The Interaction of FcαRI with IgA and Its Implications for Ligand Binding by Immunoreceptors of the Leukocyte Receptor Cluster¹

Immunoglobulin A is the major Ig of mucosal secretions and is also the second most abundant circulating Ig (1). It has a role as a first line of defense in protection from infection at the mucosa and as a second line of defense for the elimination of pathogens that have successfully eluded destruction at the mucosa (2). In addition to its role in host defense, IgA is involved in pathological immunity, especially in IgA nephritis (3, 4, 5, 6, 7, 8, 9).

In humans, the induction of cellular effector functions by IgA is dependent on the interaction with its specific receptor, FcαRI, also called CD89 (10, 11, 12). FcαRI is expressed on neutrophils, eosinophils, monocytes, and macrophages (10, 12, 13, 14, 15, 16, 17); Kupffer cells (2); and alveolar macrophages (18), and there have been several reports of possible expression on mesangial cells in the kidney (3, 4, 5, 6). IgA-dependant activation of cells via FcαRI is believed to be a key factor in host defense, and responses include phagocytosis, respiratory burst, degranulation, and cytokine release (2, 14, 15, 16, 17, 18, 19). Interestingly, despite the capacity to bind both serum and secreted IgA (9, 10, 11) the form of IgA also influences FcαRI-dependant responses, as serum, but not secreted IgA, triggers phagocytosis (2). In addition, FcαRI can act cooperatively with other arms of host resistance, including complement where CR3 is required for Ag-dependent cellular cytotoxicity via FcαRI (19). Recent studies including the use of FcαRI transgenic mice also suggest that it plays a major role in IgA nephritis (20).

Although FcαRI is clearly an Fc receptor, and like FcεRI, FcγRIIa, FcγRIIb, FcγRIIc, FcγRIIIa, and FcγRIIIb, is composed of two extracellular Ig-like domains (ectodomain 1 (EC1)³ and EC2), it is more closely related to members of the leukocyte receptor cluster: killer Ig-like receptors (KIR), Ig-like transcripts (Ig-like transcript-1 (ILT) or leukocyte Ig-like receptor), and the paired Ig-like receptors (21). The FcαRI shows greater amino acid sequence identity with members of the leukocyte receptor cluster (30–40%) compared with 20% with FcγRII, FcγRIII, and FcεRI. In addition FcαRI contains the 5-aa interdomain linker that is conserved in the leukocyte receptor cluster and is distinct from other FcRs, FcγRII, FcγRIII, and FcεRI, which have a 2-aa linker. This linker is essential for a relative spatial orientation of EC1 to EC2 of the KIR, which differs by 120° compared with the orientation of EC1 and EC2 of the FcγRII, FcγRIII, and FcεRI (22, 23, 24). In addition, the FcαRI gene maps within this cluster whose genes are closely linked on chromosome 19q13.4. Thus, it is likely that FcαRI would exhibit other differences from the other leukocyte Fc receptors. Indeed, in the limited analyses performed, comparison of the FcγRI:IgA interaction showed considerable differences from the well-defined FcγR:IgG and FcεRI:IgE interactions (25, 26). Unlike other Fc receptors, in FcαRI the ligand binding site appears to be in the first domain, not the second, and in IgA, unlike IgG or IgE, the receptor binding site is located at the interface between CH2 and CH3, not the lower hinge of CH2 as for IgG or its equivalent area in IgE Cε2 (27, 28).

Furthermore, in EC1 of FcαRI, histidine 85 and arginine 82 were identified as necessary for IgA binding and are located in the putative F-G region of EC1 (25). In other Fc receptors the F-G region of the second domain (EC2) is used as well as additional adjacent areas of EC2. In the study described herein additional areas of FcαRI were examined for their contributions to the binding of IgA. The F-G region was completely scanned for binding site residues by mutagenesis, and the B-C, C′-E, and N-terminal regions were also mutated. In addition, a homology model of the ectodomains of FcαRI was constructed using the related KIR structure as a template. Thus, knowledge of the FcαRI:IgA interaction may have broader implications for receptor-ligand interactions generally in the members of leukocyte receptor cluster on chromosome 19.

Two constructs encoding the two EC domains and the EC membrane proximal region of human FcαRI were generated by PCR of the pHuIgAR (11), using the polymerase Pwo (Roche, Castle Hill, Australia), and the forward primer oBW21 (5′-CCCGGGGAATTCCAGGAAGGGGACTTTCCC-3′) and either reverse primer oBW32 (5′-GGCCTAGGCCCATTCAGATCCTCTTCTGAGATGAGTTTTTGTTCTGCCCCGGGCCCGATCAAGTTCTGCGTCGTG-3′), which encodes a c-Myc tag for the rsFcαRI C-terminal, or reverse primer oBW35 (5′-GGCCTAGGGTGATGATGGTGATGATGTGAGCTGCTCCCGGGCCCGATCAAGTTCTGCGTCGTG3′), which encodes a six-histidine tag for the rsFcαRI C-terminal. The PCR product was cloned into the EcoRI and AvrII (New England Biolabs, Beverly, MA) sites of the Pichia expression vector pPIC-9 (Invitrogen, San Diego, CA), producing constructs pBAR62 and pBAR66, respectively. The vector pBAR66 was digested with SmaI and XbaI, liberating a fragment encoding the hexahistidine sequence. This fragment was ligated into the NotI (cut and then filled in using the Klenow fragment of DNA polymerase) and XbaI sites of pBAR62 to give pBAR151, a Pichia vector expressing rsFcαRI with a dual Myc and hexahistidine tag at the C-terminal. For expression in baculovirus, the PvuII/NotI fragment from pBAR 141, a pFastBac1 vector (Life Technologies, Gaithersburg, MD) encoding the ectodomains EC1 and EC2 of FcαRI (rsFcαRI) (25) was released and replaced with the PvuII/NotI fragment from pBAR 151. This construct (pBAR152) encoding Myc-histidine-tagged rsFcαRI was used as a template for alanine mutagenesis using a QuikChange mutagenesis kit (Stratagene, La Jolla, CA). Constructs were confirmed using a Thermosequenase dye terminator cycle sequencing kit and an ABI Prism 377 DNA sequencer (Applied Biosystems, Foster City, CA). The mutant rsFcαRs in the F-G region were Y81A, R82A, I83A, G84A, H85A, Y86A, and R87A, and those in the B-C region were Q29A, I31A, R32A, E33A, and Y35A. In the putative C′-E region, two mutants, R52A and R53A, were constructed. Normal and mutant rsFcαRI proteins were produced in baculovirus using the Fastbac system according to the manufacturer’s instructions (Life Technologies). Nickel agarose chromatography (Qiagen, Melbourne, Australia) was used to purify the recombinant soluble receptors as described previously (25). For mammalian cell surface expression of FcαRI, PCR was performed on the plasmid pHuIgAR (11), with the polymerase Pwo (Roche), using the forward primer oBW146 (5′-CCGAATTCCACGATGGACCCCAAAC-3′) and the reverse primer HT57 (5′-ACCTCCTCTAGATTACTTGCAGACACTTGG-3′) and the PCR product cloned into the vector pcDNA3 (Invitrogen) to construct pBAR234. Mutagenesis of pBAR234 to produce plasmids for expression of the Y35A, R52A, R82A, and R87A mutant receptors was performed as described above. The chimeric receptor with the N-terminal peptide of ILT1 was constructed by PCR of pBAR234 with the primers oBW198 (5′-CACCTCCCCAAGCCCACCCTCTCTGCCAAATCGAGTCCTG-3′) and oBW199 (5′-CCCTGCCTGTGCCTGAATCCTCTG-3′), using Turbo Pfu (Stratagene, La Jolla, CA) followed by phosphorylation and ligation of the linear PCR product by standard methods.

Transient expression in COS-7 cells was performed by transfection of 10⁶ cells with 5 μg of plasmid DNA and Lipofectamine Plus reagent according to the manufacturer’s instructions (Life Technologies). After 48 h the IgA binding activity of mutant FcαRI was measured by indirect immunofluorescence quantitated by flow cytometry. Cells (10⁵) were incubated with 5 μg/ml serum IgA (Calbiochem) for 30 min, washed with PBS containing 0.1% BSA, and incubated for an additional 30 min with FITC-labeled sheep Fab′₂ anti-human IgA (Silenus, Melbourne, Australia). Cells were washed, and fluorescence was measured using a FACSCalibur (Becton Dickinson, Melbourne, Australia). FcαRI expression was independently measured by immunofluorescence as described above, but using the FcαRI-specific mAb MY43 and FITC-labeled sheep Fab′₂ anti-mouse Ig (Silenus). The MY43 Ab was obtained from Medarex West (Lebanon, NH).

FcαRI binding was measured using a biosensor as described previously (25). Serum IgA (Calbiochem-Novabiochem, Alexandria, Australia) was coupled (600–1300 resonance U) to a BIAcore CM5 carboxyldextran biosensor chip (BIAcore, Melbourne, Australia) using the established carbodiimide-mediated amine reaction protocol. Binding assays were typically performed at flow rates of 10 μl/min using 20 mM HEPES, 150 mM NaCl, and 3.4 mM EDTA pH 7.4.

The sensograms for wild-type and mutant rsFcαRI binding to IgA were analyzed by taking the midpoint of each injection as the equilibrium binding response, even though the sensograms showed a small amount of additional binding that accumulated relatively slowly with increased time of injection. This small contribution of higher affinity binding was attributed to a small amount of aggregates in the receptor preparation. These equilibrium data were fitted (for the wild-type rsFcαRI) to a single binding site model, R = (B_max × [FcR])/(K_dapp + [FcR]), where R is the binding response, B_max is the response at saturation, [FcR] is the concentration of rsFcαRI, and K_dapp is the apparent K_d. The number of receptor binding sites (B_max) was constant in these experiments, as the same immobilized IgA was used to measure the binding of the normal rsFcαRI and the rsFcαRI mutants. Therefore, the value for B_max determined in the normal rsFcαRI binding analysis was set as a constant during the analyses of binding data to determine the K_d values for the mutant receptors. The measured affinities of the mutant rsFcαRI were compared with those of the wild-type rsFcαRI, and the significance of differences in affinity was determined using an unpaired t test with two-tailed p values. p < 0.05 was taken as significant, and p < 0.01 as very significant.

Different channels of the same BIAcore CM5 carboxyldextran biosensor chip were coupled with IgA and rsFcαRI as described, using carbodiimide chemistry. The binding of serum IgA to the immobilized receptor and that of soluble receptor to immobilized IgA were analyzed by global fitting of the sensograms using the program Biaevaluation 2.1 (BIAcore).

Nonredundant Protein Data Base (PDB) sequences were searched for homologues to the Fcα receptor ectodomain sequence using PSI-BLAST (29, 30) at the web site “Predicting protein 3D structures based on homologous sequence search” (http://dove.embl-heidelberg.de/3D) (30). A model of the extracellular domains of FcαRI (residues 6–195) was constructed using the coordinates of five KIR structures in the protein database (31) as templates. The PDB accession numbers of these were 1NKR (32), 2DLI, 2DL2 (33), 1EFX (34), and 1B6U. The sequence alignment obtained with PSI-BLAST (29) and the MODELLER (35) module of Insight II molecular modeling software (version 97.2, Molecular Simulations, San Diego, CA) was adjusted by inspection to maximize identities in putative strands. Ten models were constructed using MODELLER (35) without subjective intervention. An evaluation of the internal consistency of each predicted structure was obtained using the program PROFILES 3D, and the highest scoring structure was retained (36).

In our previous study two residues, R82 and H85 in EC1, were shown to be necessary for IgA binding (25) and are likely to lie in the putative F-G loop of this domain. In contrast, the binding site of other FcRs is located in EC2, but in this domain includes, in addition to the F-G region, the adjacent B-C and C-C′-E regions. To assess whether a similar configuration was retained in the EC1 of FcαRI, putative F-G, B-C, and C′-E regions were targeted for mutagenesis. In addition, as these regions may be expected to be close to the N-terminus in a typical Ig domain, the first 10 N-terminal residues of FcαRI were also mutated. All mutant proteins were then tested for their capacity to bind IgA. It should be noted that the assignment of these regions to residues was based on alignment to equivalent sequence of the related p58 KIR using the same secondary structure assignment of the p58 KIR (32). Hence, residues 81–87 were assigned to the putative F-G region, residues 29–35 to the B-C region, and C strand region and residues 52–53 were between the C′ and E strands.

Recombinant soluble normal and mutant FcαRI proteins (rsFcαR), consisting of both extracellular domains and membrane-proximal region fused to a dual c-Myc and hexahistidine tag, were produced using the baculovirus expression system and were purified by chromatography on nickel agarose (25). The purified recombinant normal and mutant soluble receptors were visualized by SDS-PAGE analysis as a tight cluster of bands with a M_r of approximately 40 kDa (Fig. 1).

There was some heterogeneity observed, which is probably related to glycosylation differences, as treatment with Endo-F reduced the size heterogeneity, although not completely (data not shown). Furthermore, no species smaller than the expected core polypeptide was observed, implying no extensive proteolysis of the proteins, and indeed all were detected by the anti-Myc Ab in Western blots (data not shown). Others have observed such carbohydrate heterogeneity for the expression of other Fc receptors using baculovirus systems (37).

Soluble wild-type and mutants of rsFcαRI in the F-G region were injected, at concentrations from 0.01 to 0.3 μM, over immobilized serum IgA, and the equilibrium binding responses were recorded using a BIAcore biosensor (Fig. 2). These data were fitted to a single site binding model, yielding values for the apparent K_d of the normal and mutant soluble receptors (Fig. 3,A). Mutation of six of the seven residues resulted in profound reductions in IgA binding (Figs. 2 and 3 A). The effect of the alanine substitutions ranged from an ablation of binding for the Y81A and R82A FcαRI proteins to large reductions in apparent affinity for, I83A (11-fold), G84A (19-fold), H85A (4-fold), and Y86A (3-fold). Only R87A had an insignificant (1.2-fold) reduction (p > 0.05) in its apparent K_d for IgA.

This analysis was performed for mutations in the adjacent B-C and C′-E regions. The substitutions Q29A, I31A, R32A, and E33A in the B-C region showed weak reductions in apparent affinity (1.3- to 1.6-fold), which were not significant (p > 0.05); however the Y35A mutant was essentially inactive (>100-fold reduced apparent affinity; p < 0.01; Fig. 3 B). The region between the C′ and E strands also contributed to IgA binding, since the R52A showed an 8-fold reduction in apparent affinity compared with wild-type FcαRI. However, R53 appeared to play no role, as IgA binding to R53A FcαRI, was indistinguishable from binding to wild-type FcαRI.

The results obtained using baculovirus-expressed soluble receptors were confirmed using mammalian cell surface expression of selected FcαRI mutants in COS cells (Fig. 4). Mutations of the B-C region (Y35), the C′-E region (R52A), and the F-G region (R82A, R87A) were introduced into the cDNA encoding the entire receptor, including the membrane-spanning and cytoplasmic regions. These mutants were then analyzed in a transient expression system at the surface of COS cells and were tested by flow cytometry for IgA binding using serum IgA and for expression at the cell surface using anti-FcαRI mAb. The mutation R82A in the F-G region and Y35A in the B-C region completely inactivated the IgA binding activity of the receptor. These proteins showed no detectable IgA binding (0.04 and 0.00% of cells stained positively for IgA) compared with 9.1% of cells transfected with the wild-type FcαRI staining positively for IgA binding, with a maximum mean fluorescent intensity of approximately 500 (Fig. 4). Thus, the IgA binding experiments with cell surface expression of FcαRI confirmed the results of the biosensor analysis using baculovirus rsFcαRI. The R52A mutant, unlike the foregoing mutants where IgA binding was completely lost, had only 8-fold reduction in IgA binding affinity in the biosensor assay. When tested in the FACS assay, 11.9% of the transfected cells stained positively for IgA binding. Thus, the activity of this mutant could not be distinguished from that of the normal receptor, but this was likely to be a consequence of the lesser sensitivity of the FACS assay compared with the BIAcore in measuring relatively small differences in binding affinities. Moreover, in the FACS assay the FITC-labeled Fab′₂ anti-IgA cross-links the bound IgA, which would enhance the apparent affinity of weak interactions. Finally, as expected, introduction of the R87A mutation, which did not significantly alter the affinity of the rsFcαRI, likewise did not measurably alter the interaction of the full-length cell surface receptor with IgA.

Generally, point mutations are well tolerated in protein structures; however, it was possible that some of these mutations had disrupted the structure of FcαRI and/or prevented its expression on the cell surface rather than altering a ligand contact. Thus, the cell surface expression of each construct was monitored by staining the transfected COS cells with the anti-FcαRI mAb MY43. The range of expression of each mutant (maximum mean fluorescent intensity, 2000) and the proportion of transfected cells (16–23%) were equivalent (Fig. 4). Hence, there is no evidence from the cell surface expression that any mutant protein was unstable compared with the normal receptor. Furthermore, since MY43 blocks IgA binding to the receptor, MY43 binding to the mutant receptors is the best available probe for the integrity of the protein in the vicinity of the binding site. The MY43 binding of the Y35 (22%) and R82 (18%) mutant receptors was equivalent to that of the normal receptor (19%), indicating that even in those receptors with no IgA binding activity, the structural integrity of the protein close to the ligand binding site was preserved.

Since the IgA binding site on the FcαRI is near the N-terminal end of EC1, as the ectodomains are Ig-like folds, we tested whether the N-terminal peptide contributed to the binding site. The N-terminal peptide, comprising the first 10 residues of the receptor, QEGDFPMPFI was altered to the corresponding sequence of ILT1, the leukocyte receptor cluster (LRC) molecule with highest overall homology with FcαRI (38% identity over both ectodomains). This sequence QAGHLPKPSL has six changes (shown underlined) from the FcαRI sequence, but maintains the residues G3, P6, and P8, which may be important in the structure of this region of the protein. Expression of this N-terminal chimeric receptor in COS cells as measured by MY43 binding was equivalent to that of normal receptor (8.0% compared with 8.8% positive cells), indicating that the chimeric protein was not less stable than the normal receptor (Fig. 5). Furthermore, as MY43 binding to its blocking epitope was not compromised, the protein structure close to the IgA binding site was presumably undisrupted. IgA binding to the normal and that to the chimeric FcαRI were also equivalent, with 7.0 and 7.4% of cells binding IgA, respectively (Fig. 5). Thus, the N-terminus of the FcαRI does not contribute to the IgA binding site.

The interpretation of how residues identified as essential for ligand binding form a binding site is best understood in the context of the three-dimensional structure of the protein. In the absence of a solved crystal structure of the FcαRI a homology-based model of the ectodomains of the receptor was constructed (Fig. 6). A sequence search of the nonredundant PDB using PSI-BLAST (29) on the web (http://dove.embl-heidelberg.de/3D) (30) predicted the KIR to be the most closely related and thus to have the best match for the fold of FcαRI. FcαRI has sequence identity of 34% to KIR (2DL1) over both ectodomains, 26% between their respective EC1s, and 44% between their EC2s. Modeling of FcαRI on KIR will correctly predict the fold of this receptor and will most accurately predict the structure of EC2. Thus, a model of the extracellular domains of the FcαRI was constructed using the coordinates of five KIR structures in the PDB as templates. The optimized sequence alignment consisted of nine blocks of homology between FcαRI (residues P8-I18, S23-C28, N44-G51, T60-C79, S91-A109, V114-S127, F132-E142, A155-V162, and G168-L190) and KIR (1NKR: residues P8-V18, T23-C28, N46-G53, S60-C79, S94-A112, T117-S130, Y134-E144, A162-V169, and G173-L200). Ten models were constructed using the MODELLER (35) modeling software (Molecular Simulations, San Diego, CA) without subjective intervention. Scores of internal consistency of each model were obtained using the program PROFILES 3D (36) and ranged from 54–73 from a theoretical best score of 86. The highest scoring structure was retained as the model for the FcαRI ectodomains.

Several important structural features found in the solved KIR structures whose amino acid sequences are found in FcαRI are appropriately reproduced in the FcαRI model, which indicates that the KIR was a valid template for the FcαRI model. Firstly, the buried central hydrophobic core of the KIR is composed of a cluster of three residues from distinct parts of the protein, with L17 and V100 packing against W188 (32). This feature is reproduced in the FcαRI model by the near identical positioning of residues V17, V97, and W183, from near the start, middle, and end of the primary sequence, indicating that the model has folded these regions correctly to bring these core residues together appropriately (Fig. 6). Secondly the linker peptide between the ectodomains is almost identical between KIR (GLYEK) and the FcαRI (GLYGK), including the pivot Leu residue. This conserved interdomain linker allied with the conserved hydrophobic core of these proteins would suggest that the FcαRI will be a bent molecule like the KIR, for which bend angles between EC1 and EC2 of 66–80 degrees have been reported (33). These structural features would not be generated by modeling FcαRI on the other Fc receptors where sequence identity is less (only ∼20%), and the orientation of EC1 to EC2 differs by 120 degrees compared with KIR. It is noteworthy that EC1 and EC2 of KIR share only 40% identity, and these structures are superimposable upon each other. It is apparent that the EC2 of FcαRI, with 44% identity to the same domain of KIR, will be well modeled on the KIR template. The 26% identity between the respective EC1s means that EC1 of the FcαRI model will be less representative of the authentic FcαRI structure than is EC2.

It is apparent that there will be some structural differences between the KIR and the FcαRI. The WSXWS motif characteristic of hemopoietic receptors is varied in EC2 of the KIRs to WSXSS, but the serine hydroxyls that hydrogen bond to the backbone of the F strand (underlined) are conserved. In EC1 of the KIR the motif is VSAPS and is a more marked deviation from the hemopoietic receptor motif, but still maintains the structural serine residues. Likewise, in the EC2 of FcαRI a slight variant of this structural motif is found, WSFPS, which is a composite of the motifs from EC1 and EC2 of the KIR. EC1 of FcαRI has no equivalent to the KIR VSAPS motif. The absence of this structural motif and the lower sequence identity between the N-terminal domains of KIR and FcαRI indicate that EC1 of the FcαRI will be the domain that differs in structure most from the KIR structure. A second structural difference apparent between KIR and FcαRI is that the KIRs have a sequence motif PGP, residues 14–16 in EC1 and residues 114–116 in EC2, the first proline of which kinks the A strand, splitting it into an A and an A′ strand. This proline kinking motif is absent in the FcαRI, suggesting that the A strand may not be split in this protein.

Displaying the binding site residues of FcαRI on the model (Fig. 7,A) predicts that these lie on one face of the receptor at the N-terminal end of the EC1. The side chains of residues essential for IgA binding are shown in red. The side chains of residues 82–87 from the F-G region and Y35 from the B-C region are closely packed and in van der Waals contact. The F-G loop forms a distinct protrusion from the domain, with the Arg⁸² and His⁸⁵, in particular, being prominent. Residue Tyr⁸¹, on the other hand, is less exposed and may play an important role in the structure of the F-G loop. If viewed with EC1 end on (Fig. 7 B), the most important binding site residues, Tyr³⁵, Tyr⁸¹, Ile⁸³, Gly⁸⁴, and Arg⁸², form a ridge flanked on one side by the less crucial residues, His⁸⁵ and Tyr⁸⁶. On the opposite side is Arg⁵² in the C′-E region, with unexamined residues in the C and C′ regions lying between Arg⁵² and Tyr³⁵. The arginyls will give the binding site a positive charged nature with His⁸⁵, Tyr³⁵, and Tyr⁸⁶ possibly contributing hydrophobic interactions.

The PSI-BLAST search for structural homologues of the FcαRI resulted in identification of Ab Fabs at low identity (10%). The site at the CH2/CH3 interface of IgG-Fc for rheumatoid factor binding is similar to the site at the CH2/CH3 interface of IgA recognized by FcαRI. Hence, the cocrystal structure of rheumatoid Fab with human IgG-Fc (38) (PDB accession no. 1ADQ) was used to produce a schematic view of how FcαRI might bind IgA-Fc. A model of IgA (39) (PDB accession no. 1IGA) was superimposed over the α carbon trace of the IgG-Fc in the Fc:rheumatoid Fab complex. The α carbon atoms of the sequence flanking the half cystines in EC1 of the FcαRI model were superimposed onto those of the V_H domain of the rheumatoid factor Fab (Fig. 8). This arrangement of FcαRI ectodomains and IgA placed the F-G region of the FcαRI model close to the essential residues (LLG 257–259, PLAF 440–443) at the CH2/CH3 interface of the IgA. The FcαRI C-terminus is in proximity to the membrane, while the ectodomains are rotated to present domain 1 away from the plasma membrane and available for binding the IgA Fc. In this representation each IgA heavy chain would appear to be able to interact independently with an FcαRI. Such an interaction would give a higher apparent affinity of binding if IgA was bound by two receptors, one on each H chain, than if bound by one receptor only. We addressed this question using purified IgA and rsFcαRI in a biosensor assay.

It is possible to measure the affinity of the intrinsic interaction of FcαRI with one H chain by measuring the binding of rsFcαRI to immobilized IgA. Conversely, it is possible to measure the affinity of potentially multiple receptor interactions with IgA by immobilizing the receptor and capturing soluble IgA. To test this, the kinetics of FcαR:IgA interactions were measured under conditions where both proteins were, in turn, immobilized separately. The binding of rsFcαRI to immobilized IgA (Fig. 9, thick line) showed a composite binding curve that included a small component of higher affinity binding to IgA. This small component of higher affinity binding was attributed to the presence of some aggregates of the receptor that bound more avidly and hence the sensogram fitted optimally to a model of competing reactions between a heterogeneous analyte. The affinity for the monomer binding reaction was estimated at 4.0 × 10⁶ M⁻¹ with a χ² of 60. The immobilization of rsFcαRI had a pronounced effect on this interaction (Fig. 9, dotted line). The binding of IgA by immobilized rsFcαRI fitted well (χ² = 6) to a single-site model with a K_a of 1.1 × 10⁸ M⁻¹. This was a 30-fold increase in the apparent affinity of the interaction over that of the monomer soluble receptor binding to the immobilized IgA. If this IgA were bound to receptor at only one heavy chain, then the binding site on the other heavy chain should be unoccupied and available to bind additional receptor. Therefore, this IgA, bound with high affinity to immobilized rsFcαRI, was tested for the capacity to bind additional soluble receptor (Fig. 9, thin line). There was no additional binding of rsFcαRI to IgA bound to immobilized rsFcαRI; in fact, the soluble receptor marginally increased the dissociation of the IgA from the layer (indicating that there is minimal rebinding of IgA to the layer in the dissociation phase). Thus, there are no free binding sites (CH2/CH3 interface) on IgA captured by immobilized FcαRI to which additional soluble rsFcαRI can bind. The simplest conclusion, since this binding is of much greater affinity than the intrinsic rsFcαR:IgA interaction, is that capture of IgA by the immobilized FcαRI involves interaction of receptors at both CH2/CH3 interfaces of the IgA.

The interaction of FcαRI with IgA is important in IgA-mediated immunity and potentially in diseases such as IgA nephropathy. In the study described herein we have made a detailed site-directed mutagenesis survey of the F-G region and other regions of FcαRI that may bind to IgA, viz., the B-C, C′-E, and N-terminal regions. Normal and mutant rsFcαRs were produced in baculovirus, and IgA binding affinities were assayed using a biosensor. The F-G region contributed six residues that interacted with IgA. Substitution of residues Y81 or R82 abolished IgA binding (K_a reduced >100-fold), while substitution of residues I83, G84, H85, and Y86 reduced the receptor apparent affinity 11-, 19-, 4-, and 3-fold, respectively. Substitution of R87 did not significantly decrease binding. Thus, the contribution of the F-G region residues to the FcαR:IgA interaction was greatest at the start of the loop and diminished as the alanine scan progressed along the loop. In addition to residues Y81 and R82 in the F-G region, Y35 in the BC region proved essential for IgA binding. Also, the R52A mutation in the putative C′-E region reduced IgA binding 8-fold, showing that the ligand binding site contains residues from at least three distinct regions of the receptor primary sequence. Different substitutions to amino acid residues with different size or charge properties would have more or less profound effects on the binding activity of mutant receptors. Thus, the definition of essential binding site residues can also vary depending on the nature of the amino acid substitutions. For example, the substitution of H85 with alanine resulted in a relatively small reduction in binding activity, but substitution with glutamate completely abolished the activity of the receptor (25).

The abrogation of IgA binding with the alanine substitution of R82 in the F-G region or Y35 in the C strand region was probably due to changing a binding contact with IgA rather than a global disruption to structure, as these results were confirmed by the expression of these mutant receptors at the surface of transfected COS cells. While IgA binding activity was completely lost, mAb MY43 binding, which blocks IgA binding to the receptor, was unaffected. Thus, the MY43 epitope does not include these IgA binding site residues. Furthermore, in these mutants the protein structure is not disrupted, as MY43 binding, near the IgA binding site, is preserved.

The fourth region tested for participation in IgA binding was the N-terminal region of FcαR. The first 10 aa of the receptor were exchanged for those of the related LRC receptor ILT1, effectively changing 6 aa residues in this N-terminal region. No effect was seen on either MY43 binding or IgA binding to this mutant receptor compared with the normal receptor. Thus, the N-terminus does not participate directly in ligand binding. Thus FcαRI is unlike the p58 KIR, where the N-terminal peptide is a binding site for a ligand (40).

Rather, the IgA binding site of FcαRI has a number of features in common with that of other Fc receptors. This is despite the fact that the principal ligand interacting domain in other FcRs is EC2 (41), while interaction occurs through EC1 in FcαRI. The interactions of other FcRs and Igs have been characterized by extensive mutagenesis (41) and recently for FcγRIII/IgG-Fc (22) and FcεRI/IgE-Fc (23) by the solution of crystal cocomplexes. These studies have shown that the B-C-C′-E regions and the F-G region comprise the ligand binding sites of these receptors. This study shows that this is also the case with the IgA binding site of FcαRI despite its low sequence relatedness (∼20% identity) to other FcRs and the location of the ligand binding site being in EC1 and not EC2. For receptors belonging to the LRC, which are more closely related to FcαRI and in which non-MHC ligands are engaged, a similar binding site might be involved.

As an aid to visualizing the mutagenesis data we constructed a homology model of the ectodomains of FcαRI based on the KIR ectodomain structures in the protein database. The basis of homology modeling lies in the fact that there are expected to be only several 1000 distinct protein folds possible for all globular proteins. There are currently approximately 10,000 solved protein structures, but these represent only 900 distinct folds. Modeling, therefore, requires identifying a correct fold to serve as a template. The identification of folds using the program PSI-BLAST (29) has a reported predictive accuracy of 98% in a test search of 685 PDB entries with <25% identity (30). PSI-BLAST identified KIR as a template for the FcαRI ectodomains, and a homology model was constructed. Several features of the FcαRI model indicated that KIR was a suitable template, and predictions based on this theoretical structure should be valid. The 44% identity between EC2 of KIR and FcαRI is indicative of an outcome from automated modeling equivalent to a low resolution x-ray structure (35). The EC1 and EC2 of KIR have 40% identity to each other, and these structures are superimposable. Although the identity between ECs1 of KIR and FcαRI was less (26%), several structural elements validate the FcαRI model. Firstly, the packing of the central hydrophobic core residues L17, V100, and W188 in KIR is emulated appropriately in the FcαRI model by the near identical packing of residues V17, V97, and W183. Since these residues come from three distinct regions of primary sequence, the overall fold of the model should be correct to achieve the packing of these three residues against each other. Secondly, the interdomain linker between EC1 and EC2 of KIR, GLYEK, was almost identical with that of FcαRI, GLYGK, including the conservation of the hinge Leu residue. Thus, the FcαRI ectodomains should be bent similar to the KIR.

The homology model of FcαRI was used to interpret the binding site residues identified by mutagenesis. The model predicts that the identified regions, while distant in primary sequence, are closely spaced in the folded protein and is suggestive of a single binding surface (Fig. 7). In particular, the residues from the B-C region and F-G loop, which contribute essential residues for ligand binding, form a single patch, with the most important residues, Y81, R82, I83, G84, and Y35, forming a central band within this patch. The less important F-G region residues H85 and Y86 lie alongside one edge of the central band of essential residues. R52 lies on the opposite side of this central band and is separated from Y35 by the C and C′ strands. As mutation of R52 also reduced binding, it may well be that residues in the intervening C and C′ regions between R52 and Y35 also contribute to ligand binding. In short, the essential residues Y35, Y81, and R82 together with the other important ligand binding residues (R52, I83, G84, and H85) form a single face near the end, but not including the N-terminus, of EC1.

Using the solved cocrystal structure of rheumatoid factor Fab and IgG-Fc, we have proposed an illustration of how FcαRI may interact with IgA-Fc. The identified ligand binding site is shown interacting at the CH2/CH3 domains of the IgA where mutagenesis of IgA determined FcαRI binding to occur (27, 28). The FcαRI residues important in IgA binding include hydrophobic residues that may complement the hydrophobic binding site residues (e.g., LLG 257–259, PLAF 440–443, Bur numbering) identified in the CH2/CH3 interface region of the IgA Fc (27, 28). There are also acidic residues in the CH2/CH3 interface critical for binding that may interact with positively charged FcαRI residues (e.g., R82) important in IgA binding. This rheumatoid Fab-based arrangement placed the FcαRI orientated to present domain 1 for binding at the IgA CH2/CH3 interface, but this is only one possible orientation. The interaction of IgG-Fc and IgE-Fc with their FcRs occurs with the Fc orientated almost upside down with respect to the FcR, such that the segmental flexibility of the Ab must be important in the simultaneous binding of Ag and Fc receptor (22, 23). The interaction of IgA with FcαRI is different to IgG and the FcγRs, in that binding occurs at the CH2/CH3 interface, and the mucin-like hinge region of IgA1 may not confer the same segmental flexibility as the hinge of IgGs. Therefore, we have represented the IgA Fc “standing up” to interact with the FcαR, as one possible configuration.

One distinctive and testable feature of this representation of the FcαR:IgA interaction is that either one or two receptors may bind to the IgA, resulting in a low 1:1 and a high 2:1 affinity interaction with IgA, respectively. Several experiments reported by others are compatible with the FcαRI binding independently to each IgA heavy chain. Firstly, a Cα3-deficient IgA myeloma protein that consisted of half molecules containing only one heavy chain bound FcαRI (10). Secondly, the binding site on IgA has been localized to the CH2/CH3 interface, so that two independent sites could exist per IgA (27, 28). Thus, although these experiments were not determinations of the stoichiometry of the interaction of intact IgA, they would suggest that a 2:1 interaction is possible between receptor and intact IgA. Such potential ligand binding behavior is likely to be significant in the biology of the receptor on leukocytes, as it has been observed that pretreatment of myeloid cell lines with IgA does not alter receptor number, but increases affinity for IgA (42). Also, the GM-CSF treatment of human neutrophils (43), cytokine treatment of eosinophils (44), or FcαRI-transfected BaF3 cells (45) increase FcαRI affinity for IgA without an increase in receptor number. This low or high affinity IgA binding behavior must depend on the organization of the receptors in the cell membrane and/or their association with other accessory surface molecules and may result from a switch from a 1:1 binding stoichiometry to a 2:1 binding stoichiometry.

This switch from low to high FcαRI binding affinity was shown in BaF3 cells to be independent of the γ subunit (45). In this report we have shown low and high affinity FcαR:monomer IgA interaction in a biosensor system using only purified receptor and IgA proteins. This involved comparing the binding of rsFcαRI to IgA immobilized to a biosensor surface with that of IgA binding to immobilized rsFcαR. The immobilization of the rsFcαRI resulted in an approximately 30-fold increase in apparent affinity over the reaction of soluble FcαRI with immobilized IgA. This corresponded to a change in K_d from 0.25 μM to 9.1 nM. The low affinity binding found with soluble receptor approximates previous studies where micromolar K_d and K_i values have been reported with soluble proteins and at the cell surface (10, 15, 25, 46). The high affinity binding observed with immobilized receptor (K_d = 9 nM) more closely approximates the result from a study in which GM-CSF priming of neutrophils determined a K_d of 6 nM, although in this study the resting K_d was still high at 21 nM (43). The location of the rsFcαRI binding site at the CH2/CH3 interface of the IgA-Fc is similar to the binding site at the CH2/CH3 interface of IgG-Fc for the neonatal FcR (FcRn). Similar to the FcαR:IgA interaction described here, the rsFcRn:IgG interaction studied using the biosensor also demonstrated high affinity binding only when the FcRn was immobilized (47, 48).

In summary, we have presented an analysis of the interaction of FcαRI with IgA at the level of individual amino acids. Mutagenesis identified residues important in IgA binding from three different regions: C strand (Y35), C′-E region (R52), and F-G loop (Y81, R82, I83, G84, H85, and Y86). A model of rsFcαRI placed these residues in a single contiguous face of domain 1. The N-terminal peptide was not involved in IgA binding. In addition, we have shown immobilized rsFcαRI binds IgA with increased affinity, a behavior compatible with IgA interaction with FcαRI at the cell surface, where low and high affinity modes of binding have been observed. These data should help define the FcαR:IgA interaction, which is of increasing interest in immunity, disease, and immunotherapy.

We thank Gary Jamieson for helpful discussions and computing, and Joe Trapani for critical reading of the manuscript.