Identification of Residues of CXCR4 Critical for Human Immunodeficiency Virus Coreceptor and Chemokine Receptor Activities*

CXCR4 is a G-coupled receptor for the stromal cell-derived factor (SDF-1) chemokine, and a CD4-associated human immunodeficiency virus type 1 (HIV-1) coreceptor. These functions were studied in a panel of CXCR4 mutants bearing deletions in the NH2-terminal extracellular domain (NT) or substitutions in the NT, the extracellular loops (ECL), or the transmembrane domains (TMs). The coreceptor activity of CXCR4 was markedly impaired by mutations of two Tyr residues in NT (Y7A/Y12A) or at a single Asp residue in ECL2 (D193A), ECL3 (D262A), or TMII (D97N). These acidic residues could engage electrostatical interactions with basic residues of the HIV-1 envelope protein gp120, known to contribute to the selectivity for CXCR4. The ability of CXCR4 mutants to bind SDF-1 and mediate cell signal was consistent with the two-site model of chemokine-receptor interaction. Site I involved in SDF-1 binding but not signaling was located in NT with particular importance of Glu14 and/or Glu15 and Tyr21. Residues required for both SDF-1 binding and signaling, and thus probably part of site II, were identified in ECL2 (Asp187), TMII (Asp97), and TMVII (Glu288). The first residues (2-9) of NT also seem required for SDF-1 binding and signaling. A deletion in the third intracellular loop abolished signaling, probably by disrupting the coupling with G proteins. The identification of CXCR4 residues involved in the interaction with both SDF-1 and HIV-1 may account for the signaling activity of gp120 and has implications for the development of antiviral compounds.

monocytes, and their progenitor cells (5,6), consistent with a role in lymphocyte homing and hematopoiesis (7)(8)(9). It has a single known receptor designated CXCR4 (10,11). Gene inactivation experiments in mice also suggest that CXCR4 and SDF-1 are involved in the embryonic development of several organs, and in particular the brain, heart, and blood vessels (8,9,12). In contrast to most other chemokines and their cognate receptors, SDF-1 and CXCR4 are constitutively expressed in many tissues (13)(14)(15)(16), which is in agreement with their broad spectrum of biological functions.
Besides its SDF-1 receptor activity, CXCR4 has been identified, along with the chemokine receptor CCR5, as a cell entry portal for the human immunodeficiency virus (HIV-1) (reviewed in Refs. [17][18][19]. Interaction of the HIV-1 surface envelope glycoprotein (gp120) with CXCR4 or CCR5 is necessary to trigger the molecular events that eventually result in virus-cell fusion and infection. In most cases, the contact of gp120 with another membrane component, CD4, is required for a functional interaction with CXCR4 or CCR5, which are therefore viewed to be CD4-associated HIV-1 coreceptors. The cell tropism of HIV-1 strains is in large part governed by the selectivity of gp120 for CXCR4 or CCR5, itself dependent upon the sequence of variable domains, and in particular the V3 loop (20 -24). Viral strains using CXCR4 are less frequently isolated than strains using CCR5, until advanced stages of infection (25,26). Their emergence could have a role in the onset of immune deficiency, possibly by allowing HIV-1 replication in more cell types (16).
Low concentrations of SDF-1 can efficiently block HIV-1 infection mediated by CXCR4 (10,11). This antiviral effect of SDF-1 is due not only to steric hindrance but also to the down-regulation of CXCR4 at the cell surface by induction of endocytosis (27). Antiviral strategies based on blocking the interaction of HIV-1 with coreceptors are now envisioned (18,28). To this end, information about the interaction of CXCR4 and CCR5 with their chemokine ligands, and with gp120, is clearly needed.
The structural elements of CXCR4 that mediate the interaction with gp120 or with the chemokine ligands have not been precisely defined. Studies with mutant or chimeric receptors have assigned the HIV-1 coreceptor activity to the extracellular domains, such as the amino-terminal domain (NT) and the second extracellular loop (ECL2) (29 -34). Although interaction with gp120 can activate CXCR4 (or CCR5) under certain conditions (35)(36)(37), mutations in the intracellular domains that uncouple these receptors with G proteins have no apparent effect on HIV-1 entry (33). It appears therefore that chemokine receptors are used by HIV-1 as docking sites at the cell surface, and not for their ability to transduce a cell signal. Experiments with chimeric receptors also showed that the amino-terminal domain of CXCR4 was sufficient for efficient binding of SDF-1, while signaling involved other extracellular domains, and in particular ECL2 (31,33). These results are compatible with a two-site model for the interaction of SDF-1 and CXCR4 (38), initially developed for the C5a chemoattractant and its receptor (39). According to this model, a discrete region in the aminoterminal domain of the receptor (site I) binds the chemokine with selectivity and high affinity, which then favors its interaction with a pocket formed by the extracellular loops and the membrane-spanning domains (site II). Activation of the receptor requires contact of the chemokine with key residues of site II.
Here we have tested a panel of CXCR4 mutants, comprising deletions in the amino-terminal domain and substitutions of amino acids in the extracellular and transmembrane domains, for their ability to mediate infection of CD4 ϩ cells by two genetically divergent HIV-1 strains, and for their capacity to bind SDF-1 and subsequently mediate a cell signal. On the basis of these experiments, we were able to identify residues of CXCR4 that are involved in these different activities.
CXCR4 Plasmids and Transfections-The wild-type (WT) and mutant CXCR4 cDNAs were subcloned downstream to the cytomegalovirus immediate-early promoter in the expression vectors pCDNA3 (CLON-TECH) or pRc/CMV (Invitrogen). Site-directed mutagenesis was performed on single-stranded templates and checked by sequencing. The position of amino acid substitutions and deletions is depicted in Fig. 1. Mutants ⌬2-9, ⌬4 -36, and N11Q (29), mutants in the ECL2 (30) and in the TMs (44) have been described. The ECL2 mutants (except D182G) bear an NH 2 -terminal Myc tag (45), which does not affect CXCR4 function. 2 The COOH-terminal intracellular domain of CXCR4 (residues 310 -354) was deleted in mutants ⌬2-9, ⌬4 -36, and N11Q. This deletion has no influence on HIV-1 infection, or on SDF-1 binding and signaling, but abolishes ligand-induced endocytosis (27). Standard calcium phosphate transfection was used to express the WT and mutant receptors in COS or U373MG-CD4 cells, and transfection with Superfect TM (Quiagen) in HEK293T cells.
Detection of CXCR4 Expression by Flow Cytometry-Detection of CXCR4 expression by immunofluorescence with the 12G5 (46) or 6H8 (47) monoclonal antibodies (mAbs) and flow cytometry was performed as described (30). Briefly, COS cells were co-transfected with WT or mutant CXCR4 vectors, and with EGFP-N1 (CLONTECH), a green fluorescent protein expression vector, in a 6:1 ratio. Cells were detached with phosphate-buffered saline, 1 mM EDTA 36 h after transfection. Approximately 2 ϫ 10 5 transfected cells were incubated for 1 h at 4°C in 100 l of cell-free supernatant from the 12G5 hybridoma (a gift from J. Hoxie, University of Pennsylvania, Philadelphia, PA), or 100 l of phosphate-buffered saline, 2% fetal calf serum with affinity-purified 6H8 mAb (10 g/ml, a gift from A. Amara, Institut Pasteur, Paris), and with SDF-1 in competition experiments. After secondary staining with phycoerythrin-conjugated goat anti-mouse serum (Dako), cells were analyzed for green and red fluorescence on an Epics Elite flow cytometer (Coultronics). Expression of CXCR4 (red fluorescence) was assessed in green fluorescent protein-positive cells. Percentages of inhibition of 12G5 or 6H8 binding by SDF-1 were calculated from mean fluorescence intensities in the presence or absence of SDF-1.
SDF-1 Binding Assay-HEK293T cells were transfected with CXCR4 vectors in 6-well trays (ϳ10 5 cells per well) using the Superfect TM reagents. After 36 h, cells were either stained for CXCR4 expression with the 12G5 mAb and analyzed by flow cytometry, as described previously, or tested for radiolabeled SDF-1 binding. In that case, cells were incubated in 700 l/well of serum-free Dulbecco's modified Eagle's medium with 1.2 unit/ml heparinase III (Sigma) for 1 h, then washed and incubated for 90 min at room temperature in 700 l of Dulbecco's modified Eagle's medium, 3% bovine serum albumin, 50 mM HEPES with 0.2 nM 125 I-SDF-1␣ (275 Ci/g). Cells were washed three times in 500 mM NaCl, 50 mM HEPES, pH 7.4, 5 mM CaCl 2 , 1 mM MgCl 2 , then lysed in phosphate-buffered saline with 2% Nonidet P-40, and total ␥-radioactivity was counted. In each experiment, counts obtained in cells transfected with pRc/CMV were subtracted as nonspecific SDF-1 binding. Results were then adjusted for the relative cell surface expression of CXCR4.
Intracellular Calcium Measurements-COS cells were transfected with CXCR4 vectors and seeded 24 h later in 35-mm glass plates. After another 24 h, cells were loaded with the fluorescent dye fura-2 (Molecular Probes, 3 M) in saline buffer (140 mM NaCl, 5 mM KCl, 10 mM HEPES, pH 7.4, 1 mM CaCl 2 1 mM MgCl 2 ) for 30 min at room temperature. The same buffer was used to wash cells before addition of 100 nM SDF-1. Single-cell Ca 2ϩ measurements were made at 37°C with a Nikon Diaphot 300 microscope and an IMSTAR imaging system as described (49).

Surface Expression of CXCR4
Mutants-Our mutagenesis of CXCR4 sought to address the functional role: (i) of the NH 2terminal extracellular domain (NT), by deletions of variable length (⌬2-9, ⌬10 -20, ⌬4 -20, and ⌬4 -36) and by replacing one or several residues, usually with alanine; (ii) of all the charged amino acids in ECL and TM, by individual substitutions with neutral residues; and (iii) of the third intracellular loop (i3) by three overlapping deletions. The position of these different mutations and deletions is depicted in Fig. 1. Besides charged residues, mutations in NT addressed the role of Tyr 7 , Tyr 12 , and Tyr 21 , which are potential sites of sulfation (50), and of the potential N-glycosylation site Asn 11 , mutated alone or with Asn 176 in ECL2.
The cell surface expression of wild-type (WT) and mutant CXCR4 was monitored in simian COS cells by flow cytometry. Transfected cells were stained with the 12G5 mAb, which recognizes a conformational epitope in ECL2 (29,30) or with the 6H8 mAb, which recognizes residues 22-25 in NT (data not shown). Similar results were obtained with the two mAbs, except when mutations disrupted the 12G5 or 6H8 epitope ( Fig.  2A and other data not shown). Staining with 12G5 was markedly reduced for cells transfected with the ⌬4 -36 deletion mutant (25% of WT), while the ⌬10 -20 and ⌬4 -20 deletions had a lesser effect (ϳ50% of WT). These deletions probably affect the transport and/or stability of CXCR4 at the cell surface. For all other CXCR4 mutants, the extent of staining was at least 75% of that of WT, suggesting that the overall conformation of the receptor was not significantly altered.
Mutations in CXCR4 Impairing HIV-1 Entry-The ability of CXCR4 mutants to mediate HIV-1 entry in the presence of CD4 (coreceptor activity) was tested in U373MG-CD4 cells. These human astroglioma cells must be transfected with a functional coreceptor in order to be permissive to HIV-1 entry (23, 51). Fig. 2, B and C, show the results of infections performed in parallel with two genetically distant HIV-1 strains, LAI (clade B) and NDK (clade D).
A markedly reduced efficiency of infection by the two HIV-1 strains (Ͻ25% of WT CXCR4) was observed upon mutation of two tyrosine residues in the NT domain (Y7A/Y12A). The Y21A mutation also reduced infection by both strains tested (Ͻ50% of WT in both cases). All other CXCR4 mutants tested (except ⌬4 -36) were able to mediate infection by at least one HIV-1 strain with relatively high efficiency (Ͼ50% of that of WT), again suggesting that the mutations did not disrupt the overall conformation of the receptor. As we already reported (29), the ⌬4 -36 mutant cannot mediate NDK infection, despite being a relatively efficient coreceptor for LAI, given its reduced cell surface expression. Mutations D171N in TMIV, D193A in ECL2, and E268A in ECL3 markedly reduced NDK infection, but had lesser effects for LAI. By contrast, the mutations D97N in TMII, D262A in ECL3, and E288Q in TMVII mainly affected infection by NDK. Several other mutations reduced the efficiency of infection by one or the other HIV-1 strain, but to a lesser extent. Mutation of the potential N-glycosylation sites Asn 11 and Asn 176 had no apparent effect on HIV-1 entry. Overall, our results confirm the importance of residues in NT and ECL2 for the HIV-1 coreceptor activity, but also point to the role of negatively charged residues in the predicted membranespanning domains.
Competition of SDF-1 with Anti-CXCR4 mAbs-The SDF-1 chemokine can prevent binding of the 12G5 or 6H8 mAbs to CXCR4, and can inhibit HIV-1 infection mediated by the CXCR4 coreceptor (27,47). These properties were used to address indirectly the ability of SDF-1 to bind to CXCR4 mutants. COS cells transfected with WT or mutant CXCR4 were stained at 4°C with the 12G5 or 6H8 mAb, in the presence of 200 nM SDF-1. For WT CXCR4, this concentration resulted in 70 -80% inhibition of 12G5 or 6H8 binding (data not shown). Table I shows the inhibition of 12G5 or 6H8 binding to cells expressing CXCR4 mutants, relative to cells expressing WT CXCR4. Overall, similar results were obtained in competition assays with the 12G5 and 6H8 mAbs. Almost no inhibition of 12G5 binding (11% of WT), and a limited inhibition of 6H8 binding (35% of WT), was observed for cells expressing the ⌬10 -20 mutant. SDF-1 also competed rather inefficiently with the 12G5 mAb when cells expressed the ⌬4 -20 or the ⌬4 -36 mutant (20 and 36% of WT, respectively), but not the ⌬2-9 mutant (Ͼ70% of WT). Efficient binding of SDF-1 therefore seems to require residues beyond position 9 in NT. In this domain, mutations at Glu 14 and Glu 15 , or at Tyr 21 markedly reduced sensitivty to SDF-1 in assays with the 12G5 mAb (ϳ30% of WT), while their effect was lesser with the 6H8 mAb (50 -60% of WT). A reduced effect of SDF-1 (Ͻ50% of WT in assays with 12G5) was also observed for the D97N (TMII) and E288Q (TMVII) mutants. Other mutations in NT (for example, D22A) or in ECL1 (K110A) had a lesser effect. These experiments suggest that residues in NT are critically required for efficient interaction between CXCR4 and SDF-1. This interaction could also be impaired by two out of four mutations in membrane-spanning domains, while few of the mutations in extracellular loops had detectable effect.
Inhibition of HIV-1 Infection by SDF-1-Infection of U373MG-CD4 cells expressing WT or mutant CXCR4 with HIV-1 (NDK strain) was performed in the presence or absence of 200 nM SDF-1. This concentration inhibits infection mediated by the WT CXCR4 by 70 -80% (Fig. 3). The inhibitory effect of SDF-1 on infection mediated by mutant CXCR4, relative to WT CXCR4 is shown in Table I. All of the NT deletions tested (⌬4 -20, ⌬10 -20, and ⌬2-9) reduced sensitivity to the antiviral effect of SDF-1 (23 to 40%). The ⌬4 -36 mutant is not a functional NDK receptor and could not be tested. In general, HIV-1 infection mediated by CXCR4 mutants with amino acid substitutions in the extracellular domains was efficiently blocked by SDF-1 (Ͼ50% of WT), suggesting that ligand binding was relatively conserved. This was in particular the case for In particular, residue Asp 97 was placed in TMII, and not in the first extracellular loop, as proposed elsewhere (34). The disulfide bridge between ECL1 and ECL2 is shown, as well as the potential N-glycosylation sites (branching symbols).
mutations at residues Glu 14 -Glu 15 and Tyr 21 , previously found to be important for interaction with SDF-1 in the mAb competition assay. By contrast, markedly lower antiviral effect of SDF-1, suggesting inefficient binding to CXCR4, was observed for the D187A mutant (40% of WT), and for the D97N and E288Q mutants (Ͻ10% of WT in both cases). These results were confirmed in dose-response experiments (Fig. 3). The SDF-1 concentration yielding 50% inhibition of infection (EC 50 ) was ϳ50 nM for WT CXCR4, and for the E14A/E15A mutant. The EC 50 of SDF-1 was higher when cells expressed the Y21A (ϳ200 nM), and moreover the D187A, E288Q, D97N, and NT deletion mutants (Ͼ500 nM).
These experiments confirmed the importance of the acidic residues Asp 187 in ECL2, Asp 97 in TMII, and Glu 288 in TMVII for efficient interaction with SDF-1. Mutations at these residues, or the ⌬2-9 deletion, resulted in resistance to the antiviral effect of SDF-1, but they had a lesser effect on the competition of SDF-1 with mAbs. The opposite was observed upon substitutions in NT, such as E14A/E15A and Y21A, while deletions ⌬4 -20 or ⌬10 -20 had an important effect in both assays. Among other experimental differences, competitions with mAbs were performed at 4°C (1 h), while infections were per-formed at 37°C (24 h). Only in the latter case could SDF-1 induce CXCR4 endocytosis. However, this phenomenon did not seem to play a significant role in our experiments, since SDF-1 completely blocked infection mediated by the N11Q mutant, which has a complete deletion of the COOH-terminal domain preventing receptor endocytosis upon ligand binding (27). A more likely explanation for discrepancies between the two types of assays is that competition with mAbs requires a high affinity interaction of SDF-1 with CXCR4, and is therefore highly dependent upon the integrity of site I in NT, while competition with HIV-1 requires interaction of SDF-1 with other residues of CXCR4, probably corresponding to site II.
Human HEK293T cells transfected with CXCR4 were left in contact for 90 min with trace amounts of 125 I-labeled SDF-1 (0.2 nM), and cell associated radioactivity was measured. Results were adjusted to account for the relative surface expression of CXCR4 mutants, and for the nonspecific binding. The E179A, D181A, and D182G mutations had no apparent effect on radiolabeled SDF-1 binding (Fig. 4). The other mutant CXCR4 tested bound SDF-1 less efficiently than WT CXCR4, with relative efficiency ranging from 20 to 30% for the NT deletion mutants, or the Y21A, D97N, and E288Q mutants, to 50 -60% for the D187A mutant. Thus, mutations that impaired SDF-1 binding in indirect competition assays also reduced SDF-1 binding in a direct assay.
Effects of CXCR4 Mutations on Activation by SDF-1-The activation of WT or mutant CXCR4 expressed in COS cells was detected by a rise in intracellular Ca 2ϩ concentration. This parameter was monitored in situ in several cells of a randomly selected field. A marked response to SDF-1 (100 nM) was observed for the WT CXCR4, and for the ⌬10 -20, E14A/E15A, Y21A, E179A, D181A, and D182G mutants (Fig. 5). There was a low intensity Ca 2ϩ response, with very few positive cells for the ⌬4 -36, ⌬4 -20, and E288Q mutants, and no detectable signal for the ⌬2-9, D97N, and D187A mutants.
Residues Glu 14 -Glu 15 and Tyr 21 were therefore required for SDF-1 binding but not for signaling, and could be part of the SDF-1 binding site I of CXCR4. Residues Asp 97 , Asp 187 , and Glu 288 were required both for SDF-1 binding and signaling, and could be part of the site II. The finding that the first residues of NT are required for signaling in response to SDF-1 and for binding was less expected. This point will be further discussed below.

Role of Third Intracellular Loop in Signal Transduction-
The mutation of an Asp-Arg-Tyr (DRY) motif in the second intracellular loop (i2) of CXCR4 abolished signaling in response to SDF-1 (31,33). This motif is conserved among chemokine receptors and probably required for their coupling to G proteins (52). Here we have addressed the possible role of the third intracellular loop (i3) in signal transduction by testing three CXCR4 mutants with overlapping deletions of 4 residues (Fig.  1). These mutants were apparently expressed normally at the cell surface (Fig. 6A), and mediated HIV-1 infection with an efficiency similar to WT CXCR4 (data not shown). Infection and reactivity with the 12G5 mAb were fully inhibited by 200 nM SDF-1 (Fig. 6B), indicating that the chemokine binds normally to the ⌬i3 mutants. Upon contact with 100 nM SDF-1, there was no intracellular Ca 2ϩ rise when cells expressed the ⌬i3-A mutant bearing the most NH 2 -terminal deletion (Fig. 6C). A Ca 2ϩ signal in response to SDF-1 was detected in cells expressing the ⌬i3-B and ⌬i3-C mutants, albeit relatively weak for ⌬i3-B. One or several residues of the Ser-His-Ser-Lys motif deleted in the ⌬i3-A mutant seem therefore critical for the coupling of CXCR4 with G proteins. Moreover, our results confirmed that signaling through CXCR4 is dispensible to HIV-1 entry, at least under these experimental conditions (33). DISCUSSION The G protein-coupled receptor CXCR4 and its ligand, the SDF-1 chemokine, seem to participate in apparently diverse biological functions, such as hematopoiesis and the embryonic development of brain or blood vessels. In addition, CXCR4 is a CD4-associated HIV-1 coreceptor, allowing cell entry in numerous cell types and probably playing a major role at the advanced stages of infection. HIV-1 entry can be efficiently blocked by SDF-1 and derived peptides (49,53), or by other CXCR4 ligands such as bicyclams (54). Dissecting the structural domains of CXCR4 supporting SDF-1 receptor and HIV-1 coreceptor activities can therefore contribute to the understanding of developmental processes and virus entry, and also provide information for the design of novel antiviral drugs. In this study, we have tested a series of mutant CXCR4 for their cell surface expression, ability to mediate infection by geneti-cally distant HIV-1 strains (LAI and NDK), and binding to SDF-1.
Determinants of CXCR4 Supporting Its HIV-1 Coreceptor Activity-Several studies have pointed to the role of two extracellular domains of CXCR4, NT, and ECL2, in its HIV-1 coreceptor activity (29,30,(32)(33)(34). We had previously reported that deletion of most of NT (⌬4 -36), or substitutions in ECL2 (e.g. D193A) impaired the coreceptor activity of CXCR4 for HIV-1 strains such as NDK, while having a lesser effect for the LAI strain (29). In the present study, we identified mutations in CXCR4 impairing the coreceptor activity to a similar extent for both strains tested (e.g. E2A, Y7A/Y12A, or Y21A, all in NT), and mutations predominantly affecting infection by NDK (D171N in TMIV), or LAI (D262A in ECL3, D97N in TMII, and E288Q in TMVII). This confirmed that HIV-1 strains can have different requirements for functional interaction with coreceptors. But we also report the importance of residues in the third extracellular loop and in the membrane-spanning domains of CXCR4.
Several types of amino acids were mutated in the NH 2terminal extracellular domain. Tyrosine residues seem to play a role in coreceptor activity, but charged residues (e.g. Glu 14 and/or Glu 15 ) could also be involved. Only charged amino acids were mutated in the extracellular loops and membrane-spanning domains. Several negatively-charged residues apparently required for coreceptor activity were identified. The importance of Tyr residues for HIV-1 coreceptor activity was also documented in the case of CCR5 (55)(56)(57). These residues have been proposed to interact with hydrophobic residues of the CCR5binding site of gp120 (58). Since these gp120 residues are relatively conserved among HIV-1 strains, they could also contribute to the interaction with CXCR4. Alternatively, Tyr residues of CXCR4 could contribute to the coreceptor activity through their sulfation, a post-translational modification resulting in addition of a negatively charged group, as proposed for CCR5 (50).
The selectivity of gp120 for CXCR4 seems due in large part to the accumulation of basic residues in variable loops, in particular V3 (20,22,24). The V3 loop is also responsible for the inability of the NDK strain to infect cells via the rat CXCR4 (23), or the human CXCR4 with mutations at Asp 193 in ECL2 (30). Electrostatical interactions could therefore take place between basic residues of gp120 and the acidic residues we have identified in extracellular and membrane-spanning domains (Asp 97 ) of CXCR4. Recently, Chabot et al. (34) also reported that mutation of Asp 97 markedly impaired the HIV-1 coreceptor activity of CXCR4. This residue, or Glu 288 in TMVII are probably relatively close from the cell surface. They may engage electrostatical interactions with a protruding region of gp120, such as the V3 loop, and contribute to its anchoring in a pocket formed by the extracellular domains of CXCR4.
Determinants of the SDF-1 Receptor Activity-The interaction of SDF-1 with CXCR4 was first addressed in competition assays involving binding of anti-CXCR4 mAbs or HIV-1 infection. A relatively high concentration of the chemokine (200 nM) was used in these assays. CXCR4 mutants whose interaction with SDF-1 was apparently impaired were tested for their ability to bind trace amounts (0.2 nM) of radiolabeled SDF-1 and to mediate a rise in intracellular Ca 2ϩ in response to SDF-1, which indicates receptor activation.
The Glu 14 and Glu 15 (mutated together) and Tyr 21 residues in NT were apparently required for SDF-1 binding but not for cell signaling, consistent with their belonging to the ligandbinding site I of CXCR4. Deletion of residues 10 -20 (⌬10 -20) also disrupted the putative site I. Aromatic (Tyr) and acidic FIG. 4. Binding of radiolabeled SDF-1 to cells expressing WT or mutant CXCR4. HEK293T cells in 6-well trays were transfected with CXCR4, incubated for 90 min with 125 I-SDF-1 (0.2 nM), and cell associated radioactivity counted in each well. Results were corrected for backgound, by subtracting radioactivity counts obtained in a parallel transfection with an empty vector, and for cell surface expression of CXCR4 tested in parallel by staining with the 12G5 mAb. Bars represent amount of bound radioactivity for cells transfected with mutant CXCR4, relative to cells transfected with WT CXCR4. Data are shown as average Ϯ S.E. from at least three independent transfections, except for the E179A, D181A, and D182A mutants, only tested once.
residues required for chemokine binding were also identified in the NT of CCR5 (56,57).
Substitutions of acidic residues in ECL2 (D187A) and in two membrane-spanning domains (D97N and E288Q) affected SDF-1 binding and impaired or abolished receptor activation, suggesting that the corresponding residues contribute to the ligand-binding site II. The first residue of SDF-1 (Lys 1 ) seems crucial for activity (38,49), and can be envisioned to interact with the acidic residues of site II, in particular those located in TM domains of CXCR4, or at the interface with ECL1 in the case of Asp 97 . The role of the ECL2, TMII, and TMVII domains of CXCR4 in SDF-1 signaling is consistent with results obtained by Doranz et al. (33) using chimeric receptors. These authors also reported that substitution of three residues at positions 179, 181, and 182 in ECL2 (QAAN mutant) abolished SDF-1 signaling. Here we observed an apparently efficient interaction with SDF-1 and signaling for the E179A, D181A, and D182G mutants. It seems unlikely that mutation of Glu 179 into Gln, or Asp 182 into Asn could have different effects, suggesting that the phenotype of the QAAN mutant results from cumulative effects of the mutations.
Deletions encompassing the first residues of NT (⌬2-9, ⌬4 -20, and ⌬4 -36) impaired SDF-1 binding and abolished signaling. This suggests that the first residues of NT can have a role in the activation of CXCR4. The assumption that the amino terminus of CXCR4 comprises different functional regions is in line with findings concerning other chemoattractant receptors. For the C5a receptor, although the ligand-binding site was assigned to residues 21-30 in NT, a deletion of residues 1-22 abolished signaling (59), like the ⌬2-9 deletion in our study. The first residues of NT of CXCR4 could be actually part of site II, possibly interacting with the extracellular loops, as has been proposed for the C5a receptor (60). Alternatively, residues 2-9, although not essential for access of the chemokine to site I, may be required for its proper orientation toward site II, and thus for signaling. Notably, a recently described CXCR4 splice variant, CXCR4-Lo, only differs in the first nine residues and yields less potent biological effects in response to SDF-1 (61).
In conclusion, the SDF-1 chemokine and the HIV-1 envelope protein gp120 seem to share certain requirements for their functional interaction with CXCR4, including residues not only in different extracellular domains, like NT and ECL2, but also in membrane-spanning domains. For example, a mutation of Asp 97 in TMII had an effect both on SDF-1 receptor and HIV-1 coreceptor activities. The role of TM residues in the interaction between CXCR4 and gp120 was unexpected and could explain the ability of gp120 to transmit a cell signal via its corresponding coreceptor, CXCR4 or CCR5 (35)(36)(37). Efficient inhibition of HIV-1 infection via CXCR4 seems to require interaction of SDF-1 with residues of CXCR4 involved in signaling (site II). Also, the bicyclam AMD3100 efficiently blocks HIV-1 entry via CXCR4 but behaves as an antagonist for this receptor (54). Antiviral strategies aimed at blocking the coreceptor activity of CXCR4 will certainly require drugs that do not interfere with the chemokine receptor activity. Preliminary studies with SDF-1-derived peptides suggest that this goal can be achieved (49). The position of the deletions is shown in Fig. 1. A, surface expression of CXCR4 was monitored by staining transfected COS cells with the 12G5 mAb, as in Fig.  1A. B, SDF-1 binding was addressed by competition with 12G5 and inhibition of infection, as in Table I. C, intracellular Ca 2ϩ measurements were performed as described in the legend to Fig. 5.