Synergistic Interactions between Chemokine Receptor Elements in Recognition of Interleukin-8 by Soluble Receptor Mimics

Abstract

Interleukin-8 (IL-8 or CXCL8), the archetypal member of the CXC chemokine subfamily, stimulates neutrophil chemotaxis by activation of the receptors CXCR1/IL8RA and CXCR2/IL8RB. Previous mutational studies have implicated both the N-terminal and third extracellular loop (E3) regions of these receptors in binding to IL-8. To investigate the interactions of these receptor elements with IL-8, we have constructed soluble proteins in which the N-terminal and E3 elements of either CXCR1 or CXCR2 are juxtaposed on a soluble scaffold protein; these are referred to as CROSS-N^X1E3^X1 and CROSS-N^X2E3^X2, respectively. Isothermal titration calorimetry (ITC) and NMR spectroscopy were used to compare the IL-8 binding properties of the receptor mimics to those of control proteins containing only the N-terminal or the E3 receptor element. CROSS-N^X2E3^X2 bound to monomeric IL-8 with the same affinity and induced the same chemical shift changes as the control protein containing only the N-terminal element of CXCR2, indicating that the E3 element of CXCR2 did not contribute to IL-8 binding. In contrast, CROSS-N^X1E3^X1 bound to IL-8 with ~10-fold increased affinity and induced different chemical shift changes compared to the control protein containing only the N-terminal element of CXCR1, suggesting that the E3 region of CXCR1 was interacting with IL-8. However, a chimeric protein containing the N-terminal region of CXCR1 and the E3 region of CXCR2 (CROSS-N^X1E3^X2) bound to IL-8 with thermodynamic properties and induced chemical shift changes indistinguishable from those of CROSS-N^X1E3^X1 and substantially different from those of CROSS-N^X2E3^X2. These results indicate that the N-terminal and E3 regions of CXCR1 interact synergistically to achieve optimal binding interactions with IL-8.

Chemokines are a family of secreted, soluble proteins that orchestrate the tissue localization of leukocytes by activation of G protein-coupled chemokine receptors expressed in the leukocyte membranes (1-3). Due to the importance of leukocyte trafficking in both homeostasis and inflammatory processes, there is widespread interest in understanding the molecular and mechanistic details of these interactions. Numerous chemokine structures have been determined and mutational studies have shown that the amino terminal and “N-loop” regions of these proteins play critical roles in their receptors interactions. The recent crystal structure of receptor CXCR4 (4) has confirmed that chemokine receptors consist of seven transmembrane helices, four extracellular elements (the N-terminus and 3 loops, referred to herein as E1, E2 and E3), and four intracellular elements (3 loops and the C-terminus). Receptor mutants and chimeras have lead to the model that chemokine ligands interact initially with the receptor N-terminus and subsequently with a second site formed by the extracellular loops and possibly including some interactions with transmembrane helices (5).

Interleukin-8 (IL-8 or CXCL8)¹, the archetypal member of the CXC chemokine subfamily, activates neutrophils by binding to the receptors CXCR1/IL8RA and CXCR2/IL8RB. Whereas CXCR1 is specific for IL-8, CXCR2 can also be activated by several other CXC chemokines, including CXCL1/GROα/MGSA and CXCL7/NAP-2 (6-8). IL-8, like many chemokines, is dimeric at high concentrations. Although recent results have implicated both monomeric and dimeric forms in chemotactic function (9, 10), the monomeric unit is sufficient for receptor activation (11). The three-dimensional structures of both wild type IL-8 and a monomeric IL-8 variant have been determined (12, 13). Mutants and synthetic variants of IL-8 have identified residues in the “N-loop” and 3₁₀-turn region (residues 10-22) as being important for binding to CXCR1 and CXCR2, whereas the N-terminal Glu-Leu-Arg (ELR) motif is critical for receptor activation (14, 15).

The N-terminal elements of CXCR1 and CXCR2 play a central role in the interactions of these receptors with chemokine ligands. Chimeric receptors consisting of the N-terminal element of one IL-8 receptor on a background of the other receptor displayed very similar ligand-binding selectivity to the receptor from which the N-terminal element was derived (16-18). In addition, site-directed mutations of several N-terminal residues caused significantly reduced binding affinity for IL-8 (19, 20). Moreover, peptides corresponding to the N-terminal elements of human or rabbit CXCR1 bind to human IL-8 (18, 21) and inhibit dimerization of the chemokine (22). NMR studies have mapped the binding sites of N-terminal peptides from human and rabbit CXCR1 on the surface of IL-8 (21, 23). Skelton et al have reported the solution structure of IL-8 bound to a peptidomimietic (CXCR1-p1) containing residues 9-14 and residues 20-29 of CXCR1 connected via a hexanoic acid linker (24). The peptidomimetic binds into a groove between the N-loop residues and the third β-strand of IL-8. In contrast, the IL-8 binding surface for the N-terminal element of CXCR2 has not been reported.

In addition to the N-terminal element, the three extracellular loops of CXCR1 and CXCR2 are also likely to play a role in receptor binding and/or activation by IL-8. A chimera consisting of the N-terminal element of receptor CCR1 (which is not a receptor for CXCR2 ligands) on a background of CXCR2 supported both binding and activation by CXCR2 ligands (17). Similarly, substitution of the C-terminal end (transmembrane region 5 onwards) of human CXCR2 by the corresponding region of rabbit CXCR1 increases affinity for IL-8 and for GRO/MGSA, suggesting that the third extracellular loop (E3) may contribute to the binding interaction (18). Moreover, extensive site-directed mutagenesis has identified residues in all three extracellular loops of CXCR1 as being necessary for IL-8 binding (19, 20). In particular, alanine mutations of several residues in the E3 loop cause substantial reductions in IL-8 affinity, although most of the mutants can support full receptor activation at high ligand concentrations. Despite the importance of the E3 loop in ligand binding, there is currently no information about the region of IL-8 that interacts with this receptor element.

We have previously developed a model system for chemokine receptors in which the N-terminal and E3 regions of a receptor are attached to a soluble scaffold protein in a manner that allows them to interact simultaneously with cognate chemokines for that receptor (25, 26). These “receptor mimics” are referred to as CROSS proteins, for Chemokine Receptor elements On a Soluble Scaffold. In the current study we describe the development of CROSS proteins corresponding to CXCR1 and CXCR2 and the application of these proteins towards characterizing the interactions of receptor N-terminal and E3 elements with IL-8. Initial results suggested that the E3 region of CXCR1 was interacting with IL-8 whereas the E3 region of CXCR2 was not. Further characterization of a chimeric receptor mimic indicated that the difference was attributable to the adjacent N-terminal element, indicating that the N-terminal and E3 elements of CXCR1 interact synergistically with IL-8.

EXPERIMENTAL PROCEDURES

Expression and Purification of Monomeric IL-8(1-66)

A plasmid containing the gene encoding a fusion protein between thioredoxin and IL-8(1-66) (27) was received as a kind gift from Dr. Krishna Rajarathnam (The University of Texas Medical Branch, Galveston, TX). The coding region for IL-8 (1-66) was subcloned between the HindIII and XhoI sites of E. coli expression vector pET11c (Novagen, Inc., Madison, WI). The resulting construct encodes an N-terminal His₆-tag, and a modified thrombin cleavage site (LVPR*SA), so that cleavage after the R*residue yields the native N-terminus (S¹A²…) of IL-8; the gene and protein sequences are given in Supporting Information Figure S1. Throughout this article, IL-8 refers to this IL-8(1-66) construct, which was used to avoid IL-8 dimerisation, a complicating factor in interpretation of binding data. BL21(DE3) cells transformed with this construct were grown in LB media (for unlabeled protein) or in M9 media supplemented with 1 g/L ¹⁵N ammonium chloride (for ¹⁵N-labeled protein) at 37 °C to an OD₆₀₀ of 0.6, induced with 1 mM IPTG, and grown for an additional 4 hours at 37 °C before harvesting. Cell pellets were washed in 20 mM Tris-HCl, pH 8.0 then resuspended in the same buffer (50 mL per L of culture). Phenylmethysulfonylfloride (PMSF) was added to a final concentration of 200 μM and the cells were lysed by two passes through a cell homogenizer. The lysate was centrifuged (30 min, 27,000g) and the pellet was suspended in 20 mM Tris-HCl, 0.75 M urea, 1% Triton X-100, pH 8.0 (50 mL per L of culture) and then centrifuged (30 min, 27,000g). The pellet was then solubilized in 6 M guanidine HCl, 100 mM NaH₂PO₄, 10 mM Tris Base, pH 8.0 (50 mL per L of culture) and clarified by centrifugation (30 min, 27,000g). The supernatant was incubated with 5-7 mL of Ni-NTA agarose resin (Qiagen, Valencia, CA), which was then poured into a bench-top column, washed successively with 6 M guanidine HCl, 100 mM NaH₂PO₄, 10 mM Tris Base, pH 8.0 and the same buffer at pH 6.3, then eluted with the same buffer at pH 4.9. Dithiothreitol was added to a final concentration of 50 mM and the sample was incubated for 3 hours at ambient temperature. The protein was then dialyzed against: 0.8 M guanidine HCl, 20 mM NaH₂PO₄, 0.5 mM oxidized glutathione, 2.5 mM reduced glutathione, pH 6.5, for 12 hours at 4 °C; 20 mM acetic acid for 8 hours at 4 °C; and finally 20 mM Tris-HCl, pH 8.5 for 12 hours at 4 °C. The dialysate was clarified by centrifugation (30 min, 27,000g, 4 °C). The N-terminal His₆-tag was removed by incubation with thrombin (≈5 U/mL, Sigma, St. Louis, MO) for 45 min at ambient temperature. The cleaved IL-8 protein was then loaded onto a HiTrap SP HP cation exchange column (Amersham Biosciences, Piscataway, New Jersey) pre-equilibrated with 20 mM Tris-HCl, 25 mM NaCl, pH 8.5 and eluted with a 0 to 2 M NaCl gradient. Purity and molecular weight were verified by SDS-PAGE and MALDI-TOF mass spectrometry.

Cloning of Expression Constructs for CROSS Proteins

The gene and protein sequences of CROSS expression constructs used in this study (summarized in Figure 1) are given in the Supporting Information Figure S2. Briefly, the N-terminal and E3-regions of the human CXCR1 and CXCR2 genes (containing the native Cys codons) were synthesized from overlapping oligonucleotides using recursive PCR, then subcloned into the gene encoding the CCR3 mimic CROSS⁵-N₃E3₃ (25, 26). Subsequently, the Cys codons were modified by Quikchange (Stratagene, Inc) site-directed mutagenesis. Each of the control constructs (see Figure 1) was obtained by one or more subcloning steps from the CROSS-N^X1E3^X1 and/or CROSS-N^X2E3^X2 constructs. The chimeric construct CROSS-N^X1E3^X2 was obtained from the CROSS-N^X1 E3^X1 construct by a single round of Quikchange mutagenesis to simultaneously introduce the Asn-282→His, and Gly284→Asp mutations.

Figure 1. Schematic diagrams of CROSS proteins and sequence alignments of N-terminal and E3 elements.

(A) Schematic diagram of CROSS-N^X1E3^X1 showing the B1 domain scaffold (black), the N-terminal and E3 loop regions of CXCR1 (red), and linkers (cyan). Arrows indicate the positions of Cys residues that form a disulfide bond in the native receptors but are mutated to Ser in the CROSS proteins used in this study. (B) Schematic representations and nomenclature of the CROSS proteins described herein, with N-terminal and E3 loop regions of CXCR1 (red) and CXCR2 (green), linkers (cyan) and B1 domain scaffold regions (black). (C) Alignment of the N-terminal sequences of CXCR1 and CXCR2 and the sequence of the CXCR1 peptidomimetic CXCR1-p1 (24). Numbering at the top corresponds to CXCR1 whereas numbering at the bottom is for CXCR1-p1. Features of CXCR1-p1 include: an N-terminal acetyl group (ac); a C-terminal amide (nh2); and a hexanoic acid linker (lin). (D) Alignment of the E3 loop sequences of CXCR1 and CXCR2; the two differences between these sequences are highlighted in violet. Numbering corresponds to the sequence of intact CXCR1. In panels C and D conserved amino acids are indicated in bold type. The Cys residues in blue boxes are mutated to Ser in the CROSS proteins used in this study.

Expression and Purification of CROSS Proteins

BL21(DE3) E. coli cells transformed with CROSS expression constructs were grown in LB media at 37 °C to an OD₆₀₀ of 0.7 to 0.8, induced with 1 mM IPTG, and grown for an additional ≈16 hours at 25 °C before harvesting. Cell pellets were resuspended in 50 mM NaH₂PO₄, 300 mM NaCl, 5 mM imidazole, pH 8.0 (20 mL per L of culture), PMSF was added to a final concentration of 500 μM and the cells were lysed by two passes through a cell homogenizer. The lysate was centrifuged (30 min, 27,000g) and the filtered supernatant was loaded onto a Ni-NTA agarose column pre-equilibrated with 50 mM NaH₂PO₄, 300 mM NaCl, 5 mM imidazole, pH 8.0, washed with the same buffer containing 20 mM imidazole, and eluted with the same buffer containing 200 mM imidazole. After dialysis against 20 mM Tris-HCl, pH 8.5 (2.5 h, ambient temperature), thrombin (≈5 U/mL, Sigma, St. Louis, MO) was added and the solution was incubated for 2.5 h at ambient temperature. The cleaved CROSS protein was then loaded onto a HiTrap Q HP anion exchange column (Amersham Biosciences, Piscataway, New Jersey) pre-equilibrated with 20 mM Tris-HCl, 25 mM NaCl, pH 8.5 and eluted with a 0 to 2 M NaCl gradient. Purity and molecular weight were verified by SDS-PAGE and MALDI-TOF mass spectrometry.

Isothermal Titration Calorimetry

Isothermal titration calorimetry (ITC) experiments were performed at 20 °C, using a VP-ITC microcalorimeter (MicroCal Inc, Northampton, MA). The CROSS and IL-8 samples were extensively dialyzed against 50 mM MOPS buffer, pH 7.0, filtered and degassed. Protein concentrations were determined from absorbance at 280 nm. Experiments were performed by titration of CROSS protein (0.2 to 0.6 mM in the injection syringe) into IL-8 (0.01 to 0.03 mM in the sample cell); the reference cell was filled with degassed distilled water. One injection of 3 μL followed by 36 injections of 8 μL were made, with 5 min equilibration periods between injections. The ITC data were analysed in Origin for ITC version 5.0 (MicroCal Inc., Northampton, MA). Data were corrected for heats of dilution, determined from the integrated peaks well beyond the inflection point of the curve, baseline corrected and integrated. The cumulative heat evolved (Q) was fit to the single-site binding equation:

$$Q=(n{P}_t\Delta H{V}_0∕2)\left\{1+L_t∕\left(n{P}_t\right)+1∕\left(n{K}_a{P}_t\right)-\sqrt{{[1+L_t∕(n{P}_t)+1∕(n{K}_a{P}_t\left)\right]}^2-4L_t∕\left(n{P}_t\right)}\right\}$$

in which: L_t and P_t are the total ligand (CROSS) and protein (IL-8) concentrations, respectively; n is the stoichiometry; K_a (=1/K_d) is the association equilibrium constant; ΔH is the binding enthalpy; and V₀ is the active initial cell volume. The initial 3 μL injection was excluded from the data fitting (28). Reported parameters are the averages from three or four replicate experiments.

NMR Spectroscopy

NMR spectra were acquired at 25 °C on a Varian 600 MHz NMR spectrometer equipped with a triple resonance Cold Probe. To monitor the binding of IL-8 to CROSS proteins, ¹⁵N HSQC spectra were recorded for samples of ¹⁵N-labeled IL-8 (initial concentration 130-200 μM in 50 mM acetate, 0.1% NaN₃, 10% D₂O, pH 5.5) containing increasing molar ratios (0, 0.2, 0.4, 0.6, 0.8, 1.0, 1.2, 1.4 and 1.6) of unlabeled CROSS. Spectra were referenced to external DSS at 0 ppm. Data were processed using NMRPipe (29) and analyzed using Sparky version 3.113 (Goddard TD, Kneller DG, University of California, San Francisco, CA). Weighted amide chemical shift changes (Δδ_wt) are defined as $\Delta {\delta }_{\mathit{wt}}=\sqrt{\Delta {\delta }_H^2+{(\Delta {\delta }_N∕5)}^2}$. The amide ¹H and ¹⁵N chemical shift assignments for IL-8(1-66) had been previously determined (27).

RESULTS

Development of Soluble Mimics of CXCR1 and CXCR2

In previous studies we have developed soluble proteins containing the N-terminal and E3 elements of CC-chemokine receptors CCR2 or CCR3 (25, 26). These “CROSS” proteins utilised a stabilised form of the B1 domain of Streptococcal protein G as a soluble scaffold. The N-terminal region of the receptor was incorporated as an N-terminal extension to the scaffold protein, whereas the receptor E3 element was inserted between two secondary structure elements (the β2-strand and α-helix) of the scaffold. The two receptor elements were connected by a native disulfide bond. In the current study we initially expressed CROSS proteins in which the N-terminal and E3 elements of CXC-chemokine receptors CXCR1 or CXCR2 were incorporated in an identical manner to those of CCR2 and CCR3 in the previous CROSS proteins. However, these proteins were expressed as heterogeneous, oligomeric mixtures and attempts to refold them to the desired monomeric form were unsuccessful (data not shown). To prevent oligomerization, we mutated the Cys residues in the stabilised scaffold back to their native B1 domain residues and the single Cys residue in each receptor element to Ser. Notably the latter mutations remove a native disulfide bond that plays an important role in maintaining native receptor structure (20). The resulting CROSS proteins were expressed as soluble monomers, could be readily purified to homogeneity and gave circular dichroism spectra consistent with their expected secondary structures (Supporting Information Figures S3 and S4). Throughout the remainder of this article, all proteins described incorporate the Cys to Ser mutations. These CROSS proteins and controls are represented schematically, along with their nomenclature, in Figures 1A and 1B. We refer to the CXCR1 and CXCR2 CROSS proteins as CROSS-N^X1E3^X1 and CROSS-N^X2E3^X2, respectively, in which the two superscripts indicate the receptors from which the N-terminal and E3 elements are derived. To identify the roles of each receptor element, we prepared four control proteins (CROSS-N^X1, CROSS-E3^X1, CROSS-N^X2 and CROSS-E3^X2), each containing only the N-terminal or only the E3 element of CXCR1 or CXCR2 (Figure 1B).

Binding of Receptor Mimics to IL-8

The affinity and thermodynamics of binding between monomeric IL-8 and each receptor mimic or control protein was monitored by ITC (Figure 2). Binding parameters are summarized in Table 1. The two control proteins containing only receptor N-terminal elements each bound exothermically to IL-8 with equilibrium dissociation constants (K_d) of approximately 10 μM (Figures 2A and 2B), whereas the control proteins containing only E3 elements showed little to no detectable binding (Figures 2C and 2D). The two CROSS proteins differed in their binding characteristics. CROSS-N^X1E3^X1 bound to IL-8 with a K_d of 0.8 ± 0.1 μM, approximately 10-fold more strongly than the CROSS-N^X1 control protein (Figure 2E). However, the affinity of CROSS-N^X2E3^X2 for IL-8 (K_d = 5.7 ± 1.7 μM) was not significantly different from the corresponding control (K_d = 7.1 ± 1.6 μM for CROSS-N^X2; Figure 2F). These results suggested that, in the context of these CROSS proteins, the E3 element of CXCR1 contributes to IL-8 binding but the E3 element of CXCR2 does not. The stoichiometry of binding estimated from ITC data was close to 1:1 for most CROSS proteins. However the data for CROSS-N^X2 indicated a stoichiometry of 0.4 ± 0.1, suggesting a possible error in the concentration of this protein; this would only have a small effect on the fitted K_d value.

Figure 2. Isothermal titration calorimetry data for binding of CROSS proteins and controls to IL-8).

Each panel shows the raw titration data (top) and the fitted binding isotherm (bottom) for a representative example from triplicate or quadruplicate experiments. Derived thermodynamic parameters are listed in Table 1.

Table 1.

Thermodynamic parameters for binding of CROSS-proteins to IL-8^a

Protein	Stoichiometry y (n)	ΔH (kcal/mol)	−TΔS (kcal/mol)	ΔG (kcal/mol)	Kd (μM)
CROSS-NX1	0.8 ± 0.1	−1.2 ± 0.3	−5.5 ± 0.4	−6.8 ± 0.1	9.6 ± 1.9
CROSS-NX2	0.4 ± 0.1	−1.5 ± 0.5	−5.5 ± 0.6	−6.9 ± 0.1	7.1 ± 1.6
CROSS-E3X1			NDb
CROSS-E3X2			NDb
CROSS- NX1E3X1	0.9 ± 0.1	−1.8 ± 0.1	−6.4 ± 0.2	−8.2 ± 0.1	0.8 ± 0.1
CROSS- NX2E3X2	0.8 ± 0.2	−1.0 ± 0.5	−6.1 ± 0.5	−7.1 ± 0.2	5.7 ± 1.7
CROSS- NX1E3X2	0.9 ± 0.1	−2.3 ± 0.3	−5.9 ± 0.3	−8.3 ± 0.1	0.7 ± 0.1

^aParameters were obtained by isothermal titration calorimetry at 20 °C. The values reported are from an average of three or four experiments.

^bBinding not detected. A very weak exothermic signal was observed.

NMR Mapping of IL-8 Residues Influenced by CROSS Binding

NMR spectroscopy was used to investigate the regions of IL-8 involved in binding to each receptor mimic and to the control proteins containing only N-terminal elements. Each protein induced linear changes in specific signals in the ¹⁵N-HSQC spectrum of IL-8 up to a molar ratio of 1:1, at which point no further changes were observed. These observations are consistent with 1:1 binding stoichiometry and affinity in the low micromolar range; K_d values in the 1-10 μM range are not readily distinguishable due to the relatively high concentrations (130-200 μM) of IL-8 used in these experiments. Figure 3 shows a detailed region of the HSQC spectrum for IL-8 alone (red in each panel) and upon binding to each CROSS or control protein. The Supporting Information includes the full spectrum of IL-8 alone and bound to CROSS-N^X1E3^X1 (Figure S5) and both a list (Table S1) and bar graph (Figure S6) of the observed chemical shift changes for each CROSS or control protein. Figures 4 and S7 show scatter plots comparing the shifts observed for each pair of CROSS or control proteins; correlation coefficients and slopes are listed in Table 2.

Figure 3. 2D NMR data for binding of CROSS proteins and controls to IL-8.

Each panel shows the same detailed region of the ¹⁵N-HSQC spectrum of IL-8 alone (red), in the presence of 0.6 molar equivalents of the indicated CROSS or control protein (green) and saturated with 1.6 molar equivalents of the indicated CROSS or control protein (blue). The peak assignments and directions of peak shifts are indicated in panel (A); in all other panels the peaks shift in the same direction, although the shifts of S44 have a larger magnitude for CROSS-N^X1E3^X1 and CROSS-N^X1E3^X2.

Figure 4. Correlations between chemical shift changes for IL-8 binding to CROSS proteins.

Each panel shows the ¹H chemical shift change (in ppm) for IL-8 binding to one CROSS protein or control protein (x-axis) plotted against the ¹H chemical shift change for another CROSS protein or control protein (y-axis). Each point represents a different residue of IL-8. Data are shown for all residues excluding L43, C50 and L51, which show intermediate exchange behaviour in spectra of IL-8 bound to CROSS-N^X1E3^X2. Solid lines show linear regression fits to the data. Similar graphs for ¹⁵N chemical shift changes are presented in Supporting Information Figure S7. Correlation coefficients and slopes for the fits are listed in Table 2.

Table 2.

Correlations between chemical shift changes for binding of CROSS-proteins to IL-8

Protein	Δδ(1H) r2	Δδ(1H) slope	Δδ(15N) r2	Δδ(15N) slope
CROSS-NX1 versus CROSS-NX2	0.97	0.74	0.97	0.73
CROSS-NX2 versus CROSS-NX2E3X2	0.95	1.05	0.99	1.09
CROSS-NX1 versus CROSS-NX1E3X1	0.40	1.13	0.58	1.33
CROSS-NX2E3X2 versus CROSS- NX1E3X1	0.38	1.39	0.58	1.26
CROSS-NX1E3X1 versus CROSS- NX1E3X2	1.00	1.08	0.99	1.12
CROSS-NX2E3X2 versus CROSS- NX1E3X2	0.59	1.69	0.83	2.29

CROSS-N^X1 and CROSS-N^X2 each induced substantial changes in numerous NH resonances of IL-8. Moreover, there are strong correlations (r² = 0.97, slope ≈ 0.7) between both the ¹H_N and ¹⁵N_H shifts induced by these two proteins (Figure 4A and Table 2), suggesting that they interact with the same region of IL-8. The structure of IL-8 is color-coded according to the weighted NH shifts (Δδ_wt) in Figure 5A. Substantial shifts are observed throughout the secondary structure elements of the chemokine.

Figure 5. Chemical shift mapping of IL-8 binding to CROSS proteins.

The structure of IL-8 bound to the peptidomimetic CXCR1-p1 is color-coded according to the weighted chemical shift changes (Δδ_wt) observed for binding to (A) CROSS-N^X1 and (B) CROSS-N^X1E3^X1 and (C) the weighted difference in chemical shifts between IL-8 bound to CROSS-N^X1 and CROSS-N^X1E3^X1. The continuous color scale is shown in the color bar; residues with Δδ_wt> 0.1 are red and residues for which data were not obtained are gray. The structurally well-defined region of CXCR1-p1 is shown as cyan sticks. The termini of CXCR1-p1 and several structural elements of IL-8 are labelled in panel B (NL=N-loop; 3₁₀= 3₁₀-turn). The cyan arrow in panel A indicates the direction in which CXCR1 residues following Pro-17 may extend to interact with the β1 and β2 strands. This figure was generated using PyMol. (D) Bar graph showing the weighted differences in chemical shifts between (red) free IL-8 and IL-8 bound to CROSS-N^X1, (yellow) free IL-8 and IL-8 bound to CROSS-N^X1E3^X1, and (cyan) IL-8 bound to CROSS-N^X1 and CROSS-N^X1E3^X1.

The soluble mimic of CXCR2, CROSS-N^X2E3^X2, also induced changes in the HSQC of IL-8 but these changes were extremely similar to those observed for the CROSS-N^X2 control protein (and also for CROSS-N^X1) as illustrated by the scatter plots in Figure 4B (r² ≥ 0.95, slope ≈ 1.1; Table 2). These results indicate that the E3 region of CXCR2 does not interact with IL-8 and does not alter the interaction of the N-terminal element with IL-8, which is consistent with the affinity measurements presented above.

In contrast to the results for CROSS-N^X2E3^X2, the chemical shift changes induced by the soluble mimic of CXCR1 (CROSS-N^X1E3^X1) differed substantially from those observed for the corresponding control protein (CROSS-N^X1) and from those observed for CROSS-N^X2E3^X2. This is evident from the scatter plots in Figures 4C and 4D (r² ≈ 0.4 to 0.6; Table 2). These comparisons indicate that the E3 region of CXCR1 either interacts with IL-8 or alters the interaction of the N-terminal element with IL-8, both of which would be consistent with the higher IL-8 affinity of CROSS-N^X1E3^X1 compared the other proteins (vide supra). Notably, the structural color map for binding of CROSS-N^X1E3^X1 to IL-8 (Figure 5B) and the difference map (Figure 5C) indicate that the same regions of IL-8 are influenced by binding to both CROSS-N^X1 and CROSS-N^X1E3^X1 (although the shifts at each specific position were different). Thus, the data did not clearly indicate a specific binding site on IL-8 for the E3 element of CXCR1.

Analysis of a Chimeric Receptor Mimic

The E3 regions of CXCR1 and CXCR2 only differ by 2 amino acids (Figure 1D). Therefore, it was surprising that the addition of the E3 element enhanced IL-8 binding for the CXCR1 mimic but not for the CXCR2 mimic. It was possible that the two variable amino residues were responsible for this difference. However, an alternative hypothesis was that the E3 region of CXCR2 behaved differently from that of CXCR1 only because of its position adjacent to the N-terminal region of CXCR2 rather than to the N-terminal region of CXCR1. To distinguish between these two possibilities, we prepared and characterized a chimeric receptor mimic (CROSS-N^X1E3^X2) containing the N-terminal region of CXCR1 and the E3 region of CXCR2 (Figure 1B).

The ITC data in Figure 2G and the thermodynamic parameters in Table 1 indicate that CROSS-N^X1E3^X2 binds to IL-8 with affinity, enthalpy and stoichiometry indistinguishable from those of CROSS-N^X1E3^X1. Moreover, CROSS-N^X1E3^X2 induces changes in the IL-8 HSQC spectrum (Figure 3E) that are highly correlated with those of CROSS-N^X1E3^X1 (r² ≈ 0.99, slope ≈ 1.1; Figure 4E and Table 2) but relatively poorly correlated with those of CROSS-N^X2E3^X2 (r² ≈ 0.6 to 0.8, slope ≈ 1.7 to 2.3; Figure 4F and Table 2). Again, the NMR shift comparisons are consistent with the variations of binding affinity observed by ITC. These data strongly suggest that CROSS-N^X1E3^X2 has the same IL-8 binding interactions as CROSS-N^X1E3^X1.

DISCUSSION

Interactions of Receptor N-terminal Elements with IL-8

In this study we have used a soluble model system to investigate the interactions of monomeric IL-8 with the N-terminal and E3 elements of both IL-8 receptors CXCR1 and CXCR2. We have shown that the N-terminal elements of both receptors, when attached to the N-terminus of the stabilized B1 domain, bind to IL-8 with affinities of 9.6 ± 1.9 and 7.1 ± 1.6 μM, respectively, and that binding of both peptides is characterized by a small favorable enthalpy change and a larger favorable entropy change. The affinities of these interactions are very similar to that previously reported for the binding of monomeric IL-8 to a peptide derived from the N-terminus of rabbit CXCR1 (K_d = 8.6 μM) (27). A similar affinity was also found for binding between the rabbit CXCR1 peptide and wild type IL-8, although in the latter case peptide binding was coupled to dissociation of the IL-8 dimer (22). Rabbit CXCR1 peptide was used in the previous studies due to practical difficulties working with peptides derived from the N-terminal region of human CXCR1 (27). The use of the B1 domain scaffold in the current study provides a way to overcome this impediment.

The NMR studies reported herein provide information about the regions of IL-8 that are influenced by interaction with the receptor N-terminal elements. Changes observed in amide chemical shifts could potentially result from direct interactions of the affected residues with the receptor elements or from indirect structural changes induced by these interactions. The amide chemical shift changes observed upon binding of CROSS-N^X1 are mapped onto the structure of IL-8 in Figure 5A. There are pronounced chemical shift changes in the N-loop (Cys-9, Ile-10, Tyr-13, Ser-14 and Phe-17), the 3₁₀-turn (Lys-20 and Phe-21) and the β3 strand (Glu-48, Leu-49 and Lys-51) of IL-8. These residues form a contiguous surface that is similar to the IL-8 binding surface for the N-terminal peptidomimetic (designated CXCR1-p1) observed previously by Skelton et al. (24). In addition, we observed substantial chemical shift changes in the N-terminal end of the α-helix (Asn-56, Trp-57, Gln-59, Arg-60, and Val-61) and for some residues in the β2-β3 turn (Leu-43 and Ser-44) and the β1 and β2 strands (Glu-24, Leu-25 and Ile-39) of IL-8. Changes in the α-helix are not surprising as the affected region interacts directly with the N-loop. Similarly, changes in the β2-β3 turn could be indirect effects related to the interactions of nearby residues in the β3 strand. The causes of the changes in the β1 and β2 strands are less obvious but may be an indication that residues of CXCR1 following Pro-17 continue to wrap around under the β-sheet as indicated by the cyan arrow in Figure 5A.

The interaction surface on IL-8 for the N-terminal element of CXCR2 has not been described previously. The correlation shown in Figure 4A indicates that binding of CROSS-N^X2 induces almost identical amide chemical shift changes of IL-8 to those observed upon binding of CROSS-N^X1. Thus, the N-terminal element of CXCR2 interacts with the same surface of IL-8 as that discussed above for the N-terminus of CXCR1. The observation that the two receptor N-terminal elements bind to the same region of IL-8 and that they do so with very similar thermodynamics is surprising considering that there are numerous differences between the amino acid sequences of these two elements. In the alignment shown in Figure 1C (which includes numerous gaps), there are only 16 amino acid identities among the 40 residues of CXCR1 (or 49 residues of CXCR2). Nevertheless, it is noteworthy that the CXCR1-p1 peptidomimetic that binds to IL-8 incorporates six of these identical residues and four of them (Pro-9, Pro-10, Asp-14, and Pro-17) lie within the residue 8-17 region that is well-defined in the structural ensemble (Figure 1C) (24). Among these, the three proline residues form numerous hydrophobic interactions with the binding groove and Asp-14 is in close proximity to IL-8 residue Lys-11. In summary, our data suggest that the N-terminal element of CXCR2 uses the few conserved amino acids to interact with IL-8 in a manner very similar to that observed for the CXCR1 peptidomimetic.

Possible Interactions of Receptor E3 Loops with IL-8

The CROSS proteins described herein provide an opportunity to explore the interactions of the E3 loop regions of CXCR1 and CXCR2 with IL-8. In the absence of the receptor N-terminal elements (i.e. for CROSS-E3^X1 and CROSS-E3^X2), we did not observe any interaction of receptor E3 elements with IL-8. Similarly, the protein containing both N-terminal and E3 elements of CXCR2 (CROSS-N^X2E3^X2) bound to IL-8 with identical affinity to CROSS-N^X2, which contains only the N-terminal element, suggesting that the E3 element of CXCR2 was not interacting with IL-8. This conclusion is supported by the excellent correlation between IL-8 chemical shift changes induced by CROSS-N^X2 and CROSS-N^X2E3^X2 (Figure 4B). In contrast, the protein containing both N-terminal and E3 elements of CXCR1 (CROSS-N^X1E3^X1) displayed a 12-fold higher affinity than CROSS-N^X1 for IL-8 and there was a much lower correlation between chemical shift changes observed for these two proteins (Figure 4C). Initially, this result suggested that the E3 element of CXCR1 was interacting with IL-8. In an effort to identify a region of IL-8 that interacts with the E3 element of CXCR1, we have color-coded the IL-8 structure with the chemical shift changes observed for binding of CROSS-N^X1 (Figure 5A) and CROSS-N^X1E3^X1 (Figure 5B). Although the binding of CROSS-N^X1E3^X1 induces larger changes in amide chemical shift (more red in Figure 5B than Figure 5A), there is no region of IL-8 that is affected by binding to CROSS-N^X1E3^X1 and not affected by binding to CROSS-N^X1. Further illustrating this point, Figures 5C and 5Dshows the chemical shift differences between IL-8 bound to CROSS-N^X1 and to CROSS-N^X1E3^X1. The regions of IL-8 showing the greatest differences are the same as those displaying the greatest changes upon binding of CROSS-N^X1. Thus, the current data do not allow us to identify a specific binding surface on IL-8 for the E3 loop of CXCR1.

Synergistic Interactions of N-terminal and E3 Elements

The E3 elements of CXCR1 and CXCR2 have highly similar sequences (Figure 1D), with only two differences in 27 residues. It was therefore surprising that the E3 element of CXCR1 enhanced IL-8 binding affinity whereas the E3 element of CXCR2 did not. The simplest possible explanation of these observations is that the E3 element of CXCR1 interacts directly with IL-8 but that the two variant residues play critical roles so the E3 element of CXCR2 cannot interact. To test this hypothesis, we constructed the chimeric protein (CROSS-N^X1E3^X2) containing the N-terminal element of CXCR1 and the E3 element of CXCR2. Both the affinity data and the chemical shift perturbation data for this variant were identical to those for CROSS-N^X1E3^X1. Thus, the ability of the E3 element to enhance IL-8 binding affinity is not reliant on the identities of the two variant residues but is dependent on the presence of the CXCR1 N-terminal element. This is a clear indication of synergistic interactions between the N-terminal and E3 elements of the CXCR1 receptor.

There are two general structural models that could account for the observation of synergistic interactions. One possibility is that the E3 element (of CXCR1 or CXCR2) interacts directly with IL-8 but that this interaction can only occur (at least in this model system) in the presence of the N-terminal element of CXCR1, not in the presence of the N-terminal element of CXCR2. An alternative explanation is that the E3 element does not interact directly with IL-8 but instead it is able to support additional or stronger interactions with IL-8 by the N-terminal element of CXCR1 (but not by the N-terminal element of CXCR2). It is not possible based on our data to differentiate unequivocally between these two models. For example, both models would be consistent with the observation that the enhanced affinity upon addition of the E3 element has both enthalpic and entropic components (ΔΔH = −0.6 kcal/mol and −TΔΔS = −0.9 kcal/mol). However, as noted above, there is no evidence from the chemical shift mapping to indicate a specific binding surface for the E3 loop. Therefore, the current data provide some implicit support for the latter model. Structural analysis of CROSS-N^X1E3^X1 bound to IL-8 could potentially reveal the structural details of the interactions. However, unfortunately we have not been able to identify appropriate conditions for either crystallization or NMR structural analysis of CROSS-N^X1E3^X1 and its complex with IL-8.

The present results have possible implications for the widely-accepted two step model for chemokine-receptor binding and activation (5). According to the two step model, in the second stage of receptor interaction the N-terminal region of IL-8 interacts with the extracellular loops and/or the transmembrane helices of the receptor. Since we could not identify a specific binding surface for the E3 loop on IL-8 and the E3 region induces little if any chemical shift perturbation in the N-terminal region of the chemokine, including the functionally critical ELR sequence, our data suggest that the E3 region alone is not sufficient to form the second binding site in the two step model. Thus, the receptor site involved in the second step of chemokine binding is likely to include regions of the receptor that are not part of the CROSS proteins described here. This conclusion is consistent with the previous observation that mutations within extracellular loop 2 of CXCR1 reduce binding affinity and eliminate activation by IL-8 (20).

Irrespective of the specific structural model, the current observation of synergistic interactions suggests that the N-terminal and E3 elements are in close proximity to each other in the IL-8 complex of CROSS-N^X1E3^X1, suggesting that they are also adjacent to each other in the intact receptor. This is consistent with expectations based on previous information. In one study of CXCR1, mutation of E3 residue Asp-265 to Ala resulted in reduced staining with an antibody raised against the receptor N-terminal element, suggesting that the E3 loop is adjacent to the N-terminus (19). Based on chemokine receptor sequence comparisons and the CXCR4 structure (4), the N-terminal and E3 element are expected to be adjacent to each other and linked by a conserved disulfide bond; mutation of the participating cysteine residues renders CXCR1 almost completely inactive (20). It is interesting that our data suggest interactions between these two receptor elements, even in the absence of this disulfide bond, which was excluded form the CROSS proteins described herein to allow isolation of homogeneous proteins. Thus, previous structural and functional data are consistent with a model in which N-terminal and E3 residues interact with each other and/or with the same regions of the chemokine ligand.

It is curious that synergistic interactions between the N-terminal and E3 elements occur in the CXCR1 model protein but not in the CXCR2 model protein. However, this difference may reflect true differences between the structures and interactions of the intact receptors, which have previously been suggested by studies of receptor internalization rates. In one study, equal doses of IL-8 caused CXCR1 to be 70% internalized in 30 minutes whereas CXCR2 was 95% internalized within 5 minutes (10). Studies of chimeric and mutant receptors clearly demonstrated that the differences in the rate and extent of internalization could be attributed to differences between the receptor extracellular elements (10, 30) suggesting that IL-8 interacts differently with the extracellular regions of CXCR1 and CXCR2, thereby inducing distinct conformational changes and different rates of internalization. The existence of different activated conformations of the receptors is supported by the observation that saturating doses of different CXCR2 ligands (or mutants of natural ligands) also stimulate receptor internalization at different rates (30).

Concluding Remarks

The structural basis of G protein-coupled receptor activation remains incompletely understood. However, the emerging picture from recent structural and pharmacological data is that these receptors have remarkable structural plasticity allowing different ligands to induce different activated states and thereby to stimulate different signalling pathways. The studies of CXCR1 and CXCR2 internalisation rates discussed above have previously suggested that these receptors can be induced to access different structural states depending on the chemokine agonist used and that the same chemokine (IL-8) can induce different structural states of the two receptors. The current study shows that IL-8 forms very similar interactions with the isolated N-terminal region of each receptor. However, the adjacent E3 region of the receptor enhances these interactions for the N-terminal region of CXCR1 but not for the corresponding region of CXCR2. These new observations are consistent with the inference from previous studies that IL-8 binds to CXCR1 and CXCR2 differently and induces different conformational changes of these two receptors. In summary, the data reported herein enhance our understanding of the structural interactions underlying differential CXCR1 and CXCR2 activation by IL-8.