Identification of Transmembrane Domain Residues Determinant in the Structure-Function Relationship of the Human Platelet-activating Factor Receptor by Site-directed Mutagenesis*

Platelet-activating factor (PAF) is a potent phospho- lipid mediator that produces a wide range of biological responses. The PAF receptor is a member of the seven- transmembrane GTP-binding regulatory protein-cou-pled receptor superfamily. This receptor binds PAF with high affinity and couples to multiple signaling pathways, leading to physiological responses that can be inhibited by various structurally distinct PAF antagonists. We have used site-directed mutagenesis and functional expression studies to examine the role of the Phe 97 and Phe 98 residues located in the third transmembrane helix and Asn 285 and Asp 289 of the seventh transmembrane helix in ligand binding and activation of the human PAF receptor in transiently transfected COS-7 cells. The double mutant FFGG (Phe 97 and Phe 98 mutated into Gly residues) showed a 3–4-fold decrease in affinity for PAF, but not for the specific antagonist WEB2086, when compared with the wild-type (WT) re- ceptor. The FFGG mutant receptor, however, displayed normal agonist activation, suggesting that these two ad- jacent Phe residues maintain the native PAF receptor conformation rather than interacting with the ligand. On the other hand, substitution of Ala for Asp 289 increased the receptor affinity for PAF but abolished PAF- dependent inositol phosphate accumulation; it did not affect WEB2086 binding. Substitution of Asn for Asp 289 , however, resulted in a mutant receptor with normal binding and activation characteristics. When Asn 285 was mutated

Platelet-activating factor (PAF) 1 is a potent phospholipid mediator that is involved in a variety of biological activities related to inflammatory and immune responses (1) as well as cardiovascular, reproductive, respiratory, and nervous system physiology (2). The PAF structure has been identified as 1-Oalkyl-2-O-acetyl-sn-glycero-3-phosphocholine (3,4). It is released by stimulated basophils, platelets, macrophages, polymorphonuclear neutrophils, and other cell types (1,2). The PAF structural requirement is highly specific for its biological actions, which are mediated through binding and activation of a specific, high-affinity receptor on the target cell surface. PAF binding has been found on several cell types, and cDNA cloning from various sources revealed that the PAF receptor belongs to the G protein-coupled receptor superfamily (5)(6)(7)(8)(9). The PAF receptor couples with various second messenger systems, including phospholipase A 2 , C, and D activation (10 -12) and the mitogen-activated protein kinase cascade (11,(13)(14)(15). PAF-dependent cellular responses can be inhibited by a variety of structurally distinct PAF antagonists (16).
The recent cloning of the PAF receptor has made possible the study of the structure-function relationship of this receptor, but little information has been published as yet. The cytoplasmic tail of the guinea pig PAF receptor has been shown not to be required for the forward signal transduction to multiple pathways but to play an essential role in the agonist-induced desensitization (17). Moreover, the human PAF receptor contains a single N-linked consensus glycosylation sequence in the putative second extracellular loop (7), in contrast to the guinea pig (18) and the rat (19) receptors, which have an additional NH 2 -terminal consensus sequence for N-glycosylation. In this context, Streptococcus pneumoniae was recently shown to use the human PAF receptor for adherence to and invasion of host cells (20). It was later shown that this N-glycosylation site enhances bacterial binding to the PAF receptor but is not required for the interaction (21). Glycosylation at this site is necessary for efficient membrane trafficking of the PAF receptor but seems to have no role in receptor affinity and activation (21). In additional structural studies, we have demonstrated that mutation of two adjacent residues, Ala 230 and Leu 231 , in the COOH-terminal region of the third intracellular loop led respectively to inactive and constitutively active phenotypes of the PAF receptor (22). Moreover, we showed that the highly conserved Asp 63 residue in the second transmembrane helix is not involved in ligand binding but is necessary for G protein coupling of the human PAF receptor (23).
Recently, a molecular model of the PAF receptor was proposed based on the bacteriorhodopsin three-dimensional structure (24). In this model, the side chains of Asp 63 , Asn 285 , and * This work was supported by the Medical Research Council of Canada. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ Supported by a studentship from the Fonds pour la Formation de Chercheurs et d'Aide à la Recherche.
¶ To whom correspondence should be addressed: Immunology Division, Faculty of Medicine, University of Sherbrooke, 3001, North 12th Ave., Sherbrooke, Quebec J1H 5N4, Canada. Tel.: 819-564-5268; Fax: 819-564-5215. 1 The abbreviations used are: PAF, platelet-activating factor; G pro-Asp 289 were adjacent to each other and oriented toward the central core of the receptor. These three residues were thought to form a negatively charged site that would attract the positively charged choline moiety of the PAF molecule by electrostatic forces. Conserved Asp residues in the third TMH, which were shown to participate in hydrophilic ligand interactions for other G protein-coupled receptors (25), are absent in the corresponding domain of the PAF receptor. Instead, two hydrophobic residues, Phe 97 and Phe 98 , are similarly located, which could possibly interact with the acetyl group or the phospholipid chain of the PAF molecule. No experimental data are yet available on residues that might be part of the binding pocket of the PAF receptor. To address this question, we mutated the Phe 97 , Phe 98 , Asn 285 , and Asp 289 residues of the PAF receptor, and properties of these mutants were compared with those of the wild-type (WT) PAF receptor in transiently transfected COS-7 cells. In this report, we suggest that the Phe 97 and Phe 98 residues may be involved in maintaining the native conformation of the PAF receptor and, moreover, that the PAF molecule does not bind to the Asn 285 and Asp 289 residues, contrary to what had been proposed (24). However, the two latter amino acids could be determinant in receptor conformation and activation.

Construction of the Mutant Receptor cDNAs and Expression
Vectors-The PAF receptor cDNA derived from Kp132 (a generous gift from Dr. Richard Ye, The Scripps Research Institute, La Jolla, CA) (8) was subcloned into the pRc-cytomegalovirus expression vector (Invitrogen). Mutated receptors were constructed by polymerase chain reaction (26) using Kp132 as a template. To create the FFGG double mutant, we made the oligonucleotide 5Ј-GTGGCTGGCTGCCTTGGCGGCATCAA-CACCTAC-3Ј and its reverse complement, which changes TTC (Phe) to GGC (Gly). Similarly, to mutate the Asn 285 to Ala and Ile, we generated the oligonucleotides 5Ј-CCTTAGCACCGCCTGTGTCTTAG-3Ј and 5Ј-CCTTAGCACCATCTGTGTCTTAG-3Ј and their reverse complements, respectively, changing AAC (Asn) to GCC (Ala) and ATC (Ile). To substitute Ala and Asn for Asp 289 , the following oligonucleotides and their reverse complements were used: 5Ј-GTCTCTTAGCGCCTGT-TATC-3Ј and 5Ј-GTCTCTTAAACCCTGTTATC-3Ј, mutating GAC (Asp) to GCG (Ala) and AAC (Asn), respectively. The FFGG double mutant polymerase chain reaction product was digested with HindIII-MScI, whereas polymerase chain reaction products for the Asn 285 and Asp 289 mutants were digested with BstEII-XbaI and subcloned into pRc-cytomegalovirus containing the WT receptor cDNA digested with the same enzymes, respectively. The region corresponding to the subcloned polymerase chain reaction fragments was sequenced on both strands by dideoxy sequencing of double-stranded DNA with Sequenase (U. S. Biochemical Corp.).
Cell Culture and Transfections-COS-7 cells were grown in Dulbecco's modified Eagle's medium (high glucose), supplemented with 10% fetal bovine serum. Cells were plated in 30-mm dishes (1.5 ϫ 10 5 cells/dish), transiently transfected with the constructions encoding the WT and mutant receptors using 5 l of LipofectAMINE (Life Technologies, Inc.) and 2 g of DNA/dish, and harvested 48 h after transfection.
Radioligand Binding Assay-Competition binding curves were done on COS-7 cells expressing the wild-type and mutant receptor species. Cells were harvested and washed twice in Hepes-Tyrode's buffer (140 mM NaCl, 2.7 mM KCl, 1 mM CaCl 2 , 12 mM NaHCO 3 , 5.6 mM D-glucose, 0.49 mM MgCl 2 , 0.37 mM NaH 2 PO 4 , and 25 mM Hepes pH 7.4) containing 0.1% (w/v) bovine serum albumin (13). Binding reactions were carried out on 5 ϫ 10 4 cells in a total volume of 0.25 ml in the same buffer with 10 nM [ 3 H]WEB2086 (DuPont NEN) and increasing concentrations of nonradioactive WEB2086 or PAF for 90 min at 25°C. Reactions were stopped by centrifugation. The cell-associated radioactivity was measured by liquid scintillation. Binding reactions involving [ 3 H]PAF (4 nM) were performed on intact adherent COS-7 cells in a total volume of 600 l of binding buffer containing 150 mM choline chloride, 10 mM Tris-HCl, 10 mM MgCl 2 , pH 7.5, and 0.25% lipid-free bovine serum albumin (27). Samples were incubated for 4 h at 4°C. Cells were washed four times with ice-cold buffer (1 ml), lysed with 0.1 N NaOH, and analyzed for radioactivity (27).
Inositol Phosphate Determination-COS-7 cells were transfected as described above with the wild-type or mutant receptors and labeled the following day for 18 -24 h with [ 3 H]myo-inositol (Amersham Corp.) at 5 Ci/ml in Dulbecco's modified Eagle's medium (high glucose, without inositol; Life Technologies). After labeling, cells were washed once in phosphate-buffered saline (PBS) and preincubated for 5 min in PBS at 37°C. At the end of this preincubation period, the PBS was removed, and cells were incubated in prewarmed Dulbecco's modified Eagle's medium (high glucose, without inositol) containing 20 mM LiCl for 5 min. Cells were then stimulated for 30 s with indicated concentrations of PAF. The reactions were terminated with the addition of perchloric acid followed by a 30-min incubation on ice. Inositol phosphates were extracted (28) and separated on Dowex AG1-X8 (Bio-Rad) columns (29). Total labeled inositol phosphates were then counted by liquid scintillation.
Flow Cytometric Studies-The N285I and WT receptors were subcloned in frame with the c-myc epitope in the pJ3M vector, kindly provided by Dr. J. Chernoff (Fox Chase Cancer Center, Philadelphia, PA) (30). The pJ3M-c-myc N285I and pJ3M-c-myc WT receptor constructs were transfected in COS-7 cells, which were harvested 48 h after transfection and subjected to flow cytometric analysis. Cells (2.5 ϫ 10 5 ) were washed twice in PBS and labeled with or without anti-c-myc antibody (9E10 hybridoma; American Type Culture Collection) at room temperature for 30 min. Cells were then washed with PBS and incubated at room temperature for an additional 30 min with fluorescein isothiocyanate-conjugated goat anti-mouse antibody (Bio/Can). All measures were performed on a FACScan flow cytometer (Becton-Dickinson).

RESULTS
To determine residues that might contribute to ligand binding and receptor activation of the human PAF receptor, we performed site-directed mutagenesis of the Phe 97 , Phe 98 , Asn 285 , and Asp 289 residues and evaluated the properties of the mutant receptors in comparison with the WT PAF receptor in transiently transfected COS-7 cells. Fig. 1 shows a representation of the putative seven membrane-spanning domain topography of the PAF receptor and indicates the amino acids that were replaced in the mutant receptors.
Characterization of the Binding Properties of the Mutant Receptors-Binding characteristics of the WT and mutant receptors are summarized in Table I. Fig. 2 shows that the affinity of the specific PAF receptor antagonist WEB2086 for the indicated mutant receptors was unaltered compared with the WT receptor in competition binding experiments. Calculated receptor densities (B max ), indicated as receptors per cell, were also similar for the different receptors, except for the N285I mutant (Table I). As is well known, PAF being a phospholipid, binding experiments using [ 3 H]PAF constitute a large nonspecific component that renders reproducibility and interpretation of the results difficult (27). Since the affinity of the antagonist was the same for all the receptors, we used competition of [ 3 H]WEB2086 by cold PAF to assess the PAF affinity for the different receptors and obtained high reproducibility between experiments. Fig. 3A  transfected cells) could be detected for this mutant in every single experiment (n ϭ 4). A 3-fold higher affinity for PAF than the WT receptor was displayed by the D289A mutant receptor, which had a K i of 12 Ϯ 2.3 nM. In contrast, the FFGG mutant receptor had almost a three-fold lower affinity for PAF, with a K i of 87 Ϯ 5.7 nM. As these results indirectly assess PAF binding by displacing the [ 3 H]WEB2086 probe, we then performed direct studies of PAF binding to further support our conclusions. Fig. 3B illustrates competition binding isotherms of PAF to the different receptors. As summarized in Table I, conclusions that can be drawn by the direct PAF binding studies are essentially the same as those from the displacement of [ 3 H]WEB2086. The WT, N285A, and D289N receptors displayed the same affinity for PAF, with respective K i values of 12 Ϯ 2.7, 9.8 Ϯ 2.4, and 12.7 Ϯ 3.1 nM. A 4-fold higher affinity for PAF than the WT receptor was displayed by the D289A mutant, whereas the FFGG double mutant showed approximately a 4-fold lower affinity for PAF than the WT receptor (Table I).
Inositol Phosphate Accumulation in Mutant Receptor-transfected Cells-The ability of the mutant receptors to transduce a signal was then tested by measuring IP accumulation following stimulation with graded concentrations of PAF. Fig. 4 illustrates the concentration-response curves of IP accumulation for the WT and mutant receptors in response to PAF concentrations from 0 to 10 Ϫ6 M. PAF concentrations higher than 10 Ϫ6 M were not used, as PAF has been shown to have nonreceptormediated effects at these concentrations (31), making the in-terpretation of results difficult. No significant difference could be detected between the activation of the FFGG, N285A, D289N, and WT receptors, in which IP production reached a plateau at 10 Ϫ7 M of agonist with similar half-maximal effective concentration values (Table I). No IPs were produced by the N285I and D289A mutants over the entire range of PAF concentrations.   Table I.

FIG. 3. Competition binding isotherms of [ 3 H]WEB2086 by PAF (A) and [ 3 H]PAF by PAF (B) in COS-7 cells. [ 3 H]WEB2086 and [ 3 H]PAF binding were determined as indicated under "Experimental
Procedures" on COS-7 cells transiently expressing the WT and the indicated mutant receptors. The results are representative of three independent experiments, the mean Ϯ S.E. values of which are reported in Table I. verify whether the phenotype observed for the N285I mutant receptor was caused by decreased cell surface expression, N285I and WT receptors were tagged with a c-myc epitope (30) and used to transfect COS-7 cells, which were then subjected to flow cytometric analysis. The tagged receptors conserved their respective binding parameters (data not shown). Fig. 5 demonstrates that both the tagged WT and N285I mutant receptors were equally well expressed at the cell surface with ϳ30% of cells displaying receptors with similar fluorescence intensity, indicating that the loss of activation and ligand binding of the N285I mutant receptor was not due to an absence of cell surface expression. DISCUSSION PAF has numerous biological activities and stimulates multiple signaling pathways through its specific G protein-coupled receptor (13). The PAF receptor was the first G protein-coupled receptor for a lipid ligand to be cloned, and, to our knowledge, no experimental data have yet been published on residues that might be involved in lipid ligand interaction with this receptor superfamily. We have previously shown that the highly conserved Asp 63 residue in the second transmembrane domain of the PAF receptor was not involved in direct binding of either the antagonist WEB2086 or PAF. The mutation of this residue to Asn increased affinity specifically for PAF but abolished G protein coupling (23). Our group has also demonstrated that the receptor affinity for PAF could be affected by mutations in the COOH-terminal region of the third intracellular loop of the PAF receptor (22). In the present report, we have used sitedirected mutagenesis to study the possible involvement of the transmembrane residues Phe 97 , Phe 98 , Asn 285 , and Asp 289 in ligand binding and activation of the human PAF receptor.
A conserved Asp residue is found in the third TMH of several G protein-coupled receptors and has been shown to participate in binding hydrophilic ligands such as the cationic neurotransmitters adrenaline, acetylcholine, dopamine, and serotonin (25), which are reminiscent of the positively charged PAF choline moiety. Asp residues are absent, however, from the third TMH of the PAF receptor. Thus, we mutated the two adjacent Phe residues (Phe 97 and Phe 98 ) of the third TMH into Gly residues (FFGG double mutant) to verify whether those hydrophobic amino acids, positioned at the locus of the conserved Asp residue mentioned above, were involved in ligand binding, pos-sibly through interaction with the acetyl group or the phospholipid chain of the PAF molecule.
Our results showed that the Phe 97 and Phe 98 residues have no role in WEB2086 binding but, when mutated, produced a 3-4-fold decrease in affinity for PAF. This change in affinity more likely reflects an increased flexibility and, consequently, a slight conformational change of the PAF receptor rather than a modification of direct interaction of these two hydrophobic residues with the PAF molecule. This is supported by the fact that the FFGG mutant produced the same level of IPs as the WT receptor following PAF binding. It has been shown that even a small modification of either the phospholipid chain, the acetyl group, or the phosphocholine moiety of the PAF molecule resulted in considerable loss of its biological activity (2). If the Phe residues were to interact with any of the chemical groups of the PAF molecule, then the FFGG mutant activation would be significantly affected. The change in agonist affinity could be caused by the substituting Gly residues, the higher degree of freedom of movement of the third TMH, or the disappearance of hydrophobic interactions in the FFGG mutant with other residues of the receptor or with the membrane. Since the double mutant displayed only a small decrease in PAF, but not in WEB2086 affinity, these two hydrophobic residues seem to be involved in maintaining the PAF receptor in its optimal conformation. These results also suggest that the agonist and antagonist are not sensitive to the same structural variation of the receptor.
An Asn residue, corresponding to position 289 of the PAF receptor, is found conserved in 95% of all G protein-coupled receptors (32). However, the PAF receptor has an Asp residue at that position, which we substituted by its isoster, Asn (D289N mutant), to ascertain whether the negative charge of the Asp 289 residue could interact with the positive charge of the PAF molecule and to study the effects of converting this residue back to the 95% conserved amino acid at this locus. In addition, the Asp 289 amino acid was mutated to Ala to study the potential involvement of residue 289 in forming hydrogen bonds, which might contribute to the structure and function of this receptor.
Substitution of Asn for Asp 289 resulted in identical binding and activation characteristics between the WT and the mutant receptors. Asp, which carries a negative charge, was replaced with Asn, an amino acid similar in size but possessing no net charge, with the rationale that this mutation would allow for assessment of the role of the negatively charged Asp at this locus. The data obtained indicate that genetic selective pressure did not introduce the Asp 289 at this position, instead of the 95% conserved Asn residue, to interact with the PAF receptor ligands; contrary to what had been proposed (24), the negative charge of Asp 289 does not seem to be required for attracting and docking the positive charge of the PAF choline moiety. The D289A mutant displayed normal WEB2086 binding parameters but showed an increased affinity for PAF and was unable to accumulate IPs following agonist stimulation. These results suggest that the residue found at position 289 is of no apparent importance in ligand binding but is critical for signal transduction. It may indicate that, although possible hydrogen bonds formed between the side chain functionalities of Asn or Asp residues at this locus with other amino acids of the PAF receptor are not important for ligand binding, such interactions could be necessary for appropriate receptor conformation and transition from the inactive to the active form, leading to IP accumulation. Other G protein-coupled receptors, such as the thrombin, thromboxane A 2 , and choriogonadotropin-releasing hormone receptors, have an Asp at this locus (32). In the choriogonadotropin-releasing hormone receptor, it was suggested that this amino acid maintained the receptor structure by forming hydrogen bonds with an Asn residue in the second TMH (33). To our knowledge, no particular function has been attributed to the cognate Asp residues of the thrombin and thromboxane A 2 receptors as yet.
In the ␤ 2 adrenergic receptor, Asn 318 , which corresponds to Asn 285 in the seventh TMH of the PAF receptor, is presumed to interact with ␤ 2 agonists (34). Therefore, we substituted Ile and Ala residues for Asn 285 (N285I and N285A mutants) to study its potential role in PAF receptor biology. The substitution of an Ala for Asn 285 showed that the presence of Asn at this position is not required for ligand binding, the maintenance of the optimal conformation, or the activation of the PAF receptor. These data may also indicate that Asn 285 is not forming essential hydrogen bonds with other amino acids of the receptor. However, mutation of Asn 285 to Ile surprisingly led to an apparent complete loss of ligand binding despite normal cell surface expression of the mutant receptor, as judged by flow cytometric analysis. It seems, with the results obtained, that the residue found at position 285 could influence receptor conformation, although it would not be involved in direct ligand binding. To explain the differences between the N285A and N285I phenotypes, we might speculate that mutation of Asn 285 to Ala would not alter the PAF receptor structure-function relationship, because both amino acids have small sizes, in contrast to Ile, which is more hydrophobic and bulky, with ␤ branching in its side chain. The Ile residue could modify the structure of the seventh TMH and/or impair the interaction of this TMH with other transmembrane domains of the PAF receptor, leading to a distortion of the receptor molecule that could prevent docking of the ligand into the binding pocket. Further experiments will be necessary to delineate the exact role of residue 285 in PAF receptor biology.
It will be interesting to examine the reactivity of PAF receptor mutants to a range of PAF receptor agonists and antagonists. In this context, preliminary data suggest that the WT, FFGG, and D289A receptors display the same affinity for the gingkolide antagonist BN52021. In contrast, binding studies revealed that the D289A mutant had a 4 -5-fold increase in affinity for the antagonist CV3988, which is a PAF analog, whereas the FFGG mutant had the same affinity when compared with the WT receptor. Other such interaction studies between PAF receptor mutants and structurally distinct ligands are presently under way in our laboratory to further analyze the sites of interaction of these molecules with the receptor.
In summary, we have reported here and in a previous study (23) that the PAF molecule does not bind to its receptor by interacting with the Asp 63 , Asn 285 , and Asp 289 residues, contrary to what had been proposed in a recent molecular model of the PAF receptor (24). Discrepancies between the molecular model and our experimental data may come from the fact that the model was based on the bacteriorhodopsin three-dimensional structure, which displays a seven-transmembrane domain topography but is not a G protein-coupled receptor and has no notable sequence homology. A projection map of rhodopsin (a G protein-coupled receptor) at a 9-Å resolution showed that the configuration of the helices of bacteriorhodopsin was different from rhodopsin (35). Such differences are likely to exist with the PAF receptor. However, we suggest that receptor conformation and transition could be influenced by the amino acids found at positions 63 (23), 97, 98, 285, and 289. Moreover, we have shown that Asp 63 (23) and the residue found at position 289 are determinant in receptor activation. Work is under way in our laboratory to identify residues directly involved in ligand binding to the PAF receptor and will hopefully contribute to a better understanding of how the PAF receptor ligands bind to their receptor and effect physiological responses.