Ligand-independent CXCR2 dimerization

GPCR: G-protein coupled -amino-3-hydroxy-5-methyl-isoxazole-4-propionic

In this paper we describe that CXCR2 forms oligomers when expressed in HEK cells and in native neuronal systems and that oligomer formation is independent of receptor activation by agonist. Attempting to define the molecular regions responsible for receptor oligomerization, we have generated several drastic deletion mutants of CXCR2. We demonstrated that most of these mutants act as dominant negative inhibitors of receptor function, in terms of cell signalling and chemotaxis. Furthermore, CXCR2 receptors mutated in different molecular domains and expressed in HEK cells do not complement the original function. From all these data, we suggest that the active form of the CXCR2 receptor is a dimer and that each individual subunit is activated with an intra-molecular mechanism (cis-activation). To our knowledge, this is the first demonstration that CXCR2 receptors function in oligomerized state.
(Hertfordshire, UK); recombinant rat CXCL2 and human CXCL8 were from Peprotech (London UK); transwell cell culture inserts were from Becton Dickinson Labware (Franklin Lakes, NJ); BCA protein assay was from Pierce (IL, USA). All culture media and monoclonal V5 Ab were purchased from Life Technology Italia (Milan, Italy).

CXCR2 forms dimers in HEK cells and in cerebellar neurons.
To examine whether CXCR2 forms oligomers, as already demonstrated for other chemokine receptors, we generated CXCR2 constructs with either a V5 (CXCR2-V5) or a GFP (GFP-CXCR2) we observed the disappearance of glycosylated monomeric CXCR2, and a shift in the putative CXCR2 dimer band (Fig. 1C). This indicates that CXCR2 receptors dimerize before they appear on the plasma membrane, during the biosynthetic pathway. To investigate whether receptor oligomer formation could be modulated by receptor activation, transfected cells were treated with the specific agonist CXCL8 for 5 min, and analysed as above. Fig. 1D indicates that the formation of CXCR2  Fig. 3B). Interestingly, the C-terminally deleted mutants A315, F183 and V142, when co-expressed together with the wild type CXCR2, retained the ability to bind the wild type (wt) receptors, as evidenced by the bands with molecular weight corresponding to the sum of the wt and deleted receptor types (here called "heterodimers" wt/A315 and wt/F183, Fig. 3A and 10 When the whole extracellular N-terminal (Y49-CXCR2) region was deleted, both homo-and heterodimers (wt/Y49) could be formed (Fig. 3D, left side). Further truncation of the CXCR2 receptor from the N-terminal region, in particular at the aminoacid D143 (described in Fig. 2) results in receptors with different properties: D143-CXCR2 mutants did not dimerize, but formed heterodimers with wt CXCR2 (Fig. 3D, right side). Intermediate N-terminal mutants obtained truncating the CXCR2 at the aminoacid A106 (A106-CXCR2) retained the ability to form both homo-and heterodimers (Fig. 3E).

CXCR2 dimer formation is impaired by GluR1 co-expression in HEK cells.
We have previously reported that CXCR2 physically interacts with various subunits of the AMPAtype glutamate receptors (GluR) both in HEK cells and cerebellar granule neurons, and that this interaction negatively modulates CXCR2-mediated cell chemotaxis (35). To investigate whether CXCR2/GluR1 co-expression interferes with CXCR2 receptor oligomerization, HEK cells were transfected with raising amounts of plasmid containing the GluR1 subunit cDNA. Fig. 4A shows that increasing the amount of GluR1 cDNA co-transfected with a fixed quantity of CXCR2 resulted in a dose-dependent reduction of CXCR2 dimers. This effect was GluR1-specific, as demonstrated when CXCR2 was co-transfected (in a cDNA ratio 1:1) together with CXCR1: in this case the levels of CXCR2 dimers are comparable with those obtained from CXCR2-transfected cells (Fig.   4B). We then investigated whether GluR1 could interact with the CXCR2 mutants. In a previous work, we have already demonstrated that GluR1 interacts with CXCR2-A315, which lacks the whole intracellular C-terminal region (35). Fig. 5A shows that also the other C-terminal deletion mutants CXCR2-F183, CXCR2-K163 and CXCR2-V142 interact with GFP-GluR1. On the other hand, Fig. 5B shows that the N-terminal mutant Y49-CXCR2 interacts with GluR1 but after a more severe N-terminal truncation, at the aminoacid D143, this interaction is lost. It is well established that CXCR2 is a GPCR that regulates the intracellular signalling through the activation of many different phospholipases and protein kinases (reviewed in 37). The CXCR2mediated activation of both the ERKs and the PI3-K/Akt pathways has been described in several cellular systems, where these pathways can be either coupled (as in neutrophils, see 38) or independent each other (as in neurons, see 39). We investigated the activation of the phosphorylation of ERK1/2 and Akt following CXCL8 treatment of HEK cells transfected with CXCR2 alone, in combination with the different CXCR2 deletion mutants or together with GluR1.
Results obtained indicate that CXCR2-transfected HEK cells respond to CXCL8 with the rapid and sustained phosphorylation of ERK1/2 and Akt, which is already detectable after 1 min and lasts up to 30 min (only shown for the time point of 10 min in Fig. 6A and B). These phosphorylations are inhibited by cell pre-treatment with pertussis toxin (PTX, 100 ng/ml, 16 h), indicating receptor coupling to PTX-sensitive G proteins (Fig. 6 A and B). In the presence of LY294002 (50 µM, 1 h), a specific PI3-K inhibitor, Akt phosphorylation was completely inhibited (Fig. 6A), while some CXCL8-induced ERK1/2 phosphorylation was still detectable (Fig. 6B), indicating the presence of both PI3-K-dependent and PI3-K-independent ERK1/2 activation in HEK cells. When CXCR2 was co-expressed, in HEK cells, together with different CXCR2 deletion mutants, CXCL8-induced signalling was greatly impaired. In particular, there was a reduction of ERKs (Fig. 7A) and Akt ( Fig. 7B) phosphorylation, which was more evident with Y49-CXCR2, but that was clearly detectable with all the mutants analysed. When the deleted CXCR2 receptors were expressed alone, in HEK cells, no obvious CXCL8-induced signalling was evident (shown in Fig. 7A and B). These data indicate that the CXCR2 deletion mutants may act as dominant negative inhibitors of CXCR2 function, reducing its signalling activities, likely due to their interaction with the wt receptor. This interaction in its turn would compete for wt receptor dimer formation.

12
When CXCR2 was co-expressed with GluR1, CXCL8 induced ERK1/2 phosphorylation was similar to CXCR2 expressing cells, while Akt phosphorylation was almost undetectable (Fig. 7A-B). The co-expression of two CXCR2 mutants, deleted in different regions, A315 and Y49 did not restore CXCL8-induced ERK1/2 and Akt phosphorylation (Fig. 7A-B), indicating that the truncated receptors do not complement their mutations. As a control, CXCR2 was co-expressed in HEK cells together with CXCR1; under these experimental conditions, CXCR2 dimer formation was not impaired (see above, Fig. 4B). To avoid a direct CXCR1 activation by CXCL8, CXCL2 at low dose (30 nM) was used as more specific agonist of CXCR2. Results obtained demonstrate that CXCR1 was only faintly activated by CXCL2 (analysed as ERK1/2 and Akt phosphorylation, Fig. 7C), and that CXCR2 activity was not impaired by its co-expression with CXCR1, in line with a specific effect of both GluR1 and the CXCR2 truncated receptors on CXCR2 oligomerization.
We have previously reported that in CGNs, obtained from newborn rats (p3), the CXCR2 receptor was not associated with GluR subunits, and that these neurons migrated in response to CXCL2 (35). In contrast, neurons from older animals (p7-p11), where the CXCR2/GluR complex was present, failed to migrate upon CXCL2 treatment, indicating an inhibitory effect of GluRs on CXCR2 function (36). In order to determine whether these differences resulted in coherent variations in cellular signalling, experiments were performed on neurons obtained from the cerebella of p3 or p7 rats. When cell lysates obtained from cerebellar neurons of p3 and p7 rats were analysed for CXCR2 oligomers, no significant differences were observed in the amount of CXCR2 oligomers (not shown). Nevertheless, when the same cells were treated with CXCL2 and analysed for ERK1/2 and Akt phosphorylation, neurons from p3 animals respond with a significantly higher frequency (Fig. 8).

Downloaded from
Since chemokine main function is to regulate cell chemotaxis, we decided to investigate whether the inhibitory behaviour of the truncated CXCR2 on signalling had any effect on the chemotactic activity of the wt receptor. With this aim, HEK cells were transfected with CXCR2 constructs alone or mixed in different combinations. Results, shown in Fig. 9, indicate that the chemotaxis induced by CXCL8 in CXCR2-expressing HEK cells was significantly reduced by the co-expression of the deletion mutants. In general, the effects of the co-expression on chemotaxis were comparable with those obtained on cellular signalling, with inhibitory effects more pronounced for the less drastic mutants Y49-CXCR2 and CXCR2-A315; while the inhibitory effect was almost undetectable for the shorter mutant analysed, CXCR2-V142. For the C-terminally truncated mutants, it is interesting to note that the progressive truncation reduces consistently the rate of chemotaxis inhibition ( Fig.   9). As already shown for cellular signalling, the expression of the CXCR2 mutants A315, Y49 and D143, alone, was not sufficient to make cells responsive to CXCL8 (Fig. 9). Similarly, the coexpression of Y49-CXCR2 and CXCR2-A315 could not produce assembled functional receptors ( Fig. 9). These results indicate that the truncated receptors behave as dominant negative inhibitors of CXCR2-mediated chemotaxis.

Discussion
In spite of a general initial scepticism, due to the deeply rooted classic view of GPCR as single working unit in the lipid bilayer, the evidence for GPCR oligomers that signal in physical conjunction with other receptors and signalling partners has been substantially accepted in recent heterodimerization, in fact, occurs during receptor biosynthesis, and is essential to produce functional receptors and to localize them properly at the plasma membrane (10). Similarly, oxytocin and vasopressin receptors form homo-and hetero-dimers during the biosynthetic pathway (19); the constitutive oligomerization of CCR5 receptors explains the inhibitory effect of the ccr5∆32 mutants on plasma membrane expression of CCR5 dimers (24,27); D2 dopaminergic receptors (4), β adrenergic receptors (41), and CXCR4 are all already found in the homo-dimeric form before any agonist treatment, although CXCL12 treatment increases CXCR4 dimer formation (28, but see also 25). On the other hand, for others GPCRs, oligomerization occurs only after agonist binding, as reported for dopamine and somatostatin complex (5), and for CCR2 and CCR5 receptors (26).
The molecular analysis of the receptor region(s) involved in CXCR2 oligomerization lead us to identify a putative "dimerization motif" in the sequence between the aminoacids A106 and K163; this sequence includes the first extracellular loop, the TM3, and the second intracellular loop. We describe that the CXCR2 deletion mutants that contain only a portion of this region, i.e. the CXCR2-V142 and the D143-CXCR2, did not form dimers when expressed in HEK cells, identifying in this region the presumed dimerization interface. However, we cannot exclude that the region responsible for the physical coupling could be only a part of the whole sequence comprised between A106 and K163. In particular, we observe that the C-terminally deleted CXCR2-V142 mutant does not dimerize, while CXCR2-K163, which only differs for the additional presence of the second intracellular loop, makes dimers. Nevertheless, the N-terminally deleted mutant D143-CXCR2, whose sequence begins with the second intracellular loop, does not dimerize, suggesting a stabilizing role, more that a direct one, for the second intracellular loop in CXCR2 oligomerization.
In addition, the SDS resistance of the physical interaction present in CXCR2 oligomers, would point to the TM3 as the main region involved in receptor assembly. On the other hand, the observation that the CXCR2-V142 and the D143-CXCR2 mutants can both form "heteromers" with wt CXCR2, even if much less efficiently in comparison with the other mutants, indicates that the presence of wt CXCR2 compensates for structural deficits of the mutated CXCR2 proteins. 16 We also report that the co-expression of the CXCR2 deletion mutants with wt CXCR2 impairs receptor function in term of cell signalling and chemotaxis. The rate of functional inhibition for most of the different mutants is inversely correlated with their physical interaction with wt CXCR2: the Y49-CXCR2, which abundantly interacts with wt CXCR2, has the strongest inhibitory effects on ERK1/2 and Akt phosphorylation, and on cell chemotaxis. On the other hand CXCR2-V142, which interacts with wt CXCR2 with the lowest efficiency, does not impair CXCL8mediated cell chemotaxis and Akt phosphorylation, and only partially affects ERK1/2 phosphorylation. The other tested mutants have intermediate behaviour both on cell signalling and chemotaxis. We hypothesize that the inhibitory effects exerted by truncated CXCR2 proteins on wt CXCR2 might reflect an abnormal trafficking of the wt/mutant complex at the plasma membrane, as already reported for other GPCRs (9)(10)(11)18,24).
We describe that, in HEK cells, the activation of the ERK1/2 and Akt pathways by CXCR2 are not completely coupled, as already shown in other cellular systems (39), but in contrast with others (38). This is proved by our observation that cell pre-treatment with LY294002, which drastically eliminated Akt phosphorylation, did not abolish CXCL8-induced ERK1/2 phosphorylation. In contrast, both pathways are coupled to PTX-sensitive G proteins as already described in other systems (42,43), but differently from others (39). It is interesting to note that the co-expression of two different CXCR2 mutations, one lacking the N-terminal (involved in chemokine binding) domain, Y49-CXCR2, and one lacking the whole C-terminal domain (involved in signalling), CXCR2-A315, did not restore cell signalling and chemotaxis, indicating that receptor activation upon agonist binding implicates an intra-molecular mechanism. This is different from what has been reported for other GPCRs, like the LH receptors (44), and indicates a mechanism of cis-activation for CXCR2 composing the dimers.
We have previously reported that the interaction of CXCR2 with GluR1, in HEK and CGNs, resulted in a drastic inhibition of cell chemotaxis upon CXCL2 treatment (35). In this paper we describe that GluR1 co-expression with CXCR2 results in a dose-dependent inhibition of CXCR2 by guest on March 24, 2020 http://www.jbc.org/ Downloaded from dimer formation, further supporting the hypothesis of CXCR2 working as functional dimer. We mapped the CXCR2 domain involved in GluR1 interaction between the aminoacids Y49 and V142.
The partial overlapping of this domain with the "dimerization motif" might explain the competition produced by GluR1 for CXCR2 dimer formation.
We observe that GluR1/CXCR2 co-expression also reduced CXCL8-mediated signalling, with a specific effect on Akt phosphorylation. This would explain the inhibition of cell chemotaxis observed in CXCR2/GluR1 co-expressing cells (35), since CXCL8-mediated chemotaxis is reported to be dependent on the PI3-K/Akt pathway (38). On the other hand, given the lack of inhibition of ERK1/2 phosphorylation in CXCR2/GluR1 transfected cells, we hypothesize that the CXCR2/GluR1 complex acquires new properties in terms of signalling and function (45). All together these results strongly indicate that the physical interaction of CXCR2 with GluR1 or with the truncated CXCR2 proteins competes for CXCR2 dimer formation and this results in a general impairment of CXCR2 functions. This interpretation is consistent with the data we obtained with rat CGNs: neurons from p7 rats were poorly or not responsive to CXCR2 agonist treatment in term of cell signalling, while CGN from p3 rats were responsive. We had previously reported that CGNs obtained from p3 rats responded to CXCR2 stimulation with cell chemotaxis, while CGNs from p7 rats, where CXCR2 co-immunoprecipitated with GluR1/2/3 subunits, were unresponsive (35).
Furthermore, the AMPA receptor antagonist CNQX impairs the neurotrophic activity of CXCR2 (45) and reverts the inhibition of cell chemotaxis in CXCR2/GluR1 expressing cells without hampering receptor co-immunoprecipitation (45). All together these data point to the existence of CXCR2/GluR1 complexes, activated by CXCR2 agonists, whose functions can be blocked with CNQX.
In conclusion, we provide evidence that CXCR2 forms dimers, and that receptors assembly occurs early in biosynthesis, before receptor glycosilation and independently of receptor activation. The region involved in receptor assembly comprises the TM3 and the adjacent extra-and intracellular regions. It is interesting to remember that CXCL8, a well-established CXCR2 agonist, is present in 18 solution both as monomer and dimer, in equilibrium at physiological, nanomolar, concentrations (46), and that these molecular forms can be equally active (47,48).