Biased Agonism at Chemokine Receptors

Abstract

In the human chemokine system, interactions between the approximately 50 known endogenous chemokine ligands and 20 known chemokine receptors (CKRs) regulate a wide range of cellular functions and biological processes including immune cell activation and homeostasis, development, angiogenesis, and neuromodulation. CKRs are a family of G protein-coupled receptors (GPCR), which represent the most common and versatile class of receptors in the human genome and the targets of approximately one third of all Food and Drug Administration-approved drugs. Chemokines and CKRs bind with significant promiscuity, as most CKRs can be activated by multiple chemokines and most chemokines can activate multiple CKRs. While these ligand-receptor interactions were previously regarded as redundant, it is now appreciated that many chemokine:CKR interactions display biased agonism, the phenomenon in which different ligands binding to the same receptor signal through different pathways with different efficacies, leading to distinct biological effects. Notably, these biased responses can be modulated through changes in ligand, receptor, and or the specific cellular context (system). In this review, we explore the biochemical mechanisms, functional consequences, and therapeutic potential of biased agonism in the chemokine system. An enhanced understanding of biased agonism in the chemokine system may prove transformative in the understanding of the mechanisms and consequences of biased signaling across all GPCR subtypes and aid in the development of biased pharmaceuticals with increased therapeutic efficacy and safer side effect profiles.

Chemokine System

Chemokine receptors (CKRs) are a subfamily of G protein-coupled receptors (GPCRs) that bind a group of small (8–12 kDa) and highly conserved chemotactic cytokines known as chemokines (1). The human chemokine system is composed of approximately 20 known CKRs and 50 known chemokines (Figure 1). The chemokines are classified into four subtypes (C, CC, CXC, CX₃C) based on the number, positioning, and spacing of conserved N-terminal cysteine residues (2). Similarly, CKRs are organized and classified according to the ligands they bind (3). Chemokines are also categorized as homeostatic chemokines, which are constitutively expressed in a variety of specific tissues and cell types, and inflammatory chemokines, which are induced during immune responses primarily to recruit leukocytes to sites of inflammation (4). Homeostatic and inflammatory classifications of chemokines are not mutually exclusive, however, as some CKRs and chemokine ligands are involved in both normal and pathologic conditions. The role of CKR activation by chemokines was first recognized in the immune response, specifically as chemoattractants to direct leukocyte migration, a process known as chemotaxis (5, 6). While the functions and roles of chemokines in leukocytes are well known, it is now appreciated that chemokines and CKRs are also produced in a variety of non-leukocyte cell types, including epithelial cells, fibroblasts, endothelial cells, and neurons (7), and play a key role in a wide range of other cellular functions and biological processes including development, angiogenesis, neuromodulation, and immune cell homeostasis (8–11). For example, the expression of neuronal chemokine ligands and receptors has recently been shown to be involved in synaptic transmission and neuronal survival (12), as well as in guidance of central nervous system (CNS) cellular interactions via neuron-astrocyte, neuron-microglia, and neuron-neuron interactions (13).

FIGURE 1: The complexity of the human chemokine system.

Chemokine receptors fall into five categories: CCRs, CXCRs, ACKRs, XCRs and CX₃CRs. Chemokine ligands fall into four categories: CCLs, CXCLs, XCLs, CX₃CLs. Lines connecting chemokine receptors to chemokines are colored for clarity.

Due to the chemokine system’s involvement in a wide variety of biological processes, it is unsurprising that chemokines and CKRs are implicated in various disease states including, but not limited to, autoimmune disorders, infectious diseases, hypersensitivity reactions, atherosclerosis, and cancer (14–18). The role of the chemokine system in chronic inflammatory diseases is particularly important and chemokines play a central role in asthma, chronic obstructive pulmonary disease, inflammatory bowel disease, arthritis, multiple sclerosis, and psoriasis (19). Additionally, certain disorders are directly associated with mutations in the genes that encode CKRs, such as the Warts, Hypogammaglobulinemia, Immunodeficiency, and Myelokathexis (WHIM) Syndrome which is driven by an autosomal dominant truncation mutant in the receptor CXCR4 (20).

While the chemokine system is known to play a significant role in many disease states, there are relatively few drugs that target it directly. Chemokines and CKRs bind with significant promiscuity, wherein most CKRs can be activated by multiple chemokine ligands and most chemokines can activate multiple CKRs (21). This promiscuity was thought to lead to “redundancy” between chemokines and their receptors, serving as a mechanism for a robust physiologic response (22). As adequate chemokine levels are imperative for immune cell function, redundant chemokine signaling would provide sufficient signals to direct leukocyte chemotaxis and function that is relatively insensitive to variations in the concentration of any individual chemokine (23). However, we now appreciate that many of these ligands can have distinct signaling profiles at the same receptor and many receptors can have distinct signaling profiles when stimulated by the same ligand, a phenomenon referred to as “biased agonism” (22, 24). Biased signaling through differences in ligands, receptors and the cellular context (system) can have important effects on chemokine signaling and implications for drug development. Here, we review the current literature on biased signaling within the chemokine system to highlight the complex and multidimensional nature of biased agonism and the biochemical mechanisms that underlie it, the importance of biased agonism within the chemokine system and its physiologic effects, and the implications of biased signaling for drug development in the chemokine system.

G Protein-Coupled Receptor Signaling and Biased Agonism

CKRs are a subfamily of the rhodopsin class of GPCRs, the most common and versatile superfamily of receptors in the human genome (25) and the target of ~34% of all Food and Drug Administration (FDA) approved pharmaceutical drugs (26). Canonical GPCR signaling starts with agonist binding, upon which a GPCR undergoes conformational changes that induce the recruitment of heterotrimeric G proteins consisting of Gα, Gβ, and Gγ subunits. The guanosine diphosphate (GDP)-bound Gα subunit undergoes nucleotide exchange for guanosine triphosphate (GTP), leading to Gα activation and dissociation of the heterotrimeric complex into its Gα and Gβγ constituents (27). The Gα subunits are classified into four families based on sequence similarity: Gα_s, Gα_i/o, Gα_q/11, and Gα_12/13 (28). The activated Gα proteins typically regulate the production and subsequent signaling of secondary messengers, such as adenosine 3’,5’-cyclic monophosphate (cAMP), intracellular calcium, and inositol triphosphate (29). Most chemokine receptors signal through Gα_i/o, which inhibits adenylyl cyclase and reduces intracellular concentrations of cAMP (30). There are various isoforms of the Gβ and Gγ subunits, and at chemokine receptors the Gβγ dimer has been shown to activate phosphoinositide-specific phospholipase Cβ (PLC) and phosphoinositide 3-kinase (PI3K). PLC then produces diacylglycerol (DAG), leading to the activation of protein kinase C (PKC), and inositol-triphosphate (IP₃), which triggers calcium mobilization (31). The signaling messengers of the Gβγ dimer in chemokine receptors have been demonstrated to play a role in the promotion of leukocyte migration, among other functions (32).

Following G protein activation, G protein-coupled receptor kinases (GRKs) are recruited to the receptor and phosphorylate the receptor C-terminus and intracellular loops. This phosphorylation promotes the interaction of the receptor with the β-arrestins, which were first described for their function in the desensitization of G protein-mediated signaling through steric hindrance of the receptor (33). β-arrestins act as multifunctional adaptor proteins and are involved in a variety of other cellular process such as receptor internalization, transactivation, trafficking, and signaling (34). β-arrestins are capable of signaling through a variety of mediators including mitogen-activated protein kinases (MAPKs), nuclear factor κB (NF-κB), protein kinase B (Akt), and Src tyrosine kinase, thereby activating distinct signaling pathways independent of G protein-mediated signaling (35–39). At some chemokine receptors, such as CXCR4 and CCR5, it has even been shown that in the absence of G protein coupling, β-arrestin-bound CKR complexes are still able to induce receptor endocytosis and signaling (40).

Biased Chemokine Signaling

With the ability of GPCRs to signal through multiple pathways such as different G proteins or β-arrestins, referred to as “pluridimensional efficacy” (41), it was soon discovered that many agonists were capable of signaling with different efficacies to their downstream effectors (42). Initially these examples were thought to be rare, but studies over the past fifteen years have demonstrated that, across a wide variety of GPCR subtypes, different ligands for the same receptor can activate specific and distinct signaling pathways downstream. These examples can range from mild bias, with slight differences in preferences for different G proteins, to completely biased ligands which preferentially activate G protein signaling pathways (G protein-biased) while not activating β-arrestin signaling (β-arrestin-biased) pathways, or vice versa (Figure 2A–C).

FIGURE 2: Biased agonism at G Protein-Coupled Receptors.

Signaling at G Protein-Coupled Receptors can be driven by ligands that are (A) balanced and equally activate both G protein and β-arrestins, (B) those that signal primarily through heterotrimeric G proteins (G protein-biased), or (C) those that signal primarily through β-arrestin adapter proteins (β-arrestin-biased). Biased agonism can be achieved through (D) biased ligands, (E) biased receptors, or (F) biased systems

Biased agonism within the chemokine subfamily was first described at CCR7, where it was demonstrated that the receptor’s two endogenous ligands—CCL19 and CCL21—both could activate G protein-dependent signaling cascades but only CCL19 was capable of activating β-arrestin dependent signaling (43–46). Since then, numerous studies have demonstrated that the chemokine system is not redundant, but rather highly specific in its cellular outputs, part of which can be attributed to biased agonism (39). Functional and structural demonstrations of bias in chemokine signaling provide support for the notion that the promiscuity of the chemokine system is not redundant, but suggests high levels of signaling specificity by allowing individual chemokines and CKRs to induce functionally distinct biochemical, cellular, and physiologic outcomes (24). Biased agonism in the chemokine system can be a result of signaling through each component of the GPCR ternary complex: the ligand, the receptor, and the transducer (“system”) elements (Figure 2D–F). With ligand bias, different ligands binding to the same receptor generate distinct responses. With receptor bias, the same ligand binding to different receptors generates distinct responses. With system bias, a different cellular context, e.g., differential transducer expression in a different cell type, results in a change in GPCR signaling. While the exact mechanisms and functional consequences of these layers of bias are incompletely understood, it is possible that this complexity allows for highly specific and coordinated physiologic responses. Additionally, the interplay between these individual sources of biased agonism can theoretically generate a multitude of permutations of immunophenotypes.

Ligand Bias

Ligand bias is arguably the most widely studied mechanism of bias in the chemokine system. Ligand bias refers to the phenomenon in which a molecule binds to a receptor and promotes an ensemble of receptor conformations that preferentially interact with some transducers over others (39). For example, the two endogenous ligands of CCR7, CCL19 and CCL21, induce G protein activation and signaling to similar extents, but CCL19 induces β-arrestin recruitment to CCR7 significantly more than CCL21 (44–46). Additionally, receptor internalization and chemotaxis induced by CCL19, but not CCL21, are reduced following siRNA knock down of β-arrestin 2 and both β-arrestin 1 and β-arrestin 2, respectively (46). Other assays assessing β-arrestin trafficking, receptor desensitization, MAPK signaling, and receptor phosphorylation are significantly different following agonist stimulation with CCL19 or CCL21 (44), consistent with CCL19 acting as a β-arrestin biased ligand relative to CCL21. Similar to the ligands of CCR7, the endogenous ligands of CXCR3 also exhibit ligand bias. In previous work performed by our group, these endogenous ligands exhibit both varying levels and distinct patterns of interactions with β-arrestin (24). When stimulating CXCR3A, CXCL11 recruits β-arrestin at a significantly higher level compared to CXCL10, even though the two ligands activate Gα_i to similar levels (47). Furthermore, CXCL11 demonstrates a class B pattern of β-arrestin recruitment, one of which promotes a stable β-arrestin:GPCR complex with trafficking of β-arrestin into endosomes, while CXCL9 and CXCL10 demonstrate a class A pattern of recruitment, which promotes a transient β-arrestin:GPCR complex limited to the plasma membrane (24). Multiple chemokine receptors and their endogenous ligands demonstrate similar degrees of biased agonism, including CCR1, CCR5, CCR10, CXCR1, and CXCR2 (24, 43, 48–52).

Ligand bias has also been demonstrated to depend on the oligomeric state of chemokines. In vivo, chemokines homo-oligomerize and exist in an equilibrium between their monomeric and dimeric forms (53), with the monomer frequently serving as the active form that binds the receptor and the dimeric form bound to glycosaminoglycans (54). Changes in the overall oligomeric state of chemokines are critical for their cellular functions, reflected by the differential signaling profiles between their monomeric and dimeric forms (55–58). Drury et al. found that at CXCR4 the monomeric form of CXCL12 is β-arrestin biased relative to its dimeric form, and this bias mediates increased levels of chemotaxis and differential MAPK activation kinetics (55). Modulation of the monomer-dimer equilibrium in vivo may consequently serve as another method for a system to finely tune chemokine signaling. Another mechanism that can regulate chemokine activity is post-translational modification, such as proteolytic processing, citrullination and glycosylation (59). While these modifications have been shown to regulate chemokine activity, it is largely unknown at this time as to whether they have an effect on ligand bias.

Beyond simple G protein vs β-arrestin bias, there are examples of bias between specific G protein subtypes or β-arrestin isoforms (60, 61). Various analogs of CCL5 analyzed by Lorenzen et al. at CCR5 signal through Gα_i/o with similar efficacies, whereas they vary widely in their ability to signal through Gα_q, suggesting that bias between G protein subtypes may affect signaling at receptors able to couple to multiple downstream transducer elements (60). While investigation into bias between Gα subtypes is difficult due to cross-talk between different G protein signaling pathways, bias between different G proteins presents an added layer of complexity to ligand bias in the chemokine system.

Receptor Bias

Receptor bias refers to a receptor that demonstrates preferential interaction with certain transducers over others (39). ACKR3, also known as CXCR7, was originally characterized as a “decoy” receptor that served to modulate CXCR4 signaling by either sequestering excess CXCL12 (as both ACKR3 and CXCR4 can bind CXCL12) or serving as a coreceptor with CXCR4, rather than activating specific downstream signaling pathways (62–65). However, it was later discovered that ACKR3 not only acted as a decoy receptor, but also was capable of independently signaling via β-arrestins (66). CXCR4 and ACKR3 signal through different pathways despite being activated by the same ligand, with CXCR4 activating both G proteins and β-arrestins and ACKR3 signaling only via β-arrestins; therefore ACKR3 acts as a β-arrestin biased receptor relative to CXCR4 (67). Additional studies have shown that ACKR3 also serves as a receptor for endogenous opioids, contributing to circadian glucocorticoid oscillations (68) and acting as a decoy for opioids in the brain (69). Interestingly, recent work has demonstrated that β-arrestin recruitment to ACKR3 is dispensable for some receptor functionality. Montpas et al. show that both siRNA knockdown of β-arrestin 1 and or β-arrestin2 in HEK293 cells does not impact ACKR3 mediated degradation of CXCL11 or CXCL12 (70). They repeated these studies in β-arrestin 1 and 2 knockout murine embryonic fibroblasts (MEFs) and also saw that ACKR3 chemokine scavenging was unaffected. Similarly, Saaber et al. demonstrate that ACKR3 chemokine scavenging and endocytosis is dependent on receptor phosphorylation but independent of β-arrestins. Utilizing a variety of transgenic mice, they further demonstrate that β-arrestin is dispensable and receptor phosphorylation is indispensable in ACKR3 mediated migration of cortical interneurons (71). These findings demonstrate that ACKR3, while technically a β-arrestin-biased receptor as it does not signal via heterotrimeric G proteins, has functionality independent of the β-arrestins, highlighting the complexity of biased agonism at GPCRs.

Alternative splicing of GPCR transcripts further increases receptor diversity and the potential for receptor bias (72). Highly truncated receptor splice variants with only three transmembrane domains have been shown to heterodimerize with their wild type counterparts and retain them in the endoplasmic reticulum, while longer splice variants with a preserved seven transmembrane architecture have been shown to promote signaling via distinct signal transduction pathways (72). The isoforms of CXCR3 present an example of how alternative splicing can lead to endogenous receptor bias (47, 73). CXCR3 has three splice variants: CXCR3A, CXCR3B, and CXCR3-alt (73). CXCR3A and CXCR3B have nearly identical sequences, with the only difference being that CXCR3B possesses an extra 51 amino acids on its extracellular N terminal domain (73). CXCR3alt has both truncated forms of the third intracellular loop and transmembrane helices 6 and 7, resulting in a highly truncated 5 transmembrane receptor (73). CXCR3alt does not recruit β-arrestin or stimulate a Gα_i response, but promotes modest extracellular signal-regulated kinase (ERK) phosphorylation and becomes internalized following ligand stimulation (73). The two functional CXCR3 splice variants CXCR3A and CXCR3B have significantly biased responses from one another (47, 73). CXCR3B is β-arrestin-biased relative to CXCR3A, as the two isoforms similarly recruit β-arrestin upon stimulation with CXCL11 but CXCR3A signals through Gα_i to a significantly greater extent (47). Their specific interactions with β-arrestin also differ, as CXCR3A’s interaction with β-arrestin demonstrates a stable class B receptor pattern while CXCR3B’s demonstrates a transient interaction, consistent with class A receptor behavior (47). Smith et al. also found differences in receptor kinetics and mechanisms driving ERK phosphorylation and receptor internalization. Finally, the two variants display differential ligand induced transcriptional activity, as stimulation of CXCR3A, but not CXCR3B, with CXCL11 resulted in a robust response using serum response element (SRE) and serum response factor-response element (SRF) transcriptional reporters. These and other examples suggest that there are a number of finely tuned mechanisms that allow receptor bias to differentially regulate the response to chemokines.

System Bias

System bias is primarily regulated through the differential expression or function of receptor and transducer elements, which consequentially promote biased cellular outputs (39). These include every element of the signal transduction cascade, from proximal effectors such as GRKs, G proteins and β-arrestins, to other proteins that regulate second messenger responses, such as Regulator of G Protein Signaling (RGS) and Activator of G Protein Signaling (AGS) proteins, to all of the downstream signal transduction machinery. This phenomenon manifests as observed differences in signaling events between cell types. For example, the functional effects of mutations on CXCR6 receptor function and selectivity of G_i/o proteins are cell-type specific (74). In an F128Y mutant of CXCR6, the receptor’s dependence on G_o proteins was diminished in HEK-293T cells, but resembled that of the wild type receptor in Jurkat E6–1 cells (74). Outside of the chemokine system, it has been shown that β-arrestin biased agonists for the dopamine receptor D₂ demonstrate different signaling effects in the striatum and prefrontal cortex due to differential expression of GRK2 and β-arrestin 2 (75). While, to our knowledge, system bias in the chemokine system has not been rigorously and explicitly studied, changes in GRK expression have been proven to alter the functionality of numerous chemokine receptors (76). For example, murine T cells with a 50% reduction of GRK2 levels experienced significantly more migration, calcium flux, protein kinase B activity, and ERK phosphorylation upon stimulation with chemokine ligands than in wild type cells (77). Similarly, Arnon et al. created GRK2^f/− mice carrying a T cell-specific or a B cell specific cre to study the effects of GRK2 on S1P receptor-1 function in lymphocyte migration (78). They found that T and B cell movement from the blood into lymphatic tissues was reduced in the absence of GRK2, and that B cells between the splenic marginal zone and follicle were also reduced. These findings have physiologic implications, as GRK expression is differentially modulated in various disease states. For example, the expression levels of GRK2 and GRK5 are elevated in the lungs of rats treated with IL-1β, while levels of GRK2 and GRK6 are reduced in mice in an experimental autoimmune encephalomyelitis (EAE) animal model (79). Furthermore, downregulated GRK2 and GRK6 have been found in patients with autoimmune diseases such as multiple sclerosis and rheumatoid arthritis, and dynamic expression of β-arrestin 1 has been found in mouse models of both EAE and an adjuvant-induced arthritis (80–84).

It is inherently difficult to accurately recapitulate the dynamic in vivo expression levels of transducer elements across various cell types in vitro. However, considering the diversity of chemokine receptors and immune cells within mammals, and the dynamic nature of GPCRs and GPCR transducer elements in inflammatory conditions, system bias likely adds a significant layer of complexity to the chemokine system’s biased functionality (76, 80).

Biochemical Mechanisms of Biased Agonism

As discussed above, the ternary complex of ligand, receptor, and transducer serves as a model for the components driving biased agonism (39). A final cellular output is highly dependent on (1) the ligand, (2) the receptor, and (3) the transducer elements in the context of the cellular system. These three elements are the ultimate facilitators of biased signaling, and understanding the structural and functional bases of their interactions is integral to understanding potential mechanisms of biased agonism.

Ligand-Receptor Interactions

At the most fundamental level, ligand binding promotes an ensemble of receptor conformations that promote differential transducer coupling, leading to the selective activation of specific signaling pathways over others. Recent crystal structures and cryoEM have changed our understanding of how chemokines bind to their receptors (85). In the classic two-site model for chemokine binding, a chemokine binds its receptor through two distinct chemokine recognition sites (CRS), CRS1 and CRS2. CRS1 consists of an interaction between the sulfated tyrosines of the receptor N-terminus with the chemokine globular core, and CRS2 consists of an interaction between the transmembrane regions of the receptor with the chemokine N-terminus. Crystal structures have demonstrated an intervening CRS1.5 between CRS1 and CRS2 at a conserved Pro-Cys of the CKR N-terminus against the conserved disulfide of the chemokines (86, 87).

Affecting CRS2 by mutating chemokine N termini has been shown to affect biased signaling responses at CCR1, with data demonstrating that this region has a role in stabilizing active receptor conformations that contribute to biased agonism (51). CCR1 is known to bind at least nine different endogenous chemokines including CCL7, CCL8, and CCL15. Sanchez et al. showed that CCL7 and CCL8 demonstrate bias towards Gα_i signaling relative to CCL15(Δ26), an N-terminal truncated version of CCL15 that renders it approximately the same length as CCL7 and CCL8. To elucidate the structural basis of this bias, Sanchez et al. created a chimeric chemokine, CCL15(N-CCL7), which consists of CCL15 with the N terminal region of CCL7. Using radioligand binding, CCL15(N-CCL7) had a binding affinity for CCR1 nearly identical to CCL7 and significantly lower than CCL15Δ(26). When looking at G protein activation, CCL15 (N-CCL7) demonstrated a signaling profile very similar to that of CCL15(Δ26) which suggests that the N-terminal regions of CCL7 and CCL15(Δ26) are equally capable of driving G protein activation. However, CCL15 (N-CCL7) shows an intermediate ability to recruit β-arrestin when compared to CCL15(Δ26) and CCL7, suggesting that the N-terminal regions of CCL7 and CCL15(Δ26) contribute to a chemokines ability to recruit β-arrestin (51). Sanchez et al. also found that shortening CCL15Δ(26) by only two residues, CCL15Δ(28), did not affect the chemokines affinity for the receptor, but drastically reduced its maximum ability to recruit β-arrestin without affecting its potency. Concurrently, this shortening increased its potency but not maximum ability to inhibit cAMP accumulation. These data demonstrate the importance of a chemokine’s N-terminal length and sequence identity not only in binding affinity, but also in stabilizing receptor conformations that differentially interact with both G proteins and β-arrestins.

Research exploring the endogenous ligands of CXCR3 suggests that each ligand exhibits differential binding interactions with the receptor, further implicating the CRS2 interface in a biased response. Cox et al. used radioligand binding studies to suggest that CXCL9 and CXCL10 bind at a separate site on the receptor than CXCL11 (88). Colvin et al. performed mutagenic studies on CXCR3 and elucidated which residues were implicated in this differential chemokine binding (89). Whereas deletion of the 16 proximal N terminal amino acids of the receptor resulted in diminished calcium flux, actin polymerization, and chemotaxis upon stimulation with CXCL10 and CXCL11, activity stimulated by CXCL9 was not affected (89). Together, these seemingly discordant results highlight the diversity in activation mechanisms at a single receptor and the importance of ligand and receptor identity, as well as the interactions between ligand, receptor, and transducer elements in determining the overall final cellular output.

Different receptors can display different binding patterns to the same ligand, further pointing to how this interaction contributes to biased signaling. For example, while the receptor CXCR4 and β-arrestin biased receptor ACKR3 are both capable of binding CXCL12, mutagenic studies have revealed that the ligand-receptor interface of CXCR4 is significantly different than that of ACKR3 (90, 91). Jaracz-Ros et al. created a variety of CXCL12 mutants and demonstrated that substitution of N-terminal amino acids decreased receptor affinity at both CXCR4 and ACKR3. However, while N-terminal truncations of CXCL12 similarly led to decreased receptor affinity at CXCR4, these truncation mutants were unable to bind to ACKR3. Interestingly, CXCL12 substitution mutants displayed a decreased ability to drive receptor internalization at CXCR4 when compared to WT CXCL12 but had no impact on internalization at ACKR3 (90). Similarly, these substitution mutants demonstrate no ability to recruit β-arrestin to CXCR4, but show slightly decreased potency and efficacy in β-arrestin recruitment at ACKR3 relative to WT CXCL12. Notably, both ACKR3 and CXCR4 bind CXCL12 through the receptor distal N-terminus wrapping around the chemokine, forming a novel CRS0.5 interface (86, 92).

A recent cryoEM structure of CCL20 bound to CCR6 further illustrates that chemokine receptors possess extremely diverse mechanisms of activation. Numerous class A GPCRs have been shown to exhibit a deep binding pocket mode of activation that involves interactions with the transmembrane receptor core and proximity to the toggle switch motif, a tryptophan residue that has been shown to contribute to conformational changes upon receptor activation (93). In contrast, CCL20 binds in a long, shallow extracellular site distal from the toggle switch (93). CCL20 does not require interactions between its N terminus and the 7 transmembrane receptor core to activate CCR6, but instead requires major interactions with extracellular loop 2 and the N terminus of the receptor (93). Wasilko et al. speculate that this differential binding pattern is due to CCL20’s short N terminal region, which consists of only 5 amino acids. Work from Riutta et al. further supports this notion as they manipulated various aspects of CCL20’s short N terminal region and determined that CCL20 binding and activation of CCR6 is highly tolerant to changes in the length and identity of the N terminus (94). Other chemokines or ligands with shorter N termini could prove to exhibit binding patterns similar to CCL20, emphasizing the notion that chemokine binding is extremely diverse and that attempting to generalize chemokine ligand-receptor interactions to a single model may be impossible. The interactions between the chemokine ligands and receptors have thus been proven to be multi-site, complex, and dynamic.

Phosphorylation Barcode

Receptor activation via ligand binding leads to a conformational change in the receptor which translates into a cellular response first through G protein recruitment and activation, followed by phosphorylation of the intracellular loops and C terminus of the receptor. This phosphorylation leads to the recruitment of the β-arrestins, which abrogate G protein signaling and also initiate β-arrestin dependent signaling cascades. Phosphorylation patterns of the receptor can vary based on the identity of the ligand used to activate the receptor and the kinases present in the cell, and these differential phosphorylation patterns can promote engagement with distinct transducer elements, and consequently generate a unique signaling cascade (Figure 3). Numerous studies support this model, known as the “barcode hypothesis,” as one mechanistic explanation that underlies biased agonism (95–98).

FIGURE 3: Phosphorylation barcode promotes differential downstream signaling.

Distinct patterns of G Protein-Coupled Receptor phosphorylation can promote differential downstream signaling cascades. (A) A phosphorylation barcode which preferentially drives β-arrestin activation of with effector “A” while (B) demonstrates a phosphorylation barcode which preferentially drives β-arrestin mediated activation of effector “B”.

Formation of the phosphorylation barcode is mediated by a number of kinases, most notably G protein-coupled receptor kinases, or GRKs (99). There are seven GRK isoforms, but only GRK2, 3, 5, and 6 are expressed ubiquitously and are primarily relevant for a discussion of bias in the chemokine system (99). GRKs are selective in the residues they phosphorylate, implicating them in the complexity required for a biased response. Peptide studies reveal that different classes of GRKs have varying affinities for phosphorylating sites within different proximities of acidic or basic residues (100). For example, GRK5 and GRK6 preferentially phosphorylate peptides with basic amino acids near the N terminal side of the target residues (101, 102). Furthermore, GRKs have also been shown to phosphorylate specific residues on the C terminal tails of chemokine receptors (103). Upon stimulation with CXCL12, CXCR4 is rapidly phosphorylated by GRK6 at Ser-324/5 and Ser-339, and slowly phosphorylated by GRK6 at Ser-330 (103). Data also suggests that phosphorylation sites at CXCR4 are hierarchical and sequential in nature, in which a certain order of phosphorylation is necessary to induce a response (104). Therefore, the pattern of phosphorylation mediated by distinct GRKs on the receptor C-terminal tail appears to be specific and coordinated.

Genetic knockdown of GRKs has connected specific classes of GRKs to distinct cellular functions both in vitro and in genetically modified mice (44, 47, 95, 105, 106). At CXCR4, for example, GRK2 and GRK6 have been shown to negatively regulate calcium mobilization. Additionally, ERK1/2 phosphorylation is negatively regulated by GRK2, but positively regulated by GRK3 and GRK6 (103). Furthermore, GRK3 is responsible for the phosphorylation of the two most distal phosphorylation sites on the CXCR4 C terminal tail, which have been implicated in β-arrestin recruitment (106). Similarly, Zidar et al. found that, at CCR7, CCL19-mediated β-arrestin recruitment was reduced following siRNA knockdown of GRK3 and GRK6, while CCL21-mediated β-arrestin recruitment was only reduced following GRK6 knockdown (44). Additionally, GRK6 knockdown led to a significant decrease in ERK 1/2 phosphorylation at both chemokines, while GRK2 and GRK3 knockdown led to an increase in ERK 1/2 phosphorylation only with CCL19 (44). Instances of differential involvement by GRKs also occur between the aforementioned splice variants of CXCR3 (47). CXCR3B driven recruitment of β-arrestin involves GRK2 and GRK3 but not GRK5 nor GRK6, while that of CXCRA involves all four kinases (47). These data are consistent with a model in which different expression levels of these kinases can influence a biased response. The utilization of different GRK isoforms by biased ligands and receptors to produce a phosphorylation barcode could lead to functionally distinct pools of β–arrestins, thus yielding important insights into the mechanism of biased agonism.

Receptor-Transducer Interactions

Upon stimulation with a ligand, a GPCR recruits transducer elements to induce signaling cascades. Potential for bias rests in the receptor-transducer interactions in a similar fashion as the ligand-receptor interaction; receptors have the potential to stabilize interactions with or conformations of transducer elements that lead to distinct signaling responses. Recent structural studies have provided important insights into how such structural changes can be linked to receptor activity and biased responses. Crystal structures of negative allosteric modulators bound to the chemokine receptors CCR2 (107) and CCR9 (108) have demonstrated an allosteric binding site on the intracellular surface. This binding site involves several conserved amino acid residues and is a common intracellular binding site in class A GPCRs (109). Notably at the type 1 angiotensin II receptor, Wingler et al. found that G-protein biased agonists stabilize an “open” conformation of the intracellular surface of the receptor, while β-arrestin biased agonists stabilize a more closed receptor intracellular conformation (110). This suggests that biased agonists promote differential interactions with transducers through the generation of specific receptor intracellular conformations, which could potentially be targeted with novel allosteric modulators. However, at this time there is little structural information on the mechanisms underlying biased agonism by chemokine receptors, although there is a large body of work that have studied how receptor-transducer interactions contribute to biased signaling.

Colvin et al. found that the endogenous ligands of CXCR3 promote signaling through different structural elements of the receptor (111). CXCL9- and CXCL10-induced internalization was dependent on the phosphorylation of the receptor C terminus, whereas CXCL11-induced internalization was dependent on phosphorylation of intracellular loop 3 (111). Similarly, the C terminal tail of CCR7 is required for chemotaxis, calcium flux, and ERK1/2 phosphorylation, but not receptor internalization nor recycling (112). Because biased ligands drive their distinct signaling profiles via the receptor, it can be postulated that biased signaling is also critically dependent on the differential interactions between the receptor and its transducers.

Distinct conformations of transducer elements can lead to varying cellular responses. Both synthetic and endogenous β-arrestin biased ligands of CXCR3A induced a conformational change in β-arrestin upon stimulation, suggesting that transducer conformation plays a role in their biased response (15, 47). Mutation of transducers has helped determine which structures are involved in specific signaling events. Cahill et al. demonstrated that β-arrestin assumes at least two distinct conformations when bound to a GPCR: a tail conformation or a core conformation (113). The “tail” conformation interacts with the C terminal tail of the receptor, while the “core” conformation involves the finger-loop region of β-arrestin and the seven transmembrane receptor core. A β-arrestin mutant lacking the finger loop region required for assuming the “core” conformation was unable to desensitize G protein mediated signaling, even though it was still able to induce receptor internalization and other β-arrestin dependent signaling effects (113). This directly connects specific conformations of transducer isoforms to cellular functions, suggesting that preferential stabilization of certain conformations could be the basis of a biased signaling response.

Furthermore, the varying functional effects of transducer conformations may be the basis of the functional selectivity of the two β-arrestin isoforms. These two isoforms, β-arrestin 1 (arrestin-2) and β-arrestin 2 (arrestin-3), despite possessing significant sequence similarity, have been shown to promote distinct cellular functions (114). For example, β-arrestin 2 is more heavily involved in receptor desensitization than β-arrestin 1 when interacting with the vasopressin 2 receptor, and β-arrestin 1 accumulates in the nucleus while β-arrestin 2 does not (115). β-arrestin1 and β-arrestin 2 have different conformations when in complex with a phosphorylated receptor, with the main difference resting in the interaction with the receptor core (115). Mutational analyses of the two isoforms suggest that the core interactions are the basis of their differential desensitization, supporting the notion of linking β-arrestin-receptor interactions to distinct functional consequences (113, 115).

Ligand-receptor interactions, receptor-transducer interactions, and the phosphorylation barcode each present unique insight into elucidating the means through which a biased response is produced. The mechanisms of biased agonism illustrate the significant potential for their involvement in the diverse functions within the chemokine system.

Functional Effects of Chemokine Bias

While much is known regarding the biochemical mechanisms underlying biased chemokine signaling, there is relatively less known about the functional consequences of such signaling. Table 1 lists selected endogenous and synthetic chemokine receptor ligands and their functional consequences at CKRs. It is difficult to completely assess ligand, receptor, and systems bias in vitro when considering the vast diversity of tissue types in which the chemokine system is expressed, as well as the variety of functions of the chemokine system under both homeostatic and inflammatory conditions. Therefore, it is important for future research to use in vivo examples of biased agonism to direct in vitro studies that aim to understand the mechanisms underlying this signaling. We highlight some examples below.

Table 1:

Selected studies of endogenous and synthetic chemokine ligands. Highlighted compounds are those at which biased agonism has been demonstrated.

Receptor	Endogenous ligands	Synthetic ligands	Synthetic ligand characterization	System	Assays	References
CXCR1	CXCL6 CXCL8			In vitro	G protein signaling β-arrestin recruitment Receptor internalization	(24, 170)
		(R)-Ketoprofen	Antagonist	In vitro	Chemotaxis	(171)
		Reparixin (Repertaxin)	Allosteric antagonist	In vitro	Chemotaxis	(172, 173)
				In vivo	Inflammation Reperfusion injury Blood pressure
		DF 2156A	Allosteric antagonist	In vitro	G protein signaling Chemotaxis Cell proliferation	(174)
				In vivo	Angiogenesis Hepatic reperfusion injury
		SX-517	Antagonist	In vitro	G protein signaling	(175)
				In vivo	Inflammation
CXCR2	CXCL1 CXCL2 CXCL3 CXCL5 CXCL6 CXCL7 CXCL8			In vitro	G protein signaling β-arrestin recruitment Receptor internalization	(24, 170)
		SB 455821	Antagonist	In vitro	Chemotaxis	(176)
				In vivo	Neutrophil migration
		Reparixin (Repertaxin)	Allosteric antagonist	In vitro	Chemotaxis	(172, 173)
				In vivo	Inflammation Reperfusion injury Blood pressure
		DF 2156A	Allosteric antagonist	In vitro	G protein signaling Chemotaxis Cell proliferation	(174)
				In vivo	Angiogenesis Hepatic reperfusion injury
		Navarixin (MK-7123)	Antagonist	In vivo	COPD Solid tumors	(177–179)
		Danirixin	Antagonist	In vitro	G protein signaling	(180–182)
				In vivo	Neutrophil migration and activation COPD
		SB225002	Antagonist	In vitro	Neutrophil chemotaxis	(183, 184)
				In vivo	Neutrophil margination Colitis Inflammatory bowel syndrome
		SX-517	Antagonist	In vitro	G protein signaling	(175)
				In vivo	Inflammation
		AZD5069	Antagonist	In vitro	Neutrophil chemotaxis	(185)
				In vivo	Acute lung inflammation
CXCR3	CXCL4 CXCL9 CXCL10 CXCL11 CCL7 (Antagonist) CCL11 (Antagonist)			In vitro	G protein signaling β-arrestin recruitment Receptor internalization Chemotaxis	(24, 89, 111, 119, 186, 187)
				In vivo	T cell polarization T cell localization
		VUF1066 VUF11418	Agonist Agonist	In vitro	G protein signaling β-arrestin recruitment Receptor internalization	(188, 189)
				In vivo	GRK engagement Chemotaxis Chemotaxis Inflammation
		FAUC1036 FAUC1104	Allosteric agonist Allosteric agonist	In vitro	G protein signaling β-arrestin recruitment Receptor internalization Chemotaxis	(190)
		BD103 BD064 cRAMX3	Allosteric modulator Allosteric modulator Antagonist	In vitro	G protein signaling β-arrestin recruitment	(191)
		SCH546738	Antagonist	In vitro	G protein signaling Chemotaxis	(192, 193)
				In vivo	Mouse collagen induced arthritis Rat experimental autoimmune encephalitis Rat Cardiac transplantation
		8-azaquinazolinone derivative	Allosteric modulator	In vitro	G protein signaling β-arrestin recruitment	(194)
CXCR 4	Monomeric/d imeric CXCL1			In vitro	G protein signaling β-arrestin recruitment Chemotaxis	(55, 195, 196)
	EPI-X4 (Antagonist)			In vivo	Cancer metastasis Hematopoietic stem cell mobilization
		Plerixafor (AMD3100)	Allosteric antagonist	In vitro	G protein signaling β-arrestin recruitment Receptor internalization Chemotaxis	(136, 137,168, 197)
		X4-2-6	Allosteric antagonist	In vivo	Hematopoietic stem cell mobilization
		TG-0054 (burixafor)	Antagonist	In vivo	Hematopoietic stem cell mobilization Choroid neovasculariz ation Inflammation Cardiac function	(198–200)
		AMD11070 (mavorixafor)	Antagonist	In vivo	X4-trophic HIV-1 infection Inflammation WHIM syndrome Melanoma	(201–205)
		MSX-122	Antagonist	In vitro	G protein signaling Chemotaxis Angiogenesis	(206, 207)
				In vivo	Inflammation Breast cancer metastasis Solid tumors
		CTCE-9908	Antagonist	In vivo	Solid tumors	(208)
		POL-6326	Antagonist	In vivo	Hematopoietic stem cell mobilization	(209)
		RSVM ASLW	Agonist Superagonist	In vitro	Chemotaxis Receptor cell surface expression	(210)
		ATI-2341	Allosteric agonist	In vitro	G protein signaling β-arrestin recruitment GRK engagement Chemotaxis Neutrophil mobilization Receptor internalization	(211–213)
		DV1-K-(DV3)	Antagonist	In vitro	G protein signaling Chemotaxis X4-tropic HIV-1 infection	(214)
		T22	Antagonist	In vitro	G protein signaling X4-tropic HIV-1 infection	(215)
		POL6326 (Balixafortide)	Antagonist	In vivo	Hematopoietic stem cell mobilization HER2 negative, locally recurrent or metastatic breast cancer	(216–218)
		T140	Inverse agonist	In vitro	X4-tropic HIV-1 infection	(219)
		CXCL12-T140 chimera	Agonist	In vitro	Chemotaxis	(220)
		CXCL12-IT1t chimera	Antagonist	In vitro	G protein signaling β-arrestin recruitment Chemotaxis	(221)
		PZ-218	Antagonist	In vitro	G protein signaling Chemotaxis	(222, 223)
				In vivo	Leukocytosis
		508MCI	Antagonist	In vitro	G protein signaling	(224)
				In vivo	Inflammation Tumor metastasis
		GSK812397	Antagonist	In vitro	Chemotaxis X4-tropic HIV-1 infection	(225)
		TIQ-15	Antagonist	In vitro	G protein signaling β-arrestin recruitment X4-tropic HIV-1 infection	(226)
		IT1t	Allosteric antagonist	In vitro	Chemotaxis	(227, 228)
				In vivo	X4-tropic HIV-1 infection
		ALX40–4C	Antagonist	In vitro In vivo	X4-tropic HIV infection	(229, 230)
CXCR5	CXCL13					(231, 232)
CXCR6	CXCL16					(233, 234)
		Small molecules with exo-[3.3.1]azabicyclononane core	Antagonist	In vitro	G protein signaling β-arrestin recruitment	(235)
				In vivo	Hepatocellular carcinoma
CCR1	CCL2 CCL3 CCL4 CCL5 CCL7 CCL8 CCL13 CCL14 CCL15 CCL16 CCL23			In vitro	G protein signaling β-arrestin recruitment Receptor internalization Chemotaxis	(24, 61)
		J113863 UCB35625	Antagonist Antagonist	In vitro	Eosinophil morphology Receptor internalization Chemotaxis	(236–239)
				In vivo	Experimental murine arthritis
		Met-CCL5 (Met-RANTES)	Antagonist	In vitro	Chemotaxis	(240–242)
				In vivo	Experimental rat and murine arthritis Rabies virus
		CCX354	Antagonist	In vitro	Chemotaxis	(243)
				In vivo	Inflammation Rheumatoid arthritis
		CP-481,715	Antagonist	In vitro	G protein signaling Chemotaxis	(244, 245)
		MLN3897	Antagonist	In vivo	Rheumatoid arthritis	(246)
		BX471	Antagonist	In vivo	Multiple sclerosis	(245)
		AZD-4818	Antagonist	In vivo	COPD	(245)
		CCX9588	Antagonist	In vitro	Chemotaxis	(247)
		LMD-559	Agonist	In vitro	G protein signaling Chemotaxis	(248)
CCR2	CCL2 CCL7 CCL8 CCL11 CCL13 CCL16				G protein signaling β-arrestin recruitment Receptor internalization	(249)
		J113863 UCB35625	Agonist Agonist	In vitro	G protein signaling β-arrestin recruitment Chemotaxis	(169)
		CAS 445479-97-0	Antagonist	In vitro	Cell proliferation Cell migration Invasion	(250)
		INCB3344	Antagonist	In vitro	G protein signaling β-arrestin recruitment Chemotaxis	(251, 252)
		N-(2-((1-(4-(3methoxyphenyl)cyclohex yl)piperidin-4-yl)amino)-2-oxoethyl)-3-(trifluoromethyl)benzami de	Antagonist			(253)
		CCR2-RA	Antagonist	In vitro	G protein signaling β-arrestin recruitment	(252)
		15a	Antagonist	In vivo	Murine atherogenesis	(254)
		MK-0812	Antagonist	In vitro	Cell morphology	(255)
				In vivo	Monocyte recruitment
CCR3	CCL5 CCL7 CCL8 CCL11 CCL13 CCL15 CCL24 CCL26 CCL28			In vitro	G protein signaling Receptor internalization Chemotaxis	(238, 256)
		J113863 UCB35625	Antagonist Antagonist	In vitro	Eosinophil morphologic change Receptor internalization CCR3-tropic HIV-1 infection Chemotaxis	(236–238)
		CH0076989	Agonist	In vitro	Eosinophil shape change Receptor internalization Chemotaxis	(238)
		SB-328437	Antagonist	In vitro	G protein signaling Eosinophil chemotaxis	(257)
		R321	Antagonist	In vitro	G protein signaling β-arrestin recruitment Chemotaxis Receptor internalization Eosinophil recruitment to lungs Murine airway hyperresponsi veness	(258)
		GW766994	Antagonist	In vivo	Asthma Eosinophilic bronchitis	(259)
CCR4	CCL17 CCL22			In vitro	G protein signaling β-arrestin recruitment Receptor internalization Chemotaxis	(52, 260–262)
				In vivo	Inflammation
		MDC67 Compounds 1–4	Antagonist Antagonist	In vitro	G protein signaling β-arrestin recruitment Receptor internalization Chemotaxis	(52)
		Compound 8a	Antagonist	In vitro	Chemotaxis	(263)
				In vivo	Allergic asthma in mice
CCR5	CCL2 CCL3 CCL3L1 CCL4 CCL5 CCL8 CCL11 CCL13 CCL14 CCL16			In vitro	G protein signaling β-arrestin recruitment	(24, 242)
				In vivo	Receptor internalization Inflammation Rabies virus
		Maraviroc	Antagonist	In vivo	R5-tropic HIV-1 infection	(134)
		Aplaviroc	Antagonist	In vitro	R5-tropic HIV-1 infection	(264, 265)
				In vivo	R5-tropic HIV-1 infection
		AOP-RANTES	Agonist	In vitro	G protein signaling Receptor internalization R5-tropic HIV-1 infection	(160)
		PSC-RANTES 5P14-RANTES 5P12-RANTES	Superagonist Agonist Agonist	In vitro	G protein signaling Receptor internalization R5-tropic HIV-1 infection	(162)
		TAK-652 TAK-779	Antagonist Antagonist	In vitro	R5-tropic HIV-1 infection	(266, 267)
		Vicriviroc	Antagonist	In vitro In vivo	R5-tropic HIV-1 infection	(268)
		J113863 UCB35625	Agonist Agonist	In vitro	G protein signaling β-arrestin recruitment Chemotaxis	(169)
		YM-370749	Agonist	In vitro	G protein signaling β-arrestin recruitment Receptor internalization Chemotaxis R5-tropic HIV-1 infection	(269)
		Met-CCL5 (Met-RANTES)	Antagonist	In vitro	Chemotaxis	(240–242)
				In vivo	Experimental rat and murine arthritis Rabies virus
		SCH-351125 (Ancriviroc)	Antagonist	In vitro	G protein signaling	(270–272)
				In vivo	Rheumatoid arthritis
		E-913	Antagonist	In vitro	G protein signaling	(270)
CCR6	CCL20			In vitro	Receptor Internalization β-arrestin recruitment ERK phosphorylatio n	(273–277)
				In vivo	Actin polymerization Graft-versus-host disease Colorectal Cancer
		CCL20 S64C	Agonist	In vitro	G protein signaling β-arrestin recruitment Receptor internalization Chemotaxis	(278)
				In vivo	IL-23 dependent murine model of psoriasis
CCR7	CCL19 CCL21			In vitro	G protein signaling β-arrestin recruitment Receptor internalization GRK engagement Chemotaxis β2 integrin activation	(44, 45, 129, 279)
				In vivo	Naïve T cell recirculation in secondary lymphoid tissue
		Cmp2105 CS-1 CS-2 Navarixin	Allosteric antagonist Allosteric antagonist Allosteric antagonist Allosteric antagonist	In vitro	Crystal structure β-arrestin recruitment Thermal shift assay	(280)
		Cosalane	Antagonist	In vitro	β-arrestin recruitment Chemotaxis	(281)
CCR8	CCL1 CCL8 CCL16 CCL18					(282, 283)
		LMD-559	Agonist	In vitro	G protein signaling Chemotaxis	(248)
		R243	Antagonist	In vitro	G protein signaling Macrophage aggregation	(284)
				In vivo	Cytokine secretion Hapten induced colitis Peritoneal adhesions
		Napthalene-sulfonamide derivatives	Allosteric antagonists	In vitro	G protein signaling Chemotaxis	(285)
CCR9	CCL25					(286)
		Vercirnon	Allosteric antagonist	In vitro	Crystal structure G protein signaling	(108, 287, 288)
				In vivo	Crohn’s disease
		CCX8037	Antagonist	In vitro	Chemotaxis	(289)
				In vivo	Intestinal T cell accumulation
		CCX282-B CCX025	Antagonist Antagonist	In vitro	G protein signaling Chemotaxis	(290, 291)
				In vivo	Crohn’s disease Ulcerative colitis
CCR10	CCL27 CCL28			In vitro	G protein signaling β-arrestin recruitment Receptor internalization Chemotaxis	(24, 292)
		Eut-22	Antagonist	In vitro	G protein signaling Chemotaxis	(293)
				In vivo	Murine contact hypersensitivity
CX3CR1	CX3CL1					(294)
		JMS-17–2	Antagonist	In vitro	Chemotaxis Pancreatic ductal adenocarcino ma motility and viability	(295, 296)
				In vivo	Breast cancer metastasis
		KAND567	Antagonist	In vivo	Atherosclerosis Myocardial infarction	(297)
XCR1/CCXCR1	XCL1 XCL2					(298)
ACKR1/DARC	CXCL5 CXCL6 CXCL8 CXCL11 CCL2 CCL5 CCL7 CCL11 CCL13 CCL14 CCL17					(299)
ACKR2	CCL2 CCL3 CCL4 CCL5 CCL7 CCL8 CCL11 CCL13 CCL14 CCL17 CCL22			In vitro	G protein signaling Receptor internalization	(300, 301)
		CCL14(9–74)	Agonist	In vitro	G protein signaling Actin rearrangement Receptor internalization	(300, 301)
ACKR3/CXCR7	CXCL11 CXCL12			In vitro	G protein signaling β-arrestin recruitment Chemotaxis	(62, 63, 66, 299)
		TC14012	Allosteric antagonist	In vitro	β-arrestin recruitment	(302)
		GSLW	Allosteric agonist	In vitro	β-arrestin recruitment Receptor internalization	(303)
		FC313	Allosteric agonist	In vitro	β-arrestin recruitment	(304)
		CXCL11_12 chimera	Agonist	In vitro	Selective ACKR3 agonist, over CXCR3 and CXCR4	(305, 306)
		AMD3100	Allosteric agonist	In vitro	β-arrestin recruitment	(307)
		VUF11207 VUF11403	Agonist	In vitro	β-arrestin recruitment Receptor internalization	(308)
		CCX77	Agonist	In vitro	β-arrestin recruitment Receptor internalization Chemotaxis	(309–312)
				In vivo	Trans-endothelial migration Tumor growth Tumor growth and metastasis Tumor angiogenesis X4-tropic HIV-1 infection c
		CCX777	Agonist	In vitro	β-arrestin recruitment	(86)
ACKR4/CCR11	CCL19 CCL20 CCL21 CCL25					(299, 313)

CXCR3

Mechanisms and functional effects of biased agonism are especially well-characterized at CXCR3 in activated CD8⁺ T cells. CXCL11 promotes significantly greater chemotactic response when compared to CXCL9 and CXCL10 (89, 116). Synthetic small molecule agonists, VUF10661 and VUF11418, similarly display biased agonism at CXCR3, with VUF10661 characterized as β-arrestin-biased and VUF11418 as G protein-biased (116, 117). Notably, Smith et al. showed in a murine model of allergic contact hypersensitivity that VUF10661 potentiates a dinitrofluorobenzene-induced inflammatory response, while VUF11418 does not (116). This increase in inflammation derived from VUF10661 was lost when these experiments were repeated in β-arrestin 2 knock out (KO) mice. Additionally, a greater number of CD3⁺ T cells and, specifically, CD3⁺ β-arrestin⁺ CXCR3⁺ T cells, were found in tissue biopsies derived from patients with allergic contact dermatitis than from nonlesional controls. However, in vitro chemotaxis assays further revealed that both β-arrestin dependent phosphorylation of Akt and G protein signaling contribute to CXCR3-mediated chemotaxis (116). The exact consequences of G protein and β-arrestin bias for chemotaxis are therefore still unclear. While VUF10661 induces the greatest chemotactic response, G protein signaling is additionally necessary for any chemotactic response, as T cell migration controlled by CXCL9, CXCL10, and CXCL11 is abrogated following pretreatment with pertussis toxin, an inhibitor of Gα_i (116, 118). It is possible that β-arrestin and G protein dependent signaling cascades that contribute to the overall chemotactic response are interrelated.

Biased signaling at CXCR3 has also been found to direct T cell polarization and function (119). In a murine model of EAE, Zohar et al. found that CXCL10 induces polarization of T helper (Th) cells into Th1/Th17 cells, promoting inflammation, whereas CXCL11 promotes Th polarization into IL-10-producing T regulatory (Treg) cells, dampening inflammation. The authors also demonstrate distinct downstream signaling cascades for the endogenous ligands, with induction of T cell polarization by CXCL10 via phosphorylation of STAT1, STAT4, and STAT5, and by CXCL11 via phosphorylation of STAT3 and STAT6 (120). Additionally, CXCR3 signaling directs T cell localization by limiting Th cells to the perivascular space in the central nervous system, thereby increasing interaction between Treg and Th cells, and attenuating inflammation, demyelination, and axonal damage (121). CXCR3−/− mice demonstrate a more disorganized perivascular distribution of Th cells and experience increased tissue damage and less recovery in EAE than WT CXCR3 mice (121). Biased signaling has not been directly implicated in this disease process; however, there is evidence demonstrating that in murine EAE, CXCL9, CXCL10, and CXCL11 are expressed in different spatial and cellular locations within the central nervous system, suggesting a functional role of biased agonism at CXCR3 (121).

CXCR1 and CXCR2

There is also evidence demonstrating different cellular signaling induced by CXCL8 at its endogenous receptors, CXCR1 and CXCR2, both of which are primarily involved in directing neutrophil chemotaxis to sites of infection and initiating its cytotoxic effects (56). Interestingly, CXCL8 can exist as a monomer, dimer, or a combination of these two states in vivo. Nasser et al. provide evidence for biased signaling in this system in that the monomeric form of CXCL8 is more effective at driving calcium mobilization, chemotaxis and exocytosis than the dimeric form. They also discovered that this ligand bias is present at CXCR1, but not at CXCR2, suggesting an interaction between both ligand and receptor bias. To our knowledge, the functional consequences of this ligand and receptor bias have not been rigorously explored in vivo.

CXCR4

Similarly, at CXCR4, which is implicated in tumor cell trafficking, differential signaling by the monomeric and dimeric forms of its endogenous ligand CXCL12 demonstrated functional consequences in a murine model of colorectal carcinoma or melanoma (55). Previous research has shown that locally produced CXCL12 limits migration potential of cells expressing CXCR4 by either desensitizing these cells to endogenous ligand or exceeding optimal levels for chemotaxis by nullifying endogenous CXCL12 gradients produced in distant tissues (122–124). Drury et al. demonstrate that the monomeric form of CXCL12 is much more efficacious in stimulating filamentousactin polymerization and inducing chemotaxis than the dimeric form (55). From this in vitro evidence, one could postulate that the dimeric form of CXCL12 would inhibit metastasis and the monomeric form would promote metastasis. However, the authors found that exogenous administration of either WT CXCL12, a preferentially monomeric form, or preferentially dimeric form all inhibited tumor metastasis to similar extents. The reasons underlying this discrepancy between in vitro and in vivo work is possibly due to systems bias where the function of exogenously administered CXCL12 may assume a different function in vivo.

Furthermore, CXCR4 also exhibits receptor bias in WHIM syndrome which is most commonly due to a gain of function receptor C terminal truncation mutant (125). The truncated CXCR4 receptor found in these patients promotes more G protein dependent signaling cascades compared to the WT receptor, presumably due to the decreased ability to recruit the GRKs and β-arrestins and desensitize the receptor. However, there is evidence demonstrating that, even in this truncated receptor, β-arrestin is still able to engage the receptor via intracellular loop 3 and that β-arrestins engage in different signaling cascades than those in the WT receptor. Lagane et al. demonstrate using CXCR4 mutants derived from patients with WHIM syndrome that receptor desensitization and internalization is regulated by the C terminal tail of CXCR4 while β-arrestin-mediated signaling, like ERK1/2 activation, is primarily mediated by its interaction with ICL3 (125). The authors also show that WT CXCR4 can homodimerize and also heterodimerize with mutant CXCR4 receptors; these heterodimers are able to further enhance chemotaxis due to reduced internalization of both the mutant and WT CXCR4 and prolonged β-arrestin dependent signaling as measured by ERK 1/2 phosphorylation (125). This study highlights the many complex levels of bias that can exist in the chemokine system, where a biased receptor can not only direct different signaling patterns, but influence those of wild-type receptors. Studies have also shown, among others, CCR5 homodimerization shortly after synthesis in the endoplasmic reticulum, CXCR4 homodimer formation and heterodimer formation with CCR2, and CXCR2 homodimerization (126–128). Future research in this area could uncover unidentified sources of biased signaling, which could explain many of the non-redundant functional effects seen within the chemokine system.

CCR7

There is evidence of biased agonism at CCR7 between its endogenous ligands CCL19 and CCL21 and their role in lymphocyte extravasation and lymphocyte trafficking to and within secondary lymphoid organs (45, 129). Förster et al. demonstrated the importance of CCR7 in these signaling pathways by generating CCR7 deficient mice, which had impaired entry and retention of naïve T cells, dendritic cells (DCs), and B cells in multiple secondary lymphoid organs (129). CCR7-dependent signaling is therefore likely necessary for effective T cell-B cell and T cell-DC interactions.

Kohout et al. further characterize biased signaling at CCR7 by examining receptor desensitization and ERK1/2 activation in response to CCL19 and CCL21 (45). The endogenous ligands promote nearly identical amounts of G protein signaling, but only CCL19 induces substantial receptor desensitization via receptor C terminal phosphorylation and β-arrestin recruitment (45). Although these ligands induce similar calcium mobilization and leukocyte migration potential in vitro, CCL19, but not CCL21, induces rapid CCR7 internalization (130). Resultingly, pretreatment of T lymphocytes with CCL19 inhibits a chemotactic response to either CCL19 or CCL21, while cells retain full chemotactic potential following pretreatment with CCL21 (130).

When comparing CCL19^−/− mice, plt/plt mice (deficient in both CCL19 and CCL21), and plt/CCL19⁻ mice (lacking CCL19 and one CCL21 allele) to WT mice, Britschgi et al. found that deficiency of both CCL19 and CCL21 is required for abnormal localization of DCs in lymph nodes (131). Interestingly, they also discovered that CCL19 was not needed for DC migration, maturation, and T-cell priming (131). These results suggest that CCL19 is dispensable for certain aspects of DC function regulated by CCR7. It is unclear if CCL19 is sufficient to support normal DC function in the absence of CCL21, and if CCL19 and CCL21 have both overlapping and distinct physiologic roles. Taken together, these studies indicate the importance of biased signaling at CCR7 in the trafficking of a wide array of immune cells, as well as their interactions within secondary lymphoid tissues.

Targeting Chemokine System Biased Agonism in Drug Development

As opposed to the other GPCRs that have one or perhaps a few endogenous ligands, most CKRs have a number of endogenous ligands, many of which are biased, have additional receptor targets, and have variable tissue expression. This primary example of naturally occurring biased agonism may unveil a variety of new drug targets which exploit this biochemical phenomenon, but it may also uncover key information regarding the mechanisms and structural determinants underlying biased agonism across all GPCRs. An enhanced understanding of the mechanisms that control biased signaling in GPCRs may lead to the development of pharmaceutical drugs with increased efficacy and reduced side effect profile by selectively activating therapeutic signaling pathways. Biased therapeutics at other GPCR subtypes, including the type 1 angiotensin II receptor (AT1R), μ-opioid receptor, κ-opioid receptor, D2 dopamine receptor have demonstrated that differential activation of G-protein and β-arrestin signaling pathways has the potential to improve efficacy and avoid unwanted side effects. However, the in vivo effects of these compounds can be difficult to predict, even with detailed studies involving knockout animals and comparisons to balanced agonists and antagonists.

Due to the wide range of biological processes and pathologies in which the chemokine system is implicated, identification of biased drugs targeting chemokine ligands and receptors is of significant interest in clinical research. Yet, the success of drug development to target the chemokine system and its associated pathologies has been limited. Although clinical trials over the past decade have attempted to target over 50% of CKRs (132), there are currently only three FDA approved drugs on the market that target CKRs. Maraviroc is a CCR5 antagonist that inhibits R5-tropic HIV-1 entry into cells (133, 134). Plerixafor (AMD3100) is a CXCR4 antagonist first identified to block X4-tropic HIV-1 infection, but is now FDA-approved for use in autologous hematopoietic stem cell (HSC) transplantation in patient with Non-Hodgkin’s lymphoma or multiple myeloma (135–138). CXCR4 signaling promotes the retention of HSCs in the bone marrow and AMD3100, as an antagonist of CXCR4, mobilizes HSCs into the peripheral blood, a crucial step for conferring the therapeutic benefit of autologous transplantation (139). Finally, mogamulizumab is a CCR4 antibody used for treating cutaneous T-cell lymphoma (CTCL) and adult T-cell leukemia/lymphoma (ATLL) (140, 141). Currently, phase 3 clinical trials are under way to evaluate additional chemokine antagonists including the CCR5 antagonist leronimab and CXCR4 antagonists balixafortide and mavorixafor (142). Studies for leronimab are investigating the safety and effectiveness of using it with and without an optimized antiretroviral therapy (ART) regimen and for the safety and effectiveness of leronimab monotherapy for the maintenance of R5-tropic HIV-1 suppression over 48 weeks (143, 144). Balixafortide is currently being studied for its efficacy, safety and tolerability when given intravenously with eribulin, a common breast cancer chemotherapy drug, versus eribulin alone in the treatment of HER2 negative, locally recurrent or metastatic breast cancer (145). Mavorixafor is being examined for its efficacy in participants with WHIM syndrome as assessed by increasing levels of circulating neutrophils compared with placebo, as well as its safety and tolerability (146). Continued studies will determine whether these drugs represent promise for therapeutic benefit in their respective disease states.

CCR5 in HIV

A large component of drug development in the chemokine system has been focused on inhibition of HIV-1 entry and infection. R5-tropic HIV-1, the predominant form of the virus during initial transmission and infection, requires CCR5 as a co-receptor to CD4 for entry into monocytes and macrophages (147–151). X4-tropic virus, which appears later and is responsible for the serious reduction in Th cell count, requires CXCR4 as a co-receptor to CD4 (147–151). A surge in research on chemokine receptor biology occurred following the findings that CCR5Δ32 homozygous mutant individuals are resistant to HIV infection, and CCR5Δ32 heterozygous mutant individuals had a slower progression to AIDS after infection (152–156). These small populations of individuals have truncated forms of CCR5 that fail to reach the cell surface and, therefore, cannot be utilized by HIV to enter the cell. This led to the development of maraviroc, a CCR5 antagonist which is currently used for combination antiretroviral therapy in adults infected with R5-tropic HIV-1 (157, 158). Drugs which induce receptor internalization of CCR5 or CXCR4, a canonically β-arrestin driven process, may therefore provide significant therapeutic benefit. Interestingly, while there have been worries that such a pharmacological approach would lead to deleterious immunological side effects, CCR5Δ32 mutant individuals do not appear to be immunocompromised (159). However, this finding does not suggest that a drug of this type would not have serious side effect, and future research should rigorously study these details to determine if this is a viable therapeutic strategy.

While endogenous CCR5 ligands CCL3, CCL4, and CCL5 (RANTES) and CCL8 display limited degrees of bias (43), many modified chemokine analogs for CCR5 do exhibit a high degree of bias. AOP-RANTES is β-arrestin-biased agonist, which rapidly induces CCR5 internalization, and inhibits recycling to the cell surface, retaining a large amount of receptor in endosomes (160). Notably, it has been shown that the potency of AOP-RANTES and CCL5 to inhibit R5-tropic HIV-1 infection was correlated with the degree of CCR5 internalization and inhibition of recycling, which supports the notion that these processes represent therapeutic targets to prevent HIV-1 infection (160). Unfortunately, AOP-RANTES simultaneously promoted the intracellular replication of HIV-1 via a mechanism downstream of viral entry that is sensitive to pertussis toxin, suggesting a role for the Gα_i subunit in viral replication (161). PSC-RANTES, is a balanced CCR5 super-agonist that is 50 times more potent than AOP-RANTES for HIV infection inhibition in vitro and acts via a mechanism of long-term intracellular sequestration of CCR5 (162). While PSC-RANTES serves as a potential topical agent to prevent HIV transmission at mucosal barriers, CCR5 induced receptor internalization occurs alongside other CCR5 direct signaling pathways which have been shown to result in intense mucosal inflammation and paradoxically enhance HIV infection (163). 5P14-RANTES is a β-arrestin-biased agonist that induces no detectable G protein activation but does induce receptor internalization, however, not to the extent of PSC-RANTES. PSC-RANTES and 5P14-RANTES also differ in their post-endocytic trafficking of CCR5, a spatiotemporal form of bias that could have significant implications for their ability to inhibit HIV infection (164). PSC-RANTES directs internalized receptor first into the endosome recycling compartment (ERC) and then into the trans-Golgi network (TGN), resulting in long term sequestration, while 5P14-RANTES directs CCR5 only to the ERC, resulting in short term sequestration (164). Bönsch et al. suggest that duration of β-arrestin association with CCR5 is key to the differential post-endocytic receptor trafficking, with PSC-RANTES inducing robust and long-term β-arrestin recruitment to CCR5 and 5P14-RANTES inducing recruitment that is more transient in nature (164).

Altogether, the past roughly 20 years of research on the importance of CCR5 and CXCR4 for HIV entry and infection of leukocytes reveals a pathological state in which biased signaling could provide significant therapeutic benefit. However, it is still not clear exactly to what extent G protein signaling and β-arrestin recruitment, and their associated downstream signaling cascades affect HIV entry and infection. Additionally, the side effect profiles of synthetic ligands which induce receptor internalization have not been determined, an important question for future research to address in order to develop clinically efficacious drugs.

CXCR3 and T cell-mediated inflammation

Similar efforts to target the biased signaling properties of the chemokine system have been focused on CXCR3. CXCR3 is an important mediator of Th1 cell function and plays a role in natural killer cell migration and inhibition of angiogenesis (165). CXCR3 is rapidly expressed on naïve T cells following their activation and remains highly expressed on Th1-type CD4⁺ T cells and effector CD8⁺ T cells (166). CXCR3, regulated by its three interferon-inducible endogenous ligands, CXCL9, CXCL10, and CXCL11, promotes T cell trafficking to peripheral sites of inflammation, as well as lymphoid tissues, facilitating the interaction of T cells with antigen presenting cells (166). CXCR3 and its endogenous ligands have been implicated in a variety of disease processes and pathological states, including allergic contact dermatitis, autoinflammation and autoimmunity, atherosclerosis, and cancer metastasis (116, 167). As previously mentioned, Smith et al. demonstrate the pharmacologic potential of biased signaling at CXCR3 using a mouse model of allergic contact hypersensitivity (116). Their finding that both G protein signaling and β-arrestin-mediated signaling are necessary for full chemotactic function suggests that drugs designed to increase migration of CXCR3⁺ T cells, such as those used in cancer immunotherapy, could promote signaling through the β-arrestin, Gα_i, or potentially both (116). On the other hand, a more complicated picture arises for development of drugs designed to inhibit migration of CXCR3⁺ T cells to reduce inflammation (116). Pertussis toxin treatment completely abrogates T cell chemotaxis, whereas Akt inhibition and β-arrestin 2 knock out (KO) T cells display only reduced chemotaxis (116). Therefore, further research is needed to determine the most efficacious mechanism of action for drugs designed to promote or inhibit migration of CXCR3⁺ T cells. Karin et al. suggest the possibility of using stabilized chemokines as biological drugs for autoimmunity and graft-versus-host disease (GVHD) (119). Endogenous chemokines have relatively short half-lives, and therefore their administration alone does not provide significant therapeutic benefit. Based on their finding that CXCL11 induces T cell polarization toward IL-10-producing Treg cells, whereas CXCL9 and CXCL10 induce T cell polarization toward Th1/Th17 cells, Karin et al. propose that exogenously stabilized chemokines could be developed as biological drugs at CXCR3 that selectively activate FOXP3⁺ Treg or Tr1 cells over proinflammatory cells, taking advantage of the phenomenon of biased agonism (119).

Other CKRs

Lastly, the pharmacologic potential of biased signaling has been implicated at various other CKRs. At CXCR4, drug tolerance has been seen to occur in response to administration of AMD3100, with increasing CXCR4 expression on the surface of hematopoietic stem cells (HSCs), promoting their rehoming to the bone marrow (168). Hitchinson et al. found that such drug tolerance was avoided in vitro with administration of X4-2-6, a β-arrestin-biased peptide antagonist (168). X4-2-6 forms a ternary complex with endogenous CXCL12 and CXCR4, causing the N terminus of CXCL12 to detach from CXCR4. This N terminal interaction is important for inducing the conformational change in CXCR4 that promotes GTP loading of Gα_i, and X4-2-6 therefore inhibits G protein signaling (168). Receptor C terminal phosphorylation, β-arrestin recruitment, and receptor internalization are retained, creating a biased signaling profile that demonstrates decreased drug tolerance (168). These findings suggest the potential of biased allosteric modulators to provide therapeutic benefit to patients who develop tolerance to unbiased antagonists. At CCR2 and CCR5, synthetic enantiomers J113863 and UCB35625 exhibit differential G protein signaling and β-arrestin recruitment, as well as functional effects on L1.2 cell (a mouse leukemia cell line) chemotaxis (169). This represents the first description of chemokine receptor ligands displaying enantioselective properties, a finding of particular pharmacologic significance as many drugs are typically racemic mixtures (169).

Conclusion

The discovery of alternative signaling pathways beyond G protein signaling, specifically those driven by β-arrestin, have led to a renewed appreciation of the complexity and non-redundant nature of the chemokine system. The number of possible cellular phenotypes driven by the chemokine signaling are staggering when considering the diversity of ligands, receptors, and cell and tissue types implicated in this system. Rather than being designed to simply produce robust outputs, numerous functional and structural studies suggest that the chemokine system is capable of driving highly specific and directed signaling processes to coordinate immune cell function. Considering that CKRs are found in many non-immune related tissue types, the complexity of the biased signaling within this system is likely underappreciated. An enhanced understanding of the pluridimensional efficacy within the chemokine system presents an opportunity to determine the fundamental principles underlying biased agonism across all GPCRs. Additionally, the phenomenon of biased signaling within the chemokine system presents a potential target for future drug development. Within the complex signaling mechanisms of the chemokine system lies the potential for greater druggability, with more focused pharmacological effects and possibly fewer harmful side effects. Although biased agonism has been widely observed using both endogenous and synthetic chemokine ligands, understanding the mechanisms and functional effects of this complex signaling in vivo must progress in order to develop promising therapeutics. Many of the problems faced in pharmaceutical research on chemokine-targeted therapeutics relate to an apparent lack of translation between cellular, animal, and human models, as well as an incomplete understanding of the biochemical and physiological implications of the chemokine system bias. Advances in our understanding of the structural and functional determinants of biased agonism may be necessary before the promise of more effective therapeutics aimed at treating CKR- and GPCR-related pathologies can be fully realized.

Highlights.

• Biased agonism within the chemokine system drives highly specific cellular outputs

• Chemokine system is a model for biased agonism at G protein-coupled receptors

• Utilizing biased agonism in the chemokine system for potential therapeutic purposes