A trapped CXCL7 dimer was designed by substituting a cysteine for Glu23 of the first β-strand that constitutes the two-fold symmetry axis. The protein was cloned, expressed, and purified as described previously (21). The gene corresponding to E23C CXCL7 was ligated into the pET 32Xa vector and expressed as a thioredoxin fusion protein with a His-tag. Proteins were expressed in the E. coli BL21 (DE3) strain either in ¹⁵N- or ¹⁵N/¹³C-enriched minimal medium. Transformed cells were grown to an A₆₀₀ of 0.6, induced with 0.2 mM isopropyl β-d-thiogalactopyranoside, and grown for 16 h at 25°C. The protein-containing supernatant was loaded on to a Ni-NTA column and eluted with the same buffer as above, except containing 250 mM imidazole. Fractions containing the fusion protein were pooled and dialyzed against the cleavage buffer (20 mM Tris, 50 mM NaCl, 2 mM CaCl₂, pH 7.4). The fusion tag was cleaved with Factor Xa, and the proteins were purified on reversed-phase high-pressure liquid chromatography (HPLC) using a gradient of acetonitrile in 0.1% heptafluorobutyric acid. The purity and molecular weight of the proteins were confirmed using analytical HPLC and matrix assisted laser desorption/ionization mass spectrometry respectively.

The recombinant CXCR2 N-terminal domain (N-domain) (residues 1–43) peptide was expressed using the same protocol as described for the trapped dimer. Heparin dp26 (degree of polymerization 26 and corresponds to a 26mer) oligosaccharide was purchased from Iduron (UK). According to the manufacturer, the oligosaccharides were purified using high resolution gel filtration chromatography, the main disaccharide unit is IdoA, 2S-GlcNS, 6S (~75%), show some variation in sulfation pattern, contain uronic acid at the non-reducing end, and a C4–C5 double bond as a result of the heparinase endolytic action.

Nuclear magnetic resonance samples were prepared in a 50-mM phosphate buffer, pH 6.0, containing 1 mM 2,2-dimethyl-2-silapentansesulfonic acid (DSS), 1 mM sodium azide, and 10% D₂O. NMR spectra were acquired on a Bruker Avance III 600 (with a QCI cryoprobe) or 800 MHz (with a TXI cryoprobe) spectrometer and processed and analyzed using either Bruker Topspin 3.2 or Sparky programs (22). Chemical shift assignments of the trapped dimer were determined at 30°C using a 400-µM sample. The ¹H and ¹⁵N chemical shifts were assigned using 3D ¹⁵N-edited NOESY and TOCSY experiments with mixing times of 150 and 80 ms, respectively. The carbon chemical shifts assignments were obtained from HNCA and CBCACONH experiments at pH 6.0.

Binding interactions of CXCR2 N-terminal domain and heparin dp26 were characterized using solution NMR spectroscopy. A series of ¹H-¹⁵N HSQC spectra were collected until there were no chemical shift changes. In the case of CXCR2 N-domain, we titrated 320 µM CXCR2 N-domain to 77 µM WT CXCL7 in 50 mM phosphate buffer at pH 6.0 and 35°C. The final molar ratio of CXCL7:CXCR2 N-domain was 1:3.5. For GAG interactions, 2.5 mM heparin dp26 was titrated into a ~100 µM sample of either WT or trapped CXCL7 in 50 mM phosphate pH 6.0 at 35°C. The final molar ratio for CXCL7:dp26 was 1:4. For all titrations, chemical shift perturbations (CSPs) were calculated as a weighted average of changes in the ¹H and ¹⁵N chemical shifts as described previously (23).

Molecular docking of heparin to the CXCL7 dimer was carried out using the HADDOCK approach as described previously (24–27). The CXCL7 dimer structure consisted of chains A and B from the tetramer structure (PDB ID:1NAP) (28) and the NMR structure of heparin 14-mer (PDB ID:1HPN) (29) were used for docking. The AB-type dimer was selected based on analysis of NMR chemical shifts that revealed a CXC-type dimer. Residues that showed NMR CSP above a threshold were included as Ambiguous Interaction Restraints. The pair-wise “ligand interface RMSD matrix” over all structures was calculated and the final structures were clustered using an RMSD cutoff value of either 4 or 7 Å for one or two GAGs, respectively. The clusters were then prioritized using RMSD and “HADDOCK score” (weighted sum of a combination of energy terms).

The CXCR2 activity was determined using a Ca²⁺ release assay as described previously (30). Stably expressing CXCR2-HL60 (CXCR2-HL60) cells were cultured in RPMI 1640 supplemented with antibiotics (Pen–Strep, Gibco) and 10% FBS (Sigma). Differentiation into the neutrophil lineage was carried out by subculturing CXCR2-HL60 cells every other day by using antibiotic-free media containing 1.25% DMSO for 6 days (31). Ca²⁺ levels were measured using a FlexStation III microplate reader using the Calcium 6 assay kit (FLIPR, Molecular Devices). On day 6, CXCR2-HL60 cells were plated at a concentration of 2 × 10⁵cells/well in a 96-well black plate (Costar). The cells were incubated with varying concentrations of CXCL7 WT or trapped dimer, and changes in fluorescence were measured every 5 s for up to 240 s at room temperature. Ca²⁺ release activity of heparin-bound CXCL7 was determined in a similar manner. 2 nM CXCL7 was mixed with different concentrations of heparin (Iduron, UK) and immediately added to dye-loaded cells. Heparin by itself did not induce any Ca²⁺ release. Statistical significance was determined using one-way ANOVA followed by Tukey’s post hoc analysis; *p < 0.05.

Both HS and heparin share a repeating disaccharide unit composed of N-acetyl glucosamine and glucuronic acid. However, heparan sulfate (HS) has a modular structure with sulfated sequences (defined as NS domain) separated by non-sulfated regions containing acetylated sequences (defined as NA domain). Heparin is preferred for structural studies for the reason it is more uniformly sulfated, commercially available, and has been shown to capture endogenous interactions (32, 33). Various cellular, ex vivo, and biophysical studies for a wide variety of chemokines have shown that the dimer and higher order oligomers, compared to the monomer, bind GAGs such as heparan sulfate with higher affinity (24, 25, 34–38). In particular, recent NMR studies of GAG heparin binding to WT CXCL1, CXCL5, and CXCL8 have shown that the dimer is the high-affinity GAG ligand and that the differences in affinities are more pronounced for longer GAGs. We had previously observed that the affinity of the heparin octasaccharide for the monomer and dimer were essentially similar (20). With this in mind, we characterized the binding of heparin 26mer (dp26) to a 120-µM sample of WT CXCL7 at pH 6.0. Under these conditions, 70% of the protein exists as the monomer and the remaining as the dimer (20). In principle, four different species will exist in solution due to coupled equilibria—both monomer and dimer in the free and dp26-bound forms. The intensities of peaks are a direct reflection of the relative populations and binding constants. During the course of the titration, the peaks corresponding to the monomer disappear and the weak peaks corresponding to the dimer gain intensity, indicating that the dimer binds dp26 with much higher affinity (Figure 1A). In addition to peak intensity changes, CSP profiles for the dimer compared to the monomer also indicates that the dimer is the high-affinity ligand (Figures 1B,C).

Knowledge of the CXCL7 dimer chemical shifts is essential to describe the molecular basis of receptor and GAG interactions. However, unlike for the CXCL7 monomer, dimer does not dominate at any solution condition making its chemical shift assignments challenging. To overcome this challenge, we used a multi-pronged strategy that included chemical shift assignments of a disulfide-linked trapped dimer, heparin binding-induced chemical shift changes in the trapped dimer, and heparin binding-induced intensity and chemical shift changes in the WT monomer and dimer. These collectively allowed assignments of 80% of the residues including all of the residues that showed perturbation on heparin binding.

A trapped dimer was designed by introducing a disulfide bond across the twofold symmetry axis of the dimer interface (39), which involved mutating the dimer interface residue E23 to cysteine. The formation of the disulfide-trapped dimer was confirmed using SDS-PAGE, mass spectrometry, and NMR spectroscopy. NMR Cβ chemical shift of the newly introduced cysteine (45.3 ppm) indicated it is in the disulfide bonded state (40) (Figure S1 in Supplementary Material). Backbone NH chemical shifts of the trapped dimer were assigned using a combination of ¹⁵N-edited NOESY, ¹⁵N-edited TOCSY, HNCA, and CBCACONH experiments.

The trapped dimer chemical shifts allowed assigning 35 native dimer residues in a straightforward manner as these residues had essentially identical nitrogen and amide chemical shifts. Assignments for another 13 residues were obtained by comparing heparin binding-induced NMR spectral changes in the WT and trapped dimer. In the WT titrations, weak dimer peaks become strong and monomer peaks become weak, which, combined with differential chemical shift changes from the monomer, allowed for unambiguous tracking of the WT dimer chemical shifts and comparison to known trapped dimer chemical shifts. Finally, we assigned six residues on the basis that the intensity of these peaks did not change during the WT titration, indicating the dimer and monomer chemical shifts of these residues are identical. These residues were, therefore, assigned on the basis of known monomer chemical shifts. We could not unambiguously assign the remaining 13 residues, but lack of this knowledge was not limiting as these residues showed minimal to no perturbation in both receptor N-domain and GAG-binding experiments.

Chemical shift perturbation profile of dp26 binding to the WT dimer revealed a contiguous surface consisting of residues from the N-loop, β₃-strand, and α-helix. Most importantly, large chemical shift changes were observed for basic residues H15, K17, R44, R54, and K57 (Figure 2A). In the case of trapped dimer, the same residues not only showed similar perturbations, but the magnitude and direction of the perturbations were also essentially similar (Figures 2B,C).

To gain insight into the binding geometry, we performed four independent HADDOCK-based calculations to ensure that both a 1:1 and 1:2 stoichiometry were covered and to avoid any inherent bias in the docking process. In run I, restraints were given between one GAG and to only one monomer of the dimer. In run II, restraints were given between one GAG and both monomers of the dimer. In run III, restraints were given from each of two GAGs to either monomer of the dimer. Finally, in run IV, restraints were given from each of two GAGs to both monomers of the dimer.

Run I essentially resulted in a single geometry that could be divided into two major clusters. In one cluster, all residues implicated from NMR studies were involved in binding, whereas the second cluster was missing interactions from R44. Comparison of the two clusters revealed that H15, K17, R54, K57, and K61 function as a core domain and R44 functions as a peripheral residue, suggesting that interactions with R44 are more transient. Runs III and IV resulted in a single geometry with one GAG binding each monomer of the dimer with geometries observed for run I. We define this binding geometry as Model-I (Figure 3). Interestingly, run II resulted in a different geometry (defined as Model-II) in which a single GAG spans both monomers of the dimer, and in this geometry, all residues except R44 from both monomers mediate binding.

Though our studies indicate that the dimer preferentially exists in the GAG-bound form, knowledge of its receptor activity and the structural basis for these interactions is necessary to fully understand the role of the dimer in the context of in vivo function. We characterized receptor activity of the CXCL7 dimer by measuring Ca²⁺ release using HL60 cells stably transfected with the CXCR2 receptor (30). The trapped dimer was as potent as the WT indicating that the activities of the monomer and dimer are similar (Figure 4). Previous studies using a trapped dimer for related CXCR2 agonists CXCL1 and CXCL8 have also shown that the dimer could be as active as the monomer (30, 41).

It is now well established that receptor activation involves binding two distinct receptor sites. The initial binding involves the receptor N-terminal domain (defined as Site-I) and subsequent binding to a site involving extracellular loops and transmembrane domain (defined as Site-II). The structural basis for Site-I interactions can be characterized by using N-domain peptides (42–44). Site-I interactions of the CXCL7 monomer binding to the CXCR2 N-domain peptide have been characterized using solution NMR spectroscopy (20). We use the same divide and conquer approach for characterizing dimer interactions.

Under the experimental conditions, CXCL7 exists as ~70% monomer and ~30% dimer. Armed with previously assigned monomer and dimer chemical shifts from our current study, we could simultaneously track binding-induced CSP for both dimer and monomer residues (20). We observe large CSP for residues M6, C7, T10, T11, G13, I14, K17, N18, I46, C47, D49, and R54 (Figures 5A,C). These residues constitute a hydrophobic pocket flanked by positively charged side chains involving the N-loop and β₃-strand residues (Figure 5B). The extent of CSP and relative peak intensities for dimer and monomer were similar, indicating Site-I interactions of monomer and dimer are essentially the same.

A central question in determining the in vivo role of dimer is how GAG binding relates to receptor activation. We observe that heparin-bound CXCL7 is impaired for CXCR2 activity (Figure 6A). NMR studies also indicated that the heparin-bound CXCL7 is unable to bind the CXCR2 N-domain peptide (Figure 6B). Independent of any binding models, we observe that there is considerable overlap between the heparin and CXCR2 binding domains, providing a structural basis as to why heparin-bound CXCL7 is unable to bind the receptor (Figure 6C).