Genetic Methods.

The first method belonging to this class is the yeast two-hybrid assay (Fields and Song, 1989). In this method, the protein of interest (the bait protein) is expressed in yeast as a fusion to the DNA-binding domain of a transcription factor lacking the transcription activation domain. To identify partners of this bait, a plasmid library that expresses cDNA-encoded proteins fused to a transcription activation domain is introduced into the yeast strain. Interaction of a cDNA-encoded protein with the bait protein results in the activation of the transcription factor and expression of a reporter gene, enabling growth on specific media or a color reaction and the identification of the cDNA encoding the target proteins. A first disadvantage is the loss of spatial-temporal localization of the interaction; in fact, the yeast two-hybrid assay only captures a snapshot of potential interactions in an artificial biologic system. A second disadvantage is that it is not possible to investigate membrane-anchored proteins since the two proteins must cross the nuclear membrane to carry the reconstituted transcription factor to the DNA. To overcome this issue, the membrane yeast two-hybrid assay (Stagljar et al., 1998) was developed. In this assay, the ubiquitin protein is split into two fragments that are fused to the two proteins of interest. The ubiquitin C-terminal fragment is then conjugated to a transcription factor that is released when the interaction occurs, and the ubiquitin protein is reformed. However, as in the yeast two-hybrid assay, the spatial-temporal localization of the interaction is lost. A second limitation is that the ubiquitin C-terminus carrying the transcription factor cannot be fused to soluble proteins because they could diffuse into the nucleus. Therefore, a mammalian version of the assay called mammalian membrane two-hybrid (Petschnigg et al., 2014) was developed. The kinase substrate sensor assay (Lievens et al., 2014), using the signal transducer and activator of transcription 3 (STAT3) as transcription factor, can also be used for investigating protein-protein interactions, including those involving cytosolic and membrane proteins in mammalian cells. However, the kinase substrate sensor assay cannot be used for studying GPCR interaction with proteins involved in the STAT3 cascade.

Biophysical Methods.

Energy transfer–based methods, such as bioluminescence and fluorescent resonance energy transfer [BRET (Xu et al., 1999) and FRET (Clegg, 1995)] assays, are targeted methods that are generally used to investigate previously reported interactions. The basis of both is the transfer of energy from a donor to a nearby acceptor (<100 Å). Their high sensitivity allows the study of weak and transient interactions. The high spatial-temporal resolution permits accurate kinetic studies for investigating interaction dynamics.

Another biophysical method, with FRET as a basis and employed for the study of protein-protein interaction, is fluorescent lifetime imaging microscopy (Sun et al., 2012). The fluorescence lifetime is the average time that a molecule spends in the excited state before returning to the ground state. Since in FRET the energy transfer from the acceptor to the donor depopulates the excited state energy of the latter, it also shortens its lifetime. Fluorescent lifetime imaging microscopy can accurately measure the shorter donor lifetime that results from FRET; thus, it allows mapping of the spatial distribution of protein-protein interactions in living cells (Sun et al., 2011). Its main advantage over intensity-based FRET is the more accurate measurement of FRET, because only donor signals are measured, which eliminates the corrections for spectral bleed-through (Sun et al., 2011). Its main disadvantages are the necessity of a live specimen and the complexity of data recording and analysis.

Biochemical Methods.

The proximity ligation assay (Fredriksson et al., 2002) is another powerful targeted method. In the direct version of the technique, two DNA oligonucleotide-conjugated antibodies are used against the proteins of interest. In the indirect version, secondary DNA-conjugated antibodies are used after the proteins of interest are targeted with an appropriate primary antibody. If the two conjugated antibodies are close enough (30–40 nm), they can bind together. The DNA connecting the two probes is then amplified and hybridized with fluorophores. This allows the visualization of the interaction in the place where it occurs, at a single-molecule resolution. The main disadvantages of the approach are the high costs and the necessity for specific antibodies, which are not always available.

In the bimolecular fluorescent complementation assay (Hu et al., 2002; Hu and Kerppola, 2003), a fluorescent protein is divided into two nonfluorescent fragments that are fused to the proteins of interest. Interaction reconstitutes the entire fluorescent protein. This method allows the direct visualization of the interaction and can be used for soluble or membrane-bound proteins. In addition, several interactions can be investigated in parallel through the use of different fluorescent proteins. Since there is a delay in fluorescence formation upon reconstitution of the fluorescent proteins, and the fluorophore formation is irreversible, these methods are usually not well suited for kinetic studies. To overcome these limitations, a novel assay called NanoBiT was developed. In this assay the nanoluciferase enzyme is divided in two subunits (LgBiT and SmBiT), with low affinity for each other, that can be brought together by the two interacting proteins. The low affinity makes the interaction reversible and therefore suitable for the investigation of kinetics (Duellman et al., 2017).

Proteomics Methods.

Proteomic methods aim to identify GIPs of a receptor of interest via the use of affinity purification coupled with mass spectrometry (AP-MS). This approach is usually employed as an unbiased method for screening virtually all the GIPs of a GPCR of interest. Targeted versions of the method also exist and rely on GIP identification by Western blotting. However, the requirement for specific antibodies seriously limits its application. Several strategies can be used for the affinity purification step. In co-immunoprecipitation (Co-IP), specific antibodies against the protein of interest are used for precipitating the bait from a protein lysate. As specific GPCR antibodies providing high immunoprecipitation (IP) yields are rarely available, epitope-tagged versions of the receptor of interest are often expressed in the cell type or the organism of interest and precipitated using antibodies against the tag. The main advantages of Co-IP are the purification of proteins interacting with the entire receptor (whenever possible the native receptor) in living cells or tissues and its ability to purify the entire associated protein complex. The main limitations are the necessity for specific antibodies to precipitate GPCRs, the loss of spatial-temporal information, and the use of detergents for cell lysis that might denaturate the GPCR of interest and, accordingly, disrupt interactions with their protein partners. For this reason, special attention must be paid to lysis conditions that efficiently solubilize the receptor while conserving the receptor’s native conformation and its interactions with GIPs. For instance, detergents such as Triton and NP-40 completely denaturate CXCR4 (Palmesino et al., 2016), whereas 3-([3-cholamidopropyl]dimethylammonio)-2-hydroxy-1-propanesulfonate, also called CHAPSO (Babcock et al., 2001), and n-dodecyl-β-d-maltopyranoside, also called DDM (Palmesino et al., 2016), yield the highest proportion of receptor in the native conformation. Despite this limitation, several CXCR4-interacting proteins have been identified using Co-IP approaches (see Table 2).

Alternatively, pull-down assays can be performed to purify GPCR partners from a cell or tissue lysate. This approach uses the receptor (or one of its domains) fused with an affinity tag (e.g., glutathione-S-transferase) and immobilized on beads as bait. Such in vitro binding assays can also be used to prove direct physical interaction between two protein partners. In this case, the bait is incubated with a purified protein instead of a cell or tissue lysate. In all methods, affinity purified proteins are systematically identified by liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS). A two-step version, named tandem affinity purification (Puig et al., 2001), has also been reported (Daulat et al., 2007) and applies to both Co-IPs or pull-downs. Although tandem affinity purification methods drastically reduce the number of false-positive identifications, they require larger amounts of starting material.

In the proximity-dependent biotin identification method (Roux et al., 2013), the bait protein is fused to a prokaryotic biotin ligase molecule that biotinylates proteins in close proximity once expressed in cells. The method can detect weak and transient interactions occurring in living cells, and detergents do not affect the results. Though the fusion of biotin ligase to the bait might alter its targeting or functions, proximity-dependent approaches were recently applied to identify a GPCR-associated protein network with a high temporal resolution. Specifically, engineered ascorbic acid peroxidase was employed in combination with quantitative proteomics to decipher β-2 adrenergic (Lobingier et al., 2017; Paek et al., 2017) and angiotensin II type 1 (Paek et al., 2017) receptor–interacting proteins.

Radioactive Labeling.

The first method that was introduced for deciphering the phosphorylation status of GPCRs is a radioactivity-based technique, consisting of culturing cells in a medium in which phosphate is replaced with its radioactive counterpart, ³²P, resulting in radioactive phosphorylated residues (Meisenhelder et al., 2001). After culture, cells are lysed and receptors are immunoprecipitated using specific antibodies, and then resolved by SDS-PAGE. Receptors can then be digested using an enzyme such as trypsin, and the resulting peptides are separated by two-dimensional migration using electrophoresis and chromatography. Radioactive peptides are then detected in-gel by autoradiography or using a phosphorimager, yielding a phosphorylation map for a given receptor in a given cell line (Chen et al., 2013). This method is very sensitive but does not give precise information on the number of phosphorylated sites nor their position. Radioactive labeling was initially employed to characterize CXCR4 phosphorylation upon agonist stimulation (Haribabu et al., 1997). These studies characterized the C-terminal domain as the preferred site for phosphorylation (Haribabu et al., 1997) and identified a serine cluster present in the C-terminal domain and containing two residues (Ser³³⁸, Ser³³⁹) phosphorylated following CXCL12 challenge (Orsini et al., 1999).

LC-MS/MS.

More recently, radioactive labeling-based methods have been progressively supplanted by the identification of phosphorylated residues by LC-MS/MS. In this method, the GPCR of interest is digested enzymatically, with one or several proteases, to generate peptides that cover a large part of the receptor sequence. The resulting peptides are then analyzed by LC-MS/MS (Dephoure et al., 2013). Although this approach can pinpoint any phosphorylated residue with high confidence, a few limitations complicate phosphorylated residue identification. First, phosphorylation can be lost during fragmentation. Second, since phosphorylation sites have a limited level of phosphorylation, only a small percentage of peptide is actually phosphorylated (Wu et al., 2011). Third, the identification of the phosphorylated residues in peptides with multiple adjacent phosphorylated residues can be challenging. For each modified site, a phosphorylation index can be estimated by dividing the ion signal intensity corresponding to the phosphorylated peptide by the sum of the ion signals of the phosphorylated peptide and its nonphosphorylated counterpart. Absolute quantification, and thus the stoichiometry of phosphorylation, can also be determined for each modified residue by spiking the sample with a known concentration of high-purity, heavy isotope–labeled peptides (AQUA peptides) that correspond to the phosphorylated peptide and not phosphorylated one. The respective ion signals of unlabeled and labeled peptides are then compared (Gerber et al., 2003). This powerful technology allowed a first comprehensive phosphorylation map of CXCR4, stimulated or not with CXCL12 in human embryonic kidney (HEK)293 cells: LC-MS/MS analyses identified six phosphorylated residues: Ser³²¹, Ser³²⁴, Ser³²⁵, one between Ser^338/339/341, one between Ser^346/347/348, and either Ser³⁵¹ or Ser³⁵² (Busillo et al., 2010).

Mutagenesis.

Another approach that can be used as a stand-alone technique or in complement with the previously described methods is to mutate potential or predicted phosphorylated residues (into alanine or aspartate to inhibit or mimic their phosphorylation, respectively) and assess functional differences between mutated and wild-type receptor (Okamoto and Shikano, 2017). Nevertheless, introducing those mutations can potentially alter expression of the receptor, its conformation, or its cellular localization. Despite these limitations, mutagenesis approaches have shown unequivocal efficiency in identifying or validating several phosphorylated residues on CXCR4 (Orsini et al., 1999; Mueller et al., 2013) in combination with a radioactive-labeling method or use of phospho-specific antibodies. Furthermore, mutating all the serine and threonine residues to alanine in the ACKR3 C-terminus abolished β-arrestin recruitment and receptor internalization, suggesting that receptor trafficking depends on the phosphorylation of some of these residues (Canals et al., 2012).

Phospho-Specific Antibody.

To be able to detect and assess phosphorylation of residues in cells or tissues, antibodies that specifically target previously identified phosphorylated residues of GPCRs can be generated by immunizing animals with purified synthetic phosphorylated peptides encompassing the phosphorylated residues (Chen et al., 2013). After selection and functional validation, those antibodies can be used in Western blot or immunohistochemistry experiments. Phosphorylation can also be indirectly detected using antibodies specific to the unphosphorylated GPCR, showing decreased binding to the target when residues are phosphorylated, and recovery of the binding by using a protein phosphatase to dephosphorylate the receptor (Hoffmann et al., 2012). Antibodies that recognize several CXCR4 phosphorylated residues [Ser^324/325, Ser³³⁰, Ser³³⁹, Ser^338/339, and Ser^346/347 (Woerner et al., 2005; Busillo et al., 2010; Mueller et al., 2013)] have been generated and used to investigate the receptor phosphorylation status in various conditions. To our knowledge, such phospho-specific antibodies are still lacking for ACKR3.

G Proteins.

CXCR4 is known to couple to the pertussis toxin–sensitive Gα_i protein family that mediates most of its signaling pathways (Busillo and Benovic, 2007). However, CXCR4 can also couple to other G proteins such as Gα_12/13 (Tan et al., 2006; Kumar et al., 2011) and Gα_q (Soede et al., 2001). Tan and colleagues (2001) observed that both Gα_i and Gα₁₃, as well as Gβγ subunits are involved in the CXCL12-dependent migration of Jurkat T cells. Specifically, Gα_i proteins promote migration through the activation of Rac, whereas Gα_12/13 proteins activate Rho. Though CXCR4 is coupled to both Gα_12/13 and Gα_i proteins in Jurkat cells, such a dual coupling has not been observed in other cell lines where the receptor specifically activates one or the other G protein family (Yagi et al., 2011). In fact, pertussis toxin inhibited the migration of nonmetastatic breast cancer cells (MCF-7), indicating that Gα_i activation is required. However, it did not prevent the migration of metastatic breast cancer cells such as MDA-MB-231 and SUM-159. In those cell lines, Gα_12/13 activation mediates CXCL12-induced migration via the activation of the Rho signaling axis (Yagi et al., 2011). Therefore, CXCR4 coupling to one or the other G protein family might depend on the abundance of GPCRs, G proteins, and downstream targets.

As an atypical chemokine receptor, ACKR3 lacks the DRYLAIV (Asp-Arg-Tyr-Leu-Ala-Ile-Val) motif necessary for interaction with G proteins. Nevertheless, using BRET, a study showed interaction of the receptor with G proteins in transfected HEK293 cells (Levoye et al., 2009). Yet, this interaction did not lead to the activation of G proteins (Levoye et al., 2009), reinforcing the common consensus that ACKR3 is unable to activate G proteins. Consistently, other studies showed that ACKR3 signals independently of G proteins through a mechanism requiring β-arrestins (Rajagopal et al., 2010; Canals et al., 2012).

Although these findings clearly indicate that ACKR3 cannot activate G proteins in most of the cell types, a report suggested that ACKR3 might activate G proteins in two specific cellular contexts, namely primary rodent astrocytes and human glioma cells (Ödemis et al., 2012). Using [³⁵S]GTPγS-binding assay, calcium mobilization, and pertussis toxin-dependent activation of downstream signaling pathways [extracellular signal–regulated kinases (ERKs 1/2 and AKT phosphorylation), this group showed an ACKR3-dependent activation of Gα_i/o proteins in primary astrocytes (Ödemis et al., 2012). They also reported pertussis toxin-dependent migration, proliferation, and activation of downstream signaling effectors in two human glioma cell lines (A764 and U343), further suggesting an ACKR3-dependent activation of Gα_i/o. So far, this is the only report suggesting a possible coupling of ACKR3 with G proteins. Though these data must be further confirmed, one possibility is that such a coupling is cell type–specific. Since ACKR3 was shown to form a heterodimer with CXCR4 in transfected cell lines (Levoye et al., 2009) and CXCR4 is well known for its coupling with G proteins (vide supra), the authors also investigated the possibility that the ACKR3-dependent activation of Gα_i/o proteins was mediated by the ACKR3/CXCR4 complex. However, constitutive suppression of CXCR4 expression in primary astrocytes did not influence the ability of ACKR3 to activate G proteins in the [³⁵S]GTPγS-binding assay (Ödemis et al., 2012). Consistently, transient suppression of CXCR4 expression did not influence the ACKR3-dependent calcium mobilization in primary cultures of astrocytes. This suggests that ACKR3 coupling to G proteins in astroglial cells, if any, occurs independently of the ACKR3/CXCR4 complex assembly.

Although in that specific case CXCR4 did not influence ACKR3 signaling, accumulating evidence supports the hypothesis that ACKR3 might conversely influence CXCR4 signaling. Specifically, the organization of CXCR4 and ACKR3 in heterodimers appears to inhibit CXCR4 interaction with G proteins in transfected HEK293T cells, as assessed by saturation BRET (Levoye et al., 2009). In accordance with a possible influence of ACKR3 on CXCR4-dependent Gα_i activation, Sierro and colleagues (2007) showed that the coexpression of ACKR3 with CXCR4 hindered the fast and G protein-dependent ERK activation triggered by CXCL12 exposure. In spite of these observations demonstrating that ACKR3 influences CXCR4-dependent G protein signaling, a direct proof of the role of the physical interaction between both receptors is still missing.

GRKs.

Agonist-occupied GPCRs are specifically phosphorylated by different GRKs, a family of seven serine/threonine kinases (Ribas et al., 2007; Penela et al., 2010a). GRKs 2, 3, 5, and 6 phosphorylate CXCR4 in the C-terminus, which contains 15 serine and three threonine residues that are potential phosphorylation sites (Fig. 1). At least six of these residues were shown to be phosphorylated following receptor activation by CXCL12 (Busillo et al., 2010; Barker and Benovic, 2011; Mueller et al., 2013). In HEK293 cells, Ser³²¹, Ser³²⁴, Ser³²⁵, Ser³³⁰, Ser³³⁹, and two sites between Ser³⁴⁶ and Ser³⁵² were shown to be phosphorylated in response to CXCL12 in the CXCR4 C-terminus using LC-MS/MS and phosphosite-specific antibodies (Busillo et al., 2010). GRK6 is able to phosphorylate Ser^324/5, Ser³³⁹ and Ser³³⁰, the latter with slower kinetics, whereas GRK2 and GRK3 phosphorylate residues between Ser³⁴⁶ and Ser³⁵² (Fig. 2) (Busillo et al., 2010), and specifically Ser^346/347 (Mueller et al., 2013). Interestingly, the latter study suggested a hierarchy in such phosphorylation events, since Ser^346/347 phosphorylation is achieved faster and is needed for the subsequent phosphorylation of Ser^338/339 and Ser^324/325. Notably, ligand washout resulted in rapid Ser^324/325 and Ser^338/339 dephosphorylation, whereas Ser^346/347 residues did not exhibit major dephosphorylation during the 60-minute period studied (Mueller et al., 2013). Phosphorylation of CXCR4 by different GRKs can elicit several molecular responses, such as fluctuations in intracellular calcium concentration and phosphorylation of ERKs 1 and 2, leading to integrated cellular responses. In HEK293 cells, calcium mobilization is negatively regulated by GRK2, GRK6, and β-arrestin2. On the other hand, GRKs 3 and 6 together with β-arrestins act as positive regulators of ERK1/2 (Busillo et al., 2010). Overall, these studies show nonoverlapping roles for the different GRKs in the regulation of CXCR4 signaling. These differential roles may explain distinct cell type–dependent responses to CXCL12. However, what dictates the specific GRK subtype recruitment still needs to be investigated. Changes in the normal CXCR4 phosphorylation pattern as a result of receptor mutations or altered GRK activity can lead to abnormal receptor expression and/or responsiveness that promotes aberrant cell signaling and thus can contribute to several pathologies. Deletion of Ser^346/347 leads to a gain of CXCR4-function and decreases receptor internalization and subsequent desensitization, indicating that mutations in the far C-terminus affect CXCR4-mediated signaling (Mueller et al., 2013). In this regard, a subpopulation of patients affected by WHIM (warts, hypogammaglobulinemia, infections, and myelokathexis) syndrome, a rare primary immunodeficiency disease, display C-terminally truncated CXCR4, leading to refractoriness to desensitization and enhanced signaling (Balabanian et al., 2008). On the contrary, increased CXCR4 phosphorylation at Ser³³⁹ is associated with poor survival in adults with B-cell acute lymphoblastic leukemia and correlates with poor prognosis in acute myeloid leukemia patients (Konoplev et al., 2011; Brault et al., 2014). Altered GRK expression/activity can also impair CXCR4 phosphorylation patterns. GRK3 suppression may contribute to abnormally sustained CXCR4 signaling in classic types of glioblastoma (Woerner et al., 2012), some WHIM patients (Balabanian et al., 2008), and in triple negative breast cancer, thus potentiating CXCR4-dependent migration, invasion, and metastasis (Billard et al., 2016; Nogués et al., 2018). It is interesting to note that, although GRKs 2 and 3 share a high homology and are able to phosphorylate the same residues in CXCR4 in model cells, their function is not redundant. Whereas both CXCR4 and GRK2 levels are increased in breast cancer patients, GRK3 is decreased, suggesting a differential role for both GRKs in a cancer context (Billard et al., 2016; Nogués et al., 2018). In fact, deregulation of GRK2 potentiates several malignant features of breast cancer cells, and its level positively correlates with tumor growth and increased metastasis occurrence (Nogués et al., 2016), but whether these roles involve changes in CXCR4 modulation is still under investigation. On the other hand, impaired chemotaxis of T and B cells toward CXCL12 is noted in the absence of GRK6, whereas GRK6 deficiency potentiates neutrophil chemotactic response to CXCL12 (Fong et al., 2002; Vroon et al., 2004), suggesting that the occurrence of highly cell type–specific mechanisms in the control of the CXCL12-CXCR4-GRK6 axis. Overall, these data indicate the complexity of CXCR4 modulation by GRKs and support the need for a better characterization of cell type– or disease-specific CXCR4-GRKs interactions.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

CXCR4 and ACKR3 residues potentially subjected to post-translational modifications. Schematic representation of the C-terminal tail of CXCR4 and ACKR3, in which serine/threonine (red), tyrosine (green), and lysine (blue) residues potentially subjected to post-translational modifications are highlighted.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

CXCR4 C-terminus phosphosites. Schematic representation of the C-terminal tail of CXCR4, in which serine residues known to be phosphorylated are highlighted in light blue. The kinases or the extracellular stimuli responsible for the phosphorylation are also specified. Hrg, heregulin.

ACKR3 has lately been the focus of many studies, in particular because of its role in cancer progression and metastasis. However, the mechanisms underlying its regulation are still not well understood, although this receptor has been shown to interact with GRKs and arrestins and to associate with other partners. The C-terminus of ACKR3 contains five serine and four threonine residues that can potentially be phosphorylated (Fig. 1). Unlike for CXCR4, little is known about their actual phosphorylation status during the activation of the receptor, as no mass spectrometry data are available to date and only few mutational studies have been conducted (Canals et al., 2012; Hoffmann et al., 2012). In fact, only one study conducted in astrocytes showed that ACKR3 is phosphorylated by GRK2, but not other GRKs, and that this phosphorylation is essential for subsequent ACKR3-operated activation of ERK1/2 and AKT pathways (Lipfert et al., 2013). This study suggests that ACKR3 is indeed phosphorylated by GRKs, but the isoform(s) involved and subsequent responses are probably cell type–dependent and remain to be investigated in detail.

Arrestins.

A study revealed that site-specific phosphorylation of CXCR4 by GRK isoforms has contrasting effects upon β-arrestin recruitment: Although receptor phosphorylation at its extreme C-terminus [two residues between Ser³⁴⁶ and Ser³⁵² (Busillo et al., 2010), and specifically Ser^346/347 (Mueller et al., 2013)] by GRK2/3 is a necessary step in β-arrestin binding; its phosphorylation by GRK6 at upstream residues [Ser^324/5, Ser³³⁰, and Ser³³⁹ (Busillo et al., 2010)] appears to inhibit arrestin recruitment to CXCR4 or results in a receptor/arrestin complex that adopts a conformation that is distinct from that induced by phosphorylation of extreme C-terminal residues (Oakley et al., 2000; Busillo et al., 2010; Mueller et al., 2013). Further supporting the importance of Ser^324/5 and Ser³³⁹ phosphorylation in β-arrestin recruitment, CXCR4 truncation mutants showing impaired phosphorylation at Ser^324/325 and Ser^338/339 also exhibit reduced CXCL12-induced receptor internalization (Mueller et al., 2013).

β-arrestins are also scaffold proteins for several signaling molecules, thus eliciting additional β-arrestin-dependent signaling pathways (Shenoy and Lefkowitz, 2011; Peterson and Luttrell, 2017). Following the recognition of β-arrestin-dependent signaling, the notion of biased ligands that preferentially induce G protein-dependent or independent signaling has emerged (Reiter et al., 2012). Biased signaling at chemokine receptors has been exhaustively reviewed elsewhere (Steen et al., 2014). For instance, a CXCR4-derived pepducin, ATI-2341, acts as a biased CXCR4 agonist that promotes Gα_i signaling but not β-arrestin signaling, in contrast to CXCL12, which activates both G protein-dependent and independent pathways (Quoyer et al., 2013; Steen et al., 2014).

Upon activation by its cognate ligands, CXCL11 and CXCL12, ACKR3 recruits β-arrestin2 both in vitro (Rajagopal et al., 2010; Benredjem et al., 2016) and in vivo (Luker et al., 2009), a process leading to receptor internalization (Canals et al., 2012), transport to lysosomes, and degradation of the receptor-bound chemokine (Luker et al., 2010; Naumann et al., 2010; Hoffmann et al., 2012). The receptor is then mainly recycled back to the plasma membrane (Luker et al., 2010) even if a partial degradation of ACKR3 can be observed (Hoffmann et al., 2012). Interestingly, the rate of receptor internalization is faster and recycling is lower in the presence of CXCL11, compared with CXCL12 (Montpas et al., 2018).

As previously mentioned, systematic mutation of C-terminal serine/threonine residues to alanine abolished ligand-induced β-arrestin2 recruitment to ACKR3, as monitored by BRET (Canals et al., 2012) and decreased ACKR3 internalization and subsequent degradation of radiolabeled CXCL12 in HEK293 cells (Hoffmann et al., 2012). Selective mutations of the two C-terminal serine/threonine clusters to alanine revealed differences in their functional properties. Mutating Ser³³⁵, Thr³³⁸, and Thr³⁴¹ (first cluster) or Ser³⁵⁰, Thr³⁵², and Ser³⁵⁵ (second cluster) to alanine decreased CXCL12 internalization only after a 5-minute challenge but not following longer agonist receptor stimulation. Yet, only mutation of the second cluster prevented CXCL12 degradation. Furthermore, ACKR3 appears to undergo ligand-independent internalization to a much greater extent than CXCR4 (Ray et al., 2012), and residues 339–362 (the two serine/threonine clusters) are essential for this peculiar cell fate in HEK293 cells. Although numerous studies have shown that ACKR3 internalization and the resulting chemokine degradation are dependent on β-arrestin, recent findings have been challenging this consensus (Montpas et al., 2018). Specifically, this study shows that β-arrestins are dispensable to chemokine degradation, suggesting that other scaffold proteins might be involved in this process.

Functional Interaction of CXCR4 with Second Messenger–Dependent Kinases and Receptor Tyrosine Kinases.

Accumulating evidence indicates that phosphorylation of GPCRs by second messenger–dependent kinases such as protein kinase A and protein kinase C (PKC) (Lefkowitz, 1993; Ferguson et al., 1996; Krupnick and Benovic, 1998), as well as by members of the receptor tyrosine kinase family (Delcourt et al., 2007), participates in the regulation of GPCR signaling. CXCR4 is phosphorylated by PKC at Ser^324/5 upon CXCL12 stimulation (Busillo et al., 2010), and this kinase has also been involved in Ser^346/7 phosphorylation (Luo et al., 2017), even though these results are not entirely consistent with a previous study using different PKC inhibitors (Mueller et al., 2013). In some glioblastoma cell types, CXCR4 is phosphorylated at Ser³³⁹ in response to the PKC activator phorbol myristate acetate (Woerner et al., 2005). This suggests that Ser³³⁹ is also a PKC phosphorylation site, and that this phosphorylation event may serve as a crosstalk mechanism between CXCR4 and GPCRs that activate Gα_q-PKC signaling. Sphingosine 1-phosphate receptors, neurokinin-1 and lysophosphatidic acid receptors may possibly be involved in glioblastoma progression via this means (Cherry and Stella, 2014). Nevertheless, the functional impact of Ser³³⁹ phosphorylation in glioblastoma remains to be established. Likewise, epidermal growth factor (EGF) through activation of its receptor can also promote CXCR4 phosphorylation at Ser³³⁹ in glioblastoma cells (Woerner et al., 2005), and both EGF and heregulin trigger Ser^324/325 and Ser³³⁰ phosphorylation in the breast cancer T47D cell line (Sosa et al., 2010). Interestingly, in MCF7 breast cancer cells, heregulin also promotes CXCR4 phosphorylation on tyrosine residues via epidermal growth factor receptor (EGFR), leading to β-arrestin2 association with CXCR4 and downstream activation of the PRex1/Rac1 axis. However, it is still unclear whether the EGFR-CXCR4 functional interaction is direct or depends on other kinases (Sosa et al., 2010). In another breast cancer line, BT-474, CXCR4 is phosphorylated on tyrosine residues in response to CXCL12 and through activation of ErbB2/ErbB3 and EGFR (Sosa et al., 2010). Although the specific Tyr residue(s) phosphorylated were not identified, it is worth noting that CXCR4 displays four intracellular Tyr residues (Ahr et al., 2005). Tyr¹⁵⁷ in the third intracellular loop has been be involved in CXCR4-dependent STAT3 signaling (Ahr et al., 2005), whereas Tyr¹³⁵, within the conserved DRY motif, might be involved in receptor coupling to G proteins (Rovati et al., 2007). Consistent with this hypothesis, EGFR-mediated phosphorylation of the equivalent Tyr in another GPCR (the μ-opioid receptor) has been reported to reduce coupling to G proteins (Clayton et al., 2010). Therefore, identification of tyrosine residues phosphorylated in CXCR4 might add some insight into the mechanisms by which growth factor–receptor tyrosine kinases modulate CXCR4 activity. The crosstalk between CXCR4 and ErbB2/ErbB3 and EGFR remains an interesting avenue for future research, given the involvement of both receptors in cancer.

Physical Interaction with Noncanonical GPCR-Interacting Proteins.

Beside canonical GIPs, CXCR4 has been shown to interact with additional proteins that modulate CXCR4 trafficking, subcellular localization and signaling, and proteins whose functions are still unknown. CXCR-interacting proteins, the methods used for the identification of these proteins, the site of their interaction in the receptor sequence, and their functional impact are summarized in Table 2.