The structure of a human type III Fcgamma receptor in complex with Fc.

Fcgamma receptors mediate antibody-dependent inflammatory responses and cytotoxicity as well as certain autoimmune dysfunctions. Here we report the crystal structure of a human Fc receptor (FcgammaRIIIB) in complex with an Fc fragment of human IgG1 determined from orthorhombic and hexagonal crystal forms at 3.0- and 3.5-A resolution, respectively. The refined structures from the two crystal forms are nearly identical with no significant discrepancies between the coordinates. Regions of the C-terminal domain of FcgammaRIII, including the BC, C'E, FG loops, and the C' beta-strand, bind asymmetrically to the lower hinge region, residues Leu(234)-Pro(238), of both Fc chains creating a 1:1 receptor-ligand stoichiometry. Minor conformational changes are observed in both the receptor and Fc upon complex formation. Hydrophobic residues, hydrogen bonds, and salt bridges are distributed throughout the receptor.Fc interface. Sequence comparisons of the receptor-ligand interface residues suggest a conserved binding mode common to all members of immunoglobulin-like Fc receptors. Structural comparison between FcgammaRIII.Fc and FcepsilonRI.Fc complexes highlights the differences in ligand recognition between the high and low affinity receptors. Although not in direct contact with the receptor, the carbohydrate attached to the conserved glycosylation residue Asn(297) on Fc may stabilize the conformation of the receptor-binding epitope on Fc. An antibody-FcgammaRIII model suggests two possible ligand-induced receptor aggregations.

Fc receptors, which are expressed on the majority of hematopoietic cells, play important roles in antibody-mediated immune responses. The binding of antigen-bound immunoglobulins (Ig) to Fc receptors activates their effector functions and leads to phagocytosis, endocytosis of IgG-opsonized particles, as well as antibody-dependent cellular cytotoxicity. The three major types of Fc receptors are Fc␥, Fc⑀, and neonatal Fc receptors. Except for the neonatal Fc receptor and Fc⑀RII (CD23), which are related structurally to class I major histocompatibility antigens and C-type lectins, respectively, all other known Fc receptors are members of the immunoglobulin superfamily (1,2). Among them, Fc␥RI and Fc⑀RI 1 are high affinity Fc receptors for IgG and IgE, respectively, with dissociation constants ranging from 10 Ϫ8 to 10 Ϫ10 M. All other receptors for IgG, such as Fc␥RII and Fc␥RIII, are low affinity receptors with dissociation constants ranging from 10 Ϫ5 to 10 Ϫ7 M (3)(4)(5). In addition to variations in affinity, each receptor displays distinct IgG subtype specificities. Unlike the high affinity receptors that can bind monomeric antibodies, the low affinity receptors preferentially bind to and are activated by immune complexes.
Human Fc␥RIII exists as two isoforms, Fc␥RIIIA and Fc␥RIIIB, that share 96% sequence identity in their extracellular immunoglobulin-binding regions. Fc␥RIIIA is expressed on macrophages, mast cells, and natural killer cells as a transmembrane receptor. In contrast, Fc␥RIIIB, present exclusively on neutrophils, is anchored by a glycosyl-phosphatidylinositol linker to the plasma membrane. Although Fc␥RIIIA associates with the immunoreceptor tyrosine-based activation motif containing Fc⑀RI ␥-chain or the T cell receptor -chain for its signaling, Fc␥RIIIB lacks a signaling component. Nevertheless, it plays an active role in triggering Ca 2ϩ mobilization and in neutrophil degranulation (6,7). In addition, Fc␥RIIIB, in conjunction with Fc␥RIIA, activates phagocytosis, degranulation, and the oxidative burst that leads to the clearance of opsonized pathogens by neutrophils. A soluble form of Fc␥RIIIB was reported to activate the CR3 complement receptordependent inflammatory process (8).
The Fc binding region on Fc␥RII and Fc␥RIII has been identified through the work of chimeric receptors with Fc⑀RI as primarily the membrane proximal domain, including both the BC and FG loops. Further site-directed mutations have revealed several residues of the receptor critical to Fc binding (9 -11). Similar regions on the ␣-chain of Fc⑀RI were also identified to be critical for IgE binding affinity (12). The receptor binding site on Fc has been located through the construction of chimeric IgG molecules and mutational analysis at the lower hinge region, residues located in the hinge region between the C H 1 and C H 2 domains and immediately adjacent to the N terminus of the C H 2 domain of IgG (13)(14)(15). In particular, residues 234 -238 (Leu-Leu-Gly-Gly-Pro) of the lower hinge of IgG1 have been implicated in the receptor binding. The corresponding region of IgE has also been implicated in the Fc⑀RI binding (16). Apart from the lower hinge region, a few residues on the C H 2 domain of an IgG2b were also suggested to interact with the receptor (17). However, with the exception of the neonatal Fc receptor, the molecular recognition between the Fc receptors and Fc remains to be elucidated (18).
The recent crystal structures of Fc⑀RI␣, Fc␥RIIA, and Fc␥RIIB have each revealed a conserved Ig-like structure, with particularly the small hinge angle between the two Ig-like domains, which is unique to the Fc receptors (19 -21). We report here the crystal structure of a human Fc␥RIII in complex with the Fc portion of a human IgG1 determined from two crystal forms.

EXPERIMENTAL PROCEDURES
Protein Expression, Purification, and Crystallization-The extracellular part of the human Fc␥RIIIb receptor, residues 1-172, was expressed as Escherichia coli inclusion bodies and then reconstituted in vitro as described previously (22). Fc fragments of human IgG1 antibody were prepared by the papain digestion as previously reported (23,24).
The complex of Fc and Fc␥RIII was prepared by mixing both components in a 1:1 molar ratio and concentrating to 8 -15 mg/ml for crystallization. Single crystals of orthorhombic and hexagonal forms were obtained by vapor diffusion in hanging drops at room temperature under slightly different crystallization conditions. Rod-shaped crystals of the orthorhombic form were grown from 10% polyethylene glycol 4000 and 50 mM Hepes at pH 6.5. They appeared after 2-3 days and grew to an average size of 0.05 ϫ 0.05 ϫ 0.2 mm in ϳ2 weeks. Crystals of the second form, hexagonal bipyramids, were crystallized from 5% polyethylene glycol 6000 and 50 mM Hepes at pH 6.0. Crystals were first observed after 4 -5 days, and reached a maximum size 0.15 ϫ 0.15 ϫ 0.4 mm after 1 month.
Structure Determination-After briefly soaking in precipitant solutions containing 25% glycerol, crystals were flash frozen at 100 K. X-ray diffraction data from single crystals of both crystal forms were collected using an ADSC Quantum IV charge-coupled device detector at the X9B beam line of the National Synchrotron Light Source at the Brookhaven National Laboratory and processed with HKL2000 (25). The hexagonal crystals belong to space group P6 5 22 with cell dimensions a ϭ b ϭ 114.9 and c ϭ 301.4 Å and diffract to 3.5 Å. Due to the large unit cell dimension along c in the hexagonal crystals, data were collected at a 300-mm crystal to detector distance with a small oscillation angle of 0.2°. The orthorhombic crystals of space group P2 1 2 1 2 1 with cell dimensions of a ϭ 76.4, b ϭ 102.8, and c ϭ 123.3 Å diffracted to 3.0 Å. Both crystal forms contain one molecule of Fc␥RIII and one molecule of Fc in the asymmetric unit.
The structure of the Fc␥RIII⅐Fc complex in both crystal forms was determined by molecular replacement. Polyalanine models of Fc (PDB accession number 1FC1) and Fc␥RIII were used in rotation and translation searches using 15-4 Å data for both crystal forms. An I/(I) cutoff of 2.0 was used for hexagonal data during search procedures. The position of the Fc molecule was determined in both crystal forms using AmoRe (26). Rotation and translation searches using data from the orthorhombic crystal resulted in an unambiguous solution with a correlation coefficient (CC) of 39% and an R-factor (R F ) of 51% (CC ϭ 51% and R F ϭ 49% after rigid-body refinement in AmoRe). Molecular replacement searches using the hexagonal crystal data yielded a solution for the Fc from the third highest rotation solution that became the highest ranking translation solution with CC ϭ 38% and R F ϭ 52% (CC ϭ 46% and R F ϭ 50% after rigid-body refinement in AmoRe). The position of Fc␥RIII was determined in both crystal forms using the program EPMR (27) with a polyalanine model of Fc␥RIII and the position of the Fc molecule fixed. Clear solutions were obtained for both crystal forms with CC ϭ 56% and R F ϭ 48% for orthorhombic crystal and CC ϭ 55% and R F ϭ 48% for hexagonal crystal, respectively. After rigid-body refinement of individual domains of the Fc␥RIII⅐Fc complex modeled as polyalanine using CNS (28) most side chains had clear electron density into which side chains were built in. Disordered side chains lacking electron density were built with occupancies set to zero. The positional and grouped B-factor refinement was carried out using maximum likelihood as a target function with CNS version 0.9. Model adjustments and rebuilding were done using the program O (29). Carbohydrate molecules were added manually using 2F o Ϫ F c electron density maps contoured at 1.0 and refined. The final model includes residues 5-172 of Fc␥RIII, residues 235-444 for one chain of Fc, and residues 233-443 for the other chain of Fc.

RESULTS AND DISCUSSION
Overall Structure of the Complex-Crystals of a human Fc␥RIII receptor in complex with a human Fc fragment of IgG1 were grown in two forms under different conditions. The orthorhombic crystals belong to the space group of P2 1 2 1 2 1 and diffract to 3.0-Å resolution, whereas the hexagonal crystals have P6 5 22 space group symmetry and diffract to 3.5-Å resolution. The structure of the complex was determined by molecular replacement in both forms and refined to their resolution limit. The final R-factors are R cryst ϭ 23.0% and R free ϭ 28.9% for the orthorhombic form and R cryst ϭ 24.9% and R free ϭ 32.6% for the hexagonal form, respectively (  contents (57% in the orthorhombic and 64% in the hexagonal form), both crystals contain one Fc␥RIII and one Fc molecule in each asymmetric unit, suggesting a 1:1 stoichiometry for the binding between the receptor and Fc (Fig. 1). This is consistent with earlier binding studies using non-equilibrium and equilibrium gel filtration experiments (22). The conformation of the Fc␥RIII⅐Fc complex is essentially identical in both crystal forms, including the conformation of the visible carbohydrate moieties on the Fc fragment ( Fig. 2A).
The Structure of Fc␥RIII-The structure of Fc␥RIII in both the orthorhombic and the hexagonal crystal forms can be readily superimposed with the structure of ligand free receptor resulting in r.m.s. differences between the individual domains of 0.6 -0.8 Å among all C␣ atoms ( Fig. 2B) (22). The hinge angle between the N-terminal (D1) and the C-terminal (D2) domains is 60°, which is slightly larger than the 50°value observed in the ligand free receptor. However, no significant change in the receptor conformation is observed upon complex formation (Fig. 2B).
The Structure of Fc-The Fc fragment of an IgG1 antibody comprises two identical chains (A and B), and each consists of two C1-type immunoglobulin domains, C H 2 and C H 3. The overall shape of the Fc fragment resembles that of a horseshoe with the two C H 3 domains packing tightly against each other at the bottom of the horseshoe and the C H 2 domains held apart by carbohydrate moieties attached to the glycosylation site Asn 297 from both chains forming the opening of the horseshoe. Well defined electron density throughout the Fc allowed for unambiguous tracing of residues Leu 234 to Ser 444 in chain A and Pro 232 to Leu 443 in chain B of Fc, including the lower hinge regions, Leu 234 -Pro 238 . The structure of the Fc fragment in complex with Fc␥RIII does not differ significantly from that observed in the structures of an unbound Fc fragment and a murine intact IgG2a antibody (30, 31) (Fig. 2C). However, the 2-fold symmetry relating the two chains of Fc in other unligated Fc structures, is no longer retained in the structure of the complex. The horseshoe-shaped Fc is slightly more open at the N-terminal end of the C H 2 domains in the Fc␥RIII⅐Fc structure compared with other known structures of Fc. The hinge angle between C H 2 and C H 3 domains of chain A (Fc-A) is 95°and 100°in the orthorhombic and the hexagonal crystals, respectively, ϳ10°larger than the corresponding angle of chain B (Fc-B) and the 84°-89°angle observed in all structures of ligand-free Fc (Fig. 2C).
The Interface between Fc␥RIII and Fc-The receptor binds to Fc at the center of the horseshoe opening making contacts to the lower hinge regions of both A and B chains of Fc (designated here as Hinge-A and Hinge-B, respectively, for the lower hinges of Fc-A and Fc-B) (Fig. 1). Such binding breaks down the dyad symmetry of the Fc, creating an asymmetric interface whereby the identical residues from Hinge-A and Hinge-B interact with different, unrelated surfaces of the receptor. Furthermore, it excludes the possibility of having a second receptor interacting with the same Fc molecule, resulting in a 1:1 stoichiometry for the receptor⅐Fc recognition. The structural implications of the activation of Fc receptors is profound. Particularly, the 1:1 receptor⅐Fc binding stoichiometry highlights the importance of antigen in the receptor aggregation. In contrast to the high affinity Fc␥RI and Fc⑀RI receptors, the binding of immunoglobulins to Fc␥RIII in the absence of antigen does not lead to receptor aggregation. It can be argued that a 1:1 receptor-ligand stoichiometry ensures the need for antigens in forming the receptor aggregation by eliminating the possibility of Fc-mediated receptor aggregation as suggested in a 2:1 stoichiometry. Precluding receptor aggregation mediated by Fc alone also eliminates the potential deleterious effect of antibodies whose concentration in vivo are often much higher than that of antigen.
The receptor⅐Fc complex buries ϳ1453 Å 2 of solvent-accessible area (Fig. 3A). The interface between Fc␥RIII and Fc molecules shows poor shape complementarity with a mean shape correlation statistic of 0.53 (32), less than those between T-cell antigen receptor and Class I major histocompatibility complex molecules, between adhesion receptor CD2 and CD58, and between antibody and antigen complexes. On the receptor side, all the contacts to Fc are made exclusively through its D2 domain. The receptor D1 domain is positioned above the Fc-B and makes no contacts with Fc (Fig. 1A). The interface of the complex consists of the hinge loop between the D1 and D2 domain of the receptor, the BC, CЈE, and FG loops, and the CЈ ␤-strand. The BC loop is positioned across the horseshoe opening making contact with residues of both Hinge-A and Hinge-B. The CЈ-strand is situated atop the Fc-A leading to the CЈE loop in contact with residues of Hinge-A. The FG loop of Fc␥RIII protrudes into the opening between the two chains of Fc (Fig.  1A). All three receptor loops (BC, CЈE, and FG) were implicated in Fc binding through earlier studies of chimeric Fc␥RII/Fc⑀RI receptors and through site-directed mutagenesis (9,10,12). On the Fc side of the complex, interactions with the receptor are dominated by residues Leu 234 -Pro 238 of the lower hinge (Table  II), consistent with results form earlier mutational studies (2). In particular, Hinge-A and -B together contribute ϳ60% of the overall receptor⅐Fc interface area (Fig. 3A). Interestingly, both Hinge-A and -B are found disordered in all known Fc structures to date, including the structure of an intact mouse IgG2a (30,(33)(34)(35). In contrast, residues of both Hinge-A and -B are clearly visible in the electron density maps from both crystal forms, suggesting that the binding of Fc␥RIII stabilizes the lower hinge conformation of Fc.
A combination of salt bridges, hydrogen bonds, and hydrophobic interactions contributes to the receptor⅐Fc recognition. Specifically, the interface between Fc␥RIII and Fc-A is dominated by hydrogen bonding interactions, whereas the hydrophobic interactions occur primarily at the interface between Fc␥RIII and Fc-B. There are a total of nine hydrogen bonds between the receptor and Fc, forming an extensive network involving both the main-chain and side-chain hydrogen bonding interactions (Fig. 3, B and C, and Table II). Seven hydrogen bonds are distributed across the receptor and Fc-A interface and two are at the receptor and Fc-B interface. Alanine mutations, such as the H134A mutant of Fc␥RII that resulted in the loss of two interface hydrogen bonds, have been shown to re- duce the receptor⅐Fc binding drastically, illustrating the importance of the interface hydrogen bonding network to the stability of the complex (10). A hydrophobic core is formed between Trp 90 , Trp 113 of the receptor, and Pro 329 of the C H 2 domain of Fc-B (Fig. 3C). This hydrophobic core extends further to include Val 158 , the aliphatic side chain of Lys 161 of the receptor and Leu 235 of Hinge-B. Mutations of both Trp 113 and Lys 161 in Fc␥RIII lead to the loss in receptor function (11,36). The side chain of Leu 235 on the Fc-B packs tightly against Gly 159 of the receptor leaving little space to accommodate any residues larger than Gly at this position. A G159A mutation on chimeric Fc␥RII resulted in the complete disruption of Fc binding, presumably due to the steric hindrance between Leu 235 and the ␤-carbon of the alanine mutant at position 159 (9). Of particular interest is Trp 113 of the receptor, which when mutated to Phe resulted in the loss of Fc binding. This residue is not only part of the interface hydrophobic core but also functions as a wedge inserted into the D1 domain to stabilize the acute interdomain hinge angle between D1 and D2 domains of Fc␥RIII. A W113F mutation would result in the loss of this wedge and lead to a disruption in binding by altering the orientation between the D1 and D2 domains.

Comparison of the Structures of Receptor⅐Fc Complexes-
Including the two crystal forms described in this work, there are a total of four Fc receptor and Fc complex structures available to date. A comparison among these structures reveals the conformation flexibility of this receptor⅐ligand complex and helps to explain the molecular interactions that differentiate the high from the low affinity receptors.
The two crystal forms of Fc␥RIII⅐Fc complexes determined in the present study are essentially identical and can be readily superimposed with a root mean square (r.m.s.) deviations of 1.1 Å among all C␣ atoms. The superposition of the hexagonal form onto the published Fc␥RIII⅐Fc complex resulted in an r.m.s deviation of 0.5 Å for all C␣ atoms (37) (Fig. 4A). An analysis of the interdomain hinge angles shows that the C H 2-C H 3 hinge angle is 10°larger in the structure of the Fc␥RIII⅐Fc complex than it is in the structure of an intact IgG2a antibody (35) or the structures of ligand-free Fc (30) (Table III) (Table II) that is not shown in the picture. Residues are colored by molecule. Important hydrogen bonds are represented by dotted lines. The BC, CЈE, and FG loops as well as the C and CЈ ␤-strands of Fc␥RIII, which play an important role in the interactions, are labeled. Some secondary structure elements of Fc lying behind and not contributing to the binding shown as semi-transparent. Carbohydrate moieties have been omitted for clarity.
indicates a well defined receptor-ligand recognition free from conformational flexibility.
The comparison between the structure of the Fc␥RIII⅐Fc complex and that of the Fc⑀RI⅐Fc complex has provided further insight into the molecular basis of the receptor affinity (38). Overall, a similar mode of receptor-ligand recognition was observed in both the Fc␥RIII⅐Fc and the Fc⑀RI⅐Fc complexes with an r.m.s. deviation of 1.5 Å between all the C␣ atoms. In fact, most of the structural difference resulted from the small variation between the C H 2-C H 3 and C⑀3-C⑀4 interdomain hinge angles (Fig. 4, B and C). This angle is ϳ10°smaller in the Fc⑀RI⅐Fc complex structure, resulting in a slightly closed conformation of Fc compared with that of the Fc␥RIII⅐Fc complex.
Detailed structural analysis shows that the interface area buried in the high affinity Fc⑀RI⅐Fc complex (1850 Å 2 ) is 400 Å 2 more than that in the low affinity Fc␥RIII⅐Fc complex (1453 Å 2 ). This is primarily due to a more extensive interaction observed between the receptor and the non-lower hinge residues of Fc in the high affinity complex than in the low affinity receptor complex. Of the total interface area of the Fc, the lower hinge and non-lower hinge regions contribute 870 and 580 Å 2 , respectively, in the Fc␥RIII⅐Fc structure. The corresponding regions contribute 740 and 1110 Å 2 , respectively, in the Fc⑀RI⅐Fc structure. This results in approximately twice as much interface area contributed by non-lower hinge residues in the high affinity receptor⅐ligand complex than in the low affinity receptor⅐ligand complex. Structurally, the lower hinge of IgE-Fc adopts a very different conformation than that of IgG-Fc in their respective receptor complexes (Fig. 4D). This conformation difference may enable the high affinity Fc⑀RI to interact more extensively with its ligand.
Although the overall pattern of the receptor⅐Fc interactions, namely a preference for hydrogen bonding in the Fc A-chain part of the interface and hydrophobic contacts in the Fc B-chain part of the interface are preserved in both the Fc␥RIII⅐Fc and Fc⑀RI⅐Fc complexes, significant differences were also observed. First, there are more extensive hydrophobic interactions between Fc⑀RI and IgE-Fc than those between Fc␥RIII and IgG-Fc. Although the tryptophan-proline sandwich formed by Trp 90 , Trp 113 of Fc␥RIII and Pro 329 of Fc-B (the corresponding Trp 87 , Trp 110 , and Pro 426 residues in Fc⑀RI⅐Fc) is preserved in both structures, additional bulky residues, such as Trp 130 , found in both Fc⑀RI and IgE-Fc may also contribute to stronger hydrophobic interactions in the Fc⑀RI⅐Fc complex compared with that of the Fc␥RIII⅐Fc complex. Second, a more extensive network of hydrogen bonds and salt bridges exists at the Fc⑀RI⅐Fc interface compared with that of Fc␥RIII⅐Fc. Furthermore, the hydrogen bonds in the Fc⑀RI⅐Fc interface are formed mostly between the side-chain atoms whereas those in the Fc␥RIII⅐Fc interface are formed primarily between the mainchain atoms or between the main-chain and side-chain atoms. There are two salt bridges Lys 117 -Asp 362 and Glu 132 -Arg 334 observed between Fc⑀RI and Fc but only one, Lys 120 -Asp 265 , is conserved in the low affinity complex between Fc␥RIII and Fc.
Our results suggest that multiple interactions contribute to the observed receptor-ligand affinity difference and that the higher affinity recognition includes more extensive hydrophobic interface area as well as more prominent electrostatic interactions.
Conserved Receptor⅐Fc Binding Interface-Of the 13 receptor interface residues, four (Trp 90 , Trp 113 , Lys 131 , and Gly 159 ) are invariant among all human Fc␥ receptors (Fig. 5A). Three of them are also conserved in the ␣-chain of Fc⑀RI. Gly 159 in Fc␥ receptors is replaced with Trp in Fc⑀RI. Because Gly 159 is in close contact with Leu 235 from the lower hinge of Fc-B, replacement of this residue with Trp may result in the observed difference of the lower hinge conformation in IgE. Three other interface residues, Lys 120 , Tyr 132 , and Val 158 , are nearly invariant among all human Fc receptors. The limited variation observed can be easily modeled into the existing interface without creating steric hindrance. It is interesting that the interface salt bridge between Lys 120 and Asp 265 of the Fc appears to be absent in Fc␥RI but conserved in all other Fc␥ receptors and in Fc⑀RI. The other six interface residues, Ile 88 , Asp 129 , His 134 , His 135 , Arg 155 , and Lys 161 are less well conserved among the receptors. Of these, variation at His 134 and His 135 may result in conformational changes in the lower hinge region of bound Fc. Overall, key features of the receptor⅐Fc interface appear to be well preserved among all the Fc receptors with possible hinge conformational adjustment for each receptor⅐Fc pair. Of particular interest is the comparison between the interface residues of Fc␥RI and those of Fc␥RIII. The binding affinity of Fc␥RIII is at least 100-fold weaker than that of Fc␥RI. Among the receptor⅐Fc interface residues, only four are different between Fc␥RIII and Fc␥RI. These are Lys 120 , Tyr 132 , Arg 155 , and Lys 161 in Fc␥RIII and Asn 120 , Phe 132 , Ser 155 , and His 161 in Fc␥RI. It is, however, not clear if any of these residues contribute to the observed variation in binding affinity.
Fc Receptor IgG Subtype Specificities-Fc␥ receptors display IgG subtype specificities. In particular, human Fc␥RIII binds tighter to IgG1 and IgG3 than it does to IgG2 and IgG4. Most of the Fc residues in contact with the receptor are conserved among the IgG sequences (Fig. 5B, residues boxed in blue and  red), suggesting a conserved binding site for all human IgGs. These binding residues, with the exception of a Glu 269 to Asp replacement, are also conserved in murine IgG2a consistent with it being a ligand for human Fc␥ receptors. The sequence differences among the IgG subclasses exist primarily at the lower hinge region. First, hIgG2 has a Val and Ala at positions 234 and 235, respectively, instead of Leu and Leu as observed in IgG1 and IgG3, and a one-residue deletion at position 237 of the corresponding IgG1. Human IgG4 has a Phe at position 234 (Fig. 5B). In addition, IgG2 and IgG4 sequences contain a three-residue deletion relative to IgG1 at the N-terminal end of the lower hinge, possibly restricting the lower hinge conformation. The length of the lower hinge has been suggested as a factor in lower receptor binding affinity of IgG2 and IgG4 (2). Among the four IgG subtypes, the length of the hinge region is longest in IgG3 and shortest in IgG2 and IgG4 (three to four residues shorter than that of IgG1). The differences in both the amino acid composition and the length of lower hinge may contribute to the observed lower receptor binding affinity of IgG2 and IgG4. The Contribution of Carbohydrate to the Fc␥ Receptor⅐Fc Binding-Both Fc␥ receptors and antibodies are glycosylated in vivo. In contrast to the Fc fragment that displays only one conserved carbohydrate attachment site located at Asn 297 , the receptor glycosylation sites vary both in number and in location among different Fc␥ receptors. For example, the glycosylation sites on the C-terminal domain of Fc␥ receptors are located at residues Asn 162 and Asn 169 on Fc␥RIII, Asn 138 and Asn 145 on Fc␥RII and asparagines 138, 145, and 149 on Fc␥RI (Fig. 5A). The influence of glycosylation on the receptor⅐Fc binding kinetics and on the receptor function has been studied extensively using both the deglycosylated receptor and Fc (4,39). These studies demonstrated that the carbohydrate attached to Asn 297 of Fc have a significant impact on the receptor binding, whereas glycosylations on the receptors appeared less critical and perhaps have more of a modulating effect on the affinity. For example, the two neutrophil antigen A alleles of Fc␥RIIIB, NA1 and NA2, differing primarily in their carbohydrate contents, display a 2-fold difference in their affinity for IgG3.
The structure of the receptor⅐Fc complex reveals potential roles for carbohydrate in receptor⅐Fc recognition. The first is the potential role of glycosylation at Asn 297 in supporting the structural framework of the Fc. The Fc fragment used in this work was generated from a human IgG1 and is therefore glycosylated. Multiple carbohydrate moieties were visible in the electron density extending from Asn 297 of both chains of Fc toward each other into the inter-chain region, referred to as the carbohydrate core region. Asn 297 is located next to the receptor binding interface. The carbohydrate moieties, however, are orientated away from the interface making no specific contacts with the receptor. The glycosylation is thus unlikely to influence the receptor⅐Fc interface directly. However, the unique arrangement between the oligosaccharide moieties and the polypeptide chains of Fc makes it possible for the carbohydrate to affect the conformational stability of the receptor binding epitopes (40). Specifically, the spacing and the orientation between the two C H 2 domains may be influenced by the presence of sugar attachments (Fig. 1A). Because the binding of the receptor to Fc requires a particular orientation of the epitopes on both chains of Fc, it makes the receptor⅐Fc interface sensitive to the relative position and orientation of the two C H 2 domains.
A Model for Fc␥RIII-IgG Recognition-On a cell surface, the Fc receptor recognizes intact immunoglobulins. The presence of the Fab portion of antibody is likely to impose restrictions to the receptor⅐Fc recognition. To date, the only structure of an intact antibody available is that of a mouse IgG2a (31). Because Fc␥RIII also recognizes mouse IgG2a, a model of this receptor⅐antibody complex was generated by superimposing the Fc part of the current structure onto the Fc of the IgG2a (Fig. 6A). This receptor⅐antibody recognition model reveals that the receptor fits tightly and is nearly engulfed by the bound antibody.
Although the current structure offers an insight to antibody⅐Fc␥ receptor recognition, the mechanism of receptor activation, namely the antigen-driven receptor clustering, remains unknown. Two receptor clustering models can be proposed based on the current structural results, a simple avidity model and an ordered receptor aggregation model (Fig. 6, B and  C). The simple avidity receptor activation model assumes that the binding of oligomeric antigens by antibodies increases the FIG. 4. Superposition of the Fc␥RIII⅐Fc complex determined from the orthorhombic crystal form onto the structures of previously determined complexes Fc␥RIII⅐Fc (46) and Fc⑀RI⅐Fc (45). A, superposition of the Fc␥RIII⅐Fc complex determined from the orthorhombic crystal form (green and cyan for Fc␥RIII and Fc, respectively) and previously determined structure of Fc␥RIII⅐Fc complex (46) (orange and red for Fc␥RIII and Fc, respectively). B, superposition of the structure of Fc␥RIII⅐Fc (green and cyan for Fc␥RIII and Fc, respectively) and Fc⑀RI⅐Fc (orange and red for Fc⑀RI and Fc, respectively) complex (45). C, a definition of the hinge angles. D, an enlarged view of superposition of Fc␥RIII⅐Fc (green and cyan) and Fc⑀RI⅐Fc (red and orange) complexes in the interface area. The lower hinge regions are labeled. The view is identical to that in B.  avidity as well as the proximity of the receptors, which is sufficient to its activation. The ordered receptor aggregation model assumes that the binding of oligomeric antigens leads to the formation of an ordered receptor-ligand aggregation, which further stabilizes the activation complexes. Recent imaging studies on T cell and NK cell receptor activation processes suggest that the formation of the so-called immune synapse is an ordered event (41,42). These results favor the structured aggregation model rather than the simple avidity model, although the molecular organization of Fc␥ receptors during their activation remain to be determined. Recently, an ordered receptor⅐ligand aggregate was observed in the crystal lattice of a natural killer cell receptor in complex with its class I major histocompatibility complex ligand (43). Such a receptor⅐ligand aggregate is not observed in the two forms of the current Fc␥RIII⅐Fc crystals. However, a parallel receptor aggregate was observed in the crystal lattice of Fc␥RIII in the absence of Fc (22). A superposition of the current complex structure onto this lattice receptor aggregate suggests that the clustering model be compatible with the structure of a receptor⅐Fc complex (Fig. 6C).
Comparison of Fc␥ Receptor with Other Ligands of Fc-The Fc region of the IgG molecule possesses multiple recognition sites for different components of immune system, including Fc␥ receptors, neonatal Fc receptor (FcRn), rheumatoid factors (RF) and components of the complement system. In addition, it is also used as a ligand by staphylococcal proteins A and G. The structures of Fc complexed to FcRn, RF, protein A, and protein G are now known (30,33,34,44). The binding of Fc by Fc␥ receptors is characteristically different from all other known Fc ligands. First, the location of the Fc receptor binding site differs from those of neonatal Fc receptor, RF, and protein A. Although Fc␥ receptors bind to the lower hinge region of Fc between the C H 1 and C H 2 domains, FcRn, RF, protein A, and protein G bind to the joint region between the C H 2 and C H 3 domains of Fc. Second, Fc␥ receptors recognize Fc in an asymmetric fashion resulting in one receptor bound to both chains of Fc whereas all other ligands bind Fc in a symmetric fashion with each chain of Fc harboring an intact binding site (Fig. 7). The distinct binding site for Fc␥ receptors suggests that it is possible to bind Fc␥ receptors simultaneously with other ligands that recognize the C H 2-C H 3 joint region on the same Fc molecule. This raises the possibility of activation of multiple immune components by the same antigen-bound immune complex.
The recognition mode of binding to the lower hinge of Fc may evolve from the unique requirement of Fc␥ receptor signaling, namely the need to have 1:1 recognition stoichiometry and to be capable of discriminating the IgG subtypes. Both C H 2 and C H 3 domains of Fc are very conserved among the subclasses of IgGs. Even the C H 2-C H 3 joint region, which is involved in binding of other Fc ligands, has near identical sequences among the IgG subclasses (Fig. 5B). The lower hinge region of IgGs, in comparison, is more variable allowing subtype-specific recognition of the receptor. The conformation of the hinge region, however, is quite flexible compared with the C H 2 and C H 3 domains of Fc (35). This hinge flexibility, which enables the Fab arms to adapt to the shape and form of antigens, may in fact hinder the binding of Fc receptors. Interestingly, there are two conserved cysteine residues forming two disulfide bonds at the N-terminal end of the lower hinge. The presence of these disulfides may stabilize the lower hinge conformation while allowing sufficient flexibility at the upper hinge region. Finally, the binding to the lower hinge region of both chains of Fc allows the receptor to monitor the integrity of the antibody.
Receptor-IgG Recognition and Autoimmune Diseases-In addition to their normal cellular functions in host immunity, Fc␥Rs, in particular Fc␥RI and Fc␥RIII, also mediate the inflammatory responses generated by cytotoxic autoantibodies and immune complex triggered inflammatory disorders (45,46). They provide a critical link to autoimmune diseases, such as rheumatoid arthritis, hemolytic anemia, and thrombocytopenia. The structure of Fc␥RIIIB in complex with IgG1-Fc reveals the molecular interface of this receptor⅐Fc recognition and thus provides new possibilities for developing therapeutic reagents to block the activation of Fc receptors by autoantibodies. For example, the lower hinge sequence of Fc may be used to generate neutralizing antibodies that could block the binding of autoantibodies to Fc␥Rs. The peptides encompassing residues of the BC and FG loops of the C-terminal domain of Fc␥ receptors could also be used to develop neutralizing antibodies against the receptors. Finally, reagents that affect the glycosylation pathway may be used to affect the carbohydrate composition of Fc and thus the conformation of the receptor binding epitope.