Sulfopeptide probes of the CXCR4/CXCL12 interface reveal oligomer-specific contacts and chemokine allostery

Abstract

Tyrosine sulfation is a post-translational modification that enhances protein-protein interactions and may identify druggable sites in the extracellular space. The G protein-coupled receptor CXCR4 is a prototypical example with three potential sulfation sites at positions 7, 12 and 21. Each receptor sulfotyrosine participates in specific contacts with its chemokine ligand in the structure of a soluble, dimeric CXCL12:CXCR4(1–38) complex, but their relative importance for CXCR4 binding and activation by the monomeric chemokine remains undefined. NMR titrations with short sulfopeptides showed that the tyrosine motifs of CXCR4 varied widely in their contributions to CXCL12 binding affinity and site specificity. Whereas the Tyr21 sulfopeptide bound the same site as in previously solved structures, the Tyr7 and Tyr12 sulfopeptides interacted nonspecifically. Surprisingly, the unsulfated Tyr7 peptide occupied a hydrophobic site on the CXCL12 monomer that is inaccessible in the CXCL12 dimer. Functional analysis of CXCR4 mutants validated the relative importance of individual CXCR4 sulfotyrosine modifications (Tyr21 > Tyr12 > Tyr7) for CXCL12 binding and receptor activation. Biophysical measurements also revealed a cooperative relationship between sulfopeptide binding at the Tyr21 site and CXCL12 dimerization, the first example of allosteric behavior in a chemokine. Future ligands that occupy the sTyr21 recognition site may act as both competitive inhibitors of receptor binding and allosteric modulators of chemokine function. Together, our data suggests that sulfation does not ubiquitously enhance complex affinity and that distinct patterns of tyrosine sulfation could encode oligomer selectivity – implying another layer of regulation for chemokine signaling.

INTRODUCTION

Chemokines are small soluble proteins that stimulate chemotactic cell migration via activation of a G protein-coupled receptor (GPCR). In addition to their vital roles in inflammation, wound healing, and stem cell homing, chemokines also contribute to many pathologies including autoimmune diseases and cancer. Interactions of the chemokine CXCL12 (stromal cell-derived factor-1/SDF-1) and its receptor CXCR4 are particularly well studied because of their participation in neurogenesis (1, 2), cardiogenesis (3), angiogenesis (4), myocardial infarction/reperfusion injury (5–7), HIV infection (8), and numerous carcinomas and sarcomas (as reviewed in (9)).

Chemokine receptor recognition and activation occurs via a two-step, two-site process (10, 11). First, the CXCR4 extracellular N-terminus binds to CXCL12 (site 1). The N-terminus of CXCL12 then recognizes the receptor transmembrane domain and activates signaling (site 2). Farzan and colleagues were the first to investigate the effect of post-translational modifications in the site 1 interaction (12). In addition to one site of O-linked glycosylation, CXCR4 possesses three tyrosine residues (Tyr7, Tyr12, and Tyr21) in the N-terminus capable of being O-sulfated in the Golgi apparatus. Mutational studies suggested that sulfation of Tyr21 enhances the binding of wild type CXCL12 (CXCL12_WT), but the level of sulfation at each tyrosine and their relative contributions to chemokine recognition were not quantified. Consistent with studies of the full-length CXCR4 receptor expressed in cells, sequential sulfation of a peptide comprising residues 1–38 (CXCR4_1–38) enhanced the affinity for CXCL12 (13). Full-length chemokine receptors CCR2b, CCR5, CCR8, CXCR3, and CX₃CR1 have since been shown to also exhibit increased ligand binding affinity upon tyrosine sulfation; replacement of tyrosines with phenylalanine residues resulted in 10 to 200-fold decreases in affinity and, in some cases, undetectable binding demonstrating the importance of sulfotyrosine recognition in chemokine signaling (12, 14–19).

While most chemokines form dimers or other oligomers, receptor activation is typically restricted to the monomeric ligand (20–24). However, structure-function studies of CXCL12 using preferentially monomeric (CXCL12_H25R) and constitutively dimeric (CXCL12₂) variants demonstrated that dimerization converts CXCL12 into a partial CXCR4 agonist that potently inhibits chemotaxis (25). As a CXCR4 ligand that stimulates intracellular calcium flux but fails to activate F-actin polymerization or β-arrestin recruitment, the CXCL12₂ dimer causes a type of ‘cellular idling’ that can block metastatic tumor formation in animal models for colorectal cancer and melanoma (26, 27). Differences in how the CXCR4 N-terminus is recognized by CXCL12 monomers and dimers may contribute to their distinct receptor activation profiles (26). For instance, in the NMR structure of the CXCL12₂:CXCR4_1–38 complex the receptor fragment wraps around both subunits of the CXCL12 dimer, placing sulfotyrosine 12 (sTyr12) and sTyr21 in distinct sites on one subunit while sTyr7 occupies a cleft at the dimer interface that would not exist in a CXCL12 monomer (25). In concordance with this structural model, it was observed that CXCR4_1–38 binding promotes CXCL12 dimerization (28). However, individual contributions of CXCR4 sulfotyrosines to the affinity and specificity of CXCL12 recognition and their impact on the monomer-dimer equilibrium remain unknown.

Site 1 contacts that contribute most to binding are potential targets for development of novel chemokine probes and antagonists. For example, we recently demonstrated that the sTyr21 binding pocket of CXCL12 can be targeted for inhibition by small molecule ligands that block CXCR4-mediated calcium signaling and chemotaxis (29, 30). This appears to be a conserved binding site, suggesting that sulfotyrosine-guided drug discovery may be a general strategy for targeting the chemokine family and other protein-protein interactions in the extracellular space (31).

Herein, a constitutively monomeric CXCL12 variant, termed CXCL12₁, was engineered to resist peptide-induced dimerization by maintaining steric repulsion of the chemokine helix. Six short CXCR4 peptides, centered on Tyr7, Tyr12, or Tyr21, were synthesized to study the contributions of individual sulfotyrosines in peptide binding and specificity. Peptides were titrated into CXCL12_WT, CXCL12₂, or CXCL12₁ and the interaction was monitored by 2D NMR. While sulfopeptides encompassing sTyr7 and sTyr12 interacted nonspecifically, an unsulfated Tyr7 peptide induced a new set of chemical shift perturbations in the CXCL12 monomer that were also observed upon binding of the intact CXCR4_1–38 N-terminal domain. In contrast, the Tyr21 peptides bound specifically to the sTyr21 recognition site in all three CXCL12 variants, but exhibited a significantly higher affinity for CXCL12₂. The sTyr21 sulfopeptide correspondingly increased the CXCL12 dimerization affinity by eight-fold, revealing an allosteric coupling between the sulfotyrosine binding site and CXCL12 dimerization.

RESULTS AND DISCUSSION

Engineering a CXCL12 constitutive monomer

The monomer-dimer equilibrium complicates the structural analysis of CXCL12 interactions. A H25R substitution at the dimer interface discourages but does not prevent CXCL12 self-association (32), as binding of the CXCR4 N-terminus induces CXCL12_H25R dimerization. In addition to Coulomb repulsion, a steric clash between the C-terminal α-helices of interacting CXCL12 monomers also limits dimer formation (33). The original NMR structure of CXCL12_WT in acetate buffer (PDB ID 1SDF; (10)) possesses a helix orientation (~55° relative to the β-sheet) incompatible with dimerization. After analyzing the proximity and geometry of backbone and C^β atoms using the program Disulfide by Design (34), we constructed a disulfide-constrained CXCL12 monomer using L55C and I58C substitutions to limit helix rearrangement and prevent dimer formation (Fig. 1A). Non-reducing SDS-PAGE, 2D NMR and SEC-MALS analyses of CXCL12₁ demonstrate a properly folded, monomeric species (Supplemental Fig. 1). Whereas CXCL12_WT self-associates with a K_d = 140 µM in 100 mM sodium phosphate (32), an HSQC dilution series of CXCL12₁ in the same conditions produced minimal chemical shift perturbations (Supplemental Fig. 1B). We concluded that the L55C/I58C disulfide stabilizes a monomeric CXCL12 conformation that is incompatible with dimer formation.

Figure 1. CXCL12₁ and CXCL12₂ have discrete CXCR4_1–38 binding sites.

(A) The CXCL12 NMR structure (PDB ID 1SDF) solved in acetate (pH 4.9) was used to identify and model the I55C/L58C mutations for disulfide formation (yellow). Dimerization is inhibited when the helix is constrained to an acute angle relative to the β-sheet (33). Chemical shift perturbations (orange) produced by CXCR4_1–3 mapped onto (B) CXCL12₁ (PDB ID 1SDF) and (C) CXCL12₂ (PDB ID 2K01). Chemical shift perturbations unique to CXCL12₁ are highlighted in ruby. Both structures are rotated 180° relative to their respective ribbon representations. (D) CXCR4_1–38 induced chemical shift perturbations fitted to a quadratic binding equation resulted in CXCL12₁ and CXCL12₂ affinities of 3.5 ± 0.1 and 0.9 ± 0.3 µM, respectively.

The CXCL12 monomer and dimer possess distinct CXCR4_1–38 binding sites

To establish a point of reference for subsequent measurements with smaller peptides, we monitored the binding of CXCR4_1–38 to the CXCL12₁ and CXCL12₂ variants by 2D NMR. Chemical shift mapping onto the CXCL12₁ model and CXCL12₂ NMR structure (Fig. 1B, C) highlighted perturbations across the β2 and β3 strands, consistent with the previous studies (25, 35), but CXCL12₁ also displayed additional contacts involving residues 58–61, 65, and 66 of the helix. Nonlinear fitting of chemical shift perturbations yielded K_d values of 3.5 ± 0.1 µM (CXCL12₁) and 0.9 ± 0.3 µM (CXCL12₂) (Fig. 1D). We recently demonstrated that perturbations induced by CXCL12_WT binding to [U-¹⁵N]-CXCR4_1–38 arise from the combination of distinct monomeric and dimeric interactions (26); in particular, CXCR4 residues 4–9 interact strongly with the preferentially monomeric CXCL12_H25R variant but weakly with CXCL12₂. Taken together, these results suggest that the CXCR4 N-terminus makes unique contacts with the helix of monomeric CXCL12 that are occluded by dimerization of the chemokine.

CXCL12 binding of short CXCR4 sulfopeptides

Sulfation of the CXCR4 N-terminal extracellular domain at either Tyr21 alone or in combination with Tyr12 and Tyr7 enhances its affinity for CXCL12 (13). To assess the respective contribution of each sulfotyrosine, we synthesized short sulfated and unsulfated CXCR4 heptapeptides centered on each of the three tyrosine residues (Fig. 2A). Each peptide was acetylated and amidated to create uncharged N- and C-termini, respectively (36, 37). A total of six peptides were titrated into each CXCL12 variant and monitored by 2D ¹H-¹⁵N HSQC experiments. For each titration, the chemical shift change of every residue was used to identify the binding site and calculate the interaction affinity. Table 1 contains a complete list of peptide binding affinities and binding energies.

Figure 2. Chemical shift perturbations of CXCR4 sulfopeptides designed from the extracellular N-terminus.

(A) The residues corresponding to the sTyr7 (cyan), sTyr12 (wheat), and sTyr21 (green) heptapeptides are indicated on the CXCR4 N-terminus amino acid sequence. The previously defined positions of sTyr7 (B), sTyr12 (C), and sTyr21 (D) heptapeptides are reproduced from the CXCL12₂:CXCR4_1–38 NMR structure (PDB ID 2K05). Chemical shift perturbations induced by sTyr7 (E, left two panels), sTyr12 (F, left two panels), and sTyr21 (G, left two panels) sulfopeptides map to the Tyr21 pocket on CXCL12₂ (PDB 2K05). In contrast, the chemical shift perturbations identify distinct binding sites for the sTyr7 (E, right two panels) and sTyr21 (F, right two panels) sulfopeptides on CXCL12₁ (PDB 2K05). Non-specific binding was observed for the sTyr12 sulfopeptide (G, right two panels).

Table 1.

Equilibrium dissociation constants and Gibbs’ free energy of binding for peptide:CXCL12 variant complexes.

Peptide	CXCL12WT		CXCL121		CXCL122
	Kd	ΔG (kcal mol−1)	Kd	ΔG (kcal mol−1)	Kd	ΔG (kcal mol−1)
CXCR41–38	4.5 ± 2.2 µM*	−7.29*	3.5 ± 0.1 µM	−7.43	0.9 ± 0.3 µM	−8.24
18SGDsYDSM24	1.8 ± 0.2 mM	−3.74	1.6 ± 0.2 mM	−3.81	211 ± 23 µM	−5.01
18SGDYDSM24	2.7 ± 0.5 mM	−3.50	1.5 ± 0.4 mM	−3.85	831 ± 137 µM	−4.20
9SDNsYTEE15	1.0 ± 0.1 mM	−4.09	1.8 ± 0.3 mM	−3.74	266 ± 38 µM	−4.87
9SDNYTEE15	2.1 ± 0.3 mM	−3.65	2.7 ± 0.8 mM	−3.50	332 ± 91 µM	−4.74
4ISIsYTSD10	4.3 ± 0.3 mM	−3.22	6.1 ± 1.3 mM	−3.02	386 ± 91 µM	−4.65
4ISIYTSD10	1.9 ± 0.3 mM	−3.71	1.1 ± 0.6 mM	−4.03	418 ± 60 µM	−4.60

ΔG = −RT ln K_d @ 298 K

^*K_d and ΔG calculated from Veldkamp et al. (32)

Non-specific binding to sY21 position

As expected, the affinity of short CXCR4 peptides for all CXCL12 variants was greatly reduced relative to the intact CXCR4_1–38 N-terminal domain, with K_d values ranging from ~ 0.2–6 mM (Table 1). Interestingly, nearly all peptides bound most tightly to the CXCL12₂ locked dimer, whereas binding to CXCL12_WT and CXCL12₁ were roughly equivalent. In titrations with each of the CXCL12 variants, chemical shift mapping indicated that peptides encompassing Tyr21 consistently bound to the previously defined sTyr21 pocket, as detailed below (Fig. 2D; Supplemental Fig. 2). In contrast, the Tyr7 and Tyr12 peptides did not localize to their corresponding interaction surfaces observed in the dimeric CXCL12₂:CXCR4_1–38 complex structure ((25); Fig. 2B,C). For example, in the NMR structure residues 4–10 of CXCR4 occupied a cleft at the dimer interface enabling sTyr7 to contact the R20 and V23 side chains of CXCL12. Here, the greatest shift perturbations (V49, T31 and F14) induced by the sTyr7 and sTyr12 sulfopeptides were more consistent with binding to the sTyr21 recognition site (Fig. 2; Supplemental Figs. 2 and 3).

Sulfopeptides bound with higher affinity than their unsulfated analogs, as observed in previous studies (13, 37, 38), with one notable exception. Binding of the unsulfated Tyr7 peptide to CXCL12₁ was roughly five-fold stronger than the sTyr7 sulfopeptide (Table 1). In contrast to the sTyr7 sulfopeptide, the unsulfated version induced substantial chemical shift changes in residues 23–25, 58–59, and 61 (Fig. 2E), matching the monomer-specific shift perturbations detected with CXCR4_1–38 (Fig. 1B). Titrations with CXCL12_WT produced a similar pattern of shift perturbations and a twofold higher affinity for the unsulfated Tyr7 peptide (Supplemental Fig. 4), but there was no preferential binding to the CXCL12₂ dimer. Unlike Tyr21 which is rapidly modified by recombinant tyrosylprotein sulfotransferase-1 (TPST-1), Tyr7 sulfation is slow and not consistently predicted by the available algorithms (13). We postulate that contacts with the CXCR4 N-terminus are predominantly hydrophobic and favored when Tyr7 is unmodified and CXCL12 is monomeric.

The Tyr21 peptide exhibits high specificity and affinity

We monitored the binding of unsulfated ₁₈SGDYDSM₂₄ and sulfated ₁₈SGDsYDSM₂₄ to each CXCL12 variant by 2D NMR. Consistent with the CXCL12₂:CXCR4_1–38 complex structure, the unsulfated peptide produced chemical shift perturbations localized to residues in the β2 strand, β3 strand and N-loop of each variant (Fig. 3A; Supplemental Fig. 2). CXCL12_WT and CXCL12₁ further exhibited shifts in residues 58 and 65–67 of the helix (Supplemental Fig. 2). Sulfation of Tyr21 strengthened the affinity of ₁₈SGDYDSM₂₄ for all three CXCL12 variants and increased the magnitude of each perturbation without altering the pattern of responding residues (Table 1; Fig. 3B). The sTyr21 sulfopeptide bound to CXCL12₂ with the highest affinity (K_d = 211 µM) compared to all other peptide:chemokine combinations – four-fold stronger than the unsulfated variant (K_d = 831 µM).

Figure 3. Sulfation of Tyr21 improves sulfopeptide binding affinity and modulates full-length receptor activity.

(A) The largest chemical shift perturbations (orange) are consistent with the putative sTyr21 sulfopeptide (green) binding site on CXCL12₂ (PDB ID 2K05). (B) Chemical shift perturbations induced by sulfated and unsulfated peptides were fitted to a quadratic binding equation to yield K_d values. The sTyr21 sulfopeptide (circles) bound CXCL12_WT with K_d = 1.8 ± 0.2 mM (black), CXCL12₁ with K_d = 1.6 ± 0.2 mM (blue) and CXCL12₂ with K_d = 211 ± 23 µM (red). The Tyr21 peptide (triangles) bound CXCL12_WT with K_d = 2.7 ± 0.5 mM (black), CXCL12₁ with K_d = 1.5 ± 0.4 mM (blue) and CXCL12₂ with K_d = 831 ± 137 µM (red). (C) The calcium response of FLAG-tagged CXCR4 variants was measured as a function of CXCL12_WT concentration. Data are representative of two experiments each performed with three replicates. (D) Four parameter fits yielded each CXCR4 variants EC₅₀ and maximum calcium response. CXCR4 variants with EC₅₀ or maximum calcium response values more than three standard deviations from the mean CXCR4_WT quantities are indicated with an asterisk.

Given that unsulfated CXCR4_1–38 (K_d = 0.9 µM) binds CXCL12₂ with a Gibbs free energy of −8.24 kcal mol⁻¹, our results suggest that the ₁₈SGDYDSM₂₄ fragment alone provides about half of the binding energy (ΔG = −4.2 kcal mol⁻¹) for the CXCR4:CXCL12₂ ‘site 1’ interaction. The identification of such hot spots, or regions of binding interfaces that contribute substantial binding energy, are of particular interest in the generation of protein-protein interface inhibitors (39). The major energetic contributions of the Tyr21 pocket justify our recent success in developing small molecules that antagonize receptor activation (29, 30). Further, identification of the hydrophobic Tyr7 pocket on monomeric CXCL12 demonstrates the utility of small receptor peptides in identifying druggable chemokine hot spots.

Tyr7 and Tyr12 are dispensable for CXCR4 activation by CXCL12

TPST enzymes catalyze the O-sulfation of Tyr21 much more efficiently than the Tyr12 or Tyr7 residues in vitro (28). Further, sTyr21 recognizes a cleft on CXCL12 that may be conserved across most members of the chemokine superfamily (25, 31). To test the hypothesis that Tyr21 sulfation is critical for receptor activation, tyrosine to alanine mutations were introduced into FLAG-tagged CXCR4 and expressed in CHO-K1 cells. CHO cells were chosen because they do not express endogenous CXCR4 and yield high levels of sulfated protein (40). Receptor activation was assessed by monitoring the calcium response as a function of increasing CXCL12_WT concentrations (Fig. 3C). The response of CXCR4(Y7A) was similar to wildtype CXCR4 whereas the potency of CXCR4(Y12A) was reduced 3-fold. In contrast, CXCR4(Y21A) was significantly impaired both in terms of EC₅₀ and efficacy. The combined mutation of Y7A, Y12A, and Y21A did not further diminish the potency relative to CXCR4(Y21A) but reduced the efficacy to ~20% of the wildtype receptor.

We hypothesize that protein misfolding is not responsible for the altered efficacies for two reasons. First, all of the mutants are surface-expressed at levels equivalent to the wildtype CXCR4 receptor (Supplemental Fig. 5). Second, the CXCR4 N-terminus is disordered and is not believed to participate in folding the overall tertiary structure of the receptor. This prompts the question of why there are efficacy changes at all. Our data suggests that the two-site model, in which site 1 is discretely responsible for chemokine binding and site 2 is specific for activation, is oversimplified and that both of these regions are ultimately required for full receptor activation. We conclude that sulfotyrosine modifications serve both to enhance CXCL12 binding affinity, and therefore potency, as well as signaling efficacy. Taken together, our results define the relative importance of individual CXCR4 sulfotyrosine modifications (Tyr21 > Tyr12 > Tyr7) for CXCL12 binding and receptor activation.

Binding at the Tyr21 site promotes dimerization

Chemokine dimerization is highly sensitive to numerous factors including GAGs, divalent anions, and pH (32). The CXCR4 N-terminus also promotes CXCL12 dimerization (28), which drastically alters the cellular response (25, 26). Interestingly, sTyr21 sulfopeptide binding to CXCL12₂ is 20-fold stronger than to CXCL12_WT or CXCL12₁ (Fig. 3B). In addition, the sulfopeptide produces large perturbations at the dimer interface of CXCL12_WT (Supplemental Fig. 2) suggesting it may allosterically induce CXCL12_WT dimerization. We used intrinsic tryptophan fluorescence polarization (FP) to determine the K_d for CXCL12_WT dimerization in the presence and absence of 3 mM sTyr21 sulfopeptide. The sulfopeptide shifts the K_d for CXCL12_WT dimerization from 15.1 ± 0.4 mM to 2.6 ± 0.4 mM. Free sulfotyrosine (Tyr-SO₃H), which also binds the sTyr21 pocket (31), also promotes CXCL12 dimerization (K_d = 9.3 ± 1.8 mM; Fig. 4A).

Figure 4. The sTyr21 binding site is allosterically linked to CXCL12 dimerization.

(A) Intrinsic tryptophan fluorescence was used to calculate the CXCL12_WT dimerization affinity alone (circles; K_d = 15.1 ± 0.4 mM), in the presence of 50 mM sulfotyrosine (squares; K_d = 8.9 ± 1.8 mM), or in the presence of 3 mM sTyr21 sulfopeptide (diamonds; K_d = 2.6 ± 0.4 mM). (B) FP and NMR derived binding affinities were used to produce a thermodynamic cycle, which illustrates that CXCL12 dimerization and sTyr21 sulfopeptide binding are coupled. After one ligand has bound, the affinity for the second ligand is enhanced with a cooperativity factor (c).

The free energy changes derived from NMR binding (ΔG_bind) and FP dimerization (ΔG_dimer) measurements at 298 K were used to construct a thermodynamic cycle diagram (Fig. 4B). In pathway 1, CXCL12 binds the sTyr21 sulfopeptide (ΔG_bind = −3.8 kcal mol⁻¹) and then dimerizes (ΔG_dimer = −3.5 kcal mol⁻¹). When the sequence is reversed (pathway 2), CXCL12 dimerizes with ΔG_dimer = −2.5 kcal mol⁻¹ and then binds the sTyr21 sulfopeptide with ΔG_bind = −5.0 kcal mol⁻¹. Analysis of both pathways yields similar coupling energies of −1.2 kcal mol⁻¹ and −1.0 kcal mol⁻¹, respectively, that link CXCL12 dimerization to ligand binding in the sTyr21 recognition pocket. Similar studies with both CCL5 and CCL2 report binding of their respective N-terminal peptides, both sulfated and unsulfated, to promote dimer dissociation (38, 41). The apparent disparity between these studies and our data is most easily explained by the spatially distinct dimerization interfaces of CXC- and CC-type chemokines. CXCL12 dimerizes through the β1 strand and α-helix whereas CC-type chemokines self-associates using the N-terminus, N-loop and β3 strand. Nonetheless, both chemokines bind sulfopeptides in the cleft formed by the N-loop and β3 strand. The close proximity of the sulfopeptide-binding site to the CC-type dimer interface is more consistent with dissociation through direct binding competition rather than an allosteric mechanism.

Conclusion

Many intracellular signal transduction complexes involve recognition of a recurring sequence motif by a protein interaction domain (42). Binding to certain short linear motifs (or SLiMs (43)) depends on post-translational modifications (PTMs) like phosphorylation, acetylation or methylation, and their corresponding recognition sites are often viewed as promising targets for drug discovery (44–46). Likewise, tyrosine sulfation enhances protein-protein interactions in the extracellular space (47), and sulfotyrosine recognition likely defines a new class of druggable extracellular targets (31). However, selective binding may require the combination of multiple SLiMs as observed for the WASP interacting protein (WIP) which uses three distinct recognition epitopes, including the conserved polyproline motif, to bind the EVH1 domain of N-WASP (48).

Many chemokine receptors contain multiple N-terminal tyrosines, which are predicted to be sulfated to different levels based on the suitability of the flanking sequences for TPST recognition (49, 50). We treated each tyrosine in the CXCR4 N-terminus as the center of a SLiM and found that the most efficient site of enzymatic sulfation (Tyr21) (13) is also the most important motif for binding to both CXCL12 monomers and dimers. Surprisingly, neither the Tyr7 or Tyr12 motif bound to a unique site on the CXCL12 dimer, and sulfation of the Tyr7 motif eliminated the site-specific binding to the CXCL12 monomer observed with the unsulfated Tyr7 motif. To our knowledge, this is the first demonstration that sulfation of chemokine receptors does not universally enhance complex affinity. In contrast, our data suggests that distinct patterns of tyrosine sulfation could encode selectivity either by enhancing or reducing affinity for unique recognition sites. Taken together this implies another layer of regulation for chemokine signaling. The structural basis for this effect awaits further study, but it appears that monosulfated (Tyr7/Tyr12/sTyr21) CXCR4 would exhibit a preference for the monomeric CXCL12 ligand while sulfation of Tyr7 would bias CXCR4 toward interactions with a CXCL12 dimer. Regardless, we conclude that the most functionally relevant mode of CXCL12-CXCR4 interaction involves specific recognition of a sulfotyrosine at position 21, and additional interactions with the receptor N-terminus that are independent of tyrosine sulfation.

Our results for CXCL12 binding to the sTyr21 motif in CXCR4 also suggest that chemokine oligomerization may be subject to allosteric control. Binding of the sTyr21 sulfopeptide, which is too short to contact both subunits of a CXCL12 dimer, significantly enhances CXCL12 dimerization by an indirect mechanism. Because dimerization converts CXCL12 into a partial agonist that potently inhibits chemotaxis and tumor metastasis (25–27), variations in the sulfation pattern of CXCR4 could in principle have functional consequences in vivo. Moreover, ligands that occupy the sTyr21 recognition site of CXCL12 may act as competitive inhibitors of receptor binding and allosteric modulators of chemokine function.

METHODS

Construction of CXCL12₁ plasmid

The CXCL12₁ variant was produced via mutagenesis of the CXCL12_WT construct using complementary primers and the QuikChange Site-Directed Mutagenesis Kit (Stratagene) as per the manufacturer’s instructions. The expression vector insert was confirmed by DNA sequencing.

Protein expression and purification

CXCL12_WT, CXCL12₁ and CXCL12₂ were expressed and purified as described previously (27). CXCR4_1–38, comprising the first 38 amino acids of CXCR4 preceded by a residual GlyMet dipeptide tag, was expressed and purified as previously described (28).

Size-exclusion chromatography multi-angle light scattering (SEC-MALS)

SEC was performed at a flow rate of 0.4 ml min⁻¹ on a Superdex 75 10/300 GL analytical column (GE Healthcare) and monitored by an 18-angle MALS detector (Dawn Heleos II). CXCL12₁ (40 mg ml⁻¹; 5 mM) and CXCL12₂ (40 mg ml⁻¹; 2.5 mM) were solubilized in 25 mM MES (pH 6.8), 650 mM NaCl, and 0.02% (w/v) NaN₃. Samples were then applied to the column in a mobile phase of identical composition. Peak dispersion and average molar mass were calculated using Astra software.

NMR spectroscopy

All NMR spectra were acquired on a Bruker DRX 600 MHz spectrometer equipped with a ¹H, ¹⁵N, ¹³C TXI cryoprobe at 298 K. Experiments were performed with either 50 µM [U-¹⁵N]-CXCL12_WT, -CXCL12₁, or -CXCL12₂ proteins in a solution containing 25 mM deuterated MES (pH 6.8), 10% (v/v) D₂O, and 0.02% (w/v) NaN₃. Full-length CXCR4_1–38 peptide titrations required 20 µM [U-¹⁵N]-CXCL12₁ and -CXCL12₂ protein samples. Sulfated CXCR4 peptides were reconstituted at 100 mM in 25 mM deuterated MES, 10% (v/v) D₂O, and 0.02% (w/v) NaN₃ buffer. Unsulfated CXCR4 peptides were similarly reconstituted to 15 mM. Two separate ₄ISIYTSD₁₀ peptide samples were reconstituted at 10 mM and 15 mM peptide in 25 mM deuterated MES, 10% (v/v) D₂O, 0.02% (w/v) NaN₃ buffer, from lyophilized powder. The 10 mM batch of ₄ISIYTSD₁₀ peptide was titrated into CXCL12_WT and CXCL12₂ and the 15 mM batch was titrated into CXCL12₁. CXCL12_WT and CXCL12₂ chemical shift assignments (¹H and ¹⁵N) were acquired from previously published sources (BMRB ID 16145 and 15633, respectively).

Peptides were titrated into CXCL12 samples and monitored by ¹H-¹⁵N Heteronuclear Single Quantum Coherence (HSQC) experiments. Total peptide additions differed for individual peptide titrations but ranged from 0 – 160 equivalencies. Spectra were processed using in-house scripts and chemical shift tracking was performed using CARA software (51). Combined ¹H/¹⁵N chemical shift perturbations were calculated as ((5 Δδ_H)²+(Δδ_NH)²)^0.5, where δ_H and δ_NH are the amide proton and nitrogen chemical shifts, respectively. Equilibrium dissociation constants (K_d) were determined by nonlinear fitting of the combined ¹H/¹⁵N_H chemical shift perturbations as a function of peptide concentration to a single-site quadratic equation (52). For a given interaction, residues with the largest chemical shift perturbations were fitted individually. The resulting K_d values and their respective fitting errors were then averaged to produce the reported affinity and standard deviation for that interaction.

Fluorescence polarization assay

Fluorescence polarization (FP) assays were performed on a PTI spectrofluorometer equipped with automated polarizers, using a time base polarization method provided by the program Felix32. Lyophilized CXCL12_WT was reconstituted in H₂O and diluted to appropriate concentrations in a 25 mM MES (pH 6.8) buffer, filtered and degassed. ₁₈SGDsYDSM₂₄ peptide and sulfotyrosine (Tyr-SO₃H) stocks were prepared separately at 25 mM and 500 mM, respectively, in a 25 mM MES (pH 6.8) buffer, filtered and degassed. Experiments were performed at 298 K and intrinsic tryptophan fluorescence was observed with emission and excitation wavelengths of 325 nm and 295 nm, respectively. FP was monitored as a function of increasing CXCL12_WT concentration (10, 25, 50, 75, 100, 250, 500, 750, 1000 and 1500 µM) alone or in the presence of 3 mM ₁₈SGDsYDSM₂₄ or 50 mM sulfotyrosine. The CXCL12_WT dimerization equilibrium dissociation constant (K_d) was determined by non-linear fitting to a three-parameter function as previously described (32). Experiments were performed in duplicate and the reported dimerization K_d values reflect the average of both experiments.

Cell culture

Chinese hamster ovary K1 (CHO-K1) cells, stably transfected with the Gα15 gene in pcDNA3.1+, were cultured in a 1:1 mixture of Dulbecco’s modified Eagle medium (DMEM) with Glutamax (Gibco):F12 nutrient mixture (Gibco) supplemented with 10% (w/v) fetal bovine serum (FBS) (Gibco). Stable expression of the Gα15 transgene was maintained by further supplementing the growth medium with 700 µg ml⁻¹ geneticin (Gibco).

Transfections

For transient transfections with the Flag-CXCR4 WT and N-terminally mutated constructs, the CHO-K1 Gα15 stable cells were lifted using 0.25% (v/v) trypsin-EDTA (Gibco), and 2 × 10⁶ cells were re-plated onto 10 cm culture dishes. To increase transfection efficiency, 0.25% (v/v) DMSO was added to the media when the cells were plated, and no geneticin was present in the plating medium. The cells were transfected 24 hours later using the Mirus TransIT-CHO Transfection Kit according to the manufacturers protocol, with the following exceptions: the amount of CHO transfection reagent was scaled up to 4 µl µg⁻¹ of DNA in the transfection mixtures, and the amount of CHO mojo reagent was scaled up to 1 µl µg⁻¹ of DNA in the mixtures. Immediately before transfection, the medium was replaced with a 1:1 mixture of DMEM with Glutamax:F12 nutrient mixture supplemented with 10% (w/v) FBS (without DMSO or geneticin).

Calcium flux

24 hours after plating, the medium was removed and the cells were washed with 5 ml PBS. The adherent cells were then incubated for 10 minutes in 3 ml of Cellstripper non-enzymatic cell dissociation solution (Cellgro). Cells were then suspended by pipetting, washed twice in Calcium Flux Buffer (Hanks Balanced Salt Solution supplemented with 20 mM HEPES) supplemented with 0.1% (w/v) bovine serum antigen (BSA) and 4 mM probenecid (Invitrogen). Each washing was carried out by centrifuging the cells at 350 × g and then re-suspending in Calcium Flux Buffer. After washing, the cell concentrations were normalized between samples and the cells were plated at 2.5 × 10⁵ cells/well in poly-D-lysine coated 96 well plates (Becton Dickinson Labware). FLIPR 4 Calcium Flux assay kit dye (Molecular Devices) was then added to each well, such that the ratio of dye to cell suspension was 1:1. The plates were then centrifuged for 3 minutes at 250 × g to ensure the cells settled onto the surface of the plates. The plated cells were incubated at 310 K for 90 min. Fluorescence was measured at 310 K using a FlexStation2 Microplate Reader with excitation and emission wavelengths at 485 nm and 515 nm, respectively. After an 18 s baseline measurement, the indicated concentrations of CXCL12 were added and the resulting calcium response was measured for an additional 50 s. Fluorescence as a function of CXCL12 concentration was fitted to a four-parameter equation. Data are representative of two experiments each performed with three replicates. CXCR4 variants with EC₅₀ or maximum calcium response values more than three standard deviations from the mean CXCR4_WT quantities were deemed significant.

Flow cytometry

For testing receptor surface expression, 3 × 10⁵ cells were set-aside during the calcium flux assay preparations. The cells were then washed twice with PBS containing 0.5% BSA (w/v) (FACS buffer). Staining was carried out in a 50X dilution of either anti-DDDDK or mouse IgG1 isotype control conjugated to SureLight APC (Columbia Biosciences) for 45 minutes on ice. Cells were then washed 3X in FACS buffer before analysis, which was carried out using a Guava bench top mini-flow cytometer.