Domain one of the high affinity IgE receptor, FcepsilonRI, regulates binding to IgE through its interface with domain two.

The high affinity receptor for IgE, FcepsilonRI, binds IgE through the second Ig-like domain of the alpha subunit. The role of the first Ig-like domain is not well understood, but it is required for optimal binding of IgE to FcepsilonRI, either through a minor contact interaction or in a supporting structural capacity. The results reported here demonstrate that domain one of FcepsilonRI plays a major structural role supporting the presentation of the ligand-binding site, by interactions generated within the interdomain interface. Analysis of a series of chimeric receptors and point mutants indicated that specific residues within the A' strand of domain one are crucial to the maintenance of the interdomain interface, and IgE binding. Mutation of the Arg(15) and Phe(17) residues caused loss in ligand binding, and utilizing a homology model of FcepsilonRI-alpha based on the solved structure of FcgammaRIIa, it appears likely that this decrease is brought about by collapse of the interface and consequently the IgE-binding site. In addition discrepancies in results of previous studies using chimeric IgE receptors comprising FcepsilonRIalpha with either FcgammaRIIa or FcgammaRIIIA can be explained by the presence or absence of Arg(15) and its influence on the IgE-binding site. The data presented here suggest that the second domain of FcepsilonRI-alpha is the only domain involved in direct contact with the IgE ligand and that domain one has a structural function of great importance in maintaining the integrity of the interdomain interface and, through it, the ligand-binding site.

The high affinity IgE receptor, Fc⑀RI, is a tetrameric complex composed of an IgE-binding ␣ subunit associated with a tetraspan ␤ subunit and homodimeric ␥ subunits and is a key player in IgE dependent effector mechanisms. The ␣ subunit, Fc⑀RI-␣, is the ligand-binding chain and is composed of two Ig-like domains. The role of the second domain has been clearly defined as containing the IgE-binding region. However, the role of the first domain is not clear in Fc⑀RI nor indeed in any Fc receptor. Analyses to date have variously indicated that domain one is necessary for optimal binding (1-3), that it has a possible role in direct interaction with IgE (4,5), and that it provides a supportive role in maintaining receptor integrity (1,2). The structural reasons for this are not apparent. Fc⑀RI, however, is related to Fc␥RIIa, and the recent description of the three dimensional structure of Fc␥RIIa (6), Fc␥RIIb (7), and Fc⑀RI-␣ (8) may provide a basis for the understanding of the roles of the individual domains in Fc⑀RI and other Fc receptors.
In the crystal structure of Fc␥RIIa the extracellular domains are "bent" to form an acute angle (52°) between domains 1 and 2. In this orientation, the IgG-binding site of domain 2 points away from the cell in such a manner as to be accessible to ligand, and domain 1 is angled away from the binding site and down toward the cell membrane. The acute angle is dictated by interactions within the interdomain interface, and the structural studies indicate that domain 1 is likely to support domain 2 providing an architectural role in the positioning of the binding site. Because Fc⑀RI and Fc␥RIIa show 40% amino acid identity and considerably higher amino acid homology, it is probable that Fc⑀RI has a similar structure to that of Fc␥RIIa, confirmed by the recent publication of the solved Fc⑀RI structure (8).
In the study described herein we have utilized a model of Fc⑀RI-␣ (see Fig. 1) based on the solved crystal structure of Fc␥RIIa (6) and undertaken a mutagenesis study of domain 1 to define its role in the interaction with IgE. The solved x-ray structure of Fc⑀RI-␣ (8) strongly resembles that of Fc␥RIIa and therefore that of the Fc⑀RI-␣ homology model. Indeed, the Fc⑀RI-␣ homology model and the x-ray structure of Fc⑀RI-␣, as described by Garman et al. (8), show compelling concurrence in comparisons of structure and molecular interactions. Here, data from the chimeric Fc receptors and alanine mutants have been used together with molecular modeling to propose a functional structure of Fc⑀RI-␣.

EXPERIMENTAL PROCEDURES
Production and Nomenclature of Fc⑀RI-␣ Chimeric cDNA Receptor Constructs-Two previously produced chimeric cDNA receptor constructs (1) were used as templates in the construction of this series of FcRs. The amino acid sequences of the chimeras and chimera nomenclature are displayed in Table I. The first template was designated ⑀⑀␥ and comprised domain one (D1) 1 and domain two (D2) of Fc⑀RI-␣ linked with the transmembrane region and cytoplasmic membrane anchor of * 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  Fc␥RIIa. The second chimeric template was based on a simple domain exchange and comprised D1 of Fc␥RIIa and D2 of Fc⑀RI-␣, also with the transmembrane region and cytoplasmic sequence of Fc␥RIIa, and was designated ␥⑀␥. Chimeric receptors were generated using the template receptor ␥⑀␥ or ⑀⑀␥. Specific loops, strands, or regions of the Fc␥RIIa D1 were replaced with the equivalent portion of Fc⑀RI (or vice versa) to produce a series of chimeric receptors using splice overlap extensionpolymerase chain reaction (SOE-PCR) using the method previously reported (9). A further template receptor was constructed with a glycosyl-phosphatidylinositol membrane anchor of Fc␥RIIIB replacing the Fc␥RIIa cytoplasmic tail of the ␥⑀␥ construct. This chimera was designated ␥⑀RIII and was generated by SOE-PCR. Substitution into domain one of the ␥⑀RIII template receptor of the A strand of Fc⑀RI D1 produced the ␥(A⑀)⑀RIII chimera. The Fc⑀RI D1 AЈ strand point mutants, R15A and F17A, were made using SOE-PCR and incorporated into ⑀⑀␥. The Fc⑀RI D1 AЈ strand point mutants N14A and R15L were constructed using the QuikChange TM site-directed mutagenesis kit (Stratagene). cDNA was purified by centrifugation in a CsCl gradient (10), and the mutations were verified by nucleotide sequencing.

Detection of IgE Activity by Enzyme-linked Immunosorbent Assay-
High bind EIA/RIA plates (Costar 3690) were coated with 8 g/ml of anti-human IgE mAb HB121 in phosphate-buffered saline overnight. The plates were blocked prior to the addition of serially diluted monoclonal IgE in phosphate-buffered saline containing 1.5% bovine serum albumin (50 l/well, 60 min). rsFc⑀RI 2 was then added (50 l/well, 60 min). Bound rsFc⑀RI was detected using horseradish peroxidase-conjugated 3B4, a nonblocking anti-human Fc⑀RI mAb (1 g/ml, 50 l/well, 60 min). The assay was carried out at 20°C, and the plates were washed seven times with water between each incubation step. Color development was with o-phenylenediamine (Sigma) and stopped after 15 min with 25 l of 4 M sulfuric acid. The optical density was measured at 490 nm.
Transfection of Mammalian Cells with cDNA-COS-7 cells were maintained (1) in Dulbecco's modified Eagles medium (Life Technologies, Inc.). For transient transfection LipofectAMINE (Life Technologies, Inc.) reagent was used, with plasmid DNA of interest, according to the manufacturer's instructions.
Immune Complex Binding-The binding of IgE or IgG immune complexes to cells transfected with chimeric or mutant cDNA was determined by erythrocyte-antibody (EA) rosetting, which was assayed and scored according to the method previously reported (1). Briefly, mouse anti-TNP IgE or IgG (moIgE or moIgG) was incubated with TNP-coated sheep red blood cells to form complexed IgE or IgG. These antibody sensitized erythrocytes were mixed with transfected cells, and the binding of these complexes to cells was determined microscopically (1). The utilization of avidity in this way permits the determination of low affinity binding.
Measurement of IgE/Fc⑀RI by Equilibrium Binding-Equilibrium binding was determined by the method previously reported (1). IgE was radioiodinated using IODO-GEN (Pierce) according to the manufacturer's instructions. The 125 I disintegrations/min were determined separately for the cell pellets (bound IgE) and the supernatant (free IgE) in a WALLAC 1470 WIZARD TM automatic ␥ counter. Nonlinear regression analysis was performed by plotting IgE free versus IgE bound in the program "Curve Expert" using the formula for single site binding, y ϭ (a * x)/(b ϩ x); where y ϭ IgE bound and x ϭ free IgE. The equilibrium binding dissociation constant (K D ) was obtained from three experiments with a correlation coefficient of Ͼ0.99 (see Table II). The maximum binding (B max ) of IgE was also determined and used to estimate receptor expression.
Detection of Membrane-bound Fc⑀RI by Monoclonal Antibodies Using Flow Cytometry-COS-7 cells were transiently transfected with rFc⑀RI cDNA as above. Approximately 40 h post-transfection the COS-7 cells were incubated with saturating amounts of antibody, on ice, for 45 min; the cells were washed, resuspended in a 1:100 dilution of anti-mouse Ig F(abЈ) 2 -fluorescein isothiocyanate (Silenus), and incubated for 30Ј on ice. The cells were washed and resuspended in phosphate-buffered saline containing 0.5% bovine serum albumin, 0.1% glucose, 3 g/ml propidium iodide and analyzed in a FACScalibur (Becton Dickinson). All washes and dilutions were in phosphate-buffered saline containing 0.5% bovine serum albumin, 0.1% glucose. Analysis was conducted on live (propidium iodide negative) cells.
Modeling of Fc⑀RI, Chimerae, and Mutants-The extracellular regions of the ␣-chain of the human Fc⑀ receptor type I (Fc⑀RI-␣) and the human Fc ␥ Receptor type II a (Fc␥RIIa) show a sequence identity of about 40% for 172 residues (this consists of a sequence identity of about 45% for the first domain and about 36% for the second domain). Fc␥RIIa is the protein most homologous to Fc⑀RI, for which the three-dimensional structure is known (6). With the significant sequence identity, even higher sequence similarity, and the conservation of several important amino acid residues between the two proteins (see the sequence alignment given in Table I), clearly Fc␥RIIa is the most appropriate three-dimensional structural template to use in modeling Fc⑀RI, more suitable than the structures of CD2 or CD4, which have been used in the past to construct models of Fc⑀RI (12,13). The recently solved crystal structure of Fc⑀RI (8) confirmed the similarity of the two structures, including the C2 sub type of the Ig-like domains and the acute FIG. 2. Fc⑀RI-␣ homology model with electrostatic potential displayed. The electrostatic potential of Fc⑀RI-␣ model was calculated and mapped onto the molecular surface, with red indicating a negative, and blue indicating a positive electrostatic potential. It can be seen that the face of Fc⑀RI-␣ comprising the C/F/G strands of domain 1 and domain 2 (A) has a considerably more negative electrostatic potential than the face comprising the A/B/E strands of domain 1 and domain 2 (B).
angle between the two domains. However, the cartesian coordinates of the crystal structure of Fc⑀RI were not available, and therefore we made use of the homology model of Fc⑀RI built in this work.
Fc⑀RI-␣ Model-Modeler (14) as implemented in the InsightII_Homology software package (Insight II (97.0), Molecular Simulations Inc., San Diego, CA) was used to build three-dimensional models of Fc⑀RI-␣ using a number of different initial sequence alignments and two structural templates of Fc␥RIIa. One of the structural templates was the three-dimensional coordinates of Fc␥RIIa where for the residues that had alternative side chain conformations (residue numbers 10, 21, 33, 57, 60, 61, 65, and 89) the conformations labeled A were selected, whereas in the other template the conformations labeled B were selected. In each Modeler run, five structural models of Fc⑀RI-␣ were generated. The following parameter values or options were used: li-brary_schedule of 1, max_var_iterations of 300, md_level of refine1Ј, repeat_optimization of 3, and max_molpdf of 10 6 . The best model from these runs had the sequence alignment given in Table I and used the structural template of Fc␥RIIa where residues 10, 21, 33, 57, 60, 61, 65, and 89 had side chains in the A conformation. The criteria for judging the best model included the lowest value of the Modeler objective function (or Ϫ1.0 ϫ ln (molecular probability density function), well behaved ProsaII (15) residue energy plot for the model (for example, negative residue energy scores throughout the sequence), and well behaved Profiles-3D (16) local 3D-1D compatibility score plot (for example, positive plot scores throughout the sequence).
Next, Modeler was used to generate 20 different structural models of Fc⑀RI-␣ using the sequence alignment and template selected above and using the parameter values and options listed above. From these, the model with the lowest Ϫ ln (molecular probability density function) value was then further improved (as measured by ProsaII, Profiles-3D, and Procheck (17)) by being selected as the template to generate structural models of the Fc⑀RI-␣ sequence in the next cycle of Modeler runs. At the end of four such cycles, the best three-dimensional model of the Fc⑀RI-␣ structure (i.e. the model with the lowest value of the Modeler objective function) was selected as the final structural model of the Fc⑀RI-␣ monomer. Secondary structure prediction performed on Fc⑀RI-␣ sequence confirmed the validity of the alignment given in Table  I and showed the pattern of ␤ strands is the same in both Fc⑀RI-␣ and Fc␥RIIa. The secondary structure prediction methods used were PHD (18) and PREDATOR (19). The model is displayed in Fig. 1.
Mutant and Chimeric Receptors-The R15A and the F17A point mutants of Fc⑀RI-␣ were modeled from the above Fc⑀RI-␣ model by mutating the R15 and F17 residues to alanines with InsightII_Homology module (MSI), adding hydrogens to the two models, and energy minimizing the structures, keeping all heavy atoms fixed except for the Ala 15 and Ala 17 residues, respectively. The program Discover v. 2.98 (MSI) was used for the energy minimization with the CFF91 force field and a distance-dependent dielectric constant of 1.0xr, and the minimization was done with the conjugate gradients method until the maximum energy gradient was less than 0.01 kcal/Å.
Three chimera structures of Fc⑀RI that were experimentally constructed and the binding to IgE investigated were modeled based on the structural template of Fc␥RIIa. The sequences of these three chimera, labeled ␥⑀␥, ␥(AЈB⑀)⑀␥, and ⑀(AЈB␥)⑀␥, respectively, are shown in Table  I. The same sequence alignment as shown in Table I and the same Modeler parameter values and options as were used to generate the model of Fc⑀RI-␣ (as described earlier) were used to construct these chimera models. Again, out of 20 models generated for each chimera, the model with the lowest Modeler objective function was selected, and the model structure was validated with ProsaII, Profiles-3D, and Procheck. Finally the electrostatic potential was calculated and mapped onto the molecular surface (Fig. 2) of the constructed Fc⑀RI-␣ model using the program GRASP (20). The PARSE3 (21) charge set was used in computing the electrostatic potential. It can be seen that one face of Fc⑀RI-␣ has a considerably more negative electrostatic potential than the other face.
The co-ordinates of the Fc⑀RI-␣ model are available on request.

Analysis of Chimeric Receptors to Establish the Role of D1 Fc⑀RI-␣ in IgE
Binding-Examination of the Fc⑀RI-␣ homology model (shown in Fig. 1 and fully described below) and the model of the ␥⑀␥ chimera indicates that the two major regions of D1 that impinge on the D1/D2 interface are, first, the A strand, AЈ strand, and AЈB loop, and second, the G strand. Utilizing a chimeric FcR (␥⑀␥) comprising Fc␥RIIa D1 and Fc⑀RI D2 (IgE-binding domain), those segments that form part of the D1/D2 interface of D1 Fc␥RIIa cDNA were replaced with the equivalent portion of Fc⑀RI cDNA (Table I). The receptors were assayed to determine which segments conferred a gain of function. Four chimeric receptors were constructed that encompassed the AЈ strand AЈB loop region, namely, ␥(ABC⑀)⑀␥, ␥(ABCCЈ⑀)⑀␥, ␥(AAЈB⑀)⑀␥, and ␥(AЈB⑀)⑀␥. Analysis of the transfected receptors by EA rosetting (1) indicated, surprisingly, that none of these chimeric receptors bound complexed moIgE as IgE-coated erythrocytes; as expected, none bound moIgG.
To determine whether the receptors were expressed on the cell surface, the chimeras were tested by flow cytometry, using a panel of four anti-Fc⑀RI-␣ mAbs that recognized separate nonoverlapping epitopes, two (mAb3B4 and mAb54) in D1 and two (mAb47 and mAb15-1) in D2. 2 Two of the four chimeras tested, ␥(ABC⑀)⑀␥ and ␥(ABCCЈ⑀)⑀␥, were not detected on the cell surface and thus were assumed not to be expressed. However, the ␥(AAЈB⑀)⑀␥ and ␥(AЈB⑀)⑀␥ chimeras were detected on the cell surface by mAb and were then tested for their ability to bind monomeric hIgE in an equilibrium binding assay (Table  II). Both chimeras failed to bind monomeric hIgE, which is consistent with their failure to bind moIgE complexes (Table  II). Thus, despite the fact that these interface sequences were derived from Fc⑀RI, the chimeras were not able to bind IgE.   The ␥(AAЈB⑀)⑀␥ (Fc⑀RI residues 1-21) and ␥(AЈB⑀)⑀␥ (Fc⑀RI residues 14 -21) chimeras differ only by the inclusion or absence of the Fc⑀RI A strand of D1 (Fc⑀RI residues 1-10), and although these chimeras were detected on the cell surface by mAb47 (which maps to an epitope in D2 near the transmembrane region), they were not detected by mAb15-1, an antibody previously reported to detect an epitope in D2 within the IgEbinding site (22). Thus, although the receptors were expressed, they were incapable of binding monomeric hIgE or moIgE complexes, implying significant disruption to the IgE-binding site. This disruption is most probably caused by the segment that is common to both receptor chimeras, that of the AЈ strand, AЈB loop (Fc⑀RI residues 14 -21), rather than the A strand, which is present only in the ␥(AAЈB⑀)⑀␥ chimera. In the structure of Fc␥RIIa and the model of Fc⑀RI, the cis-Pro 11 at the start of the AЈ strand, is essential for maintaining the conformation of this part of the interdomain interface. This would imply that the A strand segment up until the cis-Pro has little impact on the interdomain interface and was probably not involved in the disruption to IgE binding, as would be expected from the structure. This was confirmed by testing a chimera with only the A strand segment from the N terminus to the cis-Pro 11 from Fc⑀RI in D1 Fc␥RIIa, namely ␥(A⑀)⑀RIII. This A strand chimera bound hIgE with an affinity approaching that of ⑀⑀␥ (Table II) and is discussed below.
Evidently the loss of IgE binding function was related to the alteration of sequences in the interdomain interface, which implies a major role in IgE binding by domain 1. Furthermore, even though these sequences are derived from the same receptor as the IgE binding second domain, they do not provide the correct interactions unless in the context of an autologous first domain. Thus, it may be expected that IgE binding is dependent upon the AЈ strand-AЈB loop segment. To confirm that this change of function was directly related to the AЈ strand-AЈB segment and its impact on the D1/D2 interface, an additional chimera was created. This new chimera was constructed using Fc⑀RI-␣ (⑀⑀␥) as the template, (rather than ␥⑀␥) in which the AЈB strands and loop of Fc␥RIIa D1 were inserted into the corresponding position in Fc⑀RI. This ⑀(AЈB␥)⑀␥ chimera was expressed on the cell surface as measured by mAb47, but mAb15-1 again failed to detect its epitope in the IgE-binding site (Tables II and III). Moreover, the chimera did not bind monomeric hIgE or complexed moIgE, confirming that the interdomain interface has an essential role in the interaction of receptor with IgE.
The inability of the ⑀(AЈB␥)⑀␥ chimera to bind IgE is not an effect of distortion of D1, because a separate monoclonal antibody (mAb54), which binds within the BC strand region of Fc⑀RI D1, also binds to this chimera. Thus, on the basis that the D2 mAb47 and D1 mAb54 bind the receptor, and the loss of the mAb15-1 epitope, the effect of the ⑀(AЈB␥)⑀␥ mutation on IgE binding is related directly to an impact of the D1/D2 interface on the IgE-binding site.
Identification of Crucial Residues within the D1 Interface of Fc⑀RI-␣-To further define the role of the interdomain interface, two residues of the AЈ strand Fc⑀RI, Arg 15 and Phe 17 , that have substantial interactions within the interface, were mutated to alanine (R15A and F17A). In addition, Asn 14 , with backbone-backbone interactions across the interface was mutated, also to alanine. Both the R15A and F17A mutants were recognized by mAb47 and also by two mAbs with epitopes in Fc⑀RI D1, mAb54 and mAb3B4 (Table III), and neither R15A nor F17A were detected by mAb15-1 the hIgE-binding site specific antibody 15-1. Both point mutants displayed a dramatic reduction in IgE binding, implying that these mutants had altered IgE binding characteristics. The R15A mutant failed to bind monomeric mouse (data not shown) or hIgE or moIgE complexes (Table II). However, the second point mutant, F17A, was able to bind moIgE complexes but showed a substantial reduction in affinity when binding monomeric hIgE (Table II).
The alanine mutants were modeled and compared with the homology model of Fc⑀RI to determine the possible effects of mutation. In the Fc⑀RI model, Arg 15 extends outward toward solvent, whereas in the Fc␥RII crystal structure (Fc␥RII, Asn 15 ) it is constrained within the interface and oriented more toward D2. Asn 15 also forms an hydrogen bond with the Leu 90 backbone carbonyl in the Fc␥RII crystal structure. No such hydrogen bond is formed in the Fc⑀RI model with the distance between Leu 90 :c and Arg 15 :c␥ being 4.75 Å. Arg 15 participates in hydrophobic (van der Waals') contacts with Leu 89 , Phe 84 , and Leu 165 (Figs. 3 and 4) in both the x-ray structure (8) as well as the model, but the interactions with Leu 165 are lost, whereas those with Phe 84 are severely reduced in the R15A mutant model structure. Furthermore, in the Fc⑀RI model, the Glu 82 carboxylate is parallel to the guanidinium of Arg 15 , and the Arg 15 :c and Glu 82 :c␦ are 4.2 Å apart. If Arg 15 and Glu 82 exist in ionized forms in Fc⑀RI, this would lead to substantial loss of coulombic stabilization in the R15A mutant. The loss of fundamental interactions in the R15A mutant would result in destabilization of the interface and consequently the IgE-binding site above. This is consistent with the analysis of an R15L point mutant, which removes the positive charge of arginine while maintaining a similar size and displays a total loss of both hIgE and moIgE binding.
There is hydrophobic or van der Waals' contact between Phe 17 and Trp 110 in the Fc⑀RI model, which is consistent with the published structure (8). This is significant because Trp 110 is a principal residue in the B/C loop previously defined as a major contributor to the IgE-binding site (1,11,25). There are also hydrophobic contacts between Phe 17 and Leu 88 , Leu 89 , Asp 86 , and His 108 in Fc⑀RI. All of these contacts are lacking in the F17A mutant, and it is feasible that their loss would cause considerable distortion of the D1/D2 interface, as well as the binding site. The AЈB region is sensitive to change, and the presence of Arg 15 as well as Phe 17 is insufficient to allow IgE binding. This is indicated by the analysis of the ␥(AЈB⑀)⑀␥ chimera where the AЈB sequence of Fc⑀RI (NRIFKGEN) placed in ␥⑀␥, that is D1 Fc␥RIIa but D2 of Fc⑀RI surprisingly failed to bind IgE (Table I). Thus, the interface clearly maintains a series of complex interactions that work collectively to allow binding of IgE.
From the model structure and the contacts listed above, Phe 17 appears to lie beneath the IgE-binding site and has a a Receptor expression was determined by flow cytometry using mAbs. Expression was scored on a scale of 1-4 with maximum expression (4) determined after subtraction of background values (fluorescein isothiocyanate-labeled FabЈ 2 sheep anti-mouse Ig). 3 ϭ 60 -80% of maximum, 2 ϭ 40 -60% of maximum, and 1 ϭ less than 40% of maximum.
b Not detected. c ND, not determined.
critical function in maintaining organization of the linker region between the D1 G strand and D2 A strand. The linker, at the membrane distal portion of the interface, effects the display of the two domains and the ligand-binding region. Arg 15 , which plays a more crucial role in maintaining IgE binding, lies closer to the membrane. To test the possibility that distance from the linker may be a factor in determining the magnitude of the effect of mutation, Asn 14 was mutated to Ala (Table II). The N14A mutation has less effect on the binding of hIgE or moIgE, as the Fc⑀RI model suggests by the single backbone interaction of Asn 14 with Ala 92 across the interface. The analysis of these point mutants would imply that maintenance of the presentation of the ligand-binding site in Fc⑀RI is dependent upon the structure of the D1/D2 interface, which lies below the binding site, and that Arg 15 and Phe 17 are critical residues in this interaction. Is Arg 15 a Contact Residue Involved in IgE Binding?-The loss of binding by the ⑀(AЈB␥)⑀␥ chimera and the R15A and R15L point mutants could also suggest a possible hIgE contact role for the Arg 15 residue. However it is more unlikely that Arg 15 is not a contact residue because firstly, it is distant from the ligand-binding region, which is exposed to solvent on the superficial surface of the receptor (Fig. 1). Secondly, peptide inhibition 2 and mutagenesis analysis (23) have separately placed the mAb15-1 epitope close to the IgE-binding site, and mutations within, or expressed within, the D1/D2 interface have caused loss of binding of both IgE and mAb15-1 independently (23). This would confirm that the D1/D2 interface is structurally important in the presentation of the IgE-binding site and that mutations within the interface are sufficient to destroy the structure of this region. Thus, the exchange to alanine causes distortion of the receptor and not necessarily removal of a critical binding residue. Thirdly, the complete first domain of Fc␥RIIa can be substituted for the first domain of Fc⑀RI (which replaces Arg 15 with Asn) while maintaining IgE binding, although with a 2-fold loss of apparent affinity.
Substitution of the complete first domain of Fc␥RIIIA for the first domain of Fc⑀RI maintains the Arg 15 residue, and this chimera retains the ability to bind both human and mouse IgE with an equivalent affinity to that of the wild-type receptor (4). The presence of the entire Fc␥RIIa D1 may stabilize the interface region in the ␥⑀␥ chimera and compensate to some extent for the loss of the Arg15 residue. The presence or absence of this critical Arg 15 residue may also resolve previously unexplained discrepancies between studies using Fc␥RIIIA D1/ Fc⑀RI D2 and Fc␥RIIa D1/Fc⑀RI D2 chimeras.
A recently reported S162A mutant in Fc⑀RI D2 (23) causes destruction of IgE binding. Ser 162 is highly conserved within Fc receptors, and in the Fc⑀RI homology model Ser 162 interacts with Leu 89 of the D2 A strand, which in turn interacts with Arg 15 of the D1 AЈ strand (Figs. 3 and 4). The Arg 15 residue has been shown above to be of importance in maintaining the D1/D2 interface, and thus it is possible that the ablation of ligand binding is caused by changes in this linkage. The ability of a point mutation distant from the IgE-binding site to effect sufficient distortion of the receptor to destroy IgE binding further defines the importance and sensitivity of the D1/D2 interface structure in relation to IgE binding.
The expressed, bound moIgE with an avidity similar to that of ⑀⑀␥, and showed a small but reproducible increase in affinity for hIgE compared with ␥⑀␥ (Table II). Thus, although the high affinity of the wild-type ⑀⑀␥ receptor was not totally restored by these chimeras, the increase in affinity would suggest a foundation role in the presentation of the IgE-binding site.
In the case of the A strand, this role is most likely to be the structural support of the interface. The N terminus of the A strand, and indeed the ⑀ receptor, is probably located close to the cell membrane. The A strand interacts with other residues within D1 via hydrogen bonds, both in backbone interactions with the B strand and Asn 74 of the FG loop. Conformation of the A strand would assist the display of the AЈ strand in the interface so that crucial residues, such as Arg 15 and Phe 17 , are appropriately presented.
The G strand of domain 1 abuts domain 2 directly via the G strand-A strand linker and across the D1/D2 interface; it is also involved in interactions with the AЈ strand within D1. The G strand of D1 is highly conserved between Fc⑀RI and Fc␥RIIa, with few differences between the interactions of the conserved amino acids. It is therefore surprising that introducing the G strand of Fc⑀RI into the ␥⑀␥ chimera results in alterations to IgE binding. The residues in the G strand that are not conserved between Fc⑀RI and Fc␥RIIa may contribute specifically to IgE binding affinity. Glu 82 and Phe 84 interact with D1 AЈ strand residues Asn 14 , Ile 16 , and Arg 15 , and Asp 86 with Phe 17 ; these latter two interactions are with residues shown above to be critical in maintaining the D1/D2 interface. It is therefore probable that interactions of the A and G strands as well as the AЈ strand of D1 effect a role in maintaining the interface between D1 and D2 and therefore have an indirect effect on IgE binding.
In both the Fc⑀RI model and Fc␥RIIa structure, Trp 87 at the D1/D2 junction interacts with Trp 110 , which is contained within the BC loop of D2, a crucial IgE-binding region (1,11). Residues adjacent to Trp 110 , (Arg 106 and His 108 ), also interact with amino acids of the D1/D2 interface, the AЈ strand of D1, and the A strand D2, thus maintaining links between the interface and the IgE-binding region. The conservation of these residues within the FcR probably contributes to the ability to substitute the D1 of Fc␥RIIa or Fc␥RIII for the D1 of Fc⑀RI and retain IgE binding.
The Homology Model of Fc⑀RI-A homology model of Fc⑀RI (Fig. 1) based on the recently solved structure of Fc␥RIIa (6) was employed to determine the structural basis of alterations in IgE binding by the Fc⑀RI-␣ chimeras. The two structures are, therefore, very similar with some small variation at the point of sequence disparity in the region of the CЈE loop of D1 ( Table I). The pattern of ␤ strands is the same in both Fc␥RIIa and the Fc⑀RI-␣ homology model, (as stated under "Experimental Procedures"); however, the arrangement of the loops appears to depend more on the positioning of amino acids such as proline in Fc␥RIIa, whereas in Fc⑀RI-␣ there are supplementary interactions between amino acids to preserve the loop structure. In this model of Fc⑀RI-␣, one face of the molecular surface, largely comprising the juxtaposed C/F/G strands of each domain, has an overwhelmingly negative electrostatic potential (Fig. 2), unlike the opposite face of the molecule. This marked disparity in the electrostatic potential between the two faces is not observed in the case of the Fc␥RIIa molecule (results not shown) and may be of biological significance. The negative surface of Fc⑀RI-␣ would tend to sit away from the cell membrane, and as a consequence maintain the binding sites in a membrane distal position, on the upper surface of the molecule. This supports the cell surface data (detailed in Refs. 12, 24, and 25) asserting that the domains are aligned with the membrane along the long axis as shown in Fig. 1 and do not project vertically from the membrane. The ␥⑀␥ chimera (see "Experimental Procedures") was also modeled and, as expected, was shown to have a similar structure to Fc␥RIIa and the Fc⑀RI-␣ homology model.
The Fc⑀RI-␣ homology model has the features described for the x-ray crystallographic structure of Garman and co-workers (8). First, all the N-linked glycosylation sites, including the three described in the crystal structure, are solvent exposed. Second, the interchain hydrogen bonds correlate well with those disclosed in Fig. 1D of the published Fc⑀RI structure. Third, the two domains are bent relative to each other, and the model has an interdomain angle of 52°, which is similar to that depicted by Garman et al. (8). Fourth, the IgE-binding loops of domain 2 defined by Hulett et al. (1,11) are in close proximity, distal to the membrane and exposed to solvent.
By utilizing the Fc⑀RI-␣ homology model and the models of the chimeras and mutants, the authenticity of the interactions within the D1/D2 interface and the effects of the mutations on IgE binding could be defined with greater precision and fidelity. In conclusion, these data suggest that the second domain of Fc⑀RI-␣ is the only domain involved in direct contact with the IgE ligand and that domain one has a structural function of great importance in maintaining the integrity of the domain interface and, through it, the ligand-binding site.