II. The average number of binding sites per cell is between 40,000 and 200,000. values that ranged from 2.5 to 4.8 nM, which is slightly higher than that of CXCR2 (K d ϭ 0.97). As expected, the binding of IL-8 and GRO␣ in wt CXCR2 and chimeras 15A, 10A, and 5A could be heterologously competed by each other, i.e. the binding of 125 I-labeled IL-8 could be effectively competed by GRO␣, and the binding of 125 I-labeled GRO␣ could be effectively competed by cold IL-8 (data not shown).
It is worth noting the properties of the CXCR2/1 chimera, 49B. This chimera was expressed as well as wt and other chimeras as shown by the flow cytometry analysis (Fig. 4). However, cells expressing this chimera did not show any binding to IL-8 or GRO␣ in the presence of 0.1 nM radiolabeled ligands. Saturation experiments with 125 I-labeled IL-8 or 125 Ilabeled GRO␣ were carried out to determine if they had much reduced binding affinity to these chemokines. No specific binding could be detected even when up to 5 nM 125 I-labeled IL-8 or 125 I-labeled GRO␣ was used.
CXCR1/2 Chimeras Exhibit Chemotactic Responses to IL-8 and GRO␣-To further investigate the functions of the CXCR1/2 chimeras, each cell line was tested for its ability to chemotax to IL-8 and GRO␣ over a wide range of doses (Fig. 6, Table II). Both wt CXCR1 and CXCR2 transfectants showed a strong chemotactic response to IL-8. The maximum response, which exhibited a 100 -300-fold increase of migrated cells compared with the background (medium alone), was observed at 1-10 nM IL-8 although specific migration could be detected at an IL-8 concentration as low as 0.1 nM. Similar to the IL-8 response, CXCR2 transfectants also chemotaxed to GRO␣ very well with the peak response at 1-10 nM and leveled off at higher doses. CXCR1 transfectants, on the other hand, showed no response to GRO␣ up to 10 nM and showed a weak response at 100 nM, and this response increased when 1000 nM GRO␣ was used.
As shown above, the CXCR1/2 chimera 45A behaved similarly to that of wt CXCR1 with respect to antibody staining and ligand binding. This chimera chemotaxed to IL-8 similarly as wt CXCR1 in terms of the dose response but showed a much weaker response to GRO␣ even at a very high concentration (1000 nM) (Fig. 6A). Interestingly, despite un-detectable specific binding to 125 I-labeled IL-8 and 125 I-labeled GRO␣ at the concentration used (up to 5 nM), 49B did show chemotaxis to IL-8 when a higher concentration of chemokine (100 nM) was used. Moreover, it also exhibited a weak response to GRO␣ at a concentration of 100-1000 nM (Fig. 6A).
The chemotactic responses of the other CXCR1/2 chimeras (5A, 10A, and 15A) to IL-8 and GRO␣ closely correlate with their ligand binding profiles. As shown in Fig. 6B, chimeras 15A and 5A chemotaxed to IL-8 and GRO␣ similarly as did CXCR2 transfectants, with the maximum response at 1-10 nM. Chimera 10A also chemotaxed to IL-8 very well, but had a reduced response to GRO␣. This is somewhat consistent with the fact that among these three chimeras, in binding to IL-8 or GRO␣, 10A exhibited the highest K d in binding to GRO␣.

DISCUSSION
In this study, we developed and characterized a panel of anti-CXCR1 and anti-CXCR2 mAbs. Interestingly, they all map to very small regions in the N terminus of the receptor, specifically, CXCR1 residues 10 -15 (WDFDDL) for anti-CXCR1 mAbs and CXCR2 residues 6 -10 (FEDFW) for anti-CXCR2 mAbs. Previous reports showed that specific mAbs against CXCR1 or CXCR2 could also be generated using transfected 293 cells expressing each receptor subtype (20,21). Although different cell transfectants were used as immunogens in those studies, these mAbs also mapped to the N-terminal regions of the receptor. Additionally, anti-CXCR1 or anti-CXCR2 antibodies were also generated by using receptor-specific peptides, and the epitopes of these antibodies were all FIG. 6. Chemotaxis of wt and chimeric CXCR1/2 to IL-8 and GRO␣. Chemotaxis was carried out as described under "Experimental Procedures." A, chemotaxis of wt CXCR1, CXCR2, and chimeras 45A and 49B. B, chemotaxis of chimeras 15A, 10A, and 5A where CXCR2 was used as a control.

TABLE II
Summary of mAb staining, ligand binding, and chemotaxis of wt and chimeric CXCR1/2 Antibody staining, ligand binding, and chemotaxis were carried as described under "Experimental Procedures." The K d values for ligand binding are the average of at least two independent Scatchard analysis experiments, and one representative of each cell line is shown in Fig. 5. The chemotaxis data are a summary of the results in Fig. 6, "ϩϩϩ" indicates maximum chemotactic response around 1 nM of chemokine, and "ϩ/Ϫ" indicates the maximum response was only observed at a chemokine concentration of 100 -1000 nM. mapped to N-terminal regions of residues 2-11 (20) or 1-15 (22). This may suggest that the N-terminal region of the receptor is more immunogenic than the other extracellular loops, or that the other extracellular loops are inaccessible to serve as immunogens. Furthermore, none of the mAbs generated by us or by others could recognize both CXCR1 and CXCR2. This is not surprising since the N terminus is the most divergent region between the two receptors.
Most of the mAbs generated in this study are capable of effectively blocking IL-8 and/or GRO␣ ligand binding as well as chemotaxis, elastase release, and VCAM-1 binding in CXCR1/ CXCR2 transfectants and/or human neutrophils. Thus they provide valuable tools in studying receptor subtype-specific functions (30). It is interesting to note that the anti-CXCR1 mAbs generated by Chuntharapai et al. (20) map to residues 2-14 for blocking antibodies and 2-11 for non-blocking antibodies, suggesting that the binding to residues 12-14 (FDD) might be important for blocking activity. In this report we have narrowed down the binding epitope of our anti-CXCR1 mAbs to residues 10 -15, overlapping with those identified by Chuntharapai et al. (20). All the anti-CXCR2 mAbs developed in this study were mapped to CXCR2 residues 6 -10 (FEDFW). Furthermore, two others (10H2 and 4D1) generated by Chuntharapai et. al. (21), which were previously mapped to a region within residues 1-18, were also narrowed down to residues 6 -10 in this study. These results suggest that although these residues may not necessarily be the direct binding sites for IL-8, they probably play important roles in ligand binding and other functionalities of the receptors.
The most interesting finding from this study is revealed by a unique anti-CXCR1 mAb, 7D9. Although this mAb inhibited chemotaxis as effectively as other blocking mAbs, it could not inhibit IL-8 binding to CXCR1 transfectants. The inability of 7D9 to inhibit ligand binding does not seem to be caused by a lower affinity of this antibody since 7D9 and 5A12 had similar binding affinities as determined by using biotin-labeled mAbs on transfectant cells. The similar potencies of 7D9 and 5A12 in inhibiting chemotaxis also suggest that it is not due to their difference in affinity. Furthermore, the inhibition of chemotaxis by 7D9 and other mAbs did not result from receptor modulation from the cell surface during the assay (4 -5-h incubation at 37°C) because these antibodies could also effectively inhibit transient responses mediated by CXCR1 and CXCR2, such as elastase release from neutrophils and VCAM-1 binding to transfectant cells.
The N-terminal region of CXCR1 is an essential part of the epitope for 7D9 as 7D9 could recognize chimera 45A but not 49B, even though we could not completely rule out the possible involvement of other extracellular loops of the receptor since they are conserved between CXCR1 and CXCR2. However, based on the chimera staining and peptide competition results, 7D9 does not recognize the region seen by other anti-CXCR1 mAbs, i.e. the aa residues 10 -15 (WDFDDL). The inclusion of a few additional residues after this sequence is not enough for its binding either, as peptide A10 -19 could not effectively compete 7D9 binding. We noted that in Western blot analyses using these mAbs, 7D9 could not detect CXCR1, whereas 5A12 and other mAbs could. 2 These combined data support the contention that 7D9 recognizes a conformational epitope predominantly within the N terminus of the receptor, in which residues residues 10 -15 (WDFDDL) are not dominant.
In this model, the binding of 7D9 still allows IL-8 to bind, but the antibody-receptor interaction constrains the receptor conformation. Thus the receptor cannot undergo any IL-8-induced conformational changes to elicit appropriate biological responses, such as chemotaxis and degranulation. Mutational studies on adrenergic receptors have suggested that secondary structures of the receptor are important for coupling to Gproteins (31)(32)(33). It is believed that the binding of an agonist to a G-protein-coupled receptor causes a conformational change in the receptor, leading to the formation of a high-affinity agonist receptor-G-protein complex, initiating the signal transduction cascade (for a review, see Ref. 34). Results obtained from this study using blocking mAb to the receptor support this model.
Alternatively, it is possible that the binding of 7D9 to the receptor blocks the access of a signaling domain of IL-8 to the appropriate site of the receptor, therefore, leading to the failure to carry out signal transduction responses. Studies on the IL-8 structure-function relationship suggested that multiple domains of IL-8 are involved in binding and signaling (12)(13)(14). It has been shown that the interaction domains between the inflammatory protein C5a and its receptor consist of two distinct subsites and the occupation of only one of the subsites is required for receptor activation (35).
No matter which of the above two models is correct, they are all consistent with the notion that the binding and signaling of IL-8 with its receptor may occur in two discrete steps, mediated by distinct structural determinants. Although we have not completely mapped the determinant recognized by 7D9, its significance lies in the demonstration that these two processes are discrete. To our knowledge, this is the first report where a mAb against a chemokine receptor could block the signaling step but not the ligand binding. Thus it may provide a valuable tool for further dissecting of the two processes.
The results on the ligand binding and chemotaxis with the CXCR1/2 chimeras generated in the present study provide useful information on the structure-function relationship of CXCR1 and CXCR2. The two reciprocal chimeras that switched the N-terminal region of the receptors, 45A and 49B, are identical to AB1 and BA1, respectively, as described in Ahuja et al. (19), except that the transfectants were made in different cell lines (L1-2 versus HEK 293 cells). We found that some of the properties of these chimeras are the same between the two studies, and others are different. Both studies demonstrated that 45A (AB1) shows high affinity ligand binding to IL-8, and this binding cannot be competed by cold GRO␣. Our data showed that it has a slightly higher affinity than wt CXCR1 and CXCR2. Furthermore, IL-8 can induce strong signal transduction responses as measured by chemotaxis (this study) or Ca 2ϩ flux (19). As to GRO␣ responses, we could not detect any specific binding to GRO␣ even when 125 I-labeled GRO␣ was used at a concentration as high as 5 nM, nor could we observe a chemotactic response to GRO␣ unless a very high concentration of chemokine (100 nM) was used, which is similar to the response exhibited by CXCR1 transfectants. Somewhat differently, Ahuja et al. (19) reported a K i value of 13 nM to GRO␣ in ligand binding and an EC 50 of 9.5 nM for Ca 2ϩ flux for this chimera. Our results are more consistent with those of LaRosa et al. (17) and Gayle et al. (18) where the rabbit CXCR1 was used in the chimera rather than the human CXCR1, and they all suggest that the N-terminal segment of CXCR1 is a dominant determinant of receptor subtype selectivity.
For chimera 49B (BA1), neither we nor Ahuja et al. (19) could detect any specific ligand binding with 125 I-labeled IL-8. Signal transduction responses of this chimera mediated by IL-8 or GRO␣ were detectable at high concentrations of ligand (ϳ100 nM for chemotaxis and EC 50 ϭ 10 to 20 nM for Ca 2ϩ flux response). The difference between the two studies for this chimera resides in the fact that Ahuja et al. (19) was able to get a high affinity binding to GRO␣ (K i ϭ 0.7 nM) although the total binding is very low (14%), whereas we have failed to detect any 125 I-labeled GRO␣ binding. The discrepancy in the GRO␣ binding for both 45A and 49B between the two studies is not likely caused by the different level of receptor expression. For 45A, the average number of binding sites per cell was very similar (50,000 -100,000 in this study and 100,000 -200,000 by Ahuja et al. (19)). For 49B, although it is difficult to compare the receptor levels in the two studies since Ahuja et al. (19) reported 40,000 binding sites per cell and we could not calculate it due to the lack of binding to either IL-8 or GRO␣, we speculate that in our study, it was expressed at a level similar to that of wt CXCR2 (50,000 -100,000 sites/cell) since the two lines were stained by anti-CXCR2 mAb equally well. In the present study, the apparent discordance between chemotaxis and ligand binding imply that low percentage of receptor occupancy or high percentage occupancy at low affinity can still trigger a biological response. Moreover, since much higher concentrations of chemokines can be used in chemotaxis assays than in ligand binding assays, a weak or low affinity ligand binding-mediated response could be more easily detected in chemotaxis assays.
The chimeras that have smaller segments of CXCR1 sequences, 15A, 10A, and 5A, all showed properties similar to CXCR2 in ligand binding and chemotaxis. Since the first 15 aa residues between CXCR1 and CXCR2 are very divergent, these results suggest that the determinant for binding specificity to GRO␣ in CXCR2 is located distal to the first 15 aa residues. In conjunction with the data that chimera 49B does not bind GRO␣, these results support a model in which the GRO␣ binding is critically determined by sequences in the N terminus, after the first 15 aa residues, as well as other parts of the receptor. This is consistent with the conclusion drawn by Ahuja et al. (19) that the determinant of GRO␣ selectivity is located in both the N-terminal extracellular region and the TMD4 and E2 loop.
In summary, this study further demonstrates the complexity of chemokine binding to and signaling through its receptors, and suggests that a conformational change of the receptor or the ligand secondary to IL-8 binding may play an important role in ligand binding and signaling. Future experiments using mutant IL-8 and additional mutant/chimeric receptors are needed to further define the binding domain of 7D9, to determine the nature of the receptor conformational changes, and to better understand the discrete binding and signaling steps.