CXCR7 recruits β-arrestin2 with higher efficacy in a ligand-dependent manner, compared with CXCR4 and CXCR3
CXCR7 shares ligands with both CXCR4 and CXCR3 [37], but its receptor-mediated signaling pathways are not completely understood. One of the clear responses elicited by CXCR7 is ligand-dependent β-arrestin2 recruitment. To explore signaling by this chemokine receptor, β-arrestin2 recruitment assays were developed on the basis of structural complementation using NanoBiT technology. This technology uses two separate fragments of Nano Luciferase (Nluc), a small protein that catalyzes a bright luminescent reaction. Binding of the small fragment (SmBiT) with the large fragment (LgBiT) of Nano Luciferase produces a bright luminescent signal, but these fragments normally have very low affinities for each other, so their close proximity, driven by the interactions of fusion partners, can be used to study protein-protein interactions [38]. If these interactions are reversible, then both the associations and dissociations of the targeted proteins can be monitored. There is no lag time for luminescence lost, so detection is immediate, and accurate temporal dynamics of the protein interactions can be determined [39]. To monitor β-arrestin2 recruitment at chemokine receptors, four different plasmid combinations were screened for β-arrestin2 fused with NanoBiT fragments (LgBiT or SmBiT at either the N-terminus or C-terminus) and two types (LgBiT or SmBiT) of the C-terminal-fused chemokine receptors (CXCR3, 4, and 7) (Fig. 1a). The plasmid combination that showed the best response was selected for further studies. Interactions between β-arrestin2 and the CXCRs were immediately observed after ligand stimulations, reaching maximal luminescence 5 min after ligand addition. All combinations of CXCR7 and β-arrestin2 showed an increase in luminescence by SDF-1α and I-TAC with different efficacies (Fig. S1). CXCR4 combinations showed a wide variety of efficacies by SDF-1α treatment (Fig. S2). Notably, CXCR4-SmBiT and LgBiT-β-arrestin2 showed no increase in luminescence signal. For CXCR3, the response patterns from I-TAC were very similar among all combinations (Fig. S2). Neither CXCR3 nor CXCR4 showed any responses to SDF-1α or I-TAC respectively in all plasmid combinations, supporting previous reports that SDF-1α is not a ligand for CXCR3 and I-TAC is not a ligand for CXCR4 (Figs. 1b and S2). CXCR7-mediated responses to SDF-1α and to I-TAC were much higher than CXCR3 and CXCR4 responses. SDF-1α induced higher luminescence signals than I-TAC in cells expressing CXCR7, indicating that SDF-1α induces receptor conformational changes favorable to β-arrestin2 binding in comparison to I-TAC (Fig. 1b).
β-arrestin2 recruitment to the chemokine receptors precedes their internalization from the plasma membrane [40]. This internalization of CXCRs was monitored using β-arrestin2-GFP fusion proteins. As shown in Fig. 1c, β-arrestin2-GFP was localized to the cytosol in the absence of ligand. Upon SDF-1α treatment of cells expressing CXCR4 or CXCR7, GFP signals were clustered in the cytosol as aggregates, indicating that β-arrestin2 was translocated to specific regions of the cytosol, perhaps as early endosomes. CXCR4-GFP was mainly localized to the plasma membrane in the absence of ligand, and translocated to the cytosol after ligand treatment, whereas most CXCR7-GFP was detected in the cytosol regardless of ligand treatment, indicating cytosol retention that made assessment of receptor spatial locations difficult (Fig. 1d left panel). To examine receptor-only behavior in the plasma membrane, as described previously, cells expressing HA-tagged receptors were incubated with anti-HA antibodies prior to SDF-1α treatment [41]. We observed changes in the localization of CXCR4 from the plasma membrane to the cytosol after ligand treatment, and CXCR7 expression in the plasma membrane was weak but detectable, and its trafficking could be monitored after ligand treatment (Fig. 1c, right panel). Membrane expressions of chemokine receptors were further determined by HiBiT assay. Luminescence signals in cells expressing SmBiT-receptors became stronger depending on the amount of transfected plasmid. Luminescence in cells expressing CXCR7 was approximately 10-fold or 3-fold lower than luminescence in CXCR4- or CXCR3-expressing cells, respectively (Fig. 1e). According to these data, CXCR7 seems to interact strongly with β-arrestin2 compared with CXCR4, even with poor plasma membrane localization. The strong affinity of CXCR7 toward β-arrestin2 may be due to a higher binding affinity to the ligand compared with CXCR4 or CXCR3, as described previously [37,9], or the ligand-bound receptor may undergo conformational changes that are more suitable for β-arrestin2 interaction.
SDF-1α-dependent internalization of CXCR4 and CXCR7 was confirmed using two structural complementation assays based on NanoBiT technology. For the first assay, constructs receptor-LgBiT and SmBiT-FYVE domain were used, where the FYVE domain was used as an early endosome marker. The luminescence signal was significantly increased few minutes after ligand treatment, indicating that the receptor was internalized via early endosomes (Fig. S3a). The second structural complementation assay was based on the plasma membrane marker CAAX. When a receptor is expressed in the plasma membrane, its close proximity to CAAX produces high luminescence. However, when a receptor is activated by ligand, internalization occurs and the receptor is no longer in close proximity to CAAX, resulting in decreased luminescence. The luminescence signal of CXCR4-SmBiT and LgBiT-CAAX was decreased by SDF-1α (Fig. S3b, left). The luminescence signal of CXCR7-SmBiT and LgBiT-CAAX was also decreased by SDF-1α, but recovered slightly after 30 min (Fig. S3B, right), implying a dynamic localization of CXCR7 such as possibly recycling or translocation of cytosolic CXCR7. The fast recovery of CXCR7 to the membrane may have been due to dominant receptor localization to the cytosol as shown in Fig. 1d.
CXCR4 and CXCR7 compete each other for SDF-1α
Like other ACKRs, CXCR7 has been suggested as a decoy, or scavenger, for chemokines. To investigate chemokine affinities and specificities, we constructed intact forms of CXCR4 and CXCR7 in plasmids with different promoters, and applied NanoBiT assays. The SDF-1α-stimulated luminescence signals for CXCR7-LgBiT and SmBiT-β-arrestin2 decreased depending on CXCR4 expression. However, these signals were sustained at approximately half-maximum even in the presence of high CXCR4 expression (Fig. 2a and 2d). In contrast, CXCR4-LgBiT and SmBiT-β-arrestin2 signals were remarkably decreased by the overexpression of CXCR7 (Fig. 2b and 2e), suggesting a stronger affinity between CXCR7 and SDF-1α compared with CXCR4. I-TAC-stimulated luminescence for CXCR7-LgBiT and SmBiT-β-arrestin2 was not affected by CXCR4 (Fig. 2c and 2f), indicating that chemokine binding to its cognate receptor is sufficient to distinguish it from other chemokine receptors. The different promoters induced different receptor-expression efficacies, as determined by GFP expression (Fig. 2g).
CXCR7 is likely to form homodimers, but not heterodimers, with CXCR4
Previous studies have reported the possibility of heterodimerization between CXCR7 and CXCR4 [42,43]. Therefore, this possible dimerization was expected to affect I-TAC-mediated CXCR7 signaling. The luminescence signals of CXCR7-LgBiT stimulated with both SDF-1α and I-TAC should have been changed by CXCR4 overexpression, but CXCR4 overexpression had no effect on I-TAC-stimulated CXCR7 β-arrestin2 recruitment (Fig. 2c). To determine heterodimerization between CXCR4 and CXCR7, cells expressing the receptors tagged with different epitopes for co-immunoprecipitation were used with anti-FLAG antibodies. Fig. 3a shows that HA-CXCR7, but not HA-CXCR4, was co-precipitated with FLAG-CXCR7. To further confirm this, cells were incubated with a cross-linker for immunoprecipitation using RIPA buffer (for increased stringency to avoid non-specific interactions). More HA-CXCR7 was co-precipitated with the cross-linker than without it. A very weak HA-CXCR4 signal was detected at the starting line of the gel region. This might have been due to weak interactions between CXCR4 and CXCR7, or to artificial binding due to overexpression. A structural complementation assay, based on NanoBiT technology, was used to examine real-time membrane protein interactions in a living system. Both SmBiT and LgBiT forms of the receptors were co-expressed in the cell, and luminescent signals were measured at a single time point. Interestingly, combinations of the same receptor produced high luminescence, but combinations of different receptors produced low luminescence signals. These observations suggest that these receptors may be expressed in the plasma membrane as homodimers, rather than as heterodimers with other receptors (Fig. 3b).
Cells expressing both CXCR7-LgBiT and SmBiT-β-arrestin2, but with different promoter-driven intact CXCR7 expression, were treated with SDF-1α. In the presence of the intact CXCR7 under the HSV-TK promoter, SDF-1α-stimulated cells produced higher luminescent signals in comparison to those in the absence of intact CXCR7. In contrast, overexpressing intact CXCR7 with the CMV promoter decreased the luminescence (Fig. 3c). This suggests that since CXCR7s easily form homodimer in the plasma membrane (Fig. 3a and b), HSV-TK-driven intact CXCR7 may bind CXCR7-LgBiT to form dimer, which brings about increase of absolute number of CXCR7 dimer (CXCR7/CXCR7-LgBiT or CXCR7-LgBiT/CXCR7-LgBiT) being able to interact with SmBiT-β-arrestin2. However, overexpressed intact CXCR7 driven by CMV is dominant over CXCR7-LgBiT on SDF-1a binding.
Gα subunits are dispensable for ligand-stimulated β-arrestin2 recruitment at CXCR7
Chemokine receptors stimulate Gαi/o and probably Gα12/13 family members to induce cellular responses [44,45]. To understand the functional mechanisms of CXCR7 in terms of cellular responses, we investigated early signaling events mediated by heterotrimeric G-proteins. Gαi/o, activated by GPCRs, inhibits Gαs-activated adenylyl cyclase. Cyclic AMP (cAMP) production was measured by real-time luminescence in the cells transfected with the receptor and with the Glosensor-22F plasmid. HEK293 cells endogenously express β-adrenergic receptors that activate the Gαs pathway [46]. Isoproterenol-induced cAMP generation was remarkably decreased by SDF-1α pretreatment in cells expressing CXCR4, but only a slight decrease in cAMP levels was seen in cells expressing CXCR7 (Fig. 4a). Maximum isoproterenol-induced cAMP levels decreased by approximately 50% by SDF-1α in the presence of CXCR4, but by less than 20% in parent cells and CXCR7-expressing cells (Fig. 4b). This result raised the possibility that CXCR7 did not activate the Gαi/o family.
To examine the effect of G-proteins on β-arrestin2 recruitment to the receptors, cells expressing NanoBiT constructs were pre-treated with pertussis toxin and used for NanoBiT assays with SDF-1α. The luminescence signals resulting from interactions between CXCR4 and β-arresin2 were significantly decreased by pertussis toxin, whereas interactions between CXCR7 and β-arrestin2 were not affected. When the same experiments were performed in Gα12/13-knockout cells, SDF-1α-stimulated luminescence of CXCR4 towards β-arrestin2 recruitment decreased prominently compared to wild-type cells. Signal reduction by pertussis toxin was also observed, even in these knockout cells. These results suggested that SDF-1α-stimulated β-arrestin2 recruitment to CXCR4 depends on both Gαi/o and Gα12/13. The luminescence signals from interactions between CXCR7 and β-arrestin2 decreased slightly in the absence of Gα12/13 but were not affected by pertussis toxin. However, it is still unknown whether this change was due to the absence of Gα12/13 or the characteristics of cell cloning. Given the effect of Gα12/13 deficiency on interactions between CXCR4 and β-arrestin2, we suggest that both Gα families are dispensable for ligand-dependent β-arrestin2 recruitment towards CXCR7 (Fig. 4c and 4d).
Reporter gene assay is another powerful tool for detecting GPCR activation. Notably, SRE-driven luciferase expression by chemokine receptor activation that can be examined in the presence of chimeric G-protein Gαqi [47]. Increased luciferase activities were observed after SDF-1α or I-TAC treatment of cells expressing CXCR4 and CXCR3, respectively. However, neither the chemokines nor the specific agonist VUF11207 enhanced luciferase activity in cells expressing CXCR7, confirming that the Gα protein subunit is not involved in CXCR7-mediated cellular responses (Fig. 4e).
SDF-1α-stimulated ERK1/2 phosphorylation is mediated by CXCR4 but not CXCR7
Established cell lines express a wide variety of transmembrane membrane receptors to survive and to respond to extracellular stimuli. Detection of CXCR4 has previously been reported in HEK293 cells at the mRNA level [48]. As endogenously expressed GPCRs are rarely detected using antibodies (due to the amount of protein and antibody quality), we investigated CXCR4 and CXCR7 by RT-PCR. A specific CXCR4 PCR product was detected, consistent with the previous report, and one for CXCR7 was also detected, confirming that these chemokine receptors are likely expressed in HEK293 cells (Fig. 5a). To confirm SDF-1α-stimulated cellular responses, HEK293 cells were treated with SDF-1α and assessed using western blotting with anti-pERK1/2 antibodies. As shown in Fig. 5b, ERK1/2 was phosphorylated by SDF-1α in HEK293 and HeLa cells, suggesting that endogenous receptors were activated by their cognate chemokine. To identify the receptor responsible for the signaling event, cell lines deprived of each of the receptors were established by using CRISPR-Cas9 technology. SDF-1α stimulated ERK1/2 phosphorylation was still observed in CXCR7-deficient cells, but not in cells lacking CXCR4. ERK1/2 phosphorylation was not detected even in CXCR4 knockout (KO) cells transfected with the CXCR7 gene. To examine temporal patterns of ERK1/2 phosphorylation in these cells, ligand-treated cells were harvested at different time points and assessed by western blotting. In the absence of CXCR4, ERK1/2 phosphorylation was not increased, whereas the pERK1/2 bands were strong 5 min after ligand treatment, and then decreased in both wild-type and CXCR7 KO cells. Interestingly, SDF-1α-stimulated ERK1/2 phosphorylation in CXCR7 KO cells was higher than phosphorylation in wild-type cells, suggesting that endogenous CXCR4 was activated, and the signal transduced downstream without a competitor for the ligand (Fig. 5c). The inhibitory effect of SDF-1α on β-adrenergic receptor-mediated cAMP generation was prominently reproduced in CXCR7 KO cells exogenously expressing CXCR4. In contrast, this inhibition was not observed in CXCR4 KO cells expressing CXCR7 (Fig. 5d). Overall, it is reasonable to speculate that a slight cAMP reduction in wild-type cells, regardless of CXCR7 expression, may occur by endogenous CXCR4 (Fig. 4A). Our results reinforce the hypothesis that CXCR7 was not able to activate G-proteins.
Chemokine-stimulated β-arrestin2 recruitment at CXCR7 is mediated by GRKs
β-arrestin recruitment to GPCRs requires phosphorylation of intracellular domains at serine or threonine residues by GPCR kinases (GRKs) [49]. As GRK2/3 and GRK5/6 subgroups are ubiquitously expressed, GRK specificity towards CXCR4 and CXCR7 was determined using a GRK2/3-selective inhibitor (Cmpd101). Signals from both chemokine-stimulated cells expressing CXCR7 and β-arrestin2 NanoBiT constructs were downregulated in a Cmpd101 dose-dependent manner, and the signals completely disappeared at a 50 μM concentration (Fig. 6a and 6c). In the case of CXCR4 constructs, Cmpd101 decreased luminescence signals to approximately half-maximum, even at high concentrations (Fig. 6b and 6d). However, SDF-1α-stimulated ERK1/2 phosphorylation in CXCR4-expressing cells was not inhibited by Cmpd101, suggesting that the β-arrestin2 contribution towards ERK1/2 phosphorylation is minimal (Fig. 6e). These results demonstrated that GRK2 and 3 were responsible for CXCR7 phosphorylation and subsequent β-arrestin2 recruitment.
CXCR7 activates GRK2 through the β1 subunit of the heterotrimeric G-protein
To elucidate the molecular mechanism of how CXCR7 is able to activate GRK2 and 3 without G-protein activation, we further developed a structural complementation assay containing Gβ1, GRK2, and GRK5 containing the fragments LgBiT or SmBiT-tagged forms at the N- or C-terminals and chose the best combination of plasmid constructs (Fig. S4). An increase in luminescence signal was observed by the interaction of SmBiT-Gβ1 and GRK2-LgBiT when the cells were treated with SDF-1α in the presence of CXCR4 and CXCR7 (Fig. 7a, upper graphs). SDF-1α also induced an increase in luminescence by the interaction of Gβ1 and GRK5 in the presence of CXCR4, but not CXCR7 (Fig. 7a, lower graphs). This was consistent with the result that the GRK2/3-specific inhibitor affected β-arrestin2 recruitment for both receptors in a dose-dependent manner (Fig. 6a and 6b). This molecular approach provided valuable mechanistic information about how CXCR7 can recruit β-arrestin2 via GRKs activation through the β1 subunit. Luminescence due to the interaction between CXCR4 with Gβ1 was increased by SDF-1a, but for CXCR7, its interaction with Gβ1 did not elicit luminescence by the chemokine, even though SDF-1α stimulation of CXCR7 induced Gβ1 and GRK2 interactions (Fig. 7b).
Ligand-stimulated β-arrestin2 recruitment to the chemokine receptors absolutely depends on the phosphorylation of the receptors at Ser/Thr residues in the C-terminal region and in the third intracellular loop (3ICL) [49]. According to our previous results (Fig. 6a and 6b), GRK2 is likely to phosphorylate both receptors. The catalytic activity of GRKs requires physical interaction with their substrates (GPCRs). The luminescence produced by each receptor with GRK2 increased depending on chemokine stimulation. Interestingly, CXCR7 registered an increased luminescence signal only in combination with GRK2-LgBiT (Fig. 7c). This observation clearly suggests that CXCR7 interacts with the heterotrimeric G-protein in a particular way that is able to generate βγ signaling, but not Gα signaling.
Both CXCR4 and CXCR7 are necessary for SDF-1α-stimulated cell migration
Regarding SDF-1α and chemotaxis, CXCR4 is known to mediate cell migration, but similar functional information about CXCR7, another receptor for SDF-1α, is not available. CXCR7 has been reported to be highly expressed in leukemic cells and to potentiate CXCR4 responses through SDF-1α in experiments using RNA interference [50], but specific effects of CXCR7 on cell migration have not been determined. As HeLa cells express both receptors and are motile toward SDF-1α, cells lacking these receptors were established as shown in Fig. 5. These receptor deficiencies did not affect cell growth in the presence of serum or SDF-1α (Fig. 8a). Migration efficiency in parental HeLa cells was high at 100 ng/ml of SDF-1α, so this same amount of SDF-1α was added in the lower wells of the migration chambers. Cells lacking either of these receptors lost the ability to migrate toward the chemokine. When CXCR7 expression was recovered in CXCR7 KO cells, migration ability was restored. Recovery of CXCR4 in CXCR4 KO cells prominently enhanced their motility, even without SDF-1α, but SDF-1α-stimulated migration was still strongly enhanced (Fig. 8b and 8c). This result suggests that although CXCR4 is a dominant mediator for SDF-1α-stimulated migration, CXCR7 is also essential for cell migration toward the chemokine. Chemokine receptor-mediated G-protein activation, especially Gαi/o and/or Gα12/13, has been considered an indispensable process to endow cells with migration ability. However, CXCR7 does not mediate the activation of any Gα protein subunit, and yet the β and γ subunits in conjunction with β-arrestins somehow influence CXCR4-mediated signaling. The potentiation of cell migration by CXCR7 was also confirmed in this study; U397 cells expressing both receptors migrated toward SDF-1α in a dose-dependent manner. Moreover, increased CXCR7 expression potentiated cell migration without changing the sensitivity for the chemokine (Fig. 8d).