Research ArticleGPCR SIGNALING

CCR5 adopts three homodimeric conformations that control cell surface delivery

See allHide authors and affiliations

Science Signaling  08 May 2018:
Vol. 11, Issue 529, eaal2869
DOI: 10.1126/scisignal.aal2869

CCR5 delivery to the cell surface

The chemokine receptor CCR5 is a class A GPCR and co-receptor for HIV-1 infection. How class A GPCRs dimerize and how it affects their function are unclear. Jin et al. analyzed the structures of related receptors and identified CCR5 residues that mediated homodimerization. Cross-linking and energy transfer experiments and monitoring the release of receptors from the endoplasmic reticulum identified two dimeric CCR5 conformations that directed its delivery to the plasma membrane. A third CCR5 homodimer was stabilized by maraviroc, a clinically used inhibitor that binds to CCR5 and inhibits its interaction with HIV-1. These data increase our understanding of class A GPCR dimerization and provide insight into the mechanisms of inhibiting HIV-1 entry.

Abstract

Biophysical methods and x-ray crystallography have revealed that class A G protein–coupled receptors (GPCRs) can form homodimers. We combined computational approaches with receptor cross-linking, energy transfer, and a newly developed functional export assay to characterize the residues involved in the dimerization interfaces of the chemokine receptor CCR5, the major co-receptor for HIV-1 entry into cells. We provide evidence of three distinct CCR5 dimeric organizations, involving residues of transmembrane helix 5. Two dimeric states corresponded to unliganded receptors, whereas the binding of the inverse agonist maraviroc stabilized a third state. We found that CCR5 dimerization was required for targeting the receptor to the plasma membrane. These data suggest that dimerization contributes to the conformational diversity of inactive class A GPCRs and may provide new opportunities to investigate the cellular entry of HIV-1 and mechanisms for its inhibition.

INTRODUCTION

Heterotrimeric guanine nucleotide-binding protein (G protein)–coupled receptors (GPCRs) constitute a large family of proteins with seven transmembrane (TM) domains. These receptors transduce intracellular signals upon binding to their respective ligands (1). They are among the most investigated drug targets for therapy. GPCRs can form dimers or oligomers (2). Class C GPCRs, for example, metabotropic glutamate or γ-aminobutyric acid receptor type B (GABAB), form constitutive dimers, which are essential for targeting these receptors to the plasma membrane and for their activity (3). However, the structural organization of class A (rhodopsin-like family) GPCR oligomers and the functional consequences of these oligomers remain a matter of debate (4, 5). X-ray crystallographic structures of class A GPCRs identified receptor oligomers and provided information on possible interfaces (69). Two types of dimer interfaces were described: One involves the domains TM1 and helix 8, whereas another involves TM5 in combination with parts of TM4 and TM6, depending on the receptor (610). However, the functional relevance of these crystal structures and how many conformational states contribute to receptor function are still poorly understood (1113).

CCR5 belongs to the family of class A GPCRs. It is the receptor for several chemokines (including CCL3, CCL4, and CCL5) that are involved in innate immunity. This protein is also a CD4-associated co-receptor required for HIV-1 entry into host cells (14). CCR5 plays a prominent role during HIV transmission, the progression of infection, and evolution into AIDS. CCR5 is a validated therapeutic target for the inhibition of HIV-1 entry. One allosteric nonpeptidic inhibitor targeting CCR5, maraviroc (MVC), is approved for the treatment of patients infected with HIV-1 in many countries (15). Binding of MVC induces conformational changes in CCR5, inhibiting its interaction with the HIV envelope glycoprotein (16). Both biochemical and energy transfer methods suggest the existence of CCR5 homodimers (17, 18); however, their functional relevance is poorly studied (17, 19, 20). The crystal structure of the complex of CCR5 bound to MVC is a monomer (21). Although MVC could stabilize a monomeric form of CCR5, the conditions of crystallization, particularly the introduction of rubredoxin in the third intracellular loop of CCR5, may explain the lack of dimers (22). In addition, controversy exists regarding which residues are critical for CCR5 dimerization (18, 23), hampering the exploration of the role of CCR5 dimerization in receptor physiology and HIV pathogenesis.

Here, we investigated the dimeric organizations of CCR5 and addressed the effect of MVC thereon. We built models of CCR5 dimers by computationally assembling two copies of a monomeric CCR5 structure to reproduce the symmetrical dimeric organizations found in the crystal structures of two closely related class A GPCRs. On the basis of these models, we selected the residues potentially involved in the dimeric interfaces and generated receptors with mutations at these residues. We explored the dimerization of the wild-type (WT) and mutated CCR5 proteins at both molecular and functional levels by combining in silico, chemical cross-linking, and energy transfer approaches with a functional assay based on the RUSH (retention using selective hooks) system (24). We identified two dimeric states of unliganded CCR5. The binding of MVC stabilized a third dimeric organization, involving an alternative dimer interface. Functionally, we determined that CCR5 dimerization occurred in the endoplasmic reticulum (ER) and was required for the proper targeting of the receptor to the cell surface. These data provide new information on class A GPCR dimers and improve our understanding of the organization of their inactive states, their molecular dynamics (MD), and their function.

RESULTS

Characterization of the arrangement of CCR5 using cysteine cross-linking

We determined the molecular arrangement of CCR5 using a disulfide cross-linking strategy (25). We selected residues based on a model constructed from the crystal structures of two class A GPCR homodimers: the chemokine receptor CXCR4 and the μ opioid receptor (MOR) (6, 7). In CXCR4, the interface involves mostly TM5 and, to a lesser extent, TM3, TM4, and TM6. We named this interface “Interface 5” (I5). In MOR, TM5 and TM6 mainly contribute to a first interface (named I5/6), whereas TM1, TM2, and helix 8 contribute to a second interface on the opposite side of the receptor (named I1/8). In these structures, homodimers involve symmetrical interactions (6, 7) in which the interface engages the same residues in the two protomers. According to structure-based sequence alignments, the residues of CCR5 potentially involved in the three putative dimer interfaces are Q1885.32, N1925.36, L1965.40, I2005.44, and L2055.49 in I5; Q1885.32, V1995.43, V2045.48, L2085.52, L2566.56, and F2606.60 in I5/6; and S381.40, I421.44, and F3118.57 in I1/8 [superscript numbering refers to Ballesteros-Weinstein numbering (26)]. We substituted these residues one at a time with a cysteine in WT CCR5 to explore the potential proximity of the targeted residues in receptor dimers (Fig. 1A). We verified the presence of each mutant at the cell surface (Fig. 1B) and detected dimers trapped by covalent disulfide bridges by Western blotting analysis (Fig. 1, C to F, and fig. S1A). WT CCR5 and two mutants of TM5, I200C and L208C, existed mainly as monomers in the presence or absence of the oxidizing agent CuPhenanthroline (CuP), which promotes the formation of disulfide bonds (Fig. 1, C and D) (25). All other mutations in TM5 or TM6 increased the formation of dimers or higher-order oligomers, even in the absence of CuP (Fig. 1, C to F, and fig. S1A). By contrast, receptors mutated in TM1 and helix 8 behaved similarly to WT CCR5 (Fig. 1, E and F).

Fig. 1 Cysteine cross-linking identifies TM5 and TM6 in dimer interfaces.

(A) Schematic representation of human CCR5 (adapted from http://gpcrdb.org/protein/ccr5_human). Amino acid residues from the TM domains and helix 8 are represented. Amino acid residues substituted with a cysteine are shaded. (B) Relative cell surface expression of CCR5 cysteine mutants. HEK 293 cells transfected with plasmids encoding FLAG-WT-CCR5 or the indicated mutant receptors were stained with an antibody against the FLAG epitope and analyzed by flow cytometry. Bars represent staining efficiency for cells expressing FLAG-CCR5 mutants relative to cells expressing FLAG-WT-CCR5. Data are means ± SD of three experiments. (C to F) Disulfide cross-linking. (C and E) HEK 293 cells were transfected with plasmids encoding the indicated FLAG-CCR5 proteins. Cells were treated with (+) or without (−) CuP, lysed, and analyzed by Western blotting with an antibody against the FLAG epitope. Data are representative of three experiments. Bands that migrated at 36 kDa represent monomers, whereas those at 72 kDa represent dimers. (D and F) The percentage of dimers relative to the (monomers + dimers) amount of CCR5 was quantified by scanned densitometry analysis of the same film. Data are means ± SD of three experiments. *P < 0.05, **P < 0.01, ***P < 0.001 compared to WT CCR5 (without or with CuP, respectively) in an unpaired, two-tailed Student’s t test.

In the presence of the reducing agent dithiothreitol, we did not observe oligomers of WT CCR5 or the mutants (fig. S1B), suggesting that oligomerization resulted from the spontaneous formation of one or more disulfide bonds. CCR5 contains 12 cysteines, including 4 that potentially expose thiol groups at the protein surface (C581.60, C2135.57, C2245.68, and C2917.47) and are not buried, engaged in an internal disulfide bridge, or palmitoylated (Fig. 1A and fig. S1C). These cysteines may contribute to the presence of dimers observed for WT CCR5, as well as to the formation of higher-order oligomeric species (Fig. 1, C and E). On the basis of three-dimensional (3D) models of CCR5 dimer, we predicted the contributions of C213 and C224 in covalent cross-linking through residues L2055.49 and F2606.60, respectively. This was ruled out by experiments on the double mutants L205C/C213A and F260C/C224A, which behaved like the L205C and F260C single mutants, respectively (fig. S1D). These data support the existence of CCR5 dimers and highlight the role of TM5 and TM6 in CCR5 dimerization. The presence of oligomeric species also suggested multiple possibilities for protomer association.

Stabilization of the CCR5 dimer through the extracellular side of TM5

We performed resonance energy transfer–based experiments (17, 27) to confirm the role of TM5 in CCR5 dimerization by another approach. We conducted homogeneous time-resolved fluorescence (HTRF) saturation experiments (28) in live cells coexpressing a constant amount of FLAG-SNAP–tagged CCR5 (FLAG-ST-CCR5) and increasing amounts of FLAG-CLIP–tagged CCR5 (FLAG-CT-CCR5). The HTRF signal was recorded after labeling the cell surface receptors with snap-lumi4-Tb (donor) and clip-red fluorophores (acceptor). We introduced lysine at position 1965.40 (L196K) or position 2055.49 (L205K), two residues that are involved in CCR5 dimerization as determined by our cross-linking studies (Fig. 1C). We also introduced lysine at position 2005.44 (I200K) or 2085.52 (L208K), residues that were not involved in CCR5 dimerization (Fig. 1C). Introduction of a lysine is expected to prevent dimerization by electrostatic repulsion between protomers if the modified residue is part of a symmetrical dimer interface. We coexpressed the unrelated GPCR angiotensin II receptor type 2 (AT2R) with WT CCR5 as a negative control because these receptors do not interact. The L196K, I200K, and L205K mutant CCR5 proteins were present at lower amounts at the cell surface compared to those of WT CCR5 and the L208K mutant (Fig. 2A). HTRF signals varied hyperbolically with the acceptor/donor ratio (ClipTb/SnapTb) in all cases (Fig. 2B and fig. S2A), indicating specific receptor oligomerization. However, HTRF50 values (that is, the acceptor/donor ratio yielding half of the maximal HTRF signal) of the L196K, I200K, and L205K mutants were statistically significantly greater than that of WT CCR5 (Fig. 2C), suggesting a reduction in the propensity of the mutant receptors to assemble and indicating roles for these residues in CCR5 dimerization.

Fig. 2 Lysine mutagenesis destabilizes CCR5 dimer formation.

(A) Relative cell surface (nonpermeabilized) and total (permeabilized) expression of WT CCR5 and the indicated lysine mutants. HEK 293 cells transfected with plasmids encoding FLAG-ST-WT-CCR5 or the indicated FLAG-ST-CCR5 lysine mutants (L196K, I200K, L205K, or L208K) were permeabilized or not, stained with an antibody against FLAG, and analyzed by flow cytometry. Bars represent staining efficiency for cells expressing FLAG-ST-CCR5 mutants relative to cells expressing FLAG-ST-WT-CCR5. Data are means ± SD of three experiments. (B) Example of HTRF saturation experiments in cells coexpressing FLAG-ST-CCR5 as the donor (WT, 20 ng; L196K, 30 ng) and various amounts of FLAG-CT-CCR5 as the acceptor (WT, 0 to 200 ng; L196K, 0 to 300 ng) fitted according to a one-site binding model from one representative experiment. Data are means ± SD of four experiments. Results are expressed in milli-HTRF ratio units (mHTRF) plotted against the ratio of the amounts of cell surface donor and acceptor (ClipTb/Snap Tb, F620 nm). Dose-response curves (WT versus L196K) were compared by fitting nonlinear regression curves, and fits were compared by extra sum-of-squares F test. L196K curves were statistically significantly different compared to WT CCR5 curves. Data are means ± SD of four experiments (P < 0.001) (F2.32 = 95.21). (C) HTRF50 values (ClipTb/SnapTb values for half-maximal mHTRF) plotted as a function of the fluorescence intensity of the donor at 620 nm (F620), which represents the amount of FLAG-ST-CCR5 at the cell surface. (D) Plotted HTRF50 values for a range of F620 values from 900 to 1900 [arbitrary units (a.u.)]. **P < 0.01 and ***P < 0.001 comparing mutants to WT CCR5 in a Poisson regression test.

The propensity of receptor dimerization for the CCR5 lysine mutants was even lower than that observed between WT CCR5 and the unrelated AT2R (Fig. 2C). At a low receptor density, in which random collisions between receptors are limited, HTRF50 values for these mutants were two times greater than that of WT CCR5 (Fig. 2D), confirming the role of L1965.40, I2005.44, and L2055.49 in CCR5 dimerization. In contrast, the L208K mutant behaved similarly to WT CCR5 (Fig. 2, C and D), suggesting that L2085.52 is not involved in CCR5 dimerization, which is consistent with our disulfide cross-linking studies (Fig. 1, C and D). Despite the apparent role of I2005.44 in dimer formation by energy transfer experiments (Fig. 2, C and D), exchanging this residue with a cysteine did not lead to disulfide bridge formation (Fig. 1, C and D). This discrepancy could be explained by the fact that disulfide bridge formation depends not only on the proximity of the two cysteine residues but also on the appropriate positioning of the thiol groups and their local environment. Both the HTRF and cross-linking experiments suggest that CCR5 forms symmetrical dimers. Residues in the extracellular region of TM5 (L1965.40, I2005.44, and L2055.49) were involved in the formation of these symmetrical dimers.

CCR5 dimerization through TM5 in the ER

Substitution of a residue involved in dimerization with a lysine inhibited the expression of CCR5 at the cell surface (Fig. 2A), whereas substitution of the same residue with a cysteine enabled cell surface delivery (Fig. 1B). The introduction of a cysteine residue presumably stabilized an existing interface of the CCR5 dimer. We tested whether CCR5 dimerization was a prerequisite for the transport of the receptor to the cell surface, as suggested for other class A GPCRs (17, 29, 30). We quantified the transport of CCR5 using the RUSH assay (24), which enables the reversible retention of CCR5 in the ER and its synchronous release to the Golgi apparatus and the cell surface (Fig. 3A). CCR5 fused at the N terminus to enhanced green fluorescent protein (EGFP), and the streptavidin-binding peptide (SBP) (named hereafter SBP-CCR5) was retained in the ER because of its interaction with a streptavidin-fused resident protein of the ER (streptavidin-KDEL), which served as the “hook.” The addition of biotin prevented the interaction between the SBP and streptavidin, releasing SBP-CCR5 from the ER (Fig. 3A). We conducted RUSH assays in human embryonic kidney (HEK) 293 cells transfected with plasmids encoding the RUSH system for WT CCR5 or receptors with lysine mutations in TM5 (L196K, I200K, L205K, or L208K) (Fig. 3B). After the addition of biotin, SBP-WT-CCR5 expression at the cell surface increased for up to 120 min and remained constant thereafter. SBP-L208K behaved similarly, despite having a 25% lower cell surface expression relative to that of SBP-WT-CCR5. By contrast, the dimerization-compromised mutants L196K, I200K, and L205K were not efficiently transported to the cell surface after biotin addition (Fig. 3B). Confocal microscopy showed that FLAG-ST-L196K colocalized with the ER-resident protein DsRed-KDEL (Fig. 3C). This suggests that the low abundance of the lysine mutants at the cell surface was due to their intracellular retention in the ER and not due to incorrect folding or degradation of the receptor. Consistent with this, we verified that the L196K mutant and WT CCR5 bound to the chemokine CCL3 with similar affinities (fig. S2B).

Fig. 3 Lysine mutagenesis inhibits CCR5 export from the ER.

(A) Schematic representation of the principle of the RUSH assay. SBP-EGFP-CCR5 is retained in the ER through its interaction with streptavidin-KDEL. This interaction is mediated by the core streptavidin and the SBP (fused to CCR5). Release is induced by addition of biotin, which prevents the SBP-streptavidin interaction and enables trafficking of SBP-EGFP-CCR5 (SBP-CCR5) to the plasma membrane. (B) RUSH assay in HEK 293 cells transiently expressing the RUSH system (construct streptavidin-KDEL_EGFP-CCR5) for WT CCR5 or the indicated CCR5 lysine mutants (L196K, I200K, L205K, or L208K). The cell surface expression of CCR5 was monitored by flow cytometry after staining with an antibody against GFP at the indicated times after the addition of biotin (at time 0). Data are means ± SD of three experiments. (C) Micrographs of HEK 293 cells stably expressing FLAG-ST-WT-CCR5 or FLAG-ST-L196K transfected with plasmids expressing the ER marker DsRed-KDEL. FLAG-ST-CCR5 (WT or L196K) is shown in yellow. DsRed-KDEL is shown in blue. Scale bars, 5 μm. (D) Schematic representation of the principle of the dual RUSH assay. HEK 293 cells were cotransfected with plasmids expressing both the RUSH system for WT CCR5 (construct streptavidin-KDEL_EGFP-CCR5) and FLAG-CCR5-KKLV, which reside in the ER, at a ratio of 1:4. After the addition of biotin, the release of SBP-EGFP-CCR5 was detected as described in (A) in EGFP-positive cells. (E) Dual RUSH assay in HEK 293 cells coexpressing the RUSH system for WT CCR5 and pcDNA3 (ctl), FLAG-WT-CCR5-KKLV (WT-KKLV), FLAG-AT2R-KKLV (AT2AR-KKLV), or FLAG-CCR5–lysine mutants–KKLV (I42K-KKLV, L196K-KKLV, I200K-KKLV, L205K-KKLV, or L208K-KKLV). The cell surface expression of SBP-WT-CCR5 was monitored by flow cytometry at the indicated times after the addition of biotin (at time 0). Data are means ± SD of three experiments. **P < 0.01 and ***P < 0.001 compared to that of WT-CCR5-KKLV in an unpaired, two-tailed Student’s t test.

We developed a “dual RUSH assay,” which provides two retention proteins in the same cell, to assess whether CCR5 dimerization occurred in the ER and to validate the residues involved in this process. We coexpressed the SBP-WT-CCR5 together with WT CCR5 or a lysine mutant receptor bearing the ER retention signal sequence KKLV in its C terminus (CCR5-KKLV) (Fig. 3D and fig. S3A). We hypothesized that the presence of WT-CCR5-KKLV would lead to the retention of SBP-WT-CCR5 by dimerization in the ER. When coexpressed with WT-CCR5-KKLV, SBP-WT-CCR5 was not detected at the cell surface even in the presence of biotin (Fig. 3E). Thus, WT-CCR5-KKLV acted as a dominant-negative protein, preventing the transport of its dimeric partner (SBP-WT-CCR5) to the cell surface.

In contrast to WT-CCR5-KKLV, disrupting the dimer interface of the KKLV-bearing receptor rescued the trafficking of SBP-WT-CCR5 from the ER. The dimerization-compromised CCR5 mutants or the unrelated GPCR AT2R fused to KKLV enabled the transport of SBP-WT-CCR5 to the cell surface (Fig. 3E). This confirmed that WT-CCR5-KKLV retained SBP-WT-CCR5 in the ER as a result of receptor dimerization. In agreement, and similarly to WT-CCR5-KKLV, KKLV-fused CCR5 with mutations in residues outside of the dimerization interface (L208K-KKLV or I42K-KKLV) exerted a dominant-negative effect on the cell surface delivery of SBP-WT-CCR5 (Fig. 3E). Note that the L196K-KKLV, I200K-KKLV, and L205K-KKLV mutants partially restored the transport of SBP-WT-CCR5 to the cell surface. This suggests that these mutants retained some ability to associate with SBP-WT-CCR5, albeit with reduced stability. This is consistent with the HTRF studies that showed that the mutation of TM5-specific residues decreased the propensity of CCR5 to dimerize but did not fully prevent it (Fig. 2B and fig. S2A). Analysis of the total amount of the KKLV constructs by Western blotting or flow cytometry verified that they were present in similar amounts under each condition (fig. S3, B and C). These data confirm that CCR5 dimerization involves residues of TM5. However, altering the extracellular side of TM5 did not completely inhibit the dimerization of CCR5. This suggests that CCR5 forms dimers through multiple dimerization interfaces. Finally, these data suggest that CCR5 dimerization occurs in the ER and that dimerization likely controls the trafficking of CCR5 from the ER.

MVC as a pharmacological chaperone

Small–molecular weight ligands can act as pharmacological chaperones by rescuing the delivery of poorly trafficked, disease-causing GPCRs (31, 32). We tested the effect of MVC, which is a partial inverse agonist of CCR5 (33, 34), on the cell surface delivery of CCR5 dimerization-compromised mutants. We hypothesized that this compound might stabilize a conformation of the receptors suitable for export from the ER. We included two control ligands: 5P12, an analog of the chemokine CCL5 that should not cross lipid membranes, and AMD3100, a nonpeptidic antagonist of the chemokine receptor CXCR4. MVC, but not 5P12 or AMD3100, restored the cell surface delivery of the dimerization-compromised mutants SBP-L205K (Fig. 4A) and SBP-L196K (fig. S4, A and B) in RUSH assays. MVC also slightly increased the amount of SBP-WT-CCR5 that reached the cell surface (fig. S4C). By contrast, MVC had no effect on the export of an unrelated class A receptor, the β2-adrenergic receptor fused to SBP (SBP-β2AR) (fig. S4D), or SBP-CCR5 variants with the E283Q mutation that abolishes MVC binding (SBP-WT-CCR5-E283Q, SBP-L205K-E283Q, and SBP-L196K-E283Q) (fig. S4, E and F) (34).

Fig. 4 MVC favors CCR5 dimerization by changing the orientation of TM5.

(A) RUSH assay in HEK 293 cells expressing the RUSH system for L205K that were left untreated or were treated overnight with a saturating concentration (>100 nM) of MVC, 5P12, or AMD3100. The cell surface expression of SBP-L205K was monitored as described in Fig. 3B. Data are means ± SD of three experiments. (B) HTRF50 values obtained from HTRF saturation experiments performed on HEK 293 cells coexpressing FLAG-ST-WT-CCR5 and FLAG-CT-WT-CCR5 or FLAG-ST-L205K and FLAG-CT-L205K that were left untreated or were treated with 1 μM MVC overnight. Data are plotted as a function of the fluorescence intensity of the donor at 620 nm (F620). (C) Dual RUSH assay in HEK 293 cells coexpressing the RUSH system for SBP-CCR5-E283Q together with pcDNA3 (ctl), FLAG-AT2R-KKLV (ATR), FLAG-WT-CCR5-KKLV (WT), FLAG-L196K-KKLV (196), FLAG-I200K-KKLV (200), FLAG-L205K-KKLV (205), or FLAG-L208K-KKLV (208) and that were left untreated or were treated with >100 nM MVC overnight. Data are means ± SD of three experiments. *P < 0.05, **P < 0.01, and ***P < 0.001 compared to that of untreated conditions in an unpaired, two-tailed Student’s t test. (D) CuP or DSP cross-linking. HEK 293 cells were transfected with plasmids expressing FLAG-WT-CCR5. The cells were incubated overnight with (+) or without (−) 1 μM MVC, treated in the presence or absence of CuP or DSP, lysed, and analyzed by Western blotting with an antibody against the FLAG epitope. Data are representative of three experiments. (E) CuP-induced cross-linking of the indicated FLAG-CCR5 constructs (WT, S38C, I42C, F311C, N192C, or L205C) after the cells were left untreated or were treated with 1 μM MVC overnight. FLAG-CCR5 was detected as described in (D). Data are representative of two experiments. (F) DSP-induced cross-linking of the indicated FLAG-CCR5 constructs (WT, K59A, K138A, K171A, K197A, K219A, K228A, K229A, or K303A) after treatment with 1 μM MVC overnight. CCR5 was detected as described in (D). Data are representative of three experiments.

We examined whether MVC stabilized the homodimerization of the dimerization-compromised CCR5 mutant. The ability to promote the structural reorganization of dimerization-compromised mutant receptors is a proposed mechanism of action for pharmacological chaperones (31). HTRF saturation experiments revealed that MVC decreased the HTRF50 value of the L205K mutant close to that of WT CCR5 (Fig. 4B), suggesting that MVC restored the stability of L205K dimers. MVC also slightly improved the dimerization of WT CCR5 as detected by HTRF (Fig. 4B). In addition, MVC rescued the ability of dimerization-compromised, ER retention signal–tagged CCR5 mutants L196K-KKLV, L200K-KKLV, and L205K-KKLV to intracellularly retain SBP-CCR5-E283Q in the dual RUSH assay (Fig. 4C). Furthermore, the dual RUSH assays with SBP-CCR5-E283Q indicated that the binding of MVC to one protomer was sufficient to restore the overall stability of the dimer. This result suggests a distinct functional role of individual protomers within a dimer complex, consistent with the notion that even symmetrical receptor dimers form asymmetric functional units (35). Thus, the data suggest that MVC acts as a pharmacological chaperone that promotes the removal from the ER of dimerization-compromised CCR5 mutants, likely by stabilizing or reorganizing CCR5 dimers. MVC overcame the dimerization deficiency caused by mutations in TM5. These results also suggest that the lysine mutants preserved their binding capacity to MVC, confirming that these mutant receptors did not have a major folding defect, which is consistent with our earlier result (fig. S2B). Overall, these findings suggest a link between CCR5 dimerization and its trafficking from the ER to the cell surface.

The effect of MVC on the homodimeric arrangement of CCR5

To gain insight into the dimeric organization of MVC-bound CCR5, we studied the effect of MVC on the formation of covalent bonds between protomers. We compared the bridging of two neighboring cysteines in the presence of CuP to the coupling of two neighboring lysines by the amine-reactive cross-linker DSP (dithiobis succinimidyl propionate) (Fig. 4, D to F). MVC almost abolished CCR5 cross-linking promoted by CuP in experiments with WT CCR5 and single cysteine mutants (Fig. 4, D and E, and fig. S5A). The lack of formation of disulfide bridges between two CCR5 protomers in the presence of MVC contrasts with the results of HTRF experiments showing that MVC increased the propensity to dimerize of both WT CCR5 and the dimerization-compromised CCR5 mutant L205K (Fig. 4B). This suggests that MVC either decreased CCR5 dynamics and collisional cross-linking or stabilized CCR5 dimers involving an alternative dimerization interface compared to that of unliganded dimers. In agreement with this latter hypothesis, MVC increased the amount of dimeric species trapped by DSP, which occurred with a concomitant decrease in the monomeric receptor (Fig. 4D).

CCR5 contains 15 lysine residues (Fig. 1A). We substituted 8 that are potentially accessible according to the crystal structure of CCR5 (K59, K138, K171, K197, K219, K228, K229, and K303) (21) with alanine (fig. S5B) to determine which one(s) reacted with DSP in the presence of MVC. The proportion of dimers formed after DSP treatment in the presence of MVC was similar for the WT receptor and all of the mutants, except for the K197A mutant, which exhibited impaired dimerization (Fig. 4F). The same experiments performed in the absence of MVC showed that this lysine was not exposed in the unliganded form of the receptor (fig. S5C). Competition experiments with the natural CCR5 ligand CCL3 further verified that MVC binding to CCR5 was not impaired by the K197A substitution (fig. S5D). Thus, K1975.41, which is located in TM5, was targeted by DSP only in the presence of MVC. The RUSH, HTRF, and cross-linking results converge and suggest that MVC prevented the symmetrical association of TM5 in unliganded CCR5 dimers. Nevertheless, the data suggest that MVC stabilized dimerization of the receptor by promoting a different molecular organization of the two protomers.

Model of CCR5 interfaces and receptor dynamics

We built CCR5 structural models by computational approaches and compared these models with our experimental data to fine-map the dimerization interfaces. We built two 3D models of a CCR5 symmetrical dimer, reproducing the I5 and I5/6 interfaces that are present in the crystal structures of the inactivated state of CXCR4 and MOR (6, 7), respectively (Fig. 5A). The dimer models were embedded in a model hydrated lipid bilayer and submitted to three independent molecular simulation runs. Low variations over time of the root mean square deviation (rmsd) computed on atomic coordinates indicated that both receptor dimer models had similar stabilities (Fig. 5B). The model for the interface of I5 predicted the involvement of residues on TM3, TM4, and TM5, whereas the I5/6 interface was predicted to involve residues along the entirety of TM5 and a large part of TM6 (Fig. 5C and fig. S6A). Both interfaces also involved intracellular loops (ICL2 in I5 and ICL3 in I5/6) and the third extracellular loop (ECL3) (Fig. 5C and fig. S6A).

Fig. 5 3D models of unliganded CCR5 dimers.

(A) Top: Model I5, based on the crystallographic structure of the CXCR4 homodimer. Bottom: Model I5/6, based on the crystallographic structure of the MOR homodimer. The protein backbone is represented by cylinders colored by chain (light and dark gray). Lines indicate interprotomer contacts detected during MD simulation (H-bonds, ionic bonds, aromatic interactions, and contacts between hydrophobic groups). Line thickness is proportional to the frequency of occurrence. (B) MD simulations of CCR5 dimers. Top: Model I5. Bottom: Model I5/6.Time series of rmsd values for all protein atoms using the input coordinates as reference. Data represent three independent simulations. (C) Dimer model residues predicted in the I5 (top) and I5/6 (bottom) interfaces. Side view of the initial model before simulation and not minimized. The protein backbone is represented as gray ribbons; the residues of the interface are represented with Corey, Pauling, Koltun (CPK)–colored sticks with carbon atoms in gray. Asterisks indicate residues that are positive in the next panel. (D) Boxplot of interprotomer distances during MD simulations for the I5 (top) and I5/6 models (bottom). The bottom and top of each box indicate the first and third quartiles of the distribution, respectively; the ends of the bars represent the minimum and maximum of all of the distances. A distance of 4 Å is typical for two cysteine residues in a disulfide bond (red dotted line). Blue coloring indicates direct hydrophobic contacts between a residue in the first protomer with the same residue in the second protomer.

These models are consistent with the cross-linking data obtained from experiments with unliganded receptor (Fig. 1). Distances measured between the Cβ atom (first carbon atom in the side chain) of a residue in the first protomer and the Cβ atom of the same residue in the second protomer suggest that the I5 model is directly compatible with the formation of a covalent dimer for the Q188C and N192C mutants (distances of about 4 Å in Fig. 5D) and that the residues L1965.40, I2005.44, and L2055.49 were in self-proximity within the dimer (Fig. 5, C and D). In the I5/6 model, the closest inter-Cβ distances were predicted for V2045.48, L2566.56, and F2606.60 (Fig. 5D). Both models predicted average inter-Cβ distances exceeding 10 Å for both L2085.52 and V1995.43 (Fig. 5, C and D). The side chains of L2085.52 in the first and second protomers point in opposite directions in both dimer models, explaining why we detected the L208C mutant primarily as a monomer (Fig. 1C). In contrast, the side chains of V1995.43 in the first and second protomers face each other in the I5/6 model, likely explaining why the V199C substitution promoted receptor cross-linking (Fig. 1D and fig. S1A). The models also suggest that exchanging L1965.40, I2005.44, or L2055.49 with a lysine prevents the dimeric organization of the I5 interface due to electrostatic repulsion (fig. S6B). Similarly, the L196K mutant and, to a lesser extent, the I200K mutant were predicted to prevent dimer formation through the I5/6 interface (fig. S6B). Introducing a lysine at position 2055.49 was predicted to be compatible with the I5/6 interface (fig. S6B); however, this mutant failed to appear at the cell surface in the RUSH assays (Fig. 3B). Thus, together, the model and the ER export data suggest that the I5/6 dimer is not involved in receptor export from the ER. The introduction of a lysine at position 2085.52 was predicted not to generate steric collisions or electrostatic repulsions between protomers in either dimer conformation I5 or I5/6 (fig. S6B), in agreement with the experimental results obtained with this mutant (Figs. 1 to 3).

To gain insight into the effect of MVC on CCR5, we performed MD simulations of CCR5 modeled as a monomer, free or bound to MVC, embedded in a hydrated lipid bilayer (Fig. 6A and fig. S7). The presence of MVC was predicted to cause a concerted movement of the N terminus and extracellular loops (ECL2 and ECL3): ECL2 moved away from ECL3 (fig. S8, A and B), which is in agreement with published simulations (22). MVC was predicted to also weaken interhelical hydrogen bonds (H-bonds) between TM5 and TM6 (Fig. 6, B and C), slightly enlarging the distance between the extracellular parts of these two TMs. Focusing on TM5, simulations indicated that the binding of MVC to the TM cavity of CCR5 reorients the side chain of K1975.41 (Fig. 6, B and C). MVC was predicted to disrupt the H-bond between K1975.41 and A1594.56 in TM4 (Fig. 6, B and C), hence decreasing the strength of the association between TM5 and TM4 and increasing the accessibility of K1975.41, resulting in a prominent amine group (NZ) that can react with an electrophile, such as DSP. Because the MD simulations predicted that MVC modifies the ability of TM5 to interact with both TM4 and TM6, MVC would likely disrupt the dimeric organization involving I5 and I5/6. Note that the I5 and I5/6 models are not consistent with a direct bridging between K1975.41 of the two protomers by DSP (fig. S6B), explaining the lack of DSP cross-linking obtained on the unliganded receptor (Fig. 4D). Moreover, the interprotomeric distances between K1975.41 and any other lysine were greater than 12 Å in both models, ruling out cross-linking between K1975.41 in one protomer and another lysine in the other. This finding further indicates that MVC stabilizes a third mode of dimerization, which is distinct from I5 and I5/6.

Fig. 6 Effect of MVC on CCR5 structure models and dynamics.

(A) MD simulations of the CCR5 monomer when free or in complex with MVC. Simulated systems at the end of the equilibration stage. CCR5 dimer (ribbon representation) is embedded into a lipid bilayer (line representation) surrounded by water molecules (line) and the counter ions K+ and Cl (small spheres). MVC is represented with balls (magenta). CCR5 is shown free (top) and in a complex with MVC (bottom). (B) Molecular view of residues in TM5/TM6 (top) and TM4/TM5 (bottom) (gray sticks in the free receptor and pink sticks in the complex with MVC). Dotted black lines indicate intramolecular H-bonds. TM backbones are represented by gray (free CCR5) or red (MVC-bound CCR5) ribbons. (C) Frequency of occurrence of H-bonds throughout the simulation of the free (gray) or MVC-bound (red) receptor. Top: H-bond network between the extracellular parts of TM5 and TM6. Bottom: H-bond network between the extracellular parts of TM4 and TM5. In free CCR5, the K197 amine group (NZ) was engaged in two H-bonds: one with the oxygen atom of the side chain of Q194 (OE) and one with the backbone carbonyl (O) of A159 in TM4. H-bond abbreviations, main chain: donor (N), acceptor (O); side chain: donor Q(NE), K(NZ), T(OG), N(ND); acceptor N(OD), T(OG), E(OE), Q(OE). (D) 3D model of the CCR5 dimeric state bound to MVC (IMVC), as represented in Fig. 5A. (E) Boxplot of interprotomer distances during MD simulations as described for Fig. 5D. (F) BMOE cross-linking of the indicated FLAG-tagged CCR5 constructs (WT, F193C, L196C) after overnight treatment (or not) with 1 μM MVC. FLAG-CCR5 was detected as described in Fig. 1. Data are representative of three experiments.

To model this third CCR5 dimer stabilized in the presence of MVC, we used a rigid protein-protein docking approach. The docking was biased to place K1975.41 in the dimer interface. Only one solution matched with two parallel protomers properly inserted into a lipid membrane (Fig. 6D and fig. S9). As in I5, the predicted interface of this new model, named IMVC, engages TM3, TM4, and TM5 with symmetrical interactions (Fig. 6D and fig. S9, A and C). Interprotomer distances measured on this model at positions engaged in I5 and I5/6 interfaces are not compatible with the distance required for disulfide bridge formation (4 Å) (Fig. 6E), which is consistent with the lack of cysteine mutant cross-linking in the presence of MVC (Fig. 4, D and E). The predicted interprotomer distances are also consistent with MVC-induced DSP cross-linking of the dimer, which bridged K1975.41 of one protomer with K1975.41 of the second protomer (Fig. 4F and fig. S9D). The distance measured between the two amine groups was 11 Å, in agreement with the length of the spacer arm between the two reactive groups of DSP (fig. S10A). Moreover, the introduction into the model of a lysine at position 1965.40, 2005.44, 2055.49, or 2085.52 is compatible with the IMVC interface (fig. S9D).

To further validate the IMVC model, we designed a mutant predicted to form a covalent dimer. We selected position 1935.37 because the in silico mutation of F1935.37 by a cysteine in the two promoters generated a space of 8 Å between the two non-native thiol groups, which corresponds to the spacer arm of bismaleimidoethane (BMOE) (fig. S10B). Moreover, although F1935.37 is engaged in contacts between protomers in model I5 or I5/6 (fig. S6A), its position is not compatible with cross-linking using BMOE in either of the two corresponding interfaces. We used L196C as a negative control, because this residue is in the interface of the three models (figs. S6 and S9C) but never in a suitable position for the simultaneous reaction of BMOE with the residues of the two protomers. We detected dimers trapped by BMOE by Western blotting analysis. After exposure to BMOE, WT CCR5 and the L196C mutant existed mainly as monomers (Fig. 6F) with or without MVC. Under the same conditions, the F193C mutant showed only dimers independently of MVC treatment (Fig. 6F). These data support the third dimer conformation, IMVC, which we identified by modeling. The observation of F193C covalent dimers in the absence of MVC suggests that IMVC can exist with the unliganded receptor. Both modeling and cross-linking studies indicated that MVC stabilizes a third homodimer interface distinct from the I5 and I5/6 interfaces.

DISCUSSION

Despite the description of many class A GPCR homodimers and heterodimers, the existence of dimers in live cells is still under debate because of contradictory results depending on the methods used and ambiguous interpretation of the results. In addition, if dimers exist, are they playing an important role in GPCR functions? Answers to these questions are especially important given the dynamic complexity of these receptors, leading to receptor selectivity, specificity, and efficacy. Here, we combined classically used methods (cross-linking, energy transfer, and modeling) with a newly developed assay based on the RUSH export system to assess CCR5 dimerization and function in live cells. We based our predictive models on two known dimeric structures, keeping in mind that x-ray crystallography may produce artifactual packing. We obtained a convergence of results showing that CCR5 can adopt at least three dimeric conformations through distinct dimerization interfaces. We also provide evidence supporting an essential role for dimerization during receptor biogenesis.

Cross-linking and molecular modeling studies suggest that CCR5 arranges as symmetrical dimers in its ligand-free state using TM5 of both protomers. The molecular models built from two different crystallographic dimer structures (CXCR4 or MORs) explain the contribution of TM5 residues in the stabilization of the dimerization interfaces at the atomic level. The role of TM5 in dimerization interfaces has already been reported for several class A GPCRs (6, 7, 9, 36, 37). In those studies, TM5 mainly associates either with itself or with TM6, depending on the receptor. Here, we showed that for unliganded CCR5, these two modes of association may coexist. These modes being mutually exclusive, the unliganded CCR5 likely oscillates between two forms of dimers (I5 and I5/6). Our study does not exclude the possibility of other modes of association, such as I1/8, which presents a weaker association (7, 9). We showed that mutations in TM5 of CCR5 disrupted the I5 and I5/6 interfaces, although not fully preventing receptor dimerization (Figs. 2B and 3E). The multiplicity of unliganded forms of CCR5 is consistent with previous computational analyses that predicted the existence of several apo forms of CCR5 with different stabilities (33). The coexistence of several dimeric conformations is also consistent with a single-molecule fluorescence imaging study that showed preformed diverse dimeric structures of another class A GPCR (11). The transient dynamics of dimer formation, in which receptors show monomer-dimer equilibrium, may favor this process (38). More generally, the balance between multiple conformations of unliganded GPCR dimers may explain the difficulty of obtaining a consensus picture of class A GPCR dimer association.

Our experiments revealed that the presence of the inverse agonist MVC altered the interhelical contacts between TM5 and TM4/TM6 of CCR5 (Figs. 4 and 6). MVC stabilized a conformation that exposed K1975.41 at the receptor surface, thereby precluding the formation of I5 and I5/6 symmetrical dimers (Figs. 4F and 6). MVC stabilized CCR5 dimerization through a third interface (IMVC) that has not yet been described in a GPCR crystal structure (Fig. 6D and fig. S9). In addition, docking and chemical cross-linking experiments suggest that TM5 is also involved in the IMVC dimer interface (fig. S9 and Fig. 6F). The stabilization of a dimeric conformation by MVC highlights its molecular action as antiviral compound. This is consistent with the role of CCR5 oligomerization in preventing HIV-1 infection (39, 40).

In class A GPCRs, a common activation pathway involves a shift of TM7, a translation of TM3, and a rotation of TM5 and TM6 that causes a bending of TM6 and opening of the binding sites for G proteins or arrestins (9, 4146). Large conformational changes in TM5 and TM6 are especially required for the full engagement of the G protein (41, 47). Within dimers, cross-linking experiments have shown a rearrangement at TM4, TM5, and TM6 dimer interfaces upon receptor activation (25, 36, 48). The three models of CCR5 dimers proposed here each represent an inactive form of the receptor. The I5/6 organization is especially incompatible with receptor activation because of steric clashes that preclude the concerted motion of the TMs upon receptor activation (7). The coexistence of distinct inactive conformations has been reported for class A GPCRs (49, 50). In the case of the β2AR, the receptor population contains the same structural states, whether the receptor is unliganded or bound to an inverse agonist (49). Two inactive states may coexist in both the free and inverse agonist–bound β2AR, corresponding to the receptor with (state S1) or without (state S2) the ionic lock salt bridge formed between the DRY motif of TM3 (R3.50) and TM6 (D/E 6.30). A receptor without the ionic lock (state S2) is described as an intermediate state toward receptor activation (49). Here, we showed that MVC stabilized a different inactive conformation from the two ligand-free forms of CCR5. According to its sequence, CCR5 cannot form the ionic lock (51), which probably facilitates its increased dynamics. The introduction of the ionic lock by mutagenesis prevents CCR5 activation and its coupling to G proteins (51). The two unliganded forms of CCR5 (I5 and I5/6) may represent two possible CCR5 organizations of the “unlocked” state of β2AR (state S2). It is possible that MVC stabilizes a CCR5 conformation mimicking the β2AR conformational state with the ionic lock (state S1), explaining its inverse agonist property. Further studies could determine whether this concept can be generalized to the 30% of other class A GPCRs that, similarly to CCR5, lack the ionic lock.

The conformational heterogeneity of CCR5 raises questions about the role of these conformational states in receptor function. Although we cannot exclude the possibility that the amino acid introduced (lysine or cysteine) independently affected both dimerization and delivery to the plasma membrane, we propose that the assembly of either unliganded or MVC-bound CCR5 into stable dimers is required for receptor transport (Figs. 2 to 4), as suggested for other GPCRs (17, 29, 31, 32, 52). Impairing dimerization through TM5 prevents transport to the cell surface as observed for several mutations and using various assays. Furthermore, distinct inactive conformations of CCR5 enter the export pathway. Dimerization masks an ER retention sequence in class C GPCRs (3). Dimerization of class A GPCRs may also serve a chaperone-like function. This mechanism may stabilize conformers by reducing the free movement of the protein, enabling it to leave the ER, promote its interaction with specific components of the transport machinery, or both (53). CD4 is required for CCR5 delivery in primary T lymphocytes (54), suggesting a link between receptor dimerization and association with CD4 in these cells. Moreover, this CD4-CCR5 association has also been reported to occur at the plasma membrane (40, 5557). This association is likely to allosterically regulate the binding of chemokines to CCR5 and to increase the sensitivity of cells to HIV infection. This raises questions about the role of CD4 in modulating the dynamics and organization of CCR5 dimers and the role these dimers play in the pathophysiological functions of CCR5 (14). In conclusion, our data provide information on class A GPCR dimerization, its arrangement upon ligand binding, and its relevance to receptor biogenesis. CCR5 exists in different dimer organizations, either unliganded or inverse agonist–bound, which reflect distinct inactive conformations of the receptor. In addition to highlighting the structural dynamics of CCR5, these findings provide new avenues for future investigation of the role of CCR5 dimerization in HIV-1 pathogenesis and antiviral drug design.

MATERIALS AND METHODS

Plasmids and site-directed mutagenesis

FLAG-CCR5 was described previously (58). FLAG-SNAP-tag-CCR5 (FLAG-ST-CCR5), FLAG-CLIP-Tag-CCR5 (FLAG-CT-CCR5), and FLAG-AT2R were gifts from Cisbio Bioassays and C. Namhias (INSERM U981, Institut Gustave Roussy, University Paris Saclay, 94800 Villejuif, France) (59). DsRed-KDEL was purchased from Clontech (pDsRed2-ER). The constructs streptavidin-KDEL_SBP-EGFP-CCR5 (named SBP-CCR5) and streptavidin-KDEL_SBP-EGFP-β2AR (named SBP-β2AR) were derived from the streptavidin-KDEL_SBP-EGFP–E-cadherin construct (24) by substitution of the sequence encoding E-cadherin with the sequence encoding CCR5 or β2AR, respectively. The FLAG-CCR5-KKLV (named CCR5-KKLV) and FLAG-AT2R-KKLV (named AT2R-KKLV) constructs were obtained by the addition of the sequence encoding the ER retention signal KKLV to the C terminus of the sequences encoding FLAG-CCR5 or FLAG-AT2AR, respectively. This KKLV insertion as well as cysteine, lysine, and alanine substitutions were generated by site-directed mutagenesis using the QuikChange II Lightning Mutagenesis Kit (Agilent Technologies) according to the manufacturer’s instructions. The mutants were all verified by sequencing (Eurofins).

Cell lines, transfection, and stimulation

HEK 293 cells were cultured in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum and penicillin/streptomycin (Life Technologies) and were transfected with Lipofectamine 2000 (Life Technologies). To test the effects of ligands on CCR5 dimerization, the cells were preincubated for 24 hours with ligands before cross-linking, HTRF, or RUSH assays were performed. The CCR5 inverse agonist MVC (15) was obtained from the National Institutes of Health. The CCR5 antagonist 5P12 was a gift from O. Hartley (Department of Pathology and Immunology, Faculty of Medicine, University of Geneva, 1211 Geneva, Switzerland) (60). The CXCR4 antagonist AMD3100 (61) was purchased from Sigma-Aldrich. The chemokine CCL3 was purchased from Peprotech (tebu-bio), and [125I]CCL3 was obtained from PerkinElmer Life Sciences.

Cross-linking experiments

After transfection with the appropriate plasmids, HEK 293 cells, preincubated overnight (or not) with MVC, were treated with 0.5 mM dichloro(1,10-phenanthroline)copper(II) (Sigma-Aldrich) in tris-buffered saline (TBS) with 1 mM CaCl2 for 20 min at room temperature. The reaction was quenched by incubating cells with 10 mM N-ethylmaleimide (Sigma-Aldrich) for 15 min at 4°C. For DSP and BMOE cross-linking (62), the cells were incubated with 2 mM DSP (Sigma-Aldrich) or 1 mM BMOE (Sigma-Aldrich) in TBS for 20 min at room temperature. Cells were then collected and lysed in lysis buffer [0.5% n-dodecyl-β-d-maltoside, 1% iodoacetamide (50 mM), protease inhibitor in TBS] at 4°C for 30 min. Equal amounts of proteins were resolved by SDS–polyacrylamide gel electrophoresis and transferred to a membrane (Bio-Rad) for Western blotting analysis. The blots were incubated with a horseradish peroxidase (HRP)–conjugated antibody against the FLAG tag (M2, Sigma-Aldrich) for enzyme-linked chemiluminescence detection with the ECL system (Amersham Biosciences Life Sciences, GE Healthcare). A control of the protein quantity per lane was performed after an acid strip and incubation of the blots with an antibody against LDH5 (Biodesign) and a secondary HRP-conjugated anti-goat antibody (Dako). Band intensities on the same film were quantified by densitometry.

HTRF saturation assay

HEK 293 cells were transfected with a constant amount of the donor plasmid FLAG-ST-CCR5 (WT, 20 ng; mutants, 30 ng) and increasing amounts of acceptor plasmid FLAG-CT-CCR5 (WT, 0 to 200 ng; mutants, 0 to 300 ng) in separate wells. The total amount of DNA used for the transfections was maintained constant by adding appropriate amounts of pcDNA3. Twenty-four hours after transfection, the cells were detached, plated on black poly-d-lysine–treated 96-well plates (Greiner), and incubated for a further 24 hours. The cell surface expression of FLAG-ST-CCR5 or FLAG-CT-CCR5 was determined separately by labeling the cells for 2 hours at 37°C with either 100 nM Snap-Lumi4Tb or 400 nM Clip-Lumi4Tb (Cisbio Bioassays). The preparations were excited at 320 nm, and fluorescence emissions were measured at 620 nm to quantify the amount of cell surface expression of the donor or acceptor. A constant cell surface amount of the donor was verified in each experiment. In parallel, HTRF signals were obtained from another batch of cells co-labeled with 100 nM Snap-Lumi4Tb and 400 nM Clip-Red (Cisbio Bioassays) for 2 hours at 37°C. After excitation at 320 nm, fluorescence emissions were measured at 665 nm (emission wavelength of red) for 400 μs after a 50-μs delay on a Mithras LB 940 (Berthold Technologies). We calculated the mHTRF ratio as (signal at 665 nm/signal at 620 nm) × 1000 for each donor/acceptor ratio.

Fluorescence flow cytometry

Transfected HEK 293 cells were detached and incubated on ice with an antibody against the FLAG epitope (Sigma-Aldrich) followed by a phycoerythrin (PE)–conjugated anti-mouse antibody (BD) to determine the cell surface expression of FLAG-CCR5 or FLAG-ST-CCR5 and mutants. The total amount of CCR5 in the cells was determined after permeabilization of fixed cells with phosphate-buffered saline (PBS), 0.05% saponin, and staining as described earlier. The stained cells were washed and analyzed on a Gallios flow cytometer (Beckman).

RUSH assays

Transfected HEK 293 cells expressing components of the RUSH system (construct streptavidin-KDEL_SBP-EGFP-CCR5) were stimulated with 40 μM prewarmed biotin diluted in cell culture medium. At the times indicated in the figures, the culture dishes were placed on ice and the cells were detached with PBS-EDTA. The cells were incubated with an antibody against GFP (Roche), followed by a PE-conjugated anti-mouse antibody (BD) to determine the cell surface expression of SBP-EGFP-CCR5 (SBP-CCR5). Stained cells were washed and analyzed on a Gallios flow cytometer (Beckman). We verified in each experiment that the amounts of the WT CCR5 and its mutants were equivalent in the ER before they were released. In dual RUSH experiments, HEK 293 cells were cotransfected with a plasmid encoding the RUSH system for WT CCR5 and with pcDNA3 (control), FLAG-CCR5-KKLV (WT or mutants), or FLAG-AT2R-KKLV at a concentration ratio of 1:4. After staining, GFP-positive cells (corresponding to transfected cells) were analyzed by flow cytometry for SBP-CCR5 cell surface expression as described earlier. Equal intracellular amounts of the KKLV constructs were verified either by flow cytometry after cell permeabilization and staining or by Western blotting analysis.

Confocal imaging

HEK 293 cells stably expressing FLAG-ST-WT-CCR5 or FLAG-ST-L196K or transiently expressing FLAG-WT-CCR5-KKLV were transfected with a plasmid encoding DsRed-KDEL and plated on coverslips coated with poly-d-lysine (Sigma-Aldrich). Forty-eight hours later, the cells were fixed in 4% paraformaldehyde in PBS (pH 7.4), permeabilized with 0.05% saponin buffer (3% bovine serum albumin, TBS), and incubated with a polyclonal antibody against FLAG (Sigma-Aldrich) and an Alexa Fluor 488–conjugated goat anti-rabbit antibody (Invitrogen). Samples were visualized using a TCS SP5 confocal microscope (Leica Microsystems) equipped with 488 and 561 lasers and a 63×, 1.4 Plan Apochromat oil objective.

Radioligand binding assays

Membrane preparations from HEK 293 cells transfected with FLAG-WT-CCR5, FLAG-K197A, FLAG-K219A, FLAG-ST-WT-CCR5, or FLAG-ST-L196K as well as direct binding of [125I]CCL3 or displacement of [125I]CCL3 (0.1 nM) (PerkinElmer Life Sciences) in the presence of MVC were performed as previously reported (16).

Molecular modeling of CCR5

The CCR5 protomer was prepared from the crystallographic structure of the receptor [Protein Data Bank (PDB) ID: 4MBS] as previously described (63). The receptor is truncated at both termini (missing sequences: M1-S6 and K314-L352) and includes two sulfotyrosines (posttranslationally modified at Y10 and Y14). MVC was prepared from the description in the same structure. Atom types were manually verified, and hydrogen atoms were then added using MOE2014 (Chemical Computing Group); parameter and coordinate files were generated using antechamber with the AM1-BCC charge model. Two CCR5 protomers were assembled according to their respective superimposition on the first and second protein chains of the template crystallographic structure: 3OE0 for the I5 model and 4DKL for the I5/6 model. Superimposition was performed with MOE2014 to maximize the spatial overlap of the α carbons of equivalent residues, paired on the basis of sequence alignment guided by generic constraints for GPCRs. The few steric clashes observed at the protein interface were manually resolved by editing the rotameric state of residues involved in atomic collisions. The IMVC interface of the CCR5 dimer was modeled with the “easy” interface of the HADDOCK web server (http://haddock.science.uu.nl/enmr/services/HADDOCK/) (64). The two input protomers were representative structures of the trajectories computed for the MVC-free and MVC-bound forms of CCR5, respectively. K197 was indicated as an “active” residue (directly involved in the interaction). “Passive” residues (surrounding the surface residues) were defined automatically. The 191 output complexes defined six different clusters, including only one with a parallel dimer (with respect to the 7TM helix bundle axis). Free CCR5, CCR5 bound to MVC, the I5 dimer, the I5/6 dimer, and the IMVC dimer were each submitted to MD with the graphics processing unit code of AMBER14 (University of California, San Francisco, CA). The receptor (or the complex or dimer) was placed into a hydrated lipid bilayer using the input generator of CHARMM-GUI (65, 66). The membrane is composed of palmitoyloleoylphosphatidylcholine, palmitoyloleoylphosphatidylethanolamine, and cholesterol at a ratio of 2:2:1. The upper and lower leaflets have the same composition. The water layer from either side of the membrane is 12 Å thick. It includes potassium and chloride ions at a concentration of 0.15 M. After energy minimization, the system was heated to 300 K at constant volume during 75 ps while fixing the positions of all atoms using harmonic constraints of 10 kcal/mol Å2 for protein and ions and 2.5 kcal/mol Å2 for lipids and water molecules. The constraints were decreased stepwise while the system was maintained at 300 K at constant volume during 175 to 450 ps. An additional 5 ns of unconstrained MD at 300 K enabled equilibration of the periodic boundary conditions. Three independent MD productions of about 300 ns each were run for all four studied systems. Trajectories were analyzed using the cpptraj tool of AMBER14 for the measurement of interatomic distances, the detection of intramolecular H-bonds, the calculation of atomic fluctuation, and the clustering of production frames. Noncovalent interactions between the protomers in the modeled dimers were automatically detected using IChem (67). H-bond: length, ≤3.5 Å; angle, 180° ± 60°; ionic bond: length, ≤4.0 Å; aromatic interaction: distance between aromatic ring centers, ≤4.0 Å; angle between ring planes, 180° ± 30° or 90° ± 60°; hydrophobic contact: distance between nonhydrogen atoms, ≤4.5 Å. The relative accessibility of residue side chains to solvent was measured using a water probe in the savol10 routine of Sybylx2.1.1 (Tripos Inc.). 3D models of I5 and I5/6 with L196K, I200K, L205K, or L208K mutations were prepared from representative structures of the MD trajectories using MOE2016.08. Note that these models have not been refined and have not been submitted to MD.

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/11/529/eaal2869/DC1

Fig. S1. Cysteine cross-linking identifies TM5 in dimer interfaces.

Fig. S2. Lysine mutagenesis destabilizes CCR5 dimer formation without altering receptor folding.

Fig. S3. Expression of FLAG-CCR5-KKLV.

Fig. S4. MVC favors the transport of CCR5 to the cell surface.

Fig. S5. MVC causes a rearrangement of CCR5 TM5.

Fig. S6. Residues involved in interprotomer interactions.

Fig. S7. MD simulations of the CCR5 monomer when free or in complex with MVC.

Fig. S8. Influence of MVC on CCR5 dynamics.

Fig. S9. Representative model of the IMVC dimer based on rigid protein-protein docking.

Fig. S10. 3D models of the CCR5 covalent dimers obtained by cross-linking with either DSP or BMOE.

Supplementary PDB files

REFERENCES AND NOTES

Acknowledgments: We thank Cisbio Bioassays and our colleagues C. Namhias (INSERM U981, Institut Gustave Roussy, 94800 Villejuif, France) and O. Hartley (University of Geneva, 1211 Geneva, Switzerland) for the gifts of reagents. We thank R. Paul (Functional Genetics of Infectious Diseases Unit, Institut Pasteur, 75015 Paris, France) for help with statistical analysis. We thank T. Lagache (BioImage Analysis Unit, Institut Pasteur, 75015 Paris, France) for helpful discussion on HTRF experiments. We thank the flow cytometry and dynamic imaging platforms of the Institut Pasteur for technical help. The simulations were performed on computational resources supported by the High-Performance Computing facilities at the University of Strasbourg. We acknowledge the use of the GPCRdb database (www.gpcrdb.org). Funding: This work was supported by Agence Nationale de Recherches sur le SIDA et les hépatites virales (ANRS), the Fondation pour la Recherche Médicale (FRM) (DEQ20120323723), INSERM, Institut Pasteur, the French Government’s Investissement d’Avenir program, Laboratoire d’Excellence “Integrative Biology of Emerging Infectious Diseases” (grant ANR-10-LABX-62-IBEID), and Laboratoire d’Excellence “Medalis” (grant ANR-10-LABX-0034-Medalis). J.J. was supported by a fellowship from the China scholarship council. Author contributions: J.J., F.M., G.B., F.K., Z.Z., N.C., and E.K. performed experiments and analyzed data. J.J., G.B., F.A.-S., F.P., B.L., E.K., and A.B. designed research and analyzed data. B.L., E.K., and A.B. wrote the manuscript. All authors reviewed and edited the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials accessibility: All data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.
View Abstract

Navigate This Article