The IgG binding site of human FcgammaRIIIB receptor involves CC' and FG loops of the membrane-proximal domain.

Fc gamma receptors for the Fc part of IgG are the mediators for antibody effector functions. FcgammaRIII and FcgammaRII are low affinity receptors that, through the interaction with immune complexes, initiate a variety of immunological responses, such as phagocytosis, antibody-dependent cellular cytotoxicity, and release of inflammatory mediators. We set out to define the IgG binding site on human FcgammaRIII. We assumed that potential beta-turns in Ig-like domains are the most probable determinants for ligand binding, and chimeric FcgammaRIIIB/FcepsilonRI receptors as well as single residue mutants were constructed in these regions of FcgammaRIIIB. Substitution of four amino acids in the membrane-proximal domain (Gln126, Arg156, Lys162, Val164) resulted in decreased binding of human IgG1. Lys162 and Val164 were found also to be crucial for the interaction with the IgG-binding inhibitory monoclonal antibody 3G8. In a putative three-dimensional model constructed in this study, these residues map on the CC loop (Gln126), on F beta-sheet (Arg156), and on the FG loop (Lys162, Val164). Our data are consistent with the study about human FcgammaRII (Hulett, M. D., Witort, E., Brinkworth, R. I., McKenzie, I. F. C., and Hogarth, P. M. (1994) J. Biol. Chem. 269, 15287-15293), suggesting that common structural determinants, i.e. FG loop or the GFC surface of the membrane-proximal domain, can be involved in interactions with IgG by both low affinity receptor classes FcgammaRII and FcgammaRIII.

Human class III receptors are represented by two isoforms that differ in their membrane anchors (8), expression patterns, and affinities to IgG. The transmembrane Fc␥RIIIA (9) receptor is expressed on NK cells, macrophages, on subsets of monocytes, and T cells in association with dimers of the ␥-chain of Fc⑀RI (10 -12) and/or the -chain of T cell receptor (13). Expression of the glycosylphosphatidylinositol-anchored Fc␥RIIIB iso-form is restricted to neutrophils (8, 14 -16). Tissue-specific expression of the two isoforms can be regulated at the transcriptional level (17,18). Fc␥RIIIA binds IgG1 and IgG3 complexes with higher affinity (K a Ϸ 3 ϫ 10 7 M Ϫ1 ) than the B isoform (K a Յ 10 7 M Ϫ1 ) and is able to interact with monomeric IgG (19 -21). Fc␥RIIIB is represented by two allelic forms, NA1 and NA2, which can be detected with certain specific CD16 monoclonal antibodies (22,23). NA1 and NA2 (24) differ in their glycosylation patterns (25) and in their ability to trigger phagocytosis by neutrophils (26,27).
The membrane-proximal domain is crucial for IgG binding by most of the Fc receptors studied, i.e. human Fc␥RII (1,29,30), human Fc⑀RI (30 -32), mouse Fc␥RII (33,34), and rat Fc⑀RI (31). The membrane-distal domain of Fc␥RIIIB, when fused to domains 3-5 of ICAM-1, did not react with IgG (35). Thus, to identify the IgG binding sites of Fc␥RIII, we focused on the second, membrane-proximal domain of Fc␥RIIIB. We predicted potential ␤-turn regions of the second Ig-like domain with the aid of the PC Gene program, and within these regions, amino acid residues were exchanged with the equivalent ones in the ␣-chain of the human high affinity receptor for IgE, Fc⑀RI (36). The resulting chimeric Fc␥RIII/Fc⑀RI receptors should show diminished IgG binding affinity, since the extracellular part of Fc⑀RI reveals significant amino acid identity (41%) with Fc␥RIIIB, but Fc⑀RI does not interact with IgG (30).
Chimeric Fc␥RIII/Fc⑀RI receptors revealed several regions on Fc␥RIIIB (amino acids 125-127, 152-158, 160 -163), substitution of which resulted in decreased interaction with IgG. Following single residue mutagenesis and molecular modeling of the receptor indicated that amino acids critical for ligand binding are apparently located on the loops connecting C and CЈ ␤-sheets (Gln 126 ) as well as F and G ␤-sheets (Lys 162 , Val 164 ) and on the F ␤-sheet (Arg 156 ) all on the second extracellular domain of Fc␥RIIIB. These residues may constitute one discontinuous binding area on the GFC ␤-sheet surface for the ligand, IgG1 or IgG3 complexes. This was further supported by the finding that the epitope for the IgG-binding inhibitory monoclonal antibody (mAb) 1 3G8 (37) was localized on the same FG loop of the membrane-proximal domain.

MATERIALS AND METHODS
cDNAs and mAbs-The wild-type cDNA for the Fc␥RIIIB gene, cloned into the pCDM8 expression vector (14), was provided by Dr. B. Seed (Harvard Medical School, Boston, MA).
IgG Complexes, Dimers, and IgE-Human IgG1 protein was purified from the sera of a myeloma patient by (NH 4 ) 2 SO 4 precipitation and Staphylococcus aureus protein A-Sepharose CL-4B (Pharmacia, Freiburg, Federal Republic of Germany) affinity column, prepared according to the manufacturer's instructions. IgG1 was aggregated by heating at 65°C for 20 min at the concentration of 1 mg/ml resulting in about 50% of complexed protein. The complexes were purified from monomers by size exclusion high pressure liquid chromatography on a TSK-3000 SWXL column (Supelco, Bad Homburg, Germany). The fractions of separated IgG complexes were further analyzed by agarose-SDS-polyacrylamide gel electrophoresis (0.5% agarose, 1.5% polyacrylamide, 0.1% SDS) that revealed only high molecular mass (Ն1000 kDa) complexes. Dimers and smaller polymers were not detectable by this analysis. hIgG1 dimers were prepared by chemical cross-linking with bis(sulfosuccinimidyl)-suberate (Pierce, Köln, Germany) according to manufacturer's instructions. Briefly, 10 mg of hIgG1 (10 mg/ml) in phosphate-buffered saline was incubated with 4 mg of bis(sulfosuccinimidyl)-suberate for 30 min at room temperature. The reaction was terminated by adding Tris buffer (pH 7.0) up to a final concentration of 50 mM. Excess of the reagents was removed by centrifugation with the Centricon 50 microconcentrator (Amicon, Witten, Germany). Crosslinked IgG1 dimers were purified from monomers and higher polymers by size exclusion chromatography on a Bio-Gel A-5m (BIO-RAD Laboratories, Munic, Germany) column (80 ϫ 1.6 cm). Fractions containing monomers, dimers, and polymers, respectively, were collected and concentrated using Centriprep 30 (Amicon). Purity of isolated dimers was monitored by high pressure liquid chromatography analysis on the size exclusion column TSK-3000 SWXL. IgG molarities in the binding assays with dimers and the heat-aggregated complexes were calculated by assuming the molecular mass of hIgG1 monomers (150 kDa; Ref. 40), regardless of the degree of polymerization.
Construction of Chimeric Fc␥RIIIB/Fc⑀RI and Mutant Fc␥RIIIB cDNAs-Chimeric (several amino acid residues) and mutant (single residues were replaced in Fc␥RIIIB with the equivalent ones in Fc⑀RI sequence) cDNAs were created by oligonucleotide-directed mutagenesis as described in the M13 in vitro mutagenesis kit (Amersham, Braunschweig, Germany). The 568-bp SphI-KpnI fragment (nucleotides 53-620) coding for the extracellular part of Fc␥RIIIB was subcloned into M13 mp18. This clone was used as a single-stranded template for mutagenesis with specific oligonucleotides of 30 -43 bp for chimeras and 18 -20 bp for mutant receptors. "Silent" restriction sites were designed into the oligonucleotides to facilitate screening for the respective mutants. The presence of mutations was further confirmed by polymerase chain reaction sequencing with the fmol DNA sequencing kit (Promega). Mutated DNA fragments encoding for the extracellular part were excised from M13 with SphI and KpnI and cloned back into the pCDM8 expression vector containing the additional parts of the original Fc␥RIIIB cDNA.
Chimeric Fc␥RIIIA/B receptor was constructed by cloning the 536-bp SphI-HincII fragment coding for the extracellular part of the Fc␥RIIIA gene into the Fc␥RIIIB gene.
IgG Binding Assays and Flow Cytometry-Transfected cells were harvested with 1 mM EDTA in phosphate-buffered saline and washed with phosphate-buffered saline containing 0.2% bovine serum albumin (Serva, Heidelberg, Germany), and 1 ϫ 10 5 cells were assayed for binding with serial dilutions of hIgG1 complexes (6 -300 nM) or dimers (0.06 -1.2 M) at 4°C for 1 h. Incubations with CD16 mAbs and hIgE were performed under the same conditions. Immunoglobulin fractions bound were quantified with fluorescein isothiocyanate-labeled secondary antibodies to human or mouse IgG (Dianova, Hamburg, Germany) and human IgE (ICN Biochemicals, Meckenheim, Germany) using a FACScan equipped with a single argon-ion laser (Becton Dickinson, Mountain View, CA).
Molecular Modeling of Fc␥RIIIB-Sequence multialignment between Fc␥RIIIB and various immunoglobulin molecules revealed that the receptor adopts an Ig-fold. Using the alignment, we defined a secondary structure of the receptor and predicted general structural features. The three-dimensional molecular model of the two domains of Fc␥RIIIB was developed based on homology with the known structure of a human myeloma IgG molecule (44). The x-ray crystallographic structure of this antibody Fab fragment is available in the Brookhaven Protein Data Bank. Molecular modeling, minimization of energy, as well as other structural manipulations were carried out using the BIOSYM program.

Mapping of the Regions Involved in IgG Binding on the
Membrane-proximal Domain of Fc␥RIIIB-Mutational analysis of the Fc␥RIIIB was based on two hypotheses. First, ␤-turns of the Ig-like extracellular domains of the receptor could be the most probable sites for ligand binding; second, dissimilar amino acids in the extracellular parts of the highly homologous receptors, Fc␥RIIIB and Fc⑀RI␣, might perform the basis for different ligand binding properties. The membrane-distal domain of Fc␥RIIIB was shown not to be involved in ligand binding (35). Therefore, only the second, membrane-proximal domain of Fc␥RIIIB was subjected to the ␤-turn probability prediction using the Chou and Fasman algorithm (45) in the PC Gene program (Fig. 1). In the amino acid sequence comparison of Fc␥RIIIB and Fc⑀RI␣, the putative ␤-turns revealed a relatively high degree of dissimilarity, and these regions in Fc␥RIIIB were chosen for substitution with the equivalent residues of the Fc⑀RI␣ sequence (Fig. 1). The amino acid exchanges in the chimeric receptors are indicated in Fig. 1 and Table I. The chimeras are designated according to the position of the residues substituted on the Fc␥RIIIB sequence.
The wild-type and chimeric receptors were transiently expressed in 293 cells, and the structural integrity of the receptors was assessed with a panel of six CD16 mAbs (3G8, DJ130c, Gran1, MEM-154, LNK16, and B88 -9). No significant differ- ences between the interaction with wild-type Fc␥RIIIB and five of the eight chimeras were observed with the mAbs (Table I), indicating that the mutations had caused no major alterations in the structure of these chimeric receptors. In contrast, the chimera 160 -163 was not recognized by the ligand-binding inhibitory mAb 3G8, and the binding with the mAbs Gran1, MEM-154, and B88 -9 was affected (about 30 -60% of wild-type receptor). mAb DJ130c reacted at the wild-type level with all the chimeras (Table I). Recognition of two chimeric receptors, 113-118 and 134 -138, was significantly decreased by most of the CD16 mAbs (Table I). Thus, we assumed extensive structural alterations in the membrane-proximal domains of these receptors.
The epitopes for the mAbs used have not been extensively studied yet. DJ130c is considered to bind the membrane-distal domain of Fc␥RIII (46). This explains the reactivity of the antibody with all the chimeras. The other five mAbs used are directed against the membrane-proximal domain (46). 2 3G8 is known to interfere with the ligand binding of the receptor (37,46). Thus, the complete loss of binding with 3G8 indicates a major alteration within the IgG binding site on the chimera 160 -163.
In binding assays of transfected 293 cells with hIgG1 com-plexes, a K d of 53.1 nM was calculated for the wild-type Fc␥RIIIB (Table I). The affinity of the receptor to covalently linked highly purified hIgG1 dimers was about 10 times lower (K d ϭ 4.6 ϫ 10 Ϫ7 M, Table I), presumably due to the lower valency of the dimeric ligand. A Fc␥RIIIA/B chimera that maintained the extracellular domains of Fc␥RIIIA in the glycosylphosphatidylinositol-anchored molecule revealed similar to wild-type Fc␥RIIIB affinities to IgG1 complexes (K d ϭ 50.7 nM, Table I and Fig. 2) as well as to dimers (K d ϭ 2.7 ϫ 10 Ϫ7 M, Table I). Hence, we suppose that the minor differences in the amino acid sequences of the extracellular domains of these two isoforms do not account for the higher IgG binding capacity of Fc␥RIIIA (20).
293 cells bearing the chimeras 97-99, 129 -131, and 147-148 showed binding affinities similar to wt Fc␥RIIIB, (Fig. 2). In contrast, chimeras 125-127 and 152-158 bound hIgG1 complexes as well as dimers at significantly lower levels (K d values for the chimeras are shown in Table I). Chimera 160 -163 had almost completely lost the capacity to bind IgG (Fig. 2). Since this chimera was still recognized by most of the CD16 mAbs except 3G8, we speculate that we have rather replaced residues functional in ligand and 3G8 binding than destroyed the overall structure of the second domain.
Binding affinities of the receptors to chemically cross-linked IgG dimers were lower than to-heat-aggregated complexes 2 A. Tamm and R. E. Schmidt, manuscript in preparation.

TABLE I Reactivity of the chimeric and mutant receptors with CD16 mAbs and hIgG1
Amino acids exchanged in chimeric and mutant receptors are numbered according to their location in the Fc␥RIIIB sequence. Relative expression levels of the mutants were determined considering the reactivity with CD16 mAbs. Reactivity to mAbs was calculated as % of mean fluorescence intensity measured for the wt Fc␥RIIIB transfectants with the same mAb in the same experiment and shown as e for 70 -100%, v for 40 -70, and f for 0 -40% of the reactivity with the wild-type receptor. The K d values were calculated by non-linear regression analysis using KaleidaGraph 2.0.2 software. The receptor-ligand binding curve and K d calculations were carried out, according to the formula y ϭ (B max ϫ (FL Ϫ FL neg ))/(K d ϩ (FL Ϫ FL neg )), where B max is the cell receptor number, FL is anti-IgG fluorescence of the given receptor at the provided ligand concentration, FL neg is anti-IgG fluorescence in the absence of IgG, and K d is the dissociation constant for receptor-IgG interaction. The values for B max are indicated as % of B max of wt receptor in each separate experiment. The values are presented as means of three separate experiments, and significantly lowered K d values are shown in bold. ND, not detected; NT, not tested. (Table I). However, the chimeric receptors reacted with hIgG1 dimers and large complexes in the same manner, i.e. the chimeras 127-127, 152-158, and 160 -163 revealed decreased binding capacities to both the ligands, whereas the other three chimeras (97-99, 129 -131, and 147-148) showed binding affinities comparable to the wild-type Fc␥RIIIB.
Chimeras 113-118 and 134 -138 did not bind IgG as well as most of the CD16 mAbs mapped to react with the membraneproximal domain. As discussed below, the loss of IgG binding was considered to result from the destroyed structure of the domain, and these chimeras were excluded from further mutational analysis.
None of the transfectants expressing chimeric receptors were able to interact with hIgE (data not shown). Evidently, the Fc⑀RI-derived amino acid residues in the chimeras are not directly involved in IgE binding or are not sufficient for the binding detectable in our assays.

Localization of the Putative IgG Binding Regions on the Molecular Model of the Membrane-proximal Domain of
Fc␥RIIIB-Since there has been no solution to the three-dimensional structure of Fc␥RIII yet, the Fc␥RIIIB NA2 allele was computer-modeled in this study. The model demonstrates the membrane-proximal domain of the receptor as a typical Ig-like molecule with seven antiparallel ␤-sheets that are arranged in ABE and GFC surfaces (Fig. 3). The CЈ ␤-sheet between the C and E sheets may be important for connecting both the surfaces. According to the model, amino acids that were shown to be crucial for ligand binding by chimeric receptors are located on the CCЈ loop (125-127) and on the FG loop (160 -163, Fig. 3). Residues replaced in the chimera 152-158 (Ser 152 , Phe 154 , Arg 156 , and Leu 158 ) were found to be located on the F ␤-sheet and conformationally placed between the two loops involved (Fig. 3). The model is in concordance with the data of our mutation analysis, demonstrating that these three regions may constitute one conformational binding site for IgG, located on the GFC ␤-sheet surface.
Amino acids substituted in the chimeras that did not have any effect on ligand binding (97-99, 129 -131, and 147-148) are positioned apart from the binding surface on this model. Residues 97-99 stretch between the two extracellular domains, and 147-148 are located on the E ␤-sheet. Both the regions are located on the ABE surface of the membrane-proximal domain, thus on the opposite side of the potential binding area (Fig. 3). Residues 129 -131 are placed on the CCЈ loop, opposite to the 125-127 region. Replacement of these amino acids seems not to affect the function of the neighboring putative binding residues 125-127, since the substitution did not influence the interaction with IgG (Table I, Fig. 2).
The overall structure of two of the chimeric receptors (113-118 and 134 -138) was considered to be disrupted according to the monoclonal antibody data. Location of the amino acids 134 -138 on the CЈ ␤-sheet that is presumably stabilizing the two ␤-sheet surfaces would explain the disruption of the structure of the domain when these residues were substituted. According to the model, the region 113-118 within the BC loop is also connecting the two ␤-sheet surfaces and, therefore, likely to be stabilizing the conformation of the domain. In addition, this region can be involved in generating the binding site due to the close proximity to the FG loop, the potential binding structure for IgG.
The GFCCЉ surface of the membrane-proximal domain of hFc␥RII has been reported to be crucial for IgG binding (1). The key residues on hFc␥RIIA (Fig. 1) are shown to locate on the FG loop of the domain on a molecular model of this receptor (1). The FG loop is also involved in ligand binding in mouse Fc␥RII (32,33). These data support our hypothesis that the FG loop of the membrane-proximal domain is the main binding determinant in Fc␥RIII, as was demonstrated by the loss of IgG binding capacity after substituting the residues 160 -163 on the putative FG loop.
Arg/His 131 influences the interaction of hFc␥RIIA with hIgG2 (47,48). The low responsive Fc␥RIIA isoform harbors histidine in this position and interacts with human IgG2 but not with mouse IgG1, whereas arginine in the position 131 in is corresponding to the His/Arg 131 in hFc␥RIIA (Fig. 1) and is putatively positioned on the CЈ ␤-sheet. Since hFc␥RIII does not bind IgG2 and, according to our model, the CЈ ␤-sheet remains conformationally distant from the GFC surface, we suggest the residues on this ␤-sheet are not directly involved in hIgG1 binding.
Detailed Structure Analysis of the Potential Binding Site-The three amino acid regions (125-127, 152-158, and 160 -163) demonstrated to be involved in IgG binding by chimeric receptors and the molecular model were subjected to further mutational analysis. Site-directed mutagenesis of 11 single residues was carried out (Table I). Similarly to the chimeras, each residue was substituted with the equivalent one of the Fc⑀RI sequence.
Within the region 125-127 (LQN-YKD), replacement of Leu 125 with Tyr and of Gln 126 with Lys resulted in lower IgG binding capacities (K d ϭ 182 and 116 nM for IgG complexes, respectively, Table I, Fig. 4A), comparable to that of the responsive chimera (K d ϭ 150 nM). In contrast, changing Asn 127 to Asp did not alter the binding affinity of the mutant receptor (Fig. 4A, Table I). Interaction with IgG1 dimers followed the same pattern, i.e. mutants Leu 125 and Gln 126 harbored decreased ligand binding capacities as compared to the wild-type receptor. Analyzing the receptor model, we supposed that only Glu 126 and not Leu 125 is directly involved in binding. We assume that replacing the small polar side chain of leucine with that of aromatic tyrosine in the mutant Leu 125 has led to immense structural changes in the CCЈ loop and, thus, to interrupted ligand binding.
A IgG binding site of Fc␥RIIIB has been described by Hibbs et al. (35) who identified a continuous binding region on the CCЈ loop of the second domain, Gln 126 -Tyr 133 , by alanine-scanning mutagenesis. In contrast to these data, our mutational analysis revealed no linearly continuous binding sites. In our experiments, replacement of the sequence Lys 129 -Asp-Arg 131 with Glu-Ala-Leu from Fc⑀RI did not influence IgG binding. Only conversions of the polar Leu 125 and Gln 126 (Fig. 6) to aromatic Tyr or positively charged Lys, respectively, disrupted ligandreceptor interaction, whereas changing the charge of the adjacent side chain (NH 2 group of Asn 127 to negatively charged O Ϫ of Asp) did not affect the interaction at all. We suppose that substitution of every amino acid to alanine by Hibbs et al. (35) affected the structure of the CCЈ loop and, hence, the binding capacity, even when the residues neighboring to functional ones were changed. Gly 128 was also recognized to be involved in IgG binding in that study. Glycine with only one hydrogen atom as the side chain can adopt a wider range of main chain conformations than other residues (28) and should play an important role in maintaining the structure of the CCЈ loop. This might as well explain the absence of ligand binding by mutated Gly 151 , also demonstrated in that study (35).
We constructed three mutants (Phe 154 , Arg 156 , and Leu 158 ) in the second region, 152-158. Analyzing the molecular model, Ser 152 was found to be placed apart from the other residues and excluded from further studies. Replacement of the positively charged Arg 156 (Fig. 4B, Table I)  significant effects on IgG binding. Substitution of only the Arg 156 with Thr disrupted receptor-ligand interaction to a considerably higher extent than did the replacement of the longer region, 152-158 (Table I, Fig. 4B). We assume that the single residue replacement destroyed also the possible ␤-sheet structure and ablated IgG binding capacity of the receptor. In contrast, when the whole F ␤-sheet was exchanged in the chimera 152-158, IgG binding was decreased, apparently, due to the missing of the residue of direct interaction with the ligand.
Based on the chimera 160 -163, the amino acid Ser 161 was converted to Gln, Lys 162 to Leu, Asn 163 to Asp, Val 164 to Tyr, and Ser 165 to Glu. Two of the substitutions, Lys 162 to Leu and Val 164 to Tyr, resulted in decreased ligand binding capacities of the respective receptors. The mutation of Val 164 almost abolished the ability to bind IgG (Fig. 4C, Table I). Both the mutants reacted weakly with the mAbs 3G8 and B88 -9, indicating that Lys 162 and Val 164 are located also within the epitopes for these antibodies. The three other mutant receptors, Ser 161 , Asn 163 , and Ser 165 , resembled the wild-type receptor (Fig. 4C, Table I), although in every instance a substantial change in charge and/or configuration of the side chain was generated. We conclude that Lys 162 and Val 164 , which are located according to the molecular model on the FG loop, are crucial for IgG binding, either being involved in direct interactions with the ligand or determining the conformation of the FG loop.
Binding Epitope of the mAb 3G8-The chimera 160 -163 was not recognized by mAb 3G8, the mostly used ligand-binding inhibitory antibody to CD16 (Table I). Hence, these residues within the FG loop of the second domain are likely to constitute the binding epitope for the antibody. Among the single residue mutants, 3G8 reacted poorly with Lys 162 and Val 164 (Table I, Fig. 5). Other mutants of this region Ser 161 , Asn 163 , Ser 165 , as well as Gln 126 and Arg 156 , which presented disrupted IgG binding capacities, reacted at a level comparable to that of the wild-type receptor (Table I, Fig. 5). 3G8 binding studies, performed with serial dilutions of the antibody, demonstrated that the interaction was more profoundly interrupted with the mu-tant Val 164 (Fig. 5). The same was observed for the binding of IgG1 (Figs. 5 and 4C, respectively) and for another mAb, B88 -9. B88 -9 is known to interfere with 3G8-Fc␥RIII binding (46), sharing apparently an overlapping epitope with 3G8. It is likely that the residues Lys 162 and Val 164 belong to a non-linear binding epitope for the inhibitory antibody as well as for IgG. On the other hand, the FG loop may compose a conformational binding site for the ligand as well as for the monoclonal antibodies 3G8 and B88 -9, and thus, alteration of the structure of the loop could influence the binding capacity of the receptor.
hFc␥RIII-IgG Interaction-In this study, the key residues for ligand binding of the Fc␥RIIIB are assumed to be located on the GFC face of the membrane-proximal domain (Figs. 3 and 6). The GFC surface remains distant from the first, membranedistal domain on our molecular model of Fc␥RIIIB. Using the space-filling presentation of the membrane-proximal domain, the side chains of the four residues implicated (Gln 126 , Arg 156 , Lys 162 , and Val 164 ) are exposed extending to the surface of the molecule (Fig. 6). According to the model, they could belong to one discontinuous binding site for IgG. Replacement of the amino acids Lys 162 and Val 164 on the FG loop demonstrated the most profound effect on ligand binding (Table I), indicating that the FG loop is the main binding determinant on Fc␥RIII. The same GFCCЉ surface and essentially FG loop were also shown to be critical for binding on the hFc␥RII receptors (1,5), providing an additional support to our data.
On IgG1 and IgG3, the natural ligands for these receptors, the lower hinge region (amino acids 233-237) has been identified to constitute a binding pocket for all Fc␥ receptors (49 -52). The lower hinge regions are different in IgG2 and IgG4, which explains also the failure of these subclasses to interact with Fc␥ receptors. Based on these data, we suppose that the same structural elements on the low affinity Fc␥ receptors, e.g. FG loops, are interacting with the same binding site, the lower hinge region on the IgG molecules. Although the amino acid residues found to be involved in ligand binding in Fc␥RIIIB are not conserved among Fc receptors (Fig. 1), we think that the FG loops of the membrane-proximal domains of different Fc␥ receptors constitute similar binding structures for IgG, whereas different single residues are involved in direct interactions.
The extensive amino acid sequence homology between Fc⑀RI and Fc␥ receptors suggests that a similar folding pattern might be adopted by these receptors, and the FG loops could be even more widely used as an interaction site with immunoglobulins.
A relatively higher affinity to IgG has been reported for the Fc␥RIIIA receptor (19,20). The minor differences in the extracellular polypeptide sequences of the Fc␥RIII A and B were shown not to be responsible for the improved binding capacity of the A isoform, since the chimera Fc␥RIIIA/B harboring the extracellular part of the A isoform in the B gene bound hIgG1 complexes and dimers similarly to the Fc␥RIIIB. An additional binding site for Fc␥RIIIA has been proposed in the CH3 domain of IgG (53)(54)(55). Binding curves obtained in our experiments do not refer to the existence of a second binding site on Fc␥RIIIB. However, the possibility remains to be studied that on NK cells, which express Fc␥RIIIA in association with the ␥-chain of the Fc⑀RI, the receptor exposes different interaction characteristics with IgG and thus endures higher affinity to the ligand.
Understanding the molecular basis of the interactions between Fc␥ receptors and immunoglobulins is of great importance, since the use of antibodies as therapeutical agents is increasing. The low affinity Fc␥ receptors are also believed to play an important role in inducing antibody-mediated inflammation (56,57). Several chronic inflammatory diseases like rheumatoid arthritis and leukocytoclastic vasculitis are linked to constant presence of antigen-antibody complexes and con- Amino acid residues putatively involved in IgG binding are colored. The N (Leu 93 ) and C termini (Ile 172 ) of the domain are indicated by cyan. The computer model was generated based on the previously described structure of a IgG Fab fragment (44).