Structural and Functional Characterization of Human CXCR4 as a Chemokine Receptor and HIV-1 Co-receptor by Mutagenesis and Molecular Modeling Studies*

The human CXC chemokine receptor 4 (CXCR4) is a receptor for the chemokine stromal cell-derived factor (SDF-1α) and a co-receptor for the entry of specific strains of human immunodeficiency virus type I (HIV-1). CXCR4 is also recognized by an antagonistic chemokine, the viral macrophage inflammatory protein II (vMIP-II) encoded by human herpesvirus type VIII. SDF-1α or vMIP-II binding to CXCR4 can inhibit HIV-1 entry via this co-receptor. An approach combining protein structural modeling and site-directed mutagenesis was used to probe the structure-function relationship of CXCR4, and interactions with its ligands SDF-1α and vMIP-II and HIV-1 envelope protein gp120. Hypothetical three-dimensional structures were proposed by molecular modeling studies of the CXCR4·SDF-1α complex, which rationalize extensive biological information on the role of CXCR4 in its interactions with HIV-1 envelope protein gp120. With site-directed mutagenesis, we have identified that the amino acid residues Asp (D20A) and Tyr (Y21A) in the N-terminal domain and the residue Glu (E268A) in extracellular loop 3 (ECL3) are involved in ligand binding, whereas the mutation Y190A in extracellular loop 2 (ECL2) impairs the signaling mediated by SDF-1α. As an HIV-1 co-receptor, we found that the N-terminal domain, ECL2, and ECL3 of CXCR4 are involved in HIV-1 entry. These structural and mutational studies provide valuable information regarding the structural basis for CXCR4 activity in chemokine binding and HIV-1 viral entry, and could guide the design of novel targeted inhibitors.

Chemokines are a family of 8 -10-kDa small proteins that act as chemoattractants of various types of leukocytes to sites of inflammation and to secondary lymphoid organs. Based on the positions of two conserved cysteine residues in their N termini, chemokines can be divided into four subfamilies: CC, CXC, CX3C, and C (1,2). The stromal cell-derived factor-1 (SDF-1␣) 1 is one of the CXC chemokines, which plays critical roles in the migration, proliferation, and differentiation of leukocytes. SDF-1␣ is the only known natural ligand of CXCR4 receptor (3,4). CXCR4 can also be recognized by an antagonistic ligand, the viral macrophage inflammatory protein-II (vMIP-II) encoded by the Kaposi's sarcoma-associated herpesvirus (5). vMIP-II acts as a selective chemoattractant for T helper 2 cells and monocytes and is an agonist for CCR8 (6). vMIP-II displays a broader spectrum of receptor activities than any mammalian chemokine, as it binds with high affinity to a number of both CC and CXC chemokine receptors including CXCR4 and CCR5 and inhibits cell entry of human immunodeficiency virus type I (HIV-1) mediated by these receptors (7,8).
CXCR4 belongs to the family of seven transmembrane Gprotein-coupled receptors that transduce signals via heterotrimeric G-proteins (9). Recent studies with knockout mice of CXCR4 have demonstrated that this molecule plays an important role in immunomodulation, organogenesis, hematopoiesis, and derailed cerebellar neuron migration (10 -12). CXCR4 has also been identified as one of co-receptors for HIV-1 (13). CXCR4 mediates infection of T cell line tropic HIV-1 strains and has also been found to be used by human immunodeficiency virus type II (HIV-2) strains adapted to replication in CD4-negative cell lines (14).
The characterization of structural and functional determinants of CXCR4 for its ligand binding activity and HIV-1 coreceptor function is essential for understanding mechanisms of HIV-1 viral entry and developing novel therapeutic agents for HIV-1 infection. To this end, several studies have been carried out recently by a number of laboratories using chimeric chemokine receptors and mutants to demonstrate that multiple domains of CXCR4 are required for HIV-1 co-receptor activity (15)(16)(17)(18)(19)(20)(21)(22)(23). Furthermore, it has been demonstrated that the Nterminal domain of CXCR4 was a determinant in SDF-1␣ binding and the second extracellular loop (ECL2) of CXCR4 was involved in receptor signaling (17,20,22).
To better understand CXCR4 receptor and co-receptor functions, we have used an approach combining molecular modeling and experimental validation, similar to that developed by us and other groups in earlier studies of chemokine receptors interleukin-8 receptor ␤ (24) and CCR5 (25), (26), to propose plausible three-dimensional structural models of CXCR4 and its complex with SDF-1␣ or vMIP-II. These models rationalize the available data from chimeric and mutational experiments and provide a possible framework for understanding the structural basis of CXCR4 interactions with chemokine ligands and the HIV-1 envelope glycoprotein gp120. Furthermore, based on the molecular modeling results, we carried out site-directed mutagenesis studies on the N-terminal domain, the second extracellular loop (ECL2), and the third extracellular loop (ECL3) of CXCR4 to test specific structure-function hypotheses from our structural models. The chemokine ligand binding and HIV-1 membrane fusion studies of these designed CXCR4 mutants have provided valuable information about structure-activity relationships in the N-terminal domain and ECL2 of CXCR4 in chemokine binding and HIV-1 entry.

Molecular
Modeling-Molecular modeling studies of the CXCR4⅐SDF-1␣ complex were carried out by using a procedure developed from our previous study of another chemokine receptor, CCR5 (25). Energy minimization and molecular dynamics simulations were performed by using the molecular modeling package of InsightII/Discover (Biosym Technologies Inc., San Diego, CA) on Silicon Graphics Indigo2 and Silicon Graphics Onyx workstations. The structural models of CXCR4 and its complex with SDF-1␣ were constructed in several steps. First, the assignment of transmembrane segments of CXCR4 was made based on the results published by others (27). The structures of these CXCR4 segments were then generated by homology modeling based on the experimentally determined structure of the transmembrane core of bacteriorhodopsin (28,29). The structures of extracellular loops of CXCR4 were modeled by using the loop generation program of InsightII. The obtained structure of CXCR4 was energy-minimized, with two disulfide bridges Cys 28 -Cys 274 and Cys 109 -Cys 186 used as distance restraints (10). The dihedral angles of transmembrane regions were also constrained to maintain the original helical arrangement during energy minimization.
To generate the complex of CXCR4 with SDF-1␣, the crystal structure of SDF-1␣ (30), which is similar to the NMR structure (31), was manually placed on top of the extracellular domains of CXCR4, and the resulting structure was again energy-minimized. Finally, high temperature molecular dynamics simulation was performed for 200 ps to search the possible conformations of the receptor-ligand complex. Because the extracellular domains of CXCR4 are most likely the interface for ligand binding, the structure of the transmembrane helices of CXCR4 was fixed during the dynamics simulation. The conformation of SDF-1␣ (except its flexible N-terminal residues 1-7) was also constrained during the calculation to maintain its crystal structure. This facilitated a more efficient conformational searching of the complex. Because this study focused on the receptor-ligand interaction, the cytoplasmic domains of CXCR4, which are not involved in ligand binding and CXCR4 co-receptor function for HIV-1, were not included in the models. Employing a similar procedure described above, models for the CXCR4⅐vMIP-II complex were generated and analyzed using the crystal structure of vMIP-II (32), which is similar to the NMR structure (33).
Chemokines, Cells, Antibodies, and Plasmids-Recombinant human chemokines SDF-1␣ and viral chemokine vMIP-II (R&D Systems Inc.) were lyophilized and dissolved as 1 or 2.5 g/l stock solutions in sterile phosphate-buffered saline (PBS) and stored at Ϫ20°C in aliquots. They were diluted to appropriate concentrations immediately before use. The radioiodinated SDF-1␣ was purchased from PerkinElmer Life Sciences. The specific activity of 125 I-SDF-1␣ was 2200 Ci/mmol. Cell culture media and G418 were purchased from Life Technologies, Inc. Antibody 12G5 was purchased from PharMingen and 44708.111 from R&D Systems. Plasmid pCDNA-CXCR4, the human kidney cell line 293, and the murine 3T3-T4 cells were obtained through the AIDS Research and Reference Reagent Program (Division of AIDS, NIAID, National Instituted of Health, Bethesda, MD). 293 and 3T3-T4 cells were maintained in Dulbecco's modified Eagle's medium plus 10% fetal bovine serum. Recombinant vaccinia viruses encoding two envelopes of HIV-1, vSC60 (IIIB) and vBD3 (89.6), and vTF1.1 encoding T7 RNA polymerase were generous gifts from Dr. Robert W. Doms (University of Pennsylvania, Philadelphia, PA).
Site-directed Mutagenesis-All mutants of CXCR4 were prepared with the MORPH site-specific plasmid DNA mutagenesis kit (5 Prime 3 3 Prime, Inc.), according to the manufacturer's instructions. Sitedirected mutagenesis was performed on CXCR4 in a pcDNA3 vector. The mutations were confirmed by sequencing. The plasmid constructs of CXCR4 wild-type and mutants were transfected into 293 cells by the calcium phosphate precipitation method. Twenty-four hours after transfection, selection for stable expression cells was initiated by the addition of G418 (800 g/ml). Transfected cells were evaluated for expression level at the cell surface by flow cytometry, using antibodies 12G5 and 44708.111 to the second extracellular loop of CXCR4. 125 I-SDF-1␣ Binding Assay-The stably transfected cells were harvested in PBS (Ca 2ϩ -and Mg 2ϩ -free) plus 0.5 nM EDTA and washed twice with PBS. Ligand binding experiments were performed using a single concentration (0.2 nM) of 125 I-SDF-1␣ in a final volume of 100 l of binding buffer (50 nM Hepes, pH 7.4, 1 nM CaCl 2 , 5 nM MgCl 2 , 0.1% bovine serum albumin) containing 5 ϫ 10 5 cells. Nonspecific binding was determined by the addition of 100 nM unlabeled chemokines. Samples were incubated for 60 min at room temperature. The incubation was terminated by separating the cells from the binding buffer by centrifugation and washing once with 500 l of cold binding buffer. Bound ligands were determined by counting ␥ emissions. At least three independent experiments were performed.
Flow Cytometry-Sup T1 cells (2 ϫ 10 5 ) were washed with fluorescence-activated cell sorting (FACS) buffer (0.5% bovine serum albumin, 0.05% sodium azide in PBS) and incubated with an anti-CXCR4 monoclonal antibody (mAb) 12G5 (10 g/ml) for 30 min at 4°C. After washing with FACS buffer, cells were incubated with 10 g of fluorescein isothiocyanate-conjugated goat anti-mouse IgG (Sigma) for 30 min at 4°C. After washing twice with FACS buffer, cells were fixed in the fixing buffer (2% paraformaldehyde in PBS), and then analyzed on a FACScan flow cytometer (Coulter EPICS Elite, Coulter Corp., Hialeah, FL). For competition with SDF-1␣ and vMIP-II, cells were incubated with 12G5 and SDF-1␣ (50 nM) or vMIP-II (50 nM) at 4°C for 40 min. After staining, cells were fixed for FACS analysis. For detection of CXCR4 internalization, cells were first incubated with SDF-1␣ (50 nM) at 37°C for 40 min, and then washed once with acidic glycine buffer (pH 3.5), and twice with FACS buffer (34). After staining, cells were analyzed by FACS.
Gene Reporter Fusion Assay-A gene reporter fusion assay was used to determine the co-receptor activity of wild-type and mutant CXCR4 in mediating HIV-1 viral entry following a modified procedure published by others (35)(36)(37). HIV-1 Env proteins and T7 RNA polymerase were introduced into effector 293 cells by infection with recombinant vaccinia virus at a multiplicity of infection of 10 for 2 h. Infected cells were then trypsinized, washed with PBS, resuspended in medium, and incubated overnight at 32°C in the presence of rifampicin (100 g/ml). 3T3-T4 target cells expressing CD4 were co-transfected in six-well plates with plasmids encoding CXCR4 wild-type or mutants and luciferase under control of T7 promotor, using the calcium-phosphate precipitation method. Four to 6 h after transfection, cells were lifted, washed with PBS, seeded in 24-well plates, and incubated at 37°C overnight. To initiate fusion, 10 5 effector cells were added to each well and incubated at 37°C in the presence of ara-C and rifampicin. After 5 h of fusion, cells were lysed in 150 l of reporter lysis buffer (PharMingen) and assayed for luciferase activity by using commercially available reagents (PharMingen).
Intracellular Calcium Measurements-Following a modified procedure published by others (38,39), untransduced 293 cells and CXCR4 or its mutants stably transfected 293 cells were cultured in Dulbecco's modified Eagle's medium contained 10% fetal bovine serum. Cells were detached with 0.1% EDTA and washed twice with PBS. For Ca 2ϩ mobilization studies, 5 ϫ 10 6 /ml cells were loaded with the fluorescent dye, fura-2 (2 M, Molecular Probes, Eugene, OR) in Hank's balanced salt solution (140 mM NaCl, 5 mM KCl, 10 mM HEPES, pH 7.4, 1 mM CaCl 2 , 1 mM MgCl 2 , 1 mg/ml glucose, and 0.025% bovine serum albumin), for 30 min at 37°C. The cells were washed three times and resuspended at a concentration of 30 -40 ϫ 10 6 /ml, and then 1.5-2 ϫ 10 6 cells were tested in the same buffer. Intracellular Ca 2ϩ mobilization was measured using excitation at 340 and 380 nm on a fluorescence spectrometer (LS50B, PerkinElmer, Beaconsfield, England) upon the stimulation with SDF-1␣. Calibration was performed using 10% Triton X-100 for total fluorophore release and 0.5 M EGTA to chelate free Ca 2ϩ . Intracellular Ca 2ϩ concentrations were calculated using a fluorescence spectrometer measurement program.

RESULTS
Structural Models of the CXCR4⅐SDF-1␣ Complex-A model of the CXCR4⅐SDF-1␣ complex (Fig. 1A) was built by using a procedure similar to that used in our previous studies of CXCR2 (24) and CCR5 (25). Various structures extracted from the molecular dynamics simulation were subjected to further analysis. At first, all structures were minimized and the inter-action energy between CXCR4 and SDF-1␣ was calculated. The averaged interaction energy of various structures is Ϫ371 kcal/ mol and the standard deviation 13 kcal/mol. Among these structures, the conformation extracted from simulation time 130 ps had the lowest (most favorable) interaction energy of Ϫ405 kcal/mol, which compared favorably to Ϫ401 kcal/mol for the second lowest energy conformer and Ϫ371 kcal/mol for the averaged interaction energy ( Fig. 2A). Therefore, we performed further analysis on this conformation. The single residue contribution in CXCR4 to bind SDF-1␣ was calculated. Along the CXCR4 peptide chain, we found that the N terminus, ECL2, and ECL3 had the most contribution to the receptor-ligand interactions (Fig. 2B). Table I lists the residues on CXCR4 with the most contributions to these interactions among all the amino acid residues analyzed. Similarly, the structures of CXCR4⅐vMIP-II complex were also generated and a representative model is shown in Fig. 1B. We note that a structural model for CXCR4 has been shown previously by Chabot et al. (40). Because the structural details of the model by Chabot et al. are not available, we are unable to compare their model with ours.
Cell Surface Expression of CXCR4 Mutants-Based on the information of structural modeling and receptor-ligand binding energy, 12 residues at the N-terminal domain, ECL2, and ECL3 were selected for further site-directed mutagenesis studies. All of selected residues were individually substituted by alanine. The mutant constructs were transiently or stably transfected into 293 cells. Analysis of cell surface expression by flow cytometry using antibodies 12G5 and 44708.111 against CXCR4 indicated that most of CXCR4 mutants displayed an expression level comparable with wild-type CXCR4, but two constructs, W195A and Q200A, had extremely poor expression on the cell surface and were not included in further experiments (Fig. 3).
Ligand Binding Activity-All of the mutants with expression levels comparable with wild-type CXCR4 were further assayed for ligand binding activity. 293 cells, stably transfected with  wild-type and mutant CXCR4, were tested for competition binding with 125 I-labeled SDF-1␣ in the presence of unlabeled SDF-1␣ and vMIP-II. As shown as Fig. 4A, the three mutants R183A, Y184A, and Y190A in ECL2 had no effect on SDF-1␣ and vMIP-II binding, whereas the mutations Y7A, D20A, and Y21A in the N-terminal domain showed 80%, less than 50%, and less than 10% binding activities, respectively, with both SDF-1␣ and vMIP-II compared with wild-type CXCR4. Another mutation, E268A in the ECL3, retained only 50% binding activity. The monoclonal antibody 12G5, which recognizes an epitope in ECL2 (16,41), can competitively bind to CXCR4 with SDF-1␣ and vMIP-II and block HIV-1 entry supported by the CXCR4 co-receptor (42). 293 cells stably expressing wild-type or mutant CXCR4 were incubated with mAb 12G5 (10 g/ml) in the presence of 100 nM SDF-1␣ or vMIP-II. The results (Fig. 4B) obtained by FACS analysis were very similar to that of 125 Ilabeled SDF-1␣ binding experiments (Fig. 4A). Nevertheless, the ability of SDF-1␣ to compete with 12G5 was weaker than that of vMIP-II. The mutants Y7A, D20A, and Y21A in Nterminal domain and E268A in ECL3 showed a significant reduction of inhibition of 12G5 binding to CXCR4.
Activation of CXCR4 Mutants by SDF-1␣-To determine the effects of CXCR4 mutations on signal transduction in response to SDF-1␣, we measured intracellular Ca 2ϩ mobilization in stably transfected 293 cells upon stimulation with 50 nM SDF-1␣. We found that mutant Y21A in the N-terminal domain and E268A in ECL3, which exhibited lower ligand binding activity, were impaired in signaling mediated by SDF-1␣ (Fig. 5A). A low intensity Ca 2ϩ response was observed for mutants Y190A in the ECL2, although this mutant maintained binding activity comparable with wild-type CXCR4. Other CXCR4 mutants display signaling activity corresponding to their binding activity for SDF-1␣. The effect of CXCR4 mutation on SDF-1␣ signaling is dose dependent as shown by experiments of two representative CXCR4 mutants S5A and Y21A together with wild-type CXCR4 as a control (Fig. 5B).
The chemokine receptor CXCR4 can be induced to undergo rapid internalization by the ligand, SDF-1␣ (42)(43)(44). This property was used to address indirectly the ability of SDF-1␣ to activate CXCR4. We tested wild-type and mutant CXCR4 for internalization induced by SDF-1␣. After the incubation with 50 nM SDF-1␣ at 37°C for 40 min, the stably expressing cells were washed with acidic glycine buffer (34), and the cell surface CXCR4 was analyzed by flow cytometry. As shown as Fig. 6, one mutant R183A in ECL2 completely lost the potential for receptor internalization caused by SDF-1␣. D20A and Y21A in the N-terminal domain and Y184A in the ECL2 partially altered the receptor internalization by SDF-1␣.
Co-receptor Activity of CXCR4 Mutants-Co-receptor activities of wild-type and mutant CXCR4 were determined by a well characterized luciferase-based cell-cell fusion assay (35)(36)(37). As effector cells, 293 cells were co-infected with vaccinia virusencoded HIV-1 gp120 and vaccinia virus-encoded luciferase under the control of T7 promoter. 3T3-T4 target cells were co-transfected with pCDNA expression vectors with wild-type CXCR4 or mutations and T7 RNA polymerase. After overnight expression, both cell types were mixed and fusion was allowed to proceed for 4 -6 h. Finally, fusion activity was quantitatively analyzed by the luciferase assay. The results are illustrated in Fig. 7, and are presented as an average of the percentages of wild-type CXCR4. As a negative control, we used the Q200A mutant that appeared undetectable on cell surface expression. Alanine substitution of Tyr 21 in the N-terminal domain and Arg 183 in ECL2 resulted in a significant reduction (Ͼ50%) in co-receptor activity for both the T cell line-tropic IIIB and dual-tropic 89.6 HIV-1 envelopes. The mutants Y184A and Y190A in ECL2 and Q272A in ECL3 exhibited 60 -80% of wild-type activity in the fusion assay. Interestingly, the mutants E26A in the N-terminal domain and E268A in ECL3 had minor inhibitory effects on co-receptor activity for T cell linetropic IIIB, but yielded Ͼ50% and ϳ30% reduction of wild-type fusion activity, respectively, for the dual-tropic 89.6 envelope. DISCUSSION The chemokine receptor CXCR4 and its ligand SDF-1␣ have diverse biological functions, apparently different from those of other chemokine receptors and ligands. In addition, CXCR4 has been identified as an essential co-receptor for entry of T cell line tropic HIV-1 strains. Understanding the structure-activity relationships of CXCR4 with its ligand binding and HIV-1 envelope protein interaction is important for determining the molecular mechanisms by which CXCR4 functions as an HIV-1 co-receptor and, thus, for further design of novel anti-retroviral drugs. In this study, we have characterized the structurefunction relationship of CXCR4 as chemokine receptor and HIV-1 co-receptor, by combining computer modeling with sitedirected mutagenesis.
Our molecular modeling studies of the CXCR4⅐SDF-1␣ complex suggested that the N terminus, ECL2, and ECL3 of CXCR4 were involved in binding with SDF-1␣. The N-terminal domain plays a major role in ligand binding, contributing more receptor-ligand interaction energy than ECL2 and ECL3. Many residues within these segments are potentially involved in the interactions with SDF-1␣. Via the guidance of modeling studies, we selected residues for further mutagenesis analyses. Results with 125 I-SDF-1␣ and mAb 12G5 binding assays indicated that the mutants in the N terminus were found to be critical for ligand binding, and Glu 268 in ECL3 was partially involved in binding with SDF-1␣. Although the residues in ECL2 were not fully required for ligand binding, the mutant Y190A impaired the receptor activation. Of note, the role of ECL2 in signaling was observed by other groups (20,22). This suggests that our modeling study may provide a plausible structural basis to understand the functional role of ECL2. Brelot et al. (22) reported that the residue Tyr (Y21A) in the N-terminal domain was required for binding although not for signaling, but in our experiments the mutant Y21A was found to be impaired in signaling in response to SDF-1␣. The discrepancy may be due to different measurements of Ca 2ϩ mobilization. We used stably expressing cell lines and multiple cells for measuring Ca 2ϩ flux, whereas Brelot et al. employed transient expressing cells and single cells for assay.
In addition to the CXCR4⅐SDF-1␣ complex, the interaction between CXCR4 and vMIP-II was analyzed. From molecular modeling studies, vMIP-II probably recognizes CXCR4 in a manner similar to that for SDF-1␣, involving the N-terminal domain of the ligand and similar segments on the receptor. This seems to be consistent with the overall structural similarity between vMIP-II and SDF-1␣ (33). To further verify the similarity in CXCR4 binding modes of vMIP-II and SDF-1␣, we compared receptor binding activity of vMIP-II and SDF-1␣ using 125 I-SDF-1␣ and mAb 12G5 competition binding assays for each of our CXCR4 mutants and observed no major difference between SDF-1␣ and vMIP-II for any mutation. In competition assays with mAb 12G5, almost all mutations exhibited weaker ability of SDF-1␣ to compete with 12G5, than that with vMIP-II. Overall, these binding results are consistent with the modeling results and support the notion that CXCR4⅐SDF-1␣ and CXCR4⅐vMIP-II interfaces are similar or at least overlapping. The importance of the N-terminal domain of both SDF-1␣ and vMIP-II for receptor binding has been confirmed by studies of synthetic peptides derived from the N-terminal domain of SDF-1␣ (38,(45)(46)(47) and vMIP-II (48 -50).
Previous studies of the C5a chemoattractant and its receptor led to development of a two-site model of chemokine-receptor interactions (51). Recent studies on SDF-1␣ (31,38) and CXCR4 (20) proposed a model in which SDF-1␣ first binds directly to the N-terminal domain of CXCR4 with high affinity, and then further interacts with an extracellular loop-formed pocket of CXCR4, resulting in signaling. Although our model demonstrated one static complex conformation but not the dynamic nature of binding, the CXCR4⅐SDF-1␣ complex model and interaction energy distribution provided a picture in which the N-terminal domain, ECL2, and ECL3 of CXCR4 is involved in interactions with SDF-1␣. Our further mutational data suggest that the N-terminal domain of CXCR4 plays a major role in binding with SDF-1␣, whereas ECL2, and possibly ECL3 of CXCR4 are required for signaling. This is in agreement with a two-site model for CXCR4⅐SDF-1␣ interactions. As a starting point to characterize the structure-activity relationship of the CXCR4⅐SDF-1␣ complex, our model appears to be satisfactory. Furthermore, by using the present experimental results, we can build possibly more accurate models in the future to explore further information regarding interactions between chemokine receptors and their cognate chemokines.
The C terminus (52,53) and the second (54) and third (55) intracellular loops of chemokine receptors were reported to directly influence receptor internalization. However, the general details of ligand-receptor interactions involving the receptor extracellular regions that are important for receptor internalization remain poorly understood. We tested mutations of CXCR4 for SDF-1␣-induced internalization. The mutants D20A and Y21A in the N-terminal domain had low binding activity, and R183A in ECL2 impaired receptor internalization. Interestingly, a mutant Y190A in ECL2 impaired the signaling but maintained normal internalization activity. This suggests that CXCR4 internalization may be an activity somewhat independent from signaling.
Previous studies have indicated that the N-terminal domain, ECL2, and ECL3 of CXCR4 were required for HIV-1 co-receptor activity (17)(18)(19)(20)22). In this study, the mutated sites located in N-terminal domain, ECL2, and ECL3 of CXCR4 were identified by cell-cell fusion assays to impair the co-receptor activ- ity, in agreement with previous reports. This implies that the structure formed by the N-terminal domain and the extracellular loops of CXCR4 was required for interactions with the HIV-1 envelope protein. Although the complex of CXCR4 and HIV-1 gp120 is not modeled here, the models proposed for the CXCR4⅐SDF-1␣ complex could provide valuable information on the structural details of the CXCR4⅐gp120 interaction, as a recent study has shown that the ligand SDF-1␣ and HIV-1 envelope protein gp120 seem to share some common structural determinants for their functional interactions with CXCR4 (22). Whereas the N-terminal domain and ECL3 of CXCR4 appear to be common binding sites for both SDF-1␣ and the HIV-1 envelope, as indicated by mutations Y21A, E26A, and E268 (Figs. 4A and 7), it was interesting to note that mutations R183A, Y184A, and Y190A of ECL2 on CXCR4 seemed to affect CXCR4⅐gp120 interaction as indicated by decreased co-receptor activity (Fig. 7), but not CXCR4⅐SDF-1␣ binding (Fig. 4A). This finding implies a basis for the development of inhibitory molecules targeting these residues and others yet to be identified for the selective disruption of CXCR4 co-receptor function in mediating HIV-1 entry, without compromising natural ligand binding for normal function of the receptor.
The crystallographic structure of SDF-1␣ reveals a positively charged (overall ϩ8) surface that presumably binds to the negatively charged surface (Ϫ9) through electrostatic interactions (30). Meanwhile, several CXCR4 antagonists, such as ALX-40 (56), T-22 (57), and AMD3100 (58), are highly positively charged (ϩ8 or ϩ9). Mutagenesis studies also indicated that several negatively charged residues in the N-terminal domain and extracellular loops are important for ligand binding and interactions with HIV-1 envelope protein (19,21,22,59). Nevertheless, this study with structural modeling and mutagenesis suggested that Tyr residues in the N-terminal domain and ECL2 of CXCR4 play an important role in ligand binding, signaling, and co-receptor activity. This is in agreement with previous reports (22,23). Tyr residues of CCR5 were found to be involved in receptor activity (25) and in co-receptor function (22). Farzan et al. (60) reported that the tyrosine sulfation in the N-terminal domain of CCR5 played a role in gp120 binding and viral entry. The tyrosine sulfation of chemokine receptors may likely contribute to ligand binding and signaling.
In conclusion, we have used an approach combining protein structure modeling and site-directed mutagenesis to probe the structure of CXCR4 and its interactions with chemokine ligands and HIV-1 envelope protein. Hypothetical three-dimensional structures have been proposed for the complex between CXCR4 and SDF-1␣. In the absence of an experimentally determined structure of CXCR4, these proposed structures serve as a useful vehicle to rationalize data from genetic and molecular biological experiments. Based on these models, we carried out further site-directed mutagenesis studies to probe the structure-function relationships of the N-terminal domain, ECL2, and ECL3 of CXCR4. These studies have provided valuable information regarding conformational determinants of the biological activity of these sites. Together with results published by other laboratories, they further support the hypothesis that the N-terminal domain of CXCR4 is critical for ligand binding, and the ECL2 plays a major role in SDF-1␣induced signaling. As an HIV-1 co-receptor, the N-terminal domain, ECL2, and ECL3 of CXCR4 are involved in interactions with HIV-1 envelope protein. Finally, the structural models for CXCR4 receptor-ligand interactions may suggest starting points for the design and development of small molecule inhibitors of these protein interactions as novel anti-retroviral therapeutics.