Structural and Molecular Interactions of CCR5 Inhibitors with CCR5*

We have characterized the structural and molecular interactions of CC-chemokine receptor 5 (CCR5) with three CCR5 inhibitors active against R5 human immunodeficiency virus type 1 (HIV-1) including the potent in vitro and in vivo CCR5 inhibitor aplaviroc (AVC). The data obtained with saturation binding assays and structural analyses delineated the key interactions responsible for the binding of CCR5 inhibitors with CCR5 and illustrated that their binding site is located in a predominantly lipophilic pocket in the interface of extracellular loops and within the upper transmembrane (TM) domain of CCR5. Mutations in the CCR5 binding sites of AVC decreased gp120 binding to CCR5 and the susceptibility to HIV-1 infection, although mutations in TM4 and TM5 that also decreased gp120 binding and HIV-1 infectivity had less effects on the binding of CC-chemokines, suggesting that CCR5 inhibition targeting appropriate regions might render the inhibition highly HIV-1-specific while preserving the CC chemokine-CCR5 interactions. The present data delineating residue by residue interactions of CCR5 with CCR5 inhibitors should not only help design more potent and more HIV-1-specific CCR5 inhibitors, but also give new insights into the dynamics of CC-chemokine-CCR5 interactions and the mechanisms of CCR5 involvement in the process of cellular entry of HIV-1.

Highly active antiretroviral therapy has brought about a major impact on the acquired immunodeficiency syndrome (AIDS) epidemics in industrially advanced nations (1,2), however, eradication of HIV-1 2 appears to be currently impossible mainly because of the viral reservoirs remaining in blood and infected tissues (3). Successful antiviral drugs, in theory, exert their virus-specific effects by interacting with viral components such as viral genes or their transcripts without disturbing cel-lular metabolisms or functions (2). However, at present, no antiretroviral drugs or agents have been demonstrated to be completely specific for HIV-1 and devoid of toxicity or side effects in the therapy of AIDS (4). Limitations of antiviral therapy of AIDS are exacerbated by complicated regimens, emergence of drug-resistant HIV-1 variants (1), and a number of inherent adverse effects (5).
Thus, identification of new antiretroviral drugs that have unique mechanisms of action and produce no or least minimal side effects remains an important therapeutic objective (2,4). CCR5 is a member of the G protein-coupled, seven-transmembrane segment receptors, which comprise the largest superfamily of proteins in the body (6). In 1996, it was revealed that CCR5 serves as one of the two essential coreceptors for HIV-1 entry to human CD4 ϩ cells, thereby serving as an attractive target for possible intervention of HIV-1 infection (7-10). Aplaviroc (AVC; AK602/ONO4128/873140; Fig. 1), a novel spirodiketopiperazine derivative, represents a CCR5 inhibitor that specifically binds to human CCR5 with a high affinity, greatly blocks HIV-1-gp120/ CCR5 binding, and exerts potent activity against a wide spectrum of laboratory and primary R5-HIV-1 isolates including multidrug-resistant HIV-1 MDR (IC 50 values of 0.2-0.6 nM) (11). AVC, despite its much greater anti-HIV-1 activity than other previously published CCR5 inhibitors including TAK-779 and SCH-C ( Fig. 1), preserves RANTES and macrophage inflammatory protein-1␤ binding to CCR5 ϩ cells and their functions, whereas TAK-779 and SCH-C fully block the CC-chemokines/CCR5 interactions (11). AVC reportedly has an extensive and prolonged CCR5 occupancy as examined in phytohemagglutinin-activated peripheral blood mononuclear cells (t1 ⁄ 2 ϳ 9 h) (12) and in circulating lymphocytes in HIV-1-negative and HIV-1-positive individuals (t1 ⁄ 2 of 69 -152 h depending on different AVC doses) (37). In a randomized, placebo-controlled short-term monotherapy trial in patients with AIDS including those who were drug-experienced, AVC demonstrated potent antiretroviral activity and brought about significant reduction in HIV-1 viremia (by ϳ1.7 log) in AIDS patients (38).
In the present study, we examined the profile of binding to and interactions with CCR5 of three CCR5 inhibitors, AVC, SCH-C, and TAK-779. We also conducted structural analyses of the interactions of CCR5 inhibitors with CCR5, using homology modeling, robust structure refinement, and docking. Notably, the molecular modeling analyses were combined and fine-tuned with the results of the saturation binding assay using a panel of mutant CCR5-expressing cells and [ 3 H]CCR5 inhibitors and the resultant configurations and orientations of inhibitors docked within the hydrophobic cavity of CCR5 yielded structureactivity predictions and interpretations consistent with the observed experimental data. The present approach of combining the site-directed mutagenesis-based data and molecular modeling should represent a valuable strategy for gaining structural insights for membrane-bound proteins for which x-ray crystal structures are not as yet available. The present data delineating residue by residue interactions of CCR5 with CCR5 inhibitors should not only help design more potent and more HIV-1-specific CCR5 inhibitors, but also give new insights into the dynamics of CC-chemokine-CCR5 interactions and the mechanisms of CCR5 involvement in the process of cellular entry of HIV-1.
Generation of Wild-type and CCR5 MT -overexpressing Cells-A mammalian expression vector pZeoSV2 (Invitrogen) carrying the human wild-type CCR5 (CCR5 WT ) gene (pZeoSV-CCR5) (18) was a kind gift from Dr. Yosuke Maeda. A variety of plasmids carrying a mutant CCR5-encoding gene were generated using the site-directed mutagenesis technique employing the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA) as described by the manufacturer. Mutations introduced into the CCR5 gene were introduced through: (i) substitution of an amino acid(s) or (ii) deletion of an amino acid(s) at selected amino acid positions of CCR5. A mutation from Gly to Arg at position 163, where the corresponding amino acid in simian CCR5 is Arg, was also introduced. This G163R substitution has been reported to reduce the binding of R5-HIV-1-gp120 to human CCR5 and the susceptibility to HIV-1 (19). All these plasmids were confirmed to contain only the desired mutation(s) by nucleotide sequencing.
CHO cells (or U373-MAGI cells) were transfected with a plasmid containing the CCR5 WT -encoding gene or a plasmid carrying a CCR5 MT -encoding gene using Lipofectamine (Invitrogen); the transfectants were magnetically sorted, following treatment with an anti-CCR5 monoclonal antibody (2D7 or 3A9; BD Pharmingen, San Diego, CA), using Dynabeads M-450 coupled to goat anti-mouse IgG (Dynal A.S., Oslo, Norway); and the cells were cloned using the limiting dilution technique.
Determination of CCR5 Expression Levels in CHO and U373-MAGI Cells-CCR5 expression levels of the various clones described above were determined using three indicators: (i) maximal amounts of [ 3 H]AVC bound to cells (B max ); (ii) mean florescence intensity values when stained with monoclonal antibody (mAb) 3A9; and (iii) mean florescence intensity values when stained with mAb 2D7. The antigenic epitope for 2D7 is located distant from that for 3A9 (20) and 3A9 does not compete with AVC binding to CCR5 (data not shown). The expression levels were expressed as % control (CCR5 WT -expressing cells as control), and the highest value obtained was chosen as the estimated CCR5 expression level. There were no clones that had low values in all three indicators, thus sustaining that the current method used for the CCR5 expression levels was thought to be legitimate. CD4 and CCR5 expression levels were also determined using the quantitative fluorescence-activated cell sorting assay system (Quantum Simply Cellular Kit; Sigma) using 3A9 and 2D7 (12). For the HIV-1 susceptibility assay, U373-MAGI clones that expressed CD4 molecules ranging 10 -25 ϫ 10 4 antigen-binding sites were selected. CCR5 MT -expressing CHO cells were maintained in Ham's F-12 medium containing 10% FCS and 100 g/ml zeomycin. CCR5 WT -and CCR5 MT -expressing U373-MAGI cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% FCS, 200 g/ml G418, 100 g/ml hygromycin B, and 100 g/ml zeomycin.
Saturation Binding Assay Using 3 H-Labeled CCR5 Inhibitors-Saturation binding assay using 3 H-labeled CCR5 inhibitors was conducted and the K D values of CCR5 inhibitors in CCR5 WT -, or CCR5 MT -expressing CHO cells were calculated as previously described (11).
Radiolabeled Inhibitor Binding/Competition Studies-CCR5 WT -expressing cells (1.5 ϫ 10 5 ) were plated onto 48-well microculture plates, incubated for 24 h, rinsed, exposed to [ 3 H]AVC, [ 3 H]SCH-C, or [ 3 H]TAK-779 for 15 min at room temperature, subsequently exposed to various concentrations of unlabeled CCR5 inhibitors, incubated for 30 min, thoroughly washed, lysed, and the radioactivity in the lysates was counted. All experiments were performed in duplicate. The amounts of each 3 H-labeled CCR5 inhibitor bound to the cells are shown as mean % control values. Standard deviation values are indicated with the vertical lines. To obtain control values, the experiment was also performed without the addition of unlabeled CCR5 inhibitors.
Determination of CC-chemokine-and HIV-1 Gp120 Binding Affinity to CCR5 WT and CCR5 MT -Binding profiles of chemokines to CCR5 WTor CCR5 MT -expressing cells were determined using 125 I-labeled chemokines as previously reported (11) with minor modifications. In brief, CCR5 WT -or CCR5 MT -expressing cells (1.5 ϫ 10 5 ) were plated onto 48-well microculture plates, incubated for 24 h, rinsed, exposed to 5 nM 125 I-MIP-1␣, 125 I-MIP-1␤, or 125 I-RANTES at room temperature for 1 h, thoroughly washed with phosphate-buffered saline, lysed with 0.5 ml of 1 N NaOH, and the radioactivity in the lysates counted. The nonspecific binding of the labeled chemokine to the cells was determined based on the radioactivity detected in the wells plated with the same number of CCR5-negative CHO (CHO-K1) cells exposed to an equal amount of 125 I-labeled chemokine. Determination of the binding profiles of HIV-1-rgp120 to CCR5 WT or CCR5 MT was also conducted. Briefly, CCR5 ϩ CHO cells were exposed to rgp120 (5 g/ml) and sCD4 (5 g/ml: biotinylated using EZ-link sulfo-NHS-SS-biotin (Pierce)) for 1 h at 37°C. Cells were washed and binding of the rgp120-sCD4 complex to CCR5 ϩ CHO cells was determined using phycoerythrin-conjugated streptavidin (SA-PE: BD Pharmingen). Nonspecific binding was determined based on the mean florescence intensity of SA-PE with sCD4 but without rgp120. Because CCR5 expression levels vary among CCR5 clones, the % binding (occupancy) values for 125 I-chemokines and rgp120 were normalized using the following formula: % binding (occupancy) ϭ 100 ϫ (amount of 125 I-chemokine or rgp120 bound to CCR5 MT /amount of 125 I-chemokine or rgp120 bound to CCR5 WT ) ϫ (number of CCR5 WT /number of CCR5 MT ), where numbers of CCR5 WT and CCR5 MT are expressed as B max (cpm) or mean florescence intensity values as described above.
Determination of HIV-1 Susceptibility of CCR5 WT -and CCR5 MT -expressing Cells-The susceptibility of CCR5 WT -and CCR5 MT -expressing cells to the infection by an R5-HIV-1 strain, HIV-1 BaL , was determined as previously described (22). In brief, target cells (CCR5 WT -or CCR5 MT -expressing U373-MAGI cells; 10 4 /well) were plated onto 96-well flat microtiter culture plates, inoculated with 100 TCID 50 of HIV-1 BaL on the following day, cultured for 48 -72 h, stained with 400 g/ml of 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (X-gal), and all blue cells were counted. All experiments were performed in triplicate. For testing each CCR5 MT -expressing cell preparation, multiple (5 to 13) clones were examined. In each set of experiments, CCR5 WT -clone 1 was included and served as a standard. Percent infection in CCR5 WT -and CCR5 MT -expressing cells was determined using the following formula: % Infection ϭ 100 ϫ (mean blue cell number in a well)/(mean blue cell number in a well of CCR5 WT -clone 1).
Structural Modeling of the Interactions of CCR5 Inhibitors with Wildtype and Mutant CCR5 Species-An initial structural model of CCR5 was defined with homology modeling using the crystal structure of bovine rhodopsin as a template (23). This resulted in the initial placement of the helices and side chains. The CCR5-inhibitor complex structures were defined with an iterative optimization of CCR5 and inhibitor structures in the presence of each other, using software tools from Schrödinger (Schrödinger, LLC, New York), as described below. The conformational flexibility of both CCR5 and the inhibitors were taken into account. The molecular structures of AVC, SCH-C, and TAK-779, without the presence of counter-ions and solvents, were obtained by minimization using the MMFF94 force field (24). For each minimized inhibitor configuration, a set of low energy structures was generated by performing a Monte Carlo sampling of their conformations. Thus obtained structures were used as starting structures for docking calculations where their conformations were further refined.
The protonation states of CCR5 residues were assigned, and residues more than 20 Å from the active site were neutralized. In an attempt to place an inhibitor within CCR5, initially the active site was artificially enlarged by mutating Tyr 108 , Cys 178 , Glu 283 , and Met 287 to Ala. The van der Waals radii of inhibitor atoms were scaled by a factor of 0.70 to reduce steric clashes and docked into CCR5. After obtaining an initial set of CCR5-inhibitor complexes, residues 108, 178, 283, and 287 were mutated back from Ala to their original states. CCR5 atoms within 15 Å of an initially placed inhibitor were subsequently refined. It was achieved by using the rotamer library of Xiang and Honig (25) and optimizing each side chain one at a time holding all other side chains fixed. After convergence, all side chains were simultaneously energy minimized using the OPLS-AA force field (26) to remove any remaining clashes. The inhibitors were docked again and scored to estimate their relative affinity. The docked complexes with higher scores were visually examined along with the mutational data to select the best possible CCR5-inhibitor complex.
Mutated CCR5 structures were defined using the wild-type CCR5 structure and optimized using the OPLS2003 force field. Charges were taken from the force field. The minimization was carried out until the gradient was below 0.2 kJ/Å-mol. Resulting minimized structures were used as starting structures for obtaining docked complexes of mutated CCR5 with inhibitors using the protocol described above. Visualization, structural refinement, and docking were performed using Maestro 7.0, MacroModel 9.0, Prime 1.2, Glide 3.5, and IFD script from Schrödinger, LLC (New York). The extra precision mode of Glide, which has higher penalties for unfavorable and unphysical interactions, was used (27). Computations were carried out on a multiprocessor SGI Origin 3400 computer platform.
A probe radius of 1.4 Å with Connolly surfaces generated was used to define binding site cavities using the method of Exner et al. (28) as implemented in the MOLCAD tool in Sybyl 7.0 (Tripos, Inc., St. Louis, MO). Lipophilic potential was mapped onto the cavities using parameters from Viswanadhan et al. (29).

Site-directed Mutagenesis of CCR5 and Binding Affinity of CCR5
Inhibitors-We have previously reported (11) that AVC competitively blocked the binding of a monoclonal antibody, 45531, which is known to be specific against the C-terminal half (or domain B) of the second extracellular loop (ECL2B) of CCR5 (21), whereas AVC failed to block or only partially blocked the binding of two other monoclonal antibodies, 2D7 and 45523, specific for CCR5 but not for its ECL2B. When we examined three additional monoclonal antibodies, 3A9 and 45502 (both specific for NH 2 terminus) and 45549 (multidomain reactive) (20,21), none of these antibodies were replaced by AVC (data not shown). These data suggest that the potent inhibitory activity of AVC against R5 HIV-1 infection stems from its binding to ECL2B and/or its vicinity with high affinity.
In an attempt to delineate the CCR5 binding profile of the three CCR5 inhibitors, we generated a variety of CCR5 mutant-overexpressing (CCR5 MT ) CHO cells and determined the K D values of each inhibitor to mutant CCR5 species using the saturation binding assay with tritiated inhibitors. When we determined the K D value of AVC with respect to a CCR5 mutant carrying an Asp to Ala substitution at position 11 of the amino terminus domain (CCR5 D11A ), the value was 3.0 nM, virtually identical to the K D value with wild-type CCR5 (CCR5 WT ; Table 1), indicating that the D11A substitution did not affect the binding of AVC to CCR5. The K D value of AVC with respect to CCR5 Y37A was moderately greater with 7.9 nM (2.7-fold compared with the K D value with regard to CCR5 WT ). On the other hand, those of TAK-779 and SCH-C to CCR5 Y37A were 98.9 nM (3.3-fold compared with the K D with regard to CCR5 WT ) and Ͼ200 nM (Ͼ12.5-fold), respectively, in agreement with the previous reports in which both TAK-779 and SCH-C apparently failed to bind, probably explaining that these inhibitors failed to block HIV-1 infection of CCR5 Y37A -expressing cells (30,31). These data suggest that the binding of TAK-779 and SCH-C to CCR5 WT is more dependent on interactions with Tyr37 than that of AVC. We also generated a series of CCR5-overexpressing CHO cells carrying a mutation(s) at a selected amino acid position(s). As shown in Table 1, the mutations that substantially (more than 3-fold compared with AVC binding to CCR5 WT ) affected the K D values of AVC were as follows: Y108A and F113A in the third transmembrane domain (TM3) of CCR5; R168A and S180E in ECL2; K191R and K191N of the interface of ECL2B and TM5; K197A and I198A of TM5; Y251A of TM6; and M287E of TM7. The mutations that greatly diminished the binding of AVC (values of Ͼ200 nM) to CCR5 were as follows: G163R, C178A, WKNF190del, K191A, and E283A. It is worthwhile to interject that Lys 191 in ECL2 is reported to be critical for the binding of RANTES, MIP-1␣, and MIP-1␤ to CCR5 (32,33), whereas Cys 178 is presumed to form a disulfide bond with Cys 101 of ECL1 and to be critical for the conformation of CCR5 (34). Mutations that substantially affected the binding of TAK-779 and SCH-C to CCR5 were as follows: Y37A, Y108A, and E283A for TAK-779; and Y37A, WKNF190del, I198A, and E283A for SCH-C. It is noteworthy that the number of mutations that affected the binding of AVC was notably greater than those of TAK-779 and SCH-C. Thus, the CCR5 binding modes of AVC, TAK-779, and SCH-C apparently share some similar features but also have some distinct differences. Structural Analysis Locates Aplaviroc in the Interface of ECL and TM Domains-In the present study, a three-dimensional model of human CCR5-CCR5 inhibitor complex was defined by combining the results of site-directed mutagenesis-based analyses (Table 1) and molecular modeling that involved structure refinement and simultaneous docking of inhibitors to an initial structure of CCR5 based on the crystal structure of bovine rhodopsin (23). Fig. 2 illustrates the three-dimensional model of CCR5 that has a seven-transmembrane helical structure. Six hydrophobic cavities were identified in the extracellular, transmembrane, and intracellular domains of CCR5. Among them, a hydrophobic cavity to which CCR5 inhibitors highly likely to bind was identified based on its size and location (red arrowhead in Fig. 2). This cavity is the largest one among the six that can accommodate a molecule of the size of AVC and other CCR5 inhibitors and is in the region implicated to have the greatest effect on K D values with amino acid substitutions introduced in CCR5 among the six hydrophobic cavities identified. It should be noted, however, that the conformations of CCR5 without an inhibitor and CCR5 with the inhibitor can be substantially different from each other, because significant conformational changes would be expected to follow ligand binding to CCR5. Nevertheless, it is intriguing to note that this largest cavity corresponds to the binding site for the retinal ligand in bovine rhodopsin; given the propensity of this pocket to bind fairly large ligands such as the CCR5 inhibitors, it is possible that this binding cavity could accommodate some as-yet uncharacterized ligands produced in the body during the normal function of CCR5, potentially even some with regulatory roles.
Based on the set of the K D values of AVC in relation to various mutant CCR5 species overexpressed on CHO cells (Table 1), structural analysis of AVC-CCR5 interactions was conducted, which suggested that a series of intramolecular and intermolecular hydrogen bonds occur, which should stabilize the CCR5-AVC complex. We identified a significant network of hydrogen bonds among four amino acid residues: Gly 163 , Ser 180 , Lys 191 , and Thr 195 (Fig. 3A). Gly 163 is located in TM4, Ser 180 in ECL2, and Lys 191 and Thr 195 in TM5. Another network of hydrogen bonds was identified among another four amino acid residues: Tyr 37 , Glu 283 , Met 287 , and Tyr 108 (Fig. 3B). Tyr 37 is located in TM1, Tyr 108 in TM3, and Glu 283 and Met 287 in TM7. These hydrogen bond networks spanning multiple domains appear to maintain the optimal shape of the cavity for the binding of AVC. Indeed, further analysis of the cavity also revealed that the ECL regions have some hydrophilic characters (Fig. 4A, red arrowhead), whereas the rest of the cavity is mostly lipophilic (Fig. 4A). The carboxyl and hydroxymethyl of AVC interact with the hydrophilic regions of CCR5. The rest of AVC inter-  Table 1). Note that the hydrogen bond networks spanning multiple domains should maintain the optimal shape of the cavity for the binding of AVC (Fig. 4, A and B). Panel C, the predicted van der Waals contact between AVC and CCR5 residues Lys 191 and Ile 198 . AVC is shown in green spheres, whereas Lys 191 and Ile 198 in magenta.
acts with the lipophilic region of CCR5 (Fig. 4, A and B). It is of note that AVC has a molecular weight of 614.2, larger than TAK-779 (M r 531.1) and SCH-C (M r 557.5), has a substantial hydrophobic contact and fits well inside the large binding cavity within CCR5 as shown in Fig. 4A. AVC also forms hydrogen bonds with Cys 178 , Ser 180 , Lys 191 , and Thr 195 (Fig. 3, A and B). It is presumed that these interactions with residues in multiple ECLs and the interface of extracellular and transmembrane domains are responsible for the tight binding of AVC to CCR5.
Amino Acid Substitutions Are Likely to Cause Substantial Conformational Changes in ECLs-Substitutions of amino acid residues, which are involved in hydrogen bonding, appear to directly and indirectly disrupt the hydrogen bond networks observed (Fig. 3, A and B). For example, Gly 163 , located in TM4, does not interact directly with AVC, but is responsible for maintaining the shape of the binding cavity by its hydrogen bond interactions with Ser 180 and Lys 191 (Fig. 3A). Thus, the shape of the cavity is most likely altered with G163R substitution, and thereby AVC could lose critical interactions with ECL2 and TM5. In contrast, it was found that G163R exerts minimal effects on the binding of either SCH-C or TAK-779 to CCR5 (Table 1), potentially because these two inhibitors do not have direct hydrogen bond interactions with the ECL or ECL-TM interface.
As mentioned above, the carboxyl of AVC forms hydrogen bonds with Lys 191 and Ser 180 (Fig. 3A). The loss of binding with K191A substitution (Table 1) is likely because of the loss of the hydrogen bond with AVC as well as the altered shape of the cavity, which was confirmed by structural analysis of the CCR5 K191A -AVC complex. Neither SCH-C nor TAK-779 forms hydrogen bonds with Lys 191 (Fig. 5, A and B, and  Fig. 6A), hence it is thought that K191A substitution does not significantly affect the binding affinity of SCH-C or TAK-779 (Table 1). Ile 198 located in TM5 has hydrophobic interactions with AVC (Fig. 3C). SCH-C and TAK-779 were predicted to have hydrogen bond interactions with Tyr 37 located in TM1 (Fig. 5, A and B, and Fig. 6, A and B), whereas there appear to be no such interactions with AVC (Fig. 3B). This is consistent with the observation that Y37A substitution drastically changed the binding affinity of SCH-C and TAK-779 to CCR5 (Table 1). Ile 198 forms a hydrophobic contact with SCH-C (Fig. 5C), although not much with TAK-779, which well explains the reason I198A reduces the K D value of SCH-C with CCR5 but not that of TAK-   (Table 1) combined with structural analyses. CCR5 residue orientations (shown in tubes) vary from that in Fig. 4 because conformational flexibility of receptors during docking of each inhibitor was taken into account. Panel B, Tyr 37 forms hydrogen bonding interactions with SCH-C, and hence Y37A mutation reduces the binding affinity of SCH-C with CCR5. Note the location of Tyr 108 and compare its corresponding location for the complex with TAK-779. SCH-C is shown in tubes and CCR5 residues in wires. Only polar hydrogens are shown in panels A and B. The molecules are colored by atom types (carbon, gray; oxygen, red; nitrogen, blue; hydrogen, cyan; sulfur, yellow; and bromine, green). Panel C, van der Waals interactions of SCH-C (green) with Lys 191 and Ile 198 (magenta). There is tight hydrophobic binding of SCH-C with Ile 198 . Unlike AVC, Lys 191 does not form significant interactions with SCH-C. 779 with CCR5 (Table 1). It is also of note that AVC has direct interactions with residues in the extracellular domain and its proximity (Cys 178 , Ser 180 , and Lys 191 ) (Fig. 3, A and B), which should strongly affect the conformation of extracellular loops in comparison to SCH-C and TAK-779, both of which failed to directly interact with the amino acid residues in the extracellular domain. The interaction of HIV-1-gp120 with ECLs is thought to be critical for the establishment of infection, therefore, one can assume that the binding of AVC to CCR5 should result in a significant loss of interactions with gp120, which could explain the significantly greater effect of AVC in blocking viral infectivity, compared with SCH-C and TAK-779. Thus, the data above suggest that the binding pockets are in the same general area within CCR5, however, the binding interactions of the three inhibitors with CCR5 residues substantially differ from each other, in particular between AVC and other two inhibitors.
Interactions between CCR5 Inhibitors in Relation to Their Binding to CCR5-All three CCR5 inhibitors examined in this study possess the properties of allosteric antagonists of CCR5, whereas AVC exerts only partial inhibition of the binding to CCR5 and physiological function of 125 I-RANTES in comparison to TAK-779 and SCH-C (11). Moreover, although the binding pocket for these inhibitors are all located in the same hydrophobic cavity within CCR5, their binding profiles substantially differ from each other as discussed above. Therefore, we analyzed the interactions of the three inhibitors in relation to CCR5 WT . When CCR5 WT -CHO cells were exposed to [ 3 H]AVC (3 nM) for 15 min, followed by the addition of various concentrations (1 nM to 10 M) of unlabeled SCH-C, [ 3 H]AVC binding to CCR5 was reduced only moderately, by up to 32% (Fig. 7A). When the interaction between [ 3 H]AVC and unlabeled TAK-779 (0.625-62.5 M) was examined, [ 3 H]AVC binding to CCR5 was not significantly replaced by unlabeled TAK-779 (Fig.  7B). On the contrary, when [ 3 H]SCH-C was added first and then unlabeled AVC was added, the binding of [ 3 H]SCH-C to CCR5 WT -CHO cells was significantly blocked (Fig. 7C). These data suggest that AVC effectively replaces [ 3 H]SCH-C and binds to CCR5 WT . The binding of [ 3 H]TAK-779 was likewise blocked by the addition of unlabeled AVC, although the extent of replacement by AVC was lesser as compared with the case of [ 3 H]SCH-C (Fig. 7D).
Role of Amino Acid Residues of CCR5 with Which Aplaviroc Is Associated in HIV-1 Infection and CC-chemokine Binding-In an attempt to define the biological and virological roles of amino acid residues, with which AVC is associated in this binding to CCR5, we selected 15 mutant CCR5-overexpressing CHO cells and examined profiles of sCD4/gp120  (Table  1). Only polar hydrogens are shown. The molecules are colored by atom types (carbon, gray; oxygen, red; nitrogen, blue; hydrogen, cyan; sulfur, yellow). binding and CC-chemokine (RANTES, MIP-1␣, and MIP-1␤) binding profiles. Nine of the 15 mutants were chosen based on the following reasons. Tyr 108 was chosen because Tyr 108 is within the aromatic cluster seen in the proximity of TM2 and TM3 and is reported to play a crucial role in the CC-chemokine-elicited activation of CCR5 (35). Y37A and E283A were chosen because Tyr 37 and Glu 283 are involved in the binding of the three inhibitors to CCR5 and have been shown to be highly conserved in CC-chemokine receptors including CCR5, CCR2B, CCR3, CCR1, and CCR4 (6). Six more mutants (G163R, C178A, K191A, I198A, Y251A, and M287E) were also chosen, with which the binding affinity of AVC was significantly reduced compared with CCR5 WT -expressing cells (Ͼ5-fold K D difference) ( Table 1). As shown in Fig. 8A, all 9 mutations described above decreased the sCD4/gp120 binding to CCR5 compared with that to CCR5 WT . It was noted that Y108A, Y251A, and E283A substitutions resulted in the greatest reduction of sCD4/gp120 binding.
We also employed six additional mutant CCR5-overexpressing CHO cells, which showed insignificant to minimal changes (less than 3-fold differences) in their K D values compared with the K D value for wild-type CCR5 (Table 1). Two substitutions in the extracellular domain of CCR5, D11A (NH 2 terminus) and KE171AA (ECL2), induced substantial reduction in HIV-gp120/sCD4 complex binding by 65 and 70%, respectively. The F112L, F112Y (both in TM3), S180T (ECL2), and M287A (TM7) substitutions caused no significant changes in K D values (Table  1) or HIV-gp120/sCD4 complex binding (Fig. 8A). These data suggest that amino acid residues in the proximity of the hydrophobic cavity for AVC that do not change K D values generally do not affect the HIV-gp120/sCD4 complex binding to CCR5. These 6 substitutions, however, caused CC-chemokine binding inhibition at various degrees, suggesting that the profile of the HIV-gp120/sCD4 complex binding to CCR5 differs from that of CC-chemokine binding to CCR5.
We finally generated seven clonal populations of CD4 ϩ U373-MAGI cells expressing CCR5 with an amino acid substitution that caused significant reduction in the K D values of either of the three CCR5 inhibitors and examined their susceptibility to HIV-1 BaL infection (Fig. 8C). HIV-1 BaL effectively infected CCR5 WT -expressing cells, however, the susceptibility to HIV-1 BaL infection was greatly reduced in CCR5 Y37A -, CCR5 Y108A -, and CCR5 I198A -expressing cells and that in CCR5 G163R -, CCR5 C178A -, CCR5 K191A -, and CCR5 E283A -expressing cells was more greatly limited. These data indicate that the mutations examined in this experiment effectively blocked HIV-1 BaL infection, although such mutations, in particular three mutations (G163R, K191A, and I198A), allowed some degree of sCD4/gp120 binding to CCR5 ϩ cells (Fig. 8A).
As noted above, the CC-chemokine binding profile notably differed from the profile of HIV-1 BaL infection susceptibility in CCR5 G163R -and CCR5 I198A -expressing cells. Moreover, moderate levels of chemokine binding were seen in CCR5 Y108A -and CCR5 E283A -expressing cells, whereas the least infection occurred in these cells. It should be of note that when we examined what ensued upon the binding of RANTES to CCR5 G163R -expressing cells, it was found that the level of Ca 2ϩ flux that occurred in those cells was comparable with that in CCR5 WT -expressing cells following stimulation with RANTES. 3 These data, taken together, suggest that the binding affinity of AVC seen with CCR5 variants generally paralleled the HIV-gp120/sCD4 complex binding affinity to mutant CCR5s, although it is of note that further fine-tuned analysis with greater numbers of CCR5 mutants is required for full understanding of the interactions among the HIV-gp120/sCD4 complex, CC-chemokines, and CCR5 inhibitors. The data also suggest that certain conformational changes caused by amino acid substitutions (e.g. at Tyr 108 , Gly 163 , or Ile 198 residues of CCR5) might substantially reduce HIV-1 infection without significantly affecting physiological CC-chemokine-CCR5 interactions.

DISCUSSION
In the present study, we examined the structural and molecular interactions between CCR5 and three CCR5 inhibitors, AVC, SCH-C and TAK-779. When we exploited site-directed mutagenesis by generating a panel of mutant CCR5-expressing cells, and determined the K D values of each CCR5 inhibitor with each mutant CCR5 species, the K D values compiled corroborated our previous results obtained using the mAb replacement assay (11). We found that the binding affinity of AVC to wild-type CCR5 (CCR5 WT ) was the greatest with a K D value of 2.9 nM as compared with SCH-C and TAK-779 with K D values of 16.0 and 30.2 nM, respectively (Table 1). It was also noted that the number of mutations that affected the binding of AVC was greater than in the cases of TAK-779 and SCH-C (Table 1). Moreover, the amino acid substitutions with which AVC turned out to have different K D values were mostly located in the second extracellular loop (ECL2) and its interface with TM4 (Gly 163 ), or TM5 (Lys 191 ). By contrast, with regard to SCH-C and TAK-779, no substantially different K D values were obtained with amino acid substitutions in ECL2 ( Table 1). The data suggest that not only the binding interactions but also the molecular size and/or bulkiness of AVC are related to the different numbers of K D -affecting mutations and different profile of K D values obtained and that AVC forms substantial hydrophobic contacts with and fits well inside the hydrophobic cavity within CCR5 (Fig. 4A). It was also thought that such tight interactions of AVC with residues in ECLs and the interface of extracellular and transmembrane domains of CCR5 are responsible for the greater binding affinity of AVC to CCR5 as compared with two other inhibitors.
It was intriguing that certain amino acids such as Gly 163 and Lys 191 formed critical hydrogen bond networks in the interaction of AVC and CCR5 (Fig. 3, A and B). It was thought, therefore, that the amino acid substitutions in such positions dramatically altered the binding affinity of AVC to CCR5. It is also noteworthy that Cys 178 of ECL2 is presumed to form a putative disulfide bridge with Cys 101 of ECL1 and to be critical for the conformation of CCR5 (34). In the present study, C178A substitution, which should disrupt the putative disulfide bridge and cause significant conformational changes in both ECL1 and ECL2, nullified the CCR5 binding affinity of AVC, but not of SCH-C or TAK-779 (Table  1). In this respect, the lesser interactions of SCH-C and TAK-779 with extracellular domains (Table 1) are likely responsible for the lack of influence of the C178A substitution on the binding affinity of SCH-C and TAK-779, corroborating the notion that mutations in ECL2 did not alter the binding of SCH-C and TAK-779 to CCR5 (Table 1).
Two events are likely involved in the reduced binding of inhibitors to CCR5 with an amino acid substitution(s): (i) the alteration of direct interaction, if present, between the amino acid residue and the inhibitor and/or (ii) an allosteric effect(s) in which there is a significant conformational change(s) of CCR5 either near or distant from the mutated residue. In this respect, in the present work, we have shown that multiple hydrogen bond networks are important for the binding of AVC to CCR5 (Fig. 3, A and B). Some residues such as Lys 191 directly interact with AVC, and hence mutations that lose this direct interaction are very likely responsible for the loss of binding. The lack of drastic changes in the binding of SCH-C and TAK-779, which are not predicted to directly interact with Lys 191 , also supports this inference. Gly 163 , however, does not directly bind to AVC but is a critical part of the Gly 163 -Ser 180 -Lys 191 -Thr 195 hydrogen network (Fig. 3A). Our molecular modeling, however, indicates that arginine at position 163 does not form hydrogen bonds with Ser 180 or Lys 191 . This alteration in the hydrogen bond network probably changes the shape of the cavity in the ECL2-TM4-TM5 region (an allosteric effect near the mutated residue) and results in reduced binding of AVC to CCR5 G163R . Mutation at residue 283 might be an example of the one causing an allosteric conformational change(s) distant from the mutated residue. Only AVC is predicted to directly interact with Glu 283 , but a E283A mutation results in significant loss of binding for all three inhibitors examined. Thus, it is thought that certain amino acid substitutions reduce the binding of an inhibitor even though they are not directly interacting with the inhibitor. It is also of note that mutations at certain residues directly interacting with an inhibitor do not always cause significant loss of binding if CCR5 does not undergo a conformational change(s) or takes an alternate conformation that is not unfavorable for inhibitor binding.
Because a crystal structure of CCR5 is not available, unlike other targets for intervention of HIV infection including HIV reverse transcriptase (36), in the present study, in an attempt to conduct structural analyses of the interactions of CCR5 inhibitors with various mutant CCR5 species, an initial structural model of CCR5 was generated using homology modeling based on the crystal structure of bovine rhodopsin (23). The present approach of combining the site-directed mutagenesisbased data (Table 1) and molecular modeling should be a valuable strategy for gaining structural insights for membrane-bound proteins for which x-ray crystal structures are not as yet available. The CCR5-inhibitor complex structures were defined by iterative optimization of CCR5 and inhibitor structures. Mutated CCR5 structures were also defined and optimized based on the wild-type CCR5 structure, and minimized structures of various CCR5 mutants were used as starting structures for obtaining docked complexes of mutated CCR5 with inhibitors. It is of note that in the present study, our modeling study was combined and fine-tuned with the results of the saturation binding assay using a panel of mutant CCR5-expressing cells and tritiated CCR5 inhibitors and the resultant configuration and orientation of inhibitors docked within the hydrophobic cavity of CCR5 yielded a consistent analysis of the structure-activity data. The conformations of CCR5 without an inhibitor and CCR5 bound to an inhibitor are likely to be substantially different from each other because conformational changes are expected to occur upon inhibitor binding to CCR5. Thus, the conformational flexibility of both inhibitors and CCR species were taken into account in our analysis. In this regard, our ongoing analyses of binding affinity profiles using different 3 H-labeled CCR5 inhibitors and an expanded panel of mutant CCR5-expressing cell lines should further illuminate the intramolecular and intermolecular interactions of CCR5 and CCR5 inhibitors.
We also examined the binding profile of CCR5 inhibitors when one inhibitor was added to CCR5-overexpressing cells, followed by the addition of the second inhibitor (Fig. 7). Interestingly, when AVC first bound to CCR5 and SCH-C or TAK-779 was subsequently added, both SCH-C and TAK-779 only partially displaced AVC (Fig. 7, A and B). However, when SCH-C or TAK-779 first bound to CCR5 and AVC was subsequently added, SCH-C and TAK-779 were substantially displaced by AVC (Fig. 7, C and D). This ineffective displacement of AVC once bound to CCR5 WT by SCH-C and TAK-779 could be explained by the difference of their K D values (2.9 nM for AVC, 16 nM for SCH-C, and 30.2 nM for TAK-779 as illustrated in Table 1). However, it is of note that both SCH-C and TAK-779 failed to displace AVC even with very high concentrations: 3,333-fold (10 M; Fig. 7A) and 20,833-fold (62.5 M; Fig. 7B), respectively, greater than the [ 3 H]AVC concentration (3 nM). Thus, only the difference in K D values among the three inhibitors unlikely fully explains the failure of AVC replacement by SCH-C and TAK-779. In this regard, a possible explanation is that AVC induces a series of significant conformational changes in CCR5 and, upon the completion of its stable lodging within the hydrophobic cavity within CCR5 WT , the binding and/or entry into CCR5 of the second inhibitor is hindered and the interaction between AVC and the second inhibitor added is no longer merely competitive. Although the data of displacement experiments using two CCR5 inhibitors discussed above may not offer a clear mechanistic explanation at present due to the limitation of the currently available methodologies, the data may have a clinical relevance in the future because more than one CCR5 inhibitor may be simultaneously administered when multiple CCR5 inhibitors are introduced in the therapy of AIDS. It is also noteworthy that the tight CCR5 binding profile of AVC is presumably related to the extensive and prolonged CCR5 occupancy observed in phytohemagglutinin-activated peripheral blood mononuclear cells (t1 ⁄ 2 ϳ 9 h) (12) and in circulating lymphocytes in HIV-1-negative and HIV-1-positive individuals (t1 ⁄ 2 of 69 -152 h depending on different AVC doses), a potentially favorable feature that could enable once daily (QD) or twice daily (BID) administration of AVC.
In the present study, we observed that amino acid mutations within the transmembrane domains such as Y108A, G163R, and I198A exerted only minimal or moderate effects on the binding of CC-chemokines to CCR5 (Fig. 8B), whereas they caused a substantial reduction in HIV-1-gp120/CD4 complex binding to CCR5 (Fig. 8A) and a more drastic reduction in the susceptibility to HIV-1 infection (Fig. 8C). It is possible that the molecular size of the gp120/CD4 complex is far greater and its interactions with CCR5 are more extensive compared with the case of CC-chemokines, and therefore, the process for establishing HIV-1 infection is more extensively affected by such amino acid substitutions in comparison to CCR5 binding of CC-chemokines and the ensuing signal transduction. In this regard, a large body of literature has shown that CCR5 regions to which the envelope glycoproteins of HIV-1 bind are functionally and structurally quite different from those to which CC-chemokines bind (19 -21, 32, 33). Wu et al. (20) and Lee et al. (21) demonstrated that a group of NH 2 terminus-specific CCR5 monoclonal antibodies did not block the binding of or Ca 2ϩ flux induction by CCchemokines, although ECL2-specific monoclonal antibodies effectively block the binding of CC-chemokines and their Ca 2ϩ flux induction. Navenot et al. (33) subsequently confirmed these notions and further demonstrated that a CCR5 chimera with the NH 2 terminus of CXCR2 bound MIP-1␣ with an affinity similar to that of CCR5 WT . In contrast, both NH 2 terminus-and ECL2-specific mAbs reportedly blocked efficiently the binding of gp120 to CCR5 although the latter mAbs caused superior inhibition (20,21). These previously published data strongly suggest that the interactions of CCR5 with HIV-1-glycoproteins involve multiple CCR5 domains and consist of more complex processes, whereas those with CC-chemokines involve fewer CCR5 domains and potentially fewer processes. Thus, this implies that the intervention of HIV-1 infection without interrupting physiological CC-chemokine/ CCR5 interactions should be feasible. Indeed, AVC does not potently block the physiologic CC-chemokine/CCR5 interactions, although it highly efficiently suppresses the infection of HIV-1 with IC 50 values of subnanomolar concentrations (11).
Taken together, mutations associated with AVC binding to CCR5 decreased gp120 binding to CCR5 and the susceptibility to HIV-1 infection, whereas mutations in TM4 and TM5 that also decreased gp120 binding and HIV-1 infectivity had less effects on the binding of CCchemokines, suggesting that CCR5 inhibition targeting appropriate regions might render the inhibition highly HIV-1-specific, preserving the CC chemokine-CCR5 interactions. The present data should not only help design more potent and HIV-1-specific CCR5 inhibitors, but also should give new insights into the dynamics of CC-chemokine-CCR5 interactions and the mechanisms of CCR5 involvement in the process of cellular entry of HIV-1.