New paradigms in chemokine receptor signal transduction: moving beyond the two-site model

Abstract

Chemokine receptor (CKR) signaling forms the basis of essential immune cellular functions, and dysregulated CKR signaling underpins numerous disease processes of the immune system and beyond. CKRs, which belong to the seven transmembrane domain receptor (7TMR) superfamily, initiate signaling upon binding of endogenous, secreted chemokine ligands. Chemokine-CKR interactions are traditionally described by a two-step/two-site mechanism, in which the CKR N-terminus recognizes the chemokine globular core (i.e. site 1 interaction), followed by activation when the unstructured chemokine N-terminus is inserted into the receptor TM bundle (i.e. site 2 interaction). Several recent studies challenge the structural independence of sites 1 and 2 by demonstrating physical and allosteric links between these supposedly separate sites. Others contest the functional independence of these sites, identifying nuanced roles for site 1 and other interactions in CKR activation. These developments emerge within a rapidly changing landscape in which CKR signaling is influenced by receptor PTMs, chemokine and CKR dimerization, and endogenous non-chemokine ligands. Simultaneous advances in the structural and functional characterization of 7TMR biased signaling have altered how we understand promiscuous chemokine-CKR interactions. In this review, we explore new paradigms in CKR signal transduction by considering studies that depict a more intricate architecture governing the consequences of chemokine-CKR interactions.

Graphical abstract

1. Introduction

Chemokine receptors (CKRs) are cell-surface seven transmembrane domain receptors (7TMRs) that mediate a diverse repertoire of functions, such as immune surveillance and embryonic development, by directly regulating cellular migration, adhesion, growth, and survival. They are also implicated in many pathological processes such as atherosclerosis, HIV infection, tumor metastasis, and autoimmune disorders [1]. Due to their prominent roles in so many disease processes, CKRs have been the target of considerable drug development efforts since the discovery of the chemokine-CKR system in the late 1980s [2, 3].

Chemokines and CKRs demonstrate widespread promiscuity, wherein chemokines may bind multiple receptors and vice versa. Of the nearly 50 chemokines and 20 CKRs identified in humans, most bind multiple counterparts, with a minority involved in monogamous interactions. Promiscuous interactions among chemokines and their receptors are increasingly recognized as a mechanism to generate diverse signaling or other functional outcomes using a discrete set of chemokine and CKR components [4, 5]. This characteristic promiscuity may be explained, in part, by their conserved tertiary structure, composed of an unstructured N-terminus, conserved mono- or di-cysteine motif (e.g. C, CC, CXC, CX₃C, where X represents a non-cysteine residue), extended loop, three anti-parallel β-strands, and C-terminal α-helix (Fig. 1.A) [1]. One or two conserved disulfides constrain the chemokine fold by linking the cysteine motif with the β1–β2 turn (a.k.a. the 30s loop) and the β3-strand.

Fig. 1.

Chemokine tertiary structure and implications of site 1 binding modes in chemokine-CKR interactions. (A) The structure of monomeric CXCL12:CXCR4_1–38 (PDB ID 2N55) modeled into full-length CXCR4 (PDB ID 3ODU) demonstrates conserved features of chemokine tertiary structure [13]. The two steps/sites of the two-step/two-site model are also depicted. (B) Binding sites of the CKR N-termini are shown on chemokines from the two recent chemokine-CKR co-crystal structures with the chemokine represented in the same view (vMIPII:CXCR4: PDB ID 4RWS; CX3CL1:US28: PDB ID 4TX1) [8, 9]. Only TM1 from each receptor is depicted for clarity. The CXCL12:CXCR4_1–38 structure-based model from Ziarek, et al., is shown for comparison [13]. The location of the conserved Pro-Cys (PC) motif is indicated for each receptor. The orientation of the receptor N-termini varies among the three complexes, suggesting that site 1 contacts may alter subsequent chemokine-CKR interactions. (C) The CXCL12:CXCR4 model based on the CXCL12-monomer:CXCR4_1–38 structure is shown (i.e. “CXCR4:CXCL12 monomer”) along with the CXCR4 N-terminus from the CXCL12-dimer:CXCR4_1–38 structure (i.e. “CXCR4:CXCL12 dimer” ; PDB ID 2K01) [35]. CXCL12 dimerization alters the binding orientation of the CXCR4 N-terminus, which may cause unique binding modes at CXCR4 and ultimately specify functional outcomes. (D) Close-up view of CXCL12:CXCR4 contacts from (C). (E) The CXCL12 dimer is shown with one subunit in orange (aligned to the CXCL12 monomer from C) and the second subunit in grey. CXCL12 dimerization occludes the binding site of CXCR4 N-terminus residues 1–9.

CKR binding and activation is described as proceeding via a two-step/two-site mechanism, a model which dates back to the mid-1990s. This model is alternatively framed by segregating chemokine-CKR interactions functionally (i.e. two-step) and spatially (i.e. two-site). In the functional formulation, site 1 provides affinity and specificity, followed by site 2 which elicits receptor activation. In the spatial formulation, site 1 refers to interactions between the CKR N-terminus (a.k.a., chemokine recognition site 1, CRS1) and the chemokine globular core, and site 2 refers to contacts between residues in the receptor transmembrane (TM) domain (a.k.a., CRS2) and the unstructured chemokine N-terminus [3]. Notably, interactions between chemokines and CKR extracellular loops (ECLs) are variously ascribed to site 1, site 2, or not included in these models at all [1, 6–9].

Isolation of the receptor N-terminal domain has enabled structure determination of several site 1 complexes but numerous difficulties hindered the characterization of full-length receptors. Since 2007, technical innovations have made possible the purification and crystallization of over 100 family A 7TMRs including CKRs. Until recently, only apo structures or those bound to small molecule antagonists were available [10–12]. In 2015, the first structures of chemokine-CKR complexes were solved, detailing chemokine interactions in the TM domain (i.e. site 2) [8, 9]. Combination of these site 1 and site 2 structures recently enabled construction of the most detailed chemokine-CKR model to date [13]; and together, these data highlight numerous contacts that fall outside of the conventional spatial and functional definitions of sites 1 and 2. This coupled with an increased awareness of biased agonism (i.e. preferential activation of G protein or β-arrestin pathways), non-chemokine ligands (e.g. ubiquitin, β-defensins), and the expanding roles of post-translational modifications (PTMs; e.g. sulfation, polysialylation) underscores how the two-site model may overlook the complexity and diversity of CKR signaling that we now appreciate two decades after it was proposed [4, 5, 14–17].

The goal of this review is to highlight instances in which the two-site model inadequately addresses more complex features of chemokine-CKR interactions. In doing so, we hope to broaden the reader’s appreciation of the mechanistic details involved in CKR signal transduction. We emphasize that the two-site model has served as a useful framework to understand CKR activation, and in some cases may sufficiently describe binding and activation. Nevertheless, we believe it is advantageous to look beyond the functional and structural roles segregated into site 1 or site 2, to delineate new capacities for interactions that have not been well described by either site, and to include new features that have been discovered since the original conception of the two-site model.

2. Beyond Site 1

2.1. Origins of the two-site model and early studies of the site 1 interface

The two-step/two-site model of chemokine-CKR interactions was realized almost 20 years ago through the work of three contemporaneous studies [18–20]. First, Monteclaro and colleagues used a chimera of the CCR2 N-terminal domain and the CCR1 TM region to show that the receptor N-terminus was sufficient to recognize CCL2 with high affinity and recapitulate the native interaction [18]. Notably, the complementary CCR1–CCR2 chimera exhibited a 30-fold decrease in G protein signaling, demonstrating that the CCR2 N-terminus is essential for chemokine recognition but not signaling. In a follow-up study they showed that high affinity CCL2 binding was completely dependent upon the presence of the CCR2 N-terminus and could be fully recapitulated using only a membrane tethered N-terminal peptide [19]. Crump and colleagues also hypothesized a two-site mechanism through studies of the chemokine rather than the receptor. They showed that mutation of the CXCL12 N-terminus attenuated signaling activity without significant loss of affinity (e.g. 3-13-fold increase in binding K_d) [20]. Taken together, these studies suggested that the site 1 and site 2 interactions were spatially and functionally independent, with site 1 conferring receptor specificity and affinity, and site 2 mediating receptor activation. Over time, other functional studies led to the consensus that this model was broadly applicable to the chemokine-CKR system [21]. Interestingly, the two site model of chemokine-CKR interactions was predated by an analogous model described for interactions between the inflammatory protein C5a and its receptor, suggesting broad applicability of this model among other GPCRs with protein ligands [22].

At the same time other studies began to probe the site 1 interface in greater detail. Alanine scanning of CXCL8 identified residues in an extended loop between the conserved N- terminal cysteine(s) and the 3₁₀-helix (a.k.a. the N-loop) [23]. Unlike CXCL8, CXCL1 is a high affinity CXCR2 ligand with weak affinity for CXCR1. Exchange of seven CXCL1 N-loop residues with those of CXCL8, a high affinity CXCR1 ligand, resulted in a molecule capable of recognizing both receptors [24, 25]. Using a similar chimera approach, Crump and colleagues showed that insertion of the CXCL12 N-loop into unrelated CXC-family chemokines (CXCL1 and CXCL10) rendered them capable of binding and activating CXCR4 [20]. Subsequent studies expanded the importance of this region for site 1 interactions to other CC and CXC chemokines, establishing the N-loop as a critical motif for CKR recognition [19, 20, 26, 27].

While these and related functional studies have helped define functional roles for the N-loop and other chemokine domains in signal transduction, NMR titration experiments have historically been used to define structural interactions contributing to site 1 recognition. One of the most common NMR-based approaches has been to titrate unlabeled, CKR N-terminal peptides into purified, [U-¹⁵N]-labeled chemokines to identify chemokine residues that participate in direct site 1 binding interactions. This and related approaches have been used to map the chemokine site 1 binding interactions for CCL11:CCR3 [28], CCL21:CCR7 [29], CCL24:CCR3 [30], CXCL8:CXCR1 [31, 32], CXCL10:CXCR3 [26], CXCL12:CXCR4 [13, 33–37], and CX3CL1:CX3CR1 [38]. Collectively, these studies demonstrate the essential role of the chemokine N-loop in directly binding CKR N-termini, and are validated by soluble chemokine- CKR structures (discussed in section 2.4).

2.2. Site 1 interactions: chemokine allostery and conformational dynamics

While early studies of chemokine-CKR interactions suggested the functional and spatial independence of site 1 interactions, more recent studies suggest that site 1 interactions can alter functional outcomes. In a study of CXCL8 activation of CXCR2 and CXCR1, Rajarathnam and colleagues identified an important Gly-Pro (GP) sequence in the 30s-loop that had large conformational effects on CXCL8 when mutated, causing CXCL8 to activate CXCR1 and CXCR2 in unique ways [39]. While some GP mutants activated both receptors with similar potencies, two mutants (G31A and P32G) displayed a modest reduction in affinity at CXCR2 but completely lost the ability to elicit CXCR2-mediated calcium release. These same mutants lost all binding and calcium signaling at CXCR1. The study demonstrated that the change in signaling was due to intramolecular interactions between the GP motif of the 30s-loop and the conserved “ELR” motif of the CXCL8 N-terminus, which is known to be important for CXCR2 activation. The authors employed molecular dynamics (MD) simulations to show conformational switching for some CXCL8 mutants interconverted the 30s-loop between a type-I and type-II β-turn, thereby altering the conformation and orientation of the chemokine N-loop and N-terminus allosterically. They suggested that chemokines exist in conformational ensembles, and receptor binding and activation involves conformational selection on the part of the ligand and receptor. In this way, different CKR N-termini may selectively bind specific orientations of the chemokine ensemble, thereby eliciting unique functional outcomes at two separate receptors (see Section 5.1 on biased signaling).

In another study, the same group examined the role of the residue sandwiched between the conserved cysteines of CXCL8. Conversion of the CXC motif to a CC motif greatly reduced the binding affinity for both CXCR1 and CXCR2 and rendered it incapable of activating CXCR2 [40]. The mutation did not affect the chemokine fold, dimerization, or glycosaminoglycan (GAG) binding, suggesting that the attenuated binding and signaling properties were a consequence of altered intramolecular dynamics. Supported again by MD simulations, the authors suggested that allosteric site 1 interactions may in effect ‘steer’ the orientation of the chemokine N-terminus within the receptor TM bundle. Similar studies of vMIP-II and CX3CL1 cysteine motifs suggest that “conformational switching” may be a more general phenomenon among the chemokine family [41]. These studies demonstrate that subtle structural changes in one chemokine domain can significantly alter receptor activation by eliciting conformational changes in another domain. In effect, these studies challenge the structural and functional independence of site 1 and site 2 interactions.

2.3. Complex roles for receptor post-translational modifications

Farzan and colleagues expanded the scope of interactions underlying site 1 recognition by showing that sulfation of tyrosines in the CCR5 N-terminus enhanced affinity for CCL3 and CCL4 [42]. This and later studies broadened the repertoire of CKR post-translational modifications (PTMs) to include glycosylation, demonstrating that in addition to enhancing chemokine affinity, PTMs can regulate functional outcomes of site 1 interactions [17, 21].

CKRs undergo enzymatic, O-sulfate modification by tyrosyl protein sulfotransferase (TPST) during processing in the trans-Golgi network. The presence of these sulfotyrosine (sTyr or sY) PTMs at receptor N-termini enhances binding of chemokine ligands and affects receptor activation in many chemokine-CKR systems [16, 42–49]. The structural contributions of sTyr modifications to site 1 interactions were defined by the NMR solution structure of a covalently-linked CXCL12 dimer bound to the first 38 amino acids of CXCR4 that was enzymatically sulfated at three tyrosine positions. The CXCL12 locked dimer was engineered out of necessity, as CXCR4 peptide binding altered the CXCL12 monomer-dimer equilibrium, thereby hindering NMR experiments. The observation that tyrosine sulfation enhanced CXCR4 peptide binding, which in turn promotes CXCL12 dimerization, defined an allosteric model in which binding of sTyr peptides at the conserved sTyr pocket causes CXCL12 self-association. For instance, Ziarek and Getschman, et al., found that a sulfated heptapeptide corresponding to the Tyr21 region of CXCR4 specifically and preferentially binds to the CXCL12 dimer while promoting dimerization of WT-CXCL12. These studies represent the first evidence that binding at a pocket, disconnected from the CXC dimer interface, could allosterically regulate chemokine self-association [50]. Specifically, dynamic NMR studies of CXCL12 revealed that to accommodate dimer formation, the α-helix of CXCL12 rearranges to an almost 90° angle, perpendicular to the β-strands [51]. This large motion is triggered by two residues adjacent to the sulfotyrosine binding pocket that serve as a link or “microswitch” back to the C-terminal α-helix of CXCL12 [52]. Importantly, self-association of CXCL12 has significant functional effects, as the locked, dimeric version of CXCL12 elicits a unique signaling profile compared to WT-CXCL12 [36].

Glycosylation of CKR extracellular domains occurs during processing in the endoplasmic reticulum (i.e. N-linkage of asparagine residues) or Golgi (i.e. O-linkage of serine/threonine residues). Recent studies described a novel functional role for chemokine receptor glycosylation in which polysialic acid (polySia) addition to receptor glycans allows CKRs to discriminate between chemokine binding partners [17]. CCR7 is polysialyated by the enzymes ST8Sia II and ST8Sia IV on the surface of patrolling dendritic cells. This rare PTM is utilized by CCL21, which has an unusual extended C-terminal tail not shared by the other CCR7 ligand, CCL19. When CCR7 is polysialylated, CCL21 binds CCR7 with high affinity due to an interaction between the polySia of CCR7 and the C-terminus of CCL21. This interaction is thought to release CCL21’s tail from an autoinhibitory interaction with its chemokine core, freeing its N-terminus to bind and activate CCR7. In the examples described for both tyrosine sulfation and polysialylation, interactions between these receptor PTMs and chemokine site 1 domains dictate unique functional outcomes. Despite some shared features of PTMs in chemokine-CKR signal transduction (e.g. sTyr enhancement of chemokine affinity/potency [16]), many features appear to be context-specific (e.g. polySia alteration of CCL21 activity [17]) or even contradictory (e.g. sTyr promotes and inhibits chemokine dimerization for different chemokine-CKR pairs [16]), preventing the prediction of PTMs on biological activity in the absence of experimental data.

2.4. Complex interactions between receptor N-termini and the chemokine core

An early model for the interaction of CXCL8 with the CXCR1 N-terminus set the structural precedent for site 1 formation, corroborating the direct involvement of the N-loop and expanding the interface to include the chemokine cleft, formed by the N-loop and β2/β3 turn [32]. Nevertheless, this NMR structure required a biased NMR structure refinement procedure due to weak binding of their modified CXCR1 peptide, suggesting the need for more site 1 complexes to validate the site 1 interface [32]. To date, six site 1 complexes (four NMR [13, 32, 35, 53] and two crystallographic [8, 9]) have been determined in which the receptor peptide adopts three different orientations. Indeed, in all structures except the CCL11:CCR3 complex, the receptor lies nearly perpendicular to the β-sheet axis primarily contacting the N-loop, chemokine cleft and β-strands, validating the site 1 interface defined by the CXCL8:CXCR1 structure. Despite this common interface, the site 1 complexes demonstrate considerable architectural diversity. For instance, the orientation of the N- and C-termini of the CXCR4 peptide is inverted when bound to CXCL12 compared to other site 1 complex structures. While a recent review has suggested that this orientation is incompatible with chemokine N-terminal insertion into the TM domain [54], flexible docking of the monomeric CXCL12:CXCR4_1–38 structure into CXCR4 demonstrates that this distinct directionality may facilitate rotation of CXCL12 relative the CXCR4 TM bundle so that it is positioned to form extensive site 2 and intermediate-site interactions [13]. Specifically, our model predicts that CXCR4 assumes a bent conformation adjacent to the conserved N-terminal Pro-Cys (PC) motif relative to other chemokine-CKR complexes (Fig. 1B, discussed in Section 4.1) [13]. Despite their differences, all six receptors form apolar and electrostatic contacts, often through a highly conserved tyrosine with the chemokine cleft.

Dimeric CXCL12 was recently identified as a biased agonist that induces G protein signaling but is incapable of promoting β-arrestin recruitment or cellular migration [13, 36, 55]. Regardless of quaternary structure CXCR4 makes specific contacts with the N-loop and chemokine cleft, contributing 50% of the total site 1 binding energy [50], but residues of the CXCR4 N-terminal domain adopt two distinct conformations when bound to CXCL12 monomer and dimer (Fig. 1C–E). For example, CXCR4 residues 7–9 contribute an intermolecular β-strand to monomeric CXCL12 and residues 4–6 tuck into a hydrophobic pocket abutting the C-terminal helix (Fig. 1C, D) [13]. These contacts are supported by mutagenic and NMR experiments with the full-length receptor [13, 56, 57]. Self-association with a second CXCL12 molecule competes for the monomer-specific β1 and helix contacts and displaces those residues of the receptor, which instead form less stable contacts with the second CXCL12 protomer (Fig. 1E) [35]. Inspection of the CXCL12:CXCR4 hybrid model suggests dimeric CXCL12’s distinct signaling profile may result from unique contacts with the ECL and TM regions (discussed in Section 4).

It is reasonable to assert that the degree to which the receptor N-terminus wraps around the globular core modulates the chemokine’s orientation and interactions with the receptor ECL and TM regions. In the context of biased agonism, the receptor N-terminus may mask or expose epitopes to the receptor ECLs. Taken together, the site 1 interactions may generate functional complexity via unique interactions rather than simply tethering the chemokine and contributing binding energy. Classic definitions of site 1 fail to recognize the diversity of binding modes and unique domains that can interface with receptor N-termini, suggesting a greater spatial and functional complexity than the traditional model would suggest.

3. Beyond Site 2

3.1. A more complicated site 2: the major and minor binding pockets

The recent crystal and NMR structures of chemokine-CKR complexes provide clues that far from following a two-site convention, interactions are diverse and highly specific for each individual chemokine-CKR pair at the extracellular surface [8, 9, 13]. Similarly, a surfeit of 7TMR crystal structures over the past decade is defining how the conserved TM architecture recognizes diverse ligand types and triggers unique signaling outcomes [58, 59]. In particular, the early 7TMR crystal structures divided the orthosteric-binding site (i.e. the ‘main’ ligand binding pocket) into two subpockets [58, 60–62]: the major subpocket consists of the cavity defined by TMs 4, 5, and 6, and the minor subpocket by TMs 1 and 2. TMs 3 and 7 occupy the interface between the two subpockets and stabilize ligand-CKR interactions in either subpocket [3, 58, 63, 64].

A 2013 analysis of over 40 7TMR structures revealed the majority of co-crystallized family A 7TMR ligands contacted the major subpocket (especially TMs 3, 5, and 6) with few ligands forming even one contact in the minor subpocket [65]. It should be noted this analysis was enriched for antagonist contact points since inactive 7TMR structures are more abundant. Likewise, peptide-binding receptors (including CKRs) represented a small minority of the crystallized receptors in this study. Nevertheless, reviews of CKR binding determinants show a more equitable distribution of contact points among major and minor subpockets, with many CKR agonists and antagonists alike preferentially binding the minor subpocket alone [3, 7–10, 12, 64]. This trend is borne out among the five CKR-ligand co-complexes, with three of the five co-crystallized ligands primarily occupying the minor subpocket (IT1t:CXCR4, vMIP-II:CXCR4, and CX3CL1:US28), one ligand primarily occupying the major subpocket (CVX15:CXCR4), and one ligand straddling the two subpockets (maraviroc:CCR5) [8–10, 12].

CKRs possess a number of unique features that may explain why ligands more readily sample the minor binding pocket relative to other family A 7TMR subtypes. Firstly, the extracellular portion of TM1 in all CKR structures is inwardly oriented toward the center of the TM bundle, with CXCR4-IT1t displaced 9 Å relative to a prototypical family A member, β₂-adrenergic receptor (β₂AR) [10]. TM1 is positioned closer to the adjacent TM7 and creates a more contiguous helix-helix interface [8–10, 12]. Secondly, compared to the β₂AR, TM1 is 1–2 turns longer extracellularly when the receptor (CXCR4 or US28) is bound to a chemokine ligand (vMIP-II or CX3CL1) and TM7 is 1–2 turns longer regardless of the associated ligand (discussed in Section 4.1). The overall effect of the elongated TM1 and TM7 helices, and the inward orientation of TM1, is to create a larger minor pocket. Another feature that may enrich minor subpocket contacts is that chemokines almost universally bind receptor N-termini [7]. By forming extensive interactions with CKR N-termini, which themselves are linked to TM7 via a disulfide bond, chemokines are positioned directly above the minor subpocket (Fig. 2A, discussed in Section 5.1). Still, small molecule ligands also demonstrate enriched utilization of the minor binding pocket despite making no contacts with CKR N-termini, suggesting ectodomain interactions are not solely responsible [3].

Fig. 2.

(A) Chemokine (CK) -chemokine receptor (CKR) interactions encompass many more interactions than those between the chemokine core and the receptor N-terminus (i.e. site 1) and those between the chemokine N-terminus and the receptor TM domains (i.e. site 2). (B) A “multi-site”/multiple-variable model of CKR activation [103]. Functionally distinct outcomes following chemokine-CKR interactions may be generated by chemokine “selection” of a unique subset of interactions, binding modes, and conformations. These variables are listed numerically to demonstrate how pairs of chemokine-CKR interactions generate complex outcomes.

3.2. Role of subpocket specificity in receptor activation

The diversity of CKR ligand binding sites emphasizes the additional level of regulation built into receptor activation compared to major subpocket-biased 7TMRs. The major and minor subpockets contain unique sets of residues that comprise molecular switches [4, 64, 66]. Molecular switches are conserved receptor ‘hotspots’ that undergo conformational rearrangements following agonist binding, helping to drive global conformational rearrangements required for 7TMR activation [66]. To complicate matters, CKR agonists and antagonists frequently share a subset of receptor contacts, a general feature of 7TMR ligands [67]. For instance, Glu^7.39 and Trp^2.60 (Ballesteros-Weinstein nomenclature [68]) frequently serve as contact points for small molecule CKR ligands [3, 64, 69]. Notably, our recent hybrid model of CXCL12 bound to CXCR4 (based on the IT1t:CXCR4 X-ray and LM:CXCR4_1–38 NMR structures) suggests that unique positional and rotameric states of multiple CXCR4 residues, relative to antagonist-bound structures, contribute to receptor activation [13]. Consequently, site 2 binding might itself be broken down into a series of “choices” dictated by the ligand: 1) selection of the CKR binding-pocket (i.e. major subpocket, minor subpocket or a combination of both), and 2) stabilization of subpocket residues in active (or inactive conformations), both of which will have the effect of engaging (or preventing engagement of) a particular subset of molecular switch residues required to elicit the ligand-associated functional response (Fig. 2B).

The diverse binding modes in the major and minor subpockets place each ligand in the proximity of a unique subset of molecular switches. Considering the major binding pocket, the “conserved core” interaction between Pro^5.50, Ile^3.40, and Phe^6.44, directly below the ligand binding sites of the β₂AR and μ-opioid receptor (MOR), propagates conformational changes to the intracellular receptor surface for activation [70–72]. Similarly, below the minor pocket is the TxPxW motif (i.e. Thr^2.56-x-Pro^2.58-x-Trp^2.60) conserved among most chemokine receptors, although its specific role in receptor activation is not well understood [64, 73]. Pro^2.58 is important for receptor activation in multiple receptors, including CCR5 [64, 74]. Trp^2.60 has been consistently identified as a principal binding contact for CKR small molecule antagonists, and when mutated, disrupts their inhibitory effects [3, 75]. Finally, our CXCL12:CXCR4 hybrid model suggests that interactions between the TxPxW motif and residues in TMs 3 and 7 may initiate a concerted rearrangement of a group of hydrophobic residues in the TM region, resulting in receptor activation [13]. Given the unique distribution of both ligand contacts and residues involved in conserved motifs among different TM domains, it is becoming clear that ligand-specific receptor outcomes are a consequence of the stabilization of specific rotameric states in the receptor-binding pocket followed by engagement of unique subsets of molecular switches.

A chemokine’s subpocket “preference” may also depend upon its N-terminal cysteine motif. Qin and colleagues aligned multiple chemokine structures belonging to the CC and CXC subgroups, and noted that CXC chemokines display a characteristic bend immediately preceding the CXC motif, causing their N-terminus to run parallel to the N-loop [8]. In models of CXCL12 bound to CXCR4, they predicted that the bend directs the N-terminus toward the major pocket, whereas vMIP-II (a viral CC chemokine) directs its N-terminus towards the minor subpocket. Since most chemokines are thought to form interactions with receptor N-termini, CC chemokines might be predicted to preferentially utilize the minor pocket, whereas CXC chemokines would be able to take advantage of the major binding pocket by redirecting their N-termini via the CXC bend. Despite possessing a distinct “bulge” at its CX3C motif, The N-terminus of CX3CL1 inserts into US28’s minor subpocket [9, 38]. Nevertheless, the drastic deviations in chemokine orientations from those seen in recent structures and models suggests that subpocket preference may be more complicated than can be predicted by the CC/CXC/CX3C motif [8, 13]. More structures of chemokine-CKR complexes will be needed to see if the cysteine-motif elicits subpocket binding preferences.

3.3. Role of binding depth and chemokine N-terminal length in receptor activation

In addition to the ligand’s “choices” to 1) specify a binding pocket, and 2) stabilize subpocket residues, a ligand may also “choose” to bind at a particular depth within that pocket. While four of five CKR-ligand complexes bind high in the orthosteric-binding pocket relative to other 7TMRs, maraviroc binds CCR5 at a depth resembling that of many aminergic ligands [8–10, 12]. A review of mutagenic and functional studies suggests diversity in the depths at which different chemokines contact their respective receptors, with some N-termini potentially achieving depths comparable to those of deep-binding aminergic ligands [7]. Additional complex structures will be needed to validate that chemokines may contact receptors at different depths within the TM domain, however current data suggest that depth variation presents yet another level of complexity within site 2 manipulated by chemokines to achieve specific signaling outcomes.

Chemokine N-terminal length does not necessarily correlate with the chemokine’s binding depth, or its functional properties. Early chemokine structure-function studies showed that truncation of the chemokine N-terminus transforms chemokine agonists into antagonists [20, 76, 77]. Recent studies of the CCL5 N-terminus demonstrate that extension of the chemokine N-terminus produces variant-specific functional outcomes, such as receptor internalization, degradation, recycling, or biased signaling [54, 77]. Similar approaches have since been applied to other chemokines [77]. A recent study by Hanes and colleagues utilized phage display and modeling to suggest how N-terminal length influences receptor function [78]. The authors screened two phage display libraries of CXCL12 for CXCR4 antagonists: a “N-addition” library with a single amino acid addition to CXCL12 and the first four residues of the lengthened chemokine randomized, and a “N-truncation library,” where the first four residues were deleted and residues 5–8 randomized. Two results were conclusively found: 1) the N-addition library produced more antagonists, whereas the N-truncation library produced none, and 2) of the N-addition antagonists found, many bound with greater affinity than WT-CXCL12. Interestingly, the screen selected for a subset of variants possessing neutral polar and aliphatic residues, independent of amino acid sequence. The authors propose that despite the “scrambled sequence,” similar intermolecular contacts form due to the conformational dynamics of the chemokine N-terminus and receptor pocket. These results are consistent with recent NMR studies of the MOR peptide agonist dynorphin, which was highly dynamic even in a receptor-bound state [78, 79]. In sum, these studies suggest that it may be difficult to make generalizations with respect to chemokine N-terminal length as it relates to CKR activation, as examples of elongated and shortened chemokine variants demonstrate diverse outcomes. Moreover, the dynamic nature of the chemokine N-terminus suggests that elongated peptides may adopt a more folded structure in the orthosteric pocket, as opposed to “diving” more deeply into the TM bundle [78]. Indeed, comparison of the vMIP-II:CXCR4 structure and our CXCL12:CXCR4 model suggests that despite vMIP-II’s two additional N-terminal residues, both chemokines reach the same depth by virtue of vMIP-II forming a short N-terminal helix [13].

4. Beyond Site 1.5

4.1. Defining unique, non-site 1, non-site 2 interactions at the receptor surface

The first chemokine-CKR crystal structure showed that in contrast to recognizing spatially distinct receptor domains, the chemokine formed interactions spanning from the receptor N-terminus (i.e. site 1) to the receptor TM domain (i.e. site 2) [8]. Noting a region that lacked precedence as either site 1 or site 2, Qin and colleagues named an interaction between the chemokine’s CC motif and the receptor N-terminal base chemokine recognition site 1.5 (CRS1.5) (Fig. 2.A) [8]. Similar interactions were observed in the CX3CL1:US28 structure, confirming previous predictions that the N-terminal stalk region (in the context of its disulfide interaction with TM7) serves a direct and essential role in chemokine recognition [9, 80]. More recently we identified analogous sites (i.e. CRS1.5-like) in our hybrid CXCL12:CXCR4 model and experimentally validated their role in binding and activation [13]. The existence of multiple structurally-validated intermediate interfaces calls into question the assumed spatial and functional separation between sites 1 and 2, suggesting that other interactions may be overlooked by the two-site model. This section will highlight these and other intermediate chemokine-CKR interactions that do not fall into traditional spatial designations of sites 1 or 2, and will speculate on the functional implications of these interactions.

• Diverse chemokine orientations: The most striking difference between the vMIP-II:CXCR4 and CX3CL1:US28 structures is the substantial deviation in chemokine orientation relative to the orthosteric pocket of the two receptors [8, 9]. Specifically, vMIP-II and CX3CL1 are rotated ~35° about the C-terminal ends of their C-terminal α-helices (Fig. 3A, B). The CXCL12:CXCR4 hybrid model diverges even more drastically, with CXCL12 rotated ~80° relative to vMIP-II [13]. The CXCL12:CXCR4 model was produced through a combination of rigid body docking and computational relaxation such that the backbone, side chain and rigid body positions were optimized. Extensive hydrophobic, polar and charged interactions spanning nearly half of CXCL12’s total surface area supports the plausibility of a substantial deviation in orientation relative to vMIP-II. Importantly, variation in chemokine orientation allows these ligands to form unique but overlapping subsets of interactions with receptor ECLs and TM domains.

• ECL2 links a ‘disulfide network’: The orientations of CX3CL1 and vMIP-II relative to their receptor binding pockets demonstrates that CX3CL1, but not vMIP-II, is ideally positioned to interact with ECL2 (Fig. 3A, B). CX3CL1 forms multiple interactions between its 30s-loop and ECL2 of US28, which, intriguingly, completes a disulfide network spanning from TM3 to TMs 1 and 7. In addition to a family A-conserved disulfide bond between ECL2 and TM3, most CKRs possess a disulfide that connects the N-terminal stalk with TM7 to form an additional ECL, termed “ECL4” (reviewed in 52) [80]. Owing to these two CKR disulfides, rigid body motions of receptor TM domains elicited by a chemokine at one site (e.g. N-loop interactions with TM7) could be efficiently communicated to a distant receptor domain (e.g. TM3) via a third interface (i.e. ECL2-30s-loop interactions), as suggested by Rajagopalan and colleagues [6]. In short, this structure may provide a glimpse of how multiple-site coupling at extracellular domains might influence chemokine receptor conformation via lateral (through-chemokine) allostery.

• Chemokine loop engagement: The two chemokine-CKR co-crystal structures exhibit an extended TM interface (relative to other non-CKR family A 7TMRs), formed by 1–2 additional α-helical loops at the extracellular portions of TMs 1 and 7, and the disulfide bond linking TM7 to the receptor N-terminus (Fig. 2.A) [8, 9]. This extended TM1–TM7 interface may also be important for chemokine binding and orientation, analogous to the 30s-loop-ECL2 interactions of CX3CL1 and US28. For instance, Leu13 in the vMIP-II N-loop forms contacts, albeit weakly, with the extended TM1–TM7 interface (i.e. residues Cys274^7.25 and Glu277^7.28), as well as the closely positioned Gly273^ECL3. These interactions in turn may contribute to the rotation of vMIP-II relative to CX3CL1, causing the vMIP-II 30s-loop to be spatially sequestered, preventing the formation of multiple interactions with CXCR4 (i.e. vMIP-II is 30s-loop deficient). In effect, these chemokine-CKR co-crystal structures suggest that chemokines utilize different loops to stabilize intermediate (i.e. non site 1/site 2) interactions with CKRs, thereby guiding chemokine orientation and likely influencing the signaling behavior of unique chemokine-CKR complexes.

• Diverse uses of ECL2: Some descriptions of the two site model categorize chemokine interactions with ECLs as site 1 interactions due to their contribution toward specificity and affinity, as well as their interactions with the chemokine body [3, 9]. Nevertheless, ECLs also interact with chemokine N-termini and strongly influence CKR activation, as was recently shown by Chevigné and colleagues at CXCR4 [6, 7, 81, 82]. Chemokine-CKR structures also support the resistance of ECL interactions to site 1 or site 2 classification: CXCR4 preferentially uses ECL2 to stabilize the vMIP-II N-terminus, whereas US28 preferentially uses ECL2 to stabilize the CX3CL1 30s loop, making few N-terminal contacts (Fig. 3A, B) [8, 9]. Evidently, each CKR utilizes ECL2 for different purposes, likely contributing to the unique binding modes of the respective chemokines.

Fig. 3.

Unique intermediate interactions specify chemokine orientation in two chemokine-CKR complexes. (A) vMIPII binds CXCR4 such that its 30s-loop is sequestered from forming extensive interactions with extracellular or TM domains of CXCR4 (PDB ID 4RWS). ECL2 of CXCR4 interacts with the vMIPII N-terminus, boxed. (B) Unlike the vMIPII, CX3CL1 forms extensive intermediate interactions with US28 (PDB1D 4TX1) using its 30s-loop, such that CX3CL1 is stabilized extensively at two separate receptor sites: the N-terminus (i.e., site 1) and ECL2. These extensive interactions may help stabilize US28 in an active state by simultaneously drawing together extracellular domains of US28, aided by a disulfide network comprised of receptor (i.e., N-terminus-TM7 or “ECL4” and TM3-ECL2) and chemokine (CX3C-30s-loop and CX3C-N-loop) disulfides. The active-state US28 structure is overlaid with an inactive-state CXCR4 structure (PDB ID 4RWS) to demonstrate how this disulfide network could facilitate receptor activation (arrows). Unlike the vMIP-II:CXCR4 structure, ECL2 of US28 primarily stabilizes interactions with the 30s-loop of CX3CL1, boxed.

These structural examples illustrate that in some instances, spatial delineation of site 1 and site 2 may be artificial. Moreover, each chemokine-CKR complex utilizes distinct combinations of structural domains to specify unique functional outcomes. Having now seen the first comprehensive structural evidence of multiple intermediate, non-site 1/2 interactions, we will consider how receptor ECLs take on functional characteristics of sites 1 and 2 alike.

4.2. A “multi-site model” accounts for diverse, interdependent chemokine-CKR interactions

While we are only now beginning to appreciate the extent and diversity of chemokine-CKR interactions following high-resolution structural data, pharmacological evidence predating the two-site model supported a more complex “multi-site model” of CKR activation [83, 84]. Studies of CXCR1 and CXCR2 in the 1990s established a number of important principles concerning CKR recognition and activation as they relate to the receptor extracellular surface, including: 1) different chemokines utilize unique combinations of extracellular domains for the binding and activation of a single CKR [85, 86], 2) a single chemokine may utilize unique combinations of extracellular domains when binding different CKRs [85], and more generally 3) CKR binding and activation is a consequence of multiple, interdependent variables, particularly the identity of the chemokine and the simultaneous interactions it makes with all adjacent extracellular domains (i.e. N-terminus, ECLs 1–3) (Fig. 2A, B) [83–86]. Similar pharmacological and structural studies expanded these principles to other chemokine-CKR pairs, including CCR1 [81, 87, 88], CCR2 [18, 89, 90], CCR3 [91, 92], CCR5 [18, 88, 93–96], CXCR1 [97–99], CXCR2 [97, 98], CXCR3 [100], CXCR4 [82, 101], CX3CR1 [102].

CXCR3 provides an illustrative example of the interdependence of chemokine “multi-site” binding, chemokine preference for unique CKR conformations, and associated functional outcomes. Xanthou and colleagues disputed the universality of the two-site model for chemokine ligands, proposing instead a “multi-site model in which several distinct extracellular domains are required for efficient ligand binding and receptor activation” [103]. The authors created “gain-of-function” chimeras by individually swapping the N-terminus, ECL1, ECL2, and ECL3 of CXCR3 into CXCR1 (which does not share ligands with CXCR3), and vice versa to create “loss-of-function” chimeras. They found that while absence of ECL2 did not abolish chemokine binding, it completely attenuated CXCR3-mediated signaling, supporting a role beyond chemokine recognition. In addition, they showed that each chemokine required a unique subset of interactions to activate CXCR3: CXCL9 required ECL2 and ECL3; CXCL10 required all EC domains; and CXCL11 required the N-terminus, ECL1, and ECL2.

From these experiments it is clear that in addition to possessing spatial characteristics of sites 1 and 2, ECL2 may in some instances possess functional characteristics of sites 1 and 2. Moreover, this study suggests a “multi-site” mechanism by which different chemokines could, in principle, elicit functionally distinct downstream outcomes. Two lines of evidence support such a mechanism. A recent study demonstrated that the three CXCR3 ligands elicit unique patterns of Gα_i activation, β-arrestin recruitment, and internalization following CXCR3 stimulation [5]. Secondly, CXCL10 preferentially binds inactive conformations of CXCR3, whereas CXCL11 binds both inactive and active conformations [104]. Intriguingly, the distinct functional outcomes elicited by CXCL10 and CXCL11 downstream of CXCR3 can drive opposing cellular and physiologic consequences. For instance, CXCL11 signaling promotes Th1 cell polarization into an immunotolerant T cell subset, resulting in an anti-inflammatory phenotype in a mouse model of multiple sclerosis, whereas CXCL10 promotes pro-inflammatory effects [105].

In all, these data show that chemokines can preferentially bind unique subsets of EC domains and/or available receptor conformations to elicit specific functional outcomes. By illustrating the interdependence of multiple extracellular domains on CKR signal transduction, these studies undermine the functional separation of sites 1 and 2 into recognition and activation, and imply that cooperative interactions influence chemokine recognition at extracellular CKR domains and subsequent functional outcomes. Moreover, unique interactions between each chemokine-CKR pair suggest therapeutic approaches that might target non-overlapping chemokine binding sites to selectively disrupt one chemokine-CKR interaction while maintaining another. In fact, several such selective allosteric modulators have been described for CCR1 and CCR5, demonstrating the feasibility of this approach for the development of highly targeted CKR therapies [3, 106–108].

4.3. Molecular switches at the extracellular surface: allosteric coupling to the TM region

As suggested in the previous section, the role of ECLs is more nuanced than static stabilization of chemokine ligands. Dynamic interactions between ECL and TM residues may act as molecular switches regulating receptor activation, especially in the case of ECL2. While ECL2 displays high variability in sequence, length, and structure even among related receptor subtypes, an impressive number of family A receptors seem to utilize ECL2 in this capacity, including rhodopsin [109–112], the serotonin 5-HT4 receptor [113], the V(1a) vasopressin receptor (V1aR) [114], the cannabinoid receptor 1 (CB1) [115–117], the β₂AR [118], the angiotensin II type 1a receptor (ATII1aR) [119, 120], the D2 dopamine receptor (D2R) [121], the C5a complement receptor (C5aR) [122], and protease activated receptor 1 (PAR-1) [123], among others [124, 125]. Considering CXCR4, we recently suggested that manipulation of ECL2 by CXCL12 draws ECL2 toward the orthosteric pocket, thereby moving TMs 2 and 3 closer to one another [13]. These TM movements may then help initiate receptor activation, in an analogous mechanism to that predicted for CCR5 [126].

In addition to ECL2, ECLs 3 and 4 have been suggested to act as a tandem molecular switch required for CXCR4 activation [80, 101]. Using mutagenesis and a yeast-based Gα_i protein activation screen, Rana and colleagues showed that interaction between TM7 and the CXCR4 N-terminus is essential for receptor activation, and that replacement of the disulfide-bonded cysteines with an electrostatic pair (Arg-Glu) conserves CXCR4 signaling. The authors suggest a model in which the N-terminal-TM7 disulfide undergoes a conformational change during receptor activation that is transmitted to ECL3 and TM6. Comparison of the active state CX3CL1:US28 and inactive state vMIP-II:CXCR4 structures supports this model (Fig. 3.B) [8, 9]. Compared to CXCR4, US28 demonstrates an inward (i.e. toward the TM domain) motion of ECL4, seemingly driving an inward motion of ECL3 and the extracellular portion of TM6. In a well characterized mechanism, inward motion of the top of TM6 causes it to rotate about a conserved proline “kink,” resulting in substantial outward movement at the intracellular face to accommodate G protein binding [70].

Similar ECL4 motions may contribute to CXCR4 activation, as suggested by our recent CXCL12:CXCR4 model. Hydrogen bonding between the backbone carbonyl of CXCR4 Phe29^N-term, which is adjacent to the N-term-TM7 disulfide, and CXCL12 Ser6, “pulls” the extracellular domain of TM7 toward the orthosteric pocket, which in turn may facilitate rearrangement of residues in CXCR4’s TM domain during activation [13]. These examples suggest that chemokine binding to extracellular domains “primes” the receptor for activation by stabilizing an intermediate conformation, followed by chemokine N-terminal insertion into the receptor TM core and intracellular coupling of signaling effectors (i.e. G protein or β-arrestin). More structural studies will be needed to validate the role of ECL4 in the activation of CXCR4, US28, and expand its role to other CKRs. Nevertheless, it is becoming clear that chemokine-ECL interactions serve complex functional roles in CKR recognition and signaling.

5. Beyond canonical chemokine receptor signaling

5.1. Beyond 1:1 interactions: stoichiometry of chemokine-receptor interactions

Chemokine dimerization

Initially thought to be a crystallization artifact, chemokine dimerization has been reexamined over the past years in numerous structural and biochemical studies. It appears now that the vast majority of chemokines are able to form dimeric species, with the monomer-dimer equilibrium being regulated by factors such as pH, anions and interactions with glycosaminoglycans [127, 128].

Depending on the family, chemokines adopt two main oligomeric states with unique structural arrangements and interaction interfaces. CC chemokines form flexible and extended dimers mainly through residues surrounding the cysteine motif [129, 130], whereas CXC chemokines self-associate in more compact dimers via interactions involving the first β-strand [128, 131, 132]. In both types of interactions the N-termini of the two monomers are pointing in opposite directions. Chemokines of the C family, XCL1 and XCL2, have recently been shown to exist in a monomer-dimer equilibrium, unusually requiring complete protein unfolding. XCL dimers adopt a novel dimer conformation that also creates a six-stranded β-sheet [133, 134]. CX3CL1, the only member of the CX3C family, dimerizes in a similar manner to that of CC chemokines [135]. Additionally, some chemokines have been observed to form tetramers (e.g. CCL2, CCL5, CCL27, CXCL4, and CX3CL1) or higher-order oligomers [130, 135–139]. Heterodimers of two different CC or CXC chemokines as well as cross-family CC/CXC heterodimers have also been reported [140–142]. Furthermore, HMGB1 (High mobility group protein B1) protein was reported to form complexes with CXCL12 promoting different conformational rearrangements of CXCR4 from that of CXCL12 alone [143]. These findings further challenge the two-step binding model for chemokine-CKR interactions and complicate the question of which stoichiometries are capable of generating functional responses.

Immobilization of chemokines on glycosaminoglycans (GAGs) is an important step for chemokine function as it creates a gradient to direct cell migration and regulates the local chemokine concentration and availability for their receptors. Likewise, GAG binding can favor dimer formation as demonstrated for CCR2-binding chemokines [141], CXCL8 [144, 145], CXCL12 [146–149], XCL1 and XCL2 [133, 134]. Oligomerization has also been shown to increase GAG affinity by creating a more extensive surface for interactions [128].

The biological relevance of chemokine dimerization is still a matter of debate and its impact on receptor binding, stoichiometry and biased signaling remains to be unraveled [128, 132, 150, 151]. As an illustration it has been demonstrated that monomeric and dimeric CXCL12 induce different intracellular signaling responses and opposite effects on cell migration, but other recent studies suggested that this receptor interacts with CXCL12 in a 1:1 stoichiometry [36, 57].

Receptor dimerization

Throughout the past two decades, it has been assumed that CKRs exist as monomers, which behave as fully competent signaling units. This assumption, in part, forms the basis of the classical two-site binding model. However, a number of studies demonstrated that CKRs can form homodimers and/or heterodimers (Fig. 4A) [152]. CKR dimerization has been investigated by various biochemical approaches such as co-immunoprecipitation (co-IP) [153–156], protein fragment complementation (PFC) [157, 158], Förster/Bioluminescence Resonance Energy Transfer (FRET/BRET) [159–161] and GPCR Heteromer Identification Technique (GPCR HIT) [162, 163]. The first structural evidence of CKR dimerization however was provided by the first inactive-state crystal structures of CXCR4 in which the receptor was present as a dimer with the interface between the subunits located at the top of TM5 and TM6 and stabilized by hydrogen bonds [10]. CKRs from all four subfamilies (C, CC, CXC, CX3C) have now been described to form homo- or heterodimers in vitro [154, 164–166] and some of them, including CXCR4 and CCR5, were shown to interact with other families of GPCRs such as the α1A/B-adrenergic receptors [167], opioid receptors [168] or non-GPCR membrane proteins that modulate the activity of the receptor or act as coreceptor for certain non-conventional ligands (Fig. 4A) [169]. Receptor dimerization has been shown to modify ligand binding properties [155, 170] and receptor signaling [153, 167, 171, 172] as well as intracellular trafficking [158]. However, so far there is no in vivo data reporting the existence of CKR dimers and therefore their biological relevance remains controversial [152, 173].

Fig. 4.

CKR homo- and heterodimerization, models of chemokine-CKR stoichiometry, and non-cannonical CKR ligands. (A) Interactions between receptors are represented by dots. CC, CXC and atypical chemokine receptor subfamilies are represented in yellow, blue and green, respectively. Non-CKR GPCRs and non-GPCRs are represented in black and orange, respectively. Homodimers are indicated with one-color dots and heterodimers between receptors from different families by two-color dots (monomer 1/ monomer 2). (B) 1:2 stoichiometry in which a chemokine monomer binds a receptor dimer. The disulfide bridges between N-terminus/ECL3 and ECL2/TM3 are depicted as red lines. (C) 2:1 stoichiometry in which a chemokine dimer binds a receptor monomer. (D) 2:2 stoichiometry in which a receptor dimer binds a chemokines dimer. (E) 1:2* stoichiometry in which a receptor dimer interacts with a monomeric chemokine. Upon the binding of the chemokine to one monomer (receptor 1, blue), the conformational changes induced in receptor 1 are propagated to receptor 2 (green) through the dimer interface, activating receptor 2 without the need of chemokine binding. (F) Binding and activation of chemokine receptors CCR5 or CXCR4 by the HIV gp120 envelope protein require another membrane protein, CD4 (red), which acts as a primary receptor for gp120, inducing conformational rearrangements to expose the third variable loop (V3 loop). (G) Binding and activation of CXCR2, CXCR4 and CXCR7 by the pseudo-chemokine MIF also requires the presence of a primary receptor, CD74 (yellow). Like chemokine-CKR interactions, the stoichiometry of MIF:CD74:CKRs is not well established.

Stoichiometry of chemokine-receptor complex

Another poorly understood and highly debated facet of chemokine-CKR interactions is their stoichiometry in functional signaling complexes. As both chemokines and receptors can homo- and heterodimerize, novel hypotheses around the stoichiometry of their interactions have emerged, leading to more complex models than the initially proposed two-step/two-site model. Among them, the 1:2 stoichiometry model where one chemokine binds two receptors simultaneously (Fig. 4B), the 2:1 stoichiometry model in which a chemokine dimer binds one receptor (Fig. 4C), and finally, the 2:2 stoichiometry model in which both the chemokine and the receptor interact as dimers (Fig. 4D) [8, 21, 57, 174]. Complementation studies carried out with CXCR4 mutants partially deficient in site 1 (i.e. CRS1) or site 2 (i.e. CRS2) were inconsistent with a 1:2 stoichiometry model and supported CXCR4 monomers as fully competent signaling units [57]. These results were later supported by the crystal structure resolution of the viral chemokine vMIP-II in complex with CXCR4 and CX3CL1 in complex with US28, both revealing a 1:1 stoichiometry interaction and an extensive contact surface between the two partners [8, 9]. However, more recent investigation of the preferential binding of monomeric CXCL12 to either monomeric (1:1) or dimeric (1:2) CXCR4 by molecular dynamics simulations proposed that in the 1:2 stoichiometry model, the N-terminus of the chemokine could make more tight contacts with the CRS2 of the second monomer to more efficiently favor signaling than in the 1:1 stoichiometry [174]. Finally, studies of CXCR4:ACKR3 heterodimers suggest that upon CXCL11 binding to CXCR7, conformational changes propagate through the dimer interface activating CXCR4 without the need of its own ligand (1:2*stoichiometry) (Fig. 4E) [172]. Gathering structural and mechanistic information on receptor dimerization, chemokine/receptor stoichiometry and relating it to functional observations remains challenging and necessitates state-of-the-art techniques to strengthen or to invalidate the currently accepted but oversimplified two-step/two-site model [9, 21, 57].

5.2. Beyond G protein signaling: biased signaling at chemokine receptors

Once proposed to serve redundant signaling and functional roles, promiscuous chemokine-CKR interactions are widely believed to confer signaling and functional complexity to the CKR axis [1, 4–6]. Individual chemokines regulate multiple essential functions via independent CKR interactions. In turn, the consequences of each unique interaction depends upon concurrent spatial and temporal expression of both partners [175]. To complicate matters, CKRs were until recently accepted to signal exclusively through canonical G protein pathways and to couple exclusively to the Gα_i/o G protein subtype. However, mounting evidence shows that some CKRs may also signal through other G protein subtypes (Gα_s, Gα_q/11 or Gα_12/13) and activate G protein-independent signaling cascades (e.g. via β-arrestin) in ligand- and cell-specific contexts [176–178]. Analogous findings have been described for countless non-CKR 7TMRs for over a decade, signifying the new paradigm known as biased signaling or functional selectivity [179, 180]. Biased signaling appears to be ubiquitous among CKRs, and it provides a framework for understanding how a finite cast of ligands generates myriad functionally diverse outcomes. Biased signaling has been subdivided into three categories: ligand bias, receptor bias and cellular or tissue bias [4, 5, 181, 182].

Chemokine ligand bias occurs when different chemokines bind the same receptor to elicit distinct cellular responses. Ligand bias has been well documented for both CC and CXC chemokines, including CCL19 and CCL21 at CCR7 [177, 183, 184], CCL27 and CCL28 at CCR10 [5], three chemokine ligands at CXCR3 [5, 184], CXCL7 and CXCL8 at CXCR2 [185], as well as for chemokine ligands at CCR1 [5, 186], CCR2 [187] and CCR4 [188, 189]. Intriguingly, biased responses may be elicited by chemokines bearing unique PTMs, including truncation, citrullination or dimerization as reported for CCL14 [190] and CXCL12 [36]. The characteristic promiscuity of the chemokine-CKR network and poor sequence identity among chemokines, may partly explain the pervasiveness of ligand bias among chemokine ligands [5]. Indeed, variation of the structural interactions discussed in the article (summarized in Fig. 2B) such as chemokine orientation, ECL contacts, and major/minor subpocket selection likely stabilizes distinct active forms of the receptor, eliciting preferential coupling to different intracellular effectors [191]. While we are far from defining the precise structural mechanisms underpinning biased signaling, studies of other family A 7TMRs suggest a role for helical movements of and direct physical interactions with TMs 5 and 6 for G protein and TM7 for β-arrestin coupling, respectively [4, 70, 191–194].

In a peculiar twist, a chemokine agonist at one receptor can antagonize another receptor. Chemokines activating CXCR3 can also bind CCR3, blocking CCL11-induced cell migration and G protein signaling [195]. Similarly, CXCL11 and CCL7, well-characterized agonists at CXCR3 and CCR1/CCR2, respectively, were found to be antagonists at CCR5 [196–198]. In another example of dual activity chemokines, vMIP-II binds human and viral CKRs across all four families, acting as an antagonist or an agonist at different receptors [199–201]. Dual activity is often observed in cross-family interactions and could relate to differences in chemokine N-terminal orientation among CC and CXC chemokines (discussed in Section 3.2) [8]. Nevertheless, antagonism within the same family has also been reported, suggesting that other determinants may also play a role [198, 202].

Receptor bias occurs when a particular receptor preferentially or exclusively couples to a particular effector even in the context of multiple different ligands. Receptor bias has been well- characterized at atypical and viral CKRs [203]. For instance, CXCL12 binding to CXCR4 elicits both G protein and β-arrestin signaling [176, 204], whereas binding to ACKR3 (a.k.a. CXCR7) elicits G protein-independent β-arrestin signaling exclusively [205].

Cellular or tissue bias occurs when the same chemokine-CKR triggers distinct signaling pathways or cellular responses in different cellular contexts. For instance, CCL19 binding to CCR7 induces chemotaxis only in certain cell types [206, 207]. Such cellular bias is unsurprising considering the large variety of cells expressing CKRs, each of which carry unique expression profiles of signaling effectors (e.g. G protein subtypes and β-arrestin isoforms), receptor modifying enzymes (e.g. GRKs, TPSTs), as well as of other CKRs or receptor modulating partners involved in dimeric receptor interactions.

5.3. Beyond chemokines: binding and signaling by non-chemokine ligands

The chemokine receptors CCR2, CCR3, CCR5, CXCR2, CXCR4 and ACKR3 bind endogenous or virus-encoded ligands other than chemokines. These unconventional ligands vary widely in size, ranging from large proteins (e.g. > 100 kDa) to peptides, and often have no sequence or structural similarities with chemokines [15, 169, 208, 209]. Despite their structural dissimilarities, non-chemokine CKR ligands can trigger signaling pathways similar to those induced by endogenous chemokines, although in some cases they initiate unconventional signaling responses [15, 169, 208–212]. For some, binding and signaling relies on the CKR alone [15], while for others, the CKR operates in tandem with another membrane protein that usually serves as primary receptor [210, 213].

One of the best-known examples of non-chemokine CKR ligands is the HIV envelope protein gp120 (120 kDa), which uses CCR5 and CXCR4 as co-receptors for cell type-specific recognition and entry into host cells. To initiate viral membrane fusion with host cell membranes, gp120 first binds to CD4, a primary single TM segment receptor, initiating conformational changes (Fig. 4F). These changes then expose gp120’s third variable loop (i.e. V3 loop), which in turn interacts with CCR5 or CXCR4 [213–215]. The second interaction was suggested to occur in a two-site binding mechanism similar to that initially proposed for chemokines, with common interacting determinants [211]. Importantly, not only is the interaction of gp120 with CXCR4 or CCR5 required for cell-specific HIV entry but it also leads to the activation of signaling pathways such as JNK and MAPKs, facilitating the early steps of viral replication [216, 217]. Tat, the HIV-trans-activating protein (14 kDa) released extracellularly by infected cells, triggers G protein-mediated signaling and chemotaxis through CCR2 and CCR3 [218, 219] and acts as an antagonist of CXCR4 [220]. Similarly, the HIV-1 matrix protein p17 binds CXCR1 and CXCR2, inducing chemokine-like activity on monocytes through Rho/ROCK activation [221, 222].

More recently, the pseudo-chemokine MIF (macrophage migration inhibitory factor), a pleiotropic and proinflammatory chemotactic cytokine of 12.3 kDa highly expressed by tumor cells, has been identified as a ligand for CXCR2 [209], CXCR4 [169] and ACKR3 [210], inducing ERK1/2 and ZAP-70 signaling and chemotaxis. As with gp120, the binding of MIF to CKRs requires a primary receptor, CD74, a single segment membrane-spanning protein also known as HLA class II histocompatibility antigen gamma chain (Fig 4G). Although MIF possesses some chemokine-like features, including a pseudo-ELR motif (D45-X-R129) and an N-loop-like region (amino acids 48–57), it lacks the canonical cysteine motif and is therefore classified among the chemokine-like function (CLF) chemokines.

In addition to CXCL12, CXCR4 also binds extracellular ubiquitin (eUb, 8.6 kDa). The eUb-CXCR4 interaction was proposed to follow a two-site binding mode, leading to G protein signaling similar to that induced by CXCL12 [15]. Other endogenous non-chemokine ligands such as human β3-defensin (HDB-3) (5.1 kDa) [223] and EPI-X4 (1.8 kDa) a 16-amino acid peptide derived from human albumin [224] also interact with CXCR4 but fail to induce intracellular signaling. Finally, human cytosolic proteins such as histidyl- and asparginyl-tRNA synthetases, released in some inflammatory pathologies, were shown to induce leukocyte migration through CCR3 and CCR5 [225]. Similar results were reported for parasitic asparginyl-tRNA synthetases which act as agonists for CXCR1 and CXCR2 [226].

The identification of non-cognate ligands for CKRs, some exclusive to a single receptor, others interacting with several receptors across several subfamilies, further emphasizes the complexity of the chemokine receptor network, which seems now more promiscuous and predisposed to bias than initially thought. These new ligands will certainly help to uncover other important physiological and pathological functions for this family of receptors, explain past observations and provide new therapeutic opportunities to modulate chemokine/CKR activity.

6. Conclusion

The two-site model has served as a valuable conceptual framework in which to understand chemokine-CKR signal transduction over the past twenty years. Nevertheless, a growing body of evidence supports a more complex model regulating how chemokines interact with their receptors to mediate an increasingly diverse complement of outcomes. The first structures depicting the complete extracellular and TM chemokine-CKR interfaces identified new interactions resisting site 1 and site 2 categorization. In parallel, advances in our understanding of 7TMR structure, dynamics, and activation are helping to define mechanisms by which chemokine binding is translated into receptor activation. Finally, paradigm shifts from within and outside of the chemokine realm are altering how we understand the complexity of the chemokine system. A more complete, “multi-site model” [103] of how chemokine-CKR interactions elicit functional outcomes will require active-state CKR structures and complementary studies probing the dynamics and functional specificity of unique ligand-receptor pairings.