New insights into intra- and intermolecular interactions of immunoglobulins: crystal structure of mouse IgG2b-Fc at 2·1-Å resolution

Abstract

The structure of the Fc fragment of monoclonal antibody IgG2b from hybridom M75 of Mus musculus has been determined by single crystal X-ray diffraction. This is the first report of the structure of the murine immunoglobulin isotype IgG2b. The structure refined at 2·1 Å resolution provides more detailed structural information about native oligosaccharides than was previously available. High-quality Fourier maps provide a clear identification of α-l-fucose with partial occupancy in the first branch of the antennary oligosaccharides. A unique Fc:Fc interaction was observed at the C_H2-C_H3 interface.

Introduction

Immunoglobulins are large Y-shaped glycoproteins that participate in the adaptive component of immune system defence against extracellular pathogens. Their structural diversity allows them to bind specifically millions of structurally unique molecules.

Molecules of immunoglobulin G (IgG) form the major group of molecules of the immunoglobulin superfamily. IgG is composed of two light and two heavy chains that form three compact folded globulae (each of a molecular weight of about 50 kDa). Proteolytic cleavage of immunoglobulins results in fragments Fab and Fc, the latter having a highly conserved structure even between different isotypes.^1–4

The Fc fragment in its native state is glycosylated by two branched oligosaccharide chains.¹ Each oligosaccharide is covalently bound to one of the heavy chains by an N-glycosidic bond between an N-acetylglucosamine (Nag) unit of the oligosaccharide and a side chain of an asparagine residue of the protein. In its native state, one population of an antibody isotype is composed of various mixtures of different glycoforms.⁵

There are several specific ligand binding sites on Fc fragment: Fc-receptor binding site located in a groove between two C_H2 domains (near the hinge region), and two chemically and structurally identical regions at or in the close vicinity of the C_H2-C_H3 interface with broader ligand specificity.^6–8

The three-dimensional structure of the Fc fragment has been studied extensively. The Protein Data Bank (PDB) contains 34 records with experimental information on the structure of the Fc fragment of immunoglobulin.⁹ In many cases accurate interpretation of oligosaccharides bound to the Fc fragment was difficult because of the low reliability of electron density in oligosaccharide regions. Many antibody structures deposited in the PDB do not model complete oligosaccharides present in the molecule. Either the quality and resolution of the analyses did not allow the model to be completed in the region of glycosylation or electron density indicated the presence of additional saccharide units in the chain but these were not included in the final structure for unknown reasons.^10,11 The expression ‘natively glycosylated’ will be used throughout to describe the native state of the Fc fragment as a mixture of forms variably deglycosylated within an organism without any intended removal of saccharide units.

Aggregating immune complexes (ICs) responsible inter alia for autoimmune diseases (rheumatoid arthritis, cryoglobulinaemia, systemic lupus erythematosus, serum sickness, diabetes, etc.^12,13) are formed in an organism when it is exposed to a large dose of an antigen.^14,15 Such ICs involve antibodies with bound antigens forming aggregates or clusters of variable size which can circulate in the blood or form a precipitate. The formation of ICs implies direct interactions between participating complexed immunoglobulin molecules, and the nature of such interactions has not yet been fully elucidated. It has been shown that specific Fc:Fc interactions (the colon denotes the formation of a non-covalent complex) play a significant role in precipitation of antigen–immunoglobulin complexes, and immunoprecipitation that is heterogeneous with respect to antibody specificity and antigen type has also been observed.^16–19 In spite of the number of three-dimensional structures containing the Fc fragment that are available from the PDB and the amount of experimental effort devoted to the role of Fc:Fc interactions in immune precipitation, to the best of our knowledge a detailed structural basis of the role of the Fc fragment in IC formation has not been proposed to date.

We report the first three-dimensional X-ray structure of the Fc fragment of mouse IgG2b. Murine monoclonal antibody IgG2b from hybridom M75 is specific in its adhesion to carbonic anhydrase IX produced in tumour tissues, and potential applications in cancer therapy have been proposed.²⁰ The structure reveals Fc:Fc interactions of a type consistent with current knowledge concerning formation of ICs.

Materials and methods

Protein production

The protocol used for production of IgG2b monoclonal antibody from hybridom M75 was described previously.²¹ Digestion of the hinge region was carried out with papain at a concentration of 50 mg/ml, and the sample was purified in a protein A Sepharose column (Bio-Rad, Prague, Czech Republic) and finally concentrated to 8 mg/ml of Fc fragment for crystallization.

Crystallization and data collection

Crystallization was performed using the hanging-drop vapour-diffusion method. The protein solution contained 0·1 m phosphate-buffered saline (PBS), pH 7·2, and 0·05% sodium azide. The crystal used for X-ray data collection was crystallized from a drop with an initial ratio of protein to reservoir solutions of 1 : 1; the reservoir contained 0·1 m HEPES, pH 7·5, and 20% [weight/volume (w/v)] polyethylene glycol (PEG) 2000. Triangular plates grew at a temperature of 301 K. X-ray diffraction data were collected at the European Synchrotron Radiation Facility, beamline ID29 in Grenoble, France; 30% glycerol (v/v) was used as a cryoprotectant, and 320 oscillation images were collected. The diffraction data were processed using the hkl program package at 2·1-Å resolution.²² The crystal belongs to space group C2. Further details are given in Table 1.

Table 1.

Summary of crystallographic data and structural refinement details

X-ray source, beamline	ESRF (Grenoble, France), ID29
Wavelength (Å)	0·9168
Number of oscillation images	320
Total oscillation range	160°
Resolution range (Å)	20·0–2·1 (2·16–2·10)
Number of observed reflections	112 843
Number of unique reflections	34 029
Number of rejected reflections	481
Data completeness (%)	99·5 (100)
Average redundancy	3·3 (3·3)
Mosaicity (°)	0·73
Space group	C2
Unit cell parameters
a, b, c (Å)	135·73, 62·75, 69·81
β (°)	103·35
Rmerge	0·113 (0·449)
<I>/<σI>	10·3 (2·9)
Wilson B (Å2)	25
Number of monomers per asymmetric unit	2
Number of residues (number of non-H atoms)	413 (3368)
Number of monosaccharides (number of non-H atoms)	18 (226)
Number of water molecules	533
Number of all non-H atoms	4127
RMSD bond lengths from ideal (Å)	0·013
RMSD bond angles from ideal (°)	1·6
Average B-factors (Å2)
Protein atoms	24
Oligosaccharide atoms	28
Water oxygen	37
All atoms	26
R (Rfree)	0·202 (0·212)
Number of reflections for Rfree calculation (%)	1693 (5%)

Data in parentheses refer to the highest resolution shell, if not otherwise specified. RMSD, root mean square deviation. , where I_hkl are observed diffraction intensities with Miller indices hkl. , where F_o and F_c are observed and calculated structure factor amplitudes, respectively. R_free is the R factor calculated on a random subset of 1693 reflections (5%) excluded from refinement. The final structure refinement was performed on all observed structure factors.

The crystals used for data collection were able to withstand prolonged exposure to air at room temperature after the X-ray diffraction experiment and their overall robustness and hardness (in crash tests without the presence of the mother liquor) were above average for protein crystals, implying high stability of the crystal lattice.

Structure determination and model refinement

The phase problem was solved by molecular replacement using amore²³ in the ccp4 program package;²⁴ one chain of the Fc fragment from the structure of human IgG1 with PDB id 1IGT was used as a search model.³ The monomer was fitted into two positions. Despite the high homology between the search model and the target structure, extensive manual rebuilding would have been required on the basis of the first inspection of the electron density and the state of the model. Automated model building was therefore performed with arp/warp.²⁵ The model was refined with refmac5.²⁶ 2F_o–F_c and F_o–F_c electron-density Fouriers were used as guidelines for manual corrections using the program xfit from the xtalview program package (F_o and F_c are observed and calculated structure factor amplitudes, respectively).²⁷ The progress of structure refinement was monitored using the statistic R_free, which was calculated for 5% of all reflections.

Fourteen residues were modelled in alternative conformations. All disulfide bridges (Cys261–Cys321 and Cys367–Cys425) were modelled with 80% occupancy because of radiation damage. Finally, water molecules were added in a multi-step process, at first automatically by XtalView, and then using peakmax and watpeak from ccp4.²⁴

A summary of the crystallographic data and refinement results is given in Table 1.

Quality of the final model

The model quality and data parameters were assessed using programs procheck and sfcheck.^28,29 One hundred per cent of the residues were found in the favoured region of the Ramachandran plot as analysed by program rampage.³⁰

The co-ordinates and structure factors for the Fc fragment of monoclonal antibody IgG2b from Mus musculus have been deposited in the Protein Data Bank at the Rutgers Research Collaboratory for Structural Bioinformatics under accession code 2RGS.

Results

Overall structure of the Fc fragment of IgG2b

The overall structure of the Fc fragment of monoclonal antibody IgG2b is similar in its tertiary structure to all structures containing an Fc fragment deposited in the PDB (Fig. 1a). The Fc-fragment dimer (equal to the asymmetric unit of the crystal) is formed by two fragments of heavy chains cleaved from the antibody. Each chain forms two antiparallel β-sandwich domains (C_H2, residues Gly237–Lys338 and C_H3, Val343–Ser442) connected by a short flexible linker (Ile339–Leu342). The β-sandwich structure is stabilized by a conserved disulfide bond between two strands (Cys261–Cys321 in C_H2 and Cys367–Cys425 in domain C_H3). Two C_H3 domains form a compact non-covalent dimer with a buried surface area of 1050 Ų (calculated with areaimol).²⁴

Figure 1.

Structure and intermolecular interactions of the Fc fragment of immunoglobulin G2b (IgG2b). (a) Schematic view of the structure of the Fc fragment cleaved with papain from mouse monoclonal antibody IgG2b M75. The secondary structure of the protein is shown, and oligosaccharides are represented by sticks. C^α atoms of Asn297 are shown as cyan spheres, cysteins forming disulfide bridges as blue sticks, and short flexible linkers connecting domains C_H2 and C_H3 as magenta loops. Secondary structure elements were assigned using dssp.³¹ (b) Overall view of two interacting dimers of the Fc fragment; two regions of interaction are marked, and are shown in detail in (c) and (d). (c) The loops of domain C_H2 binding to the C_H2–C_H3 interface; (d) an interaction between domains C_H3 and C_H2 of neighbouring Fc fragments. The detailed views show C^α trace protein chains and individual interacting residues as sticks coloured according to atom type; the colour of the carbon atoms differentiates between interacting neighbours. Contact distances for hydrogen bonds are given in Å. The graphics were created with pymol.³²

Sixty per cent identity was found for alignment of the amino acid sequence of the Fc fragment of mouse IgG2b with that of the Fc fragment of human IgG1, the most extensively studied isotype (alignment not shown), implying a high level of similarity for the overall structure, but also indicating regions of local differences, which are discussed further below.

Structural details

N- and C-termini

Based on the papain specificity of immunoglobulin cleavage, Cys229 is assumed to be the first residue of the protein chain present in the crystal.³³ Gly237 is the first residue localized in electron density for both chains. Eight residues of the hinge region (Cys229–Gly236) are not localized in the structure (similar to all other structures of the Fc fragment in the PDB) probably as a result of high mobility of the hinge segment, giving the chain a number of possible conformations.

Crystal packing affects the structure of the C-termini of both chains. Close contacts between chain B and a symmetry-related protein chain (these contacts are lacking for chain A) result in localization of Arg443 of chain B, for which there is insufficient electron density in chain A. The surroundings of the C-termini also differ in hydrogen bonds forming the secondary structure.

Glycosylation

Accurate determination of the structure of the oligosaccharides is important because it has been shown many times that the composition of oligosaccharides in the Fc fragment is associated with the efficiency of immunoglobulin function and that pathological states (such as tuberculosis) are accompanied by defects in glycosylation.⁵

Nine monosaccharide units of the branched oligosaccharide chains for both protein–oligosaccharide chains were unambiguously localized in electron density (Fig. 2). The oligosaccharide chains branch on Man4. A longer branch winds along the same protein chain ending with galactose unit Gal7, which forms three hydrogen bonds to Thr260, Lys258 and Pro244. A shorter branch extends into the bulk solvent region in the space between two C_H2 domains, and therefore no electron density was observed for the terminal galactose unit in this branch.

Figure 2.

Schematic representation of the oligosaccharide structure and of interactions in the Fc fragment of mouse immunoglobulin G2b (IgG2b). Hydrogen bonds are shown as broken lines; the positions of individual monosaccharide subunits are numbered.

α-l-Fucose is bound to Nag1 by a 6-1 type of glycosidic link. Refinement of the model with fucose fully occupied produces strong difference negative maxima of electron density in this region. Fucose with 80% occupancy and an alternative water molecule (hydrogen bond to Nag1/O⁶; occupancy 20%) explain well the electron density in both oligosaccharide chains. The remaining parts of the oligosaccharides were refined with full occupancy of atoms.

Four hydrogen bonds between opposite oligosaccharide chains provide the only direct link between the C_H2 domains. All hydrogen bonds between the oligosaccharides and between the oligosaccharides and the protein are shown in Fig. 2. The pattern of interaction between the oligosaccharides is unique and has not previously been observed in any X-ray structure containing an Fc fragment in the PDB. None, one or two hydrogen bonds between the oligosaccharides are found in the majority of the published structures and the interacting saccharide units vary to such an extent that no single hydrogen bond occurs repeatedly in the whole set of structures. In many cases no hydrogen bonds between the oligosaccharides are observed. There are various reasons for this (e.g. structure quality and ligand binding) and such bonds are also missing in some structures of intact antibodies.³

Intermolecular interactions of symmetry-related molecules

Crystals of the Fc fragment of mouse IgG2b belong to space group C2. This is the first occurrence of space group C2 in any of the structures of the Fc fragment in the PDB. A specific interaction was found between two symmetry-related molecules, where two loops of one C_H2 domain interact with the C_H2–C_H3 interface of a symmetry-related molecule (type C_H2:C_H2–C_H3; Fig. 1c). This contact between the molecules is further stabilized by another interacting site (type C_H3–C_H2; Fig. 1d). Molecules of the Fc fragment joined by these types of interaction form extended linear assemblies of antiparallel-oriented molecules (Fig. 1b). The characteristics of these interactions are summarized in Table 2.

Table 2.

Observed interactions between neighbouring molecules of the Fc fragment

Residue/atom	Distance (Å)	Residue/ atom	Interaction class
Contacts between domain CH2 and CH2–CH3 interface
Glu268/Oε2	2·7	Asn434/Nδ2	S:S
Glu268/Oε1	3·1	Asn434/Oδ1	S:S
Asp295/Oδ1	2·9	Asn434/N	S:M
Tyr296/Oζ	2·7	Val250/O	S:M
Tyr296	–	Ile253	CH-π
Contacts between domains CH3 and CH2
Asp401/O	3·2	His285/N	M:M
Arg344/Nζ1	2·6	His285/O	S:M
Arg344/Nε	2·9	His285/O	S:M
Arg344/Nζ2	2·8	Ala287/O	S:M
Arg344/Nζ1	3·1	Gln288/Oε1	S:S

Two interaction sites are described; distances are given only for hydrogen bonds between acceptors and donors. The class of interaction specifies whether side-chain (S) or main-chain (M) atoms of a residue are involved in hydrogen bond formation.

Discussion

Although this structure bears high sequence and structural similarity to other PDB records containing the Fc fragment of immunoglobulin, careful structural analysis reveals several features worth discussing in a broader context. These include fucose geometry, fucose occupancy, and identification of interesting Fc:Fc interactions potentially relevant for formation of ICs.

Fucose

The structural results show the fucose moieties of the Fc fragment to have 80% occupancy. For steric reasons, the probability of the remaining 20% of the fucose occupancy being spread in disordered or alternative conformations is very low. This suggests that the Fc fragment used for crystallization was approximately 80% fucosylated. The refinement result is in rough agreement with other experimental results.^5,34 Based on this structure it is also evident that the level of fucosylation does not affect the overall structure significantly, which is also supported by a previous comparison of fucosylated and totally defucosylated glycoforms of the Fc fragment.³⁵

The crystallographic model contains α-l-fucose with a good fit to electron density and meaningful contacts with its hydrogen bonding partners. It should be noted that β-l-fucose, the incorrect saccharide isoster,³⁴ is present in most structures of the Fc fragment of immunoglobulin,^36,37 mainly in structures determined at higher resolution, such as PDB records 1L6X (1·65 Å resolution), and 2DTQ (2·0 Å resolution).^7,35 Inspection of these data shows discrepancies between structure and electron density in this region, and also distorted geometry of β-l-fucose, the incorrect saccharide isoster.

Intermolecular interactions of the Fc fragment

The C_H2–C_H3 interface is one of the ligand-binding sites of the Fc fragment, and to the best of our knowledge this is the first observation of Fc-fragment non-covalent binding to the ligand-binding interface of another molecule of Fc fragment, with a set of specific interactions.

The C_H2–C_H3 interface exhibits high adaptability to bind different ligands. It interacts with at least four different natural scaffolds,⁸ for example two α-helices (minimized version of the B-domain of protein A called Z34C; PDB code 1L6X) and two β-strands (engineered 13-residue peptide DCAWHLGELVWCT-NH₂; PDB code 1DN2).^7,8 We have performed a detailed analysis of Fc:Fc interactions in available structures including all crystal contacts (data not shown), and the results suggest that the contacts observed in our structure reveal a new type of interaction of this ligand-binding site that has not been observed before. Here, the C_H2–C_H3 interface interacts with a loop containing glycosylated Asn297 and another nearby loop of the same neighbouring molecule. Oligosaccharides are not involved in direct contacts with symmetry-related molecules. The complex of two (or more) antiparallel-oriented dimers of the Fc fragment is stabilized by interactions of two types: C_H2:C_H2–C_H3 (Fig. 1c) and C_H3:C_H2 (Fig. 1d).

The high adaptability of the C_H2–C_H3 interface and the high sequence similarity of immunoglobulins even across different isotypes suggest that the observed type of Fc:Fc interaction would not necessarily be limited to the type of antibody studied here. The key amino acids involved in formation of the crystal contacts of mouse IgG2b are conserved in terms of structure and sequence in human IgG1 with one exception, Glu268 (mIgG2b)→His268 (hIgG1), which would potentially affect one of the 10 interactions.

Immunocomplexes

Formation of pathogenic ICs in connection with disease has been studied by various methods [light microscopy, rate-zonal ultracentrifugation, immunoprecipitation, enzyme-linked immunosorbent assay (ELISA), electron microscopy, etc.].^{12,16–19,38} Studies focused on the role of the Fc fragment in the formation of ICs revealed that immunoprecipitation of intact antibodies complexed with their specific antigens is enhanced compared with the case of Fab:antigen complexes alone, and this finding was confirmed using various approaches.^16–19 Møller et al. (1979)^16,17 explained the observed differences by postulating a specific role of the Fc fragment in immunoprecipitation, suggesting some Fc:Fc interactions to play a major role in increased aggregation. However, no molecular model of such interactions was proposed.

The reported structure provides detailed evidence for Fc:Fc type contacts, which involve the glycosylation loop of one Fc-fragment dimer (residues Arg293–Thr299) binding to the C_H2–C_H3 interface of another Fc fragment. This interaction comprises four hydrogen bonds and one CH-π interaction, and together with another weaker interaction between domains C_H2 and C_H3 of neighbouring molecules results in an antiparallel arrangement of neighbouring Fc-fragment dimers (Fig. 1b). This is the first structural evidence of Fc:Fc interactions with participation of this Fc-fragment ligand-binding site. Given the role of complement components in inhibiting formation of immunoprecipitates and their interaction with the C_H2–C_H3 interface, the observed interaction pattern can be thought of as a prototype of contacts that could facilitate formation of ICs. In the case of intact immunoglobulin molecules this arrangement would retain a high degree of freedom of Fab and would preserve the ability to bind antigen even in large oligomeric clusters of immunoglobulins. A linear system of immunoglobulin molecules can be envisaged, where Fc-fragment dimers are repeatedly connected using the same interaction pattern as in the structure presented here (Fig. 1b).

Conclusion

The structure of the Fc fragment of mouse monoclonal antibody IgG2b from hybridom M75 was determined at 2·1-Å resolution; this is the first reported three-dimensional structure of this subtype of immunoglobulin. Comprehensive comparison of the newly determined structure with available structural data and the literature enabled a detailed analysis of several important features of this molecule to be performed.

Oligosaccharides of previously determined structures of the Fc fragment exhibit distorted geometry, mainly for fucose, which was modelled incorrectly as β-l-fucose in most of these structures. Oligosaccharides bound to Asn297 in the reported structure are clearly identified and supported in detail by good-quality electron density.

For the first time, partial fucosylation is observed in our crystal structure. Our structure also provides clear evidence for α-l-fucose, supported by X-ray diffraction results, and suggests 80% fucosylation of the Fc fragment, in rough agreement with values resulting from other studies. It has been shown that partial defucosylation does not affect the structure significantly, and the current findings provide clear evidence of the stabilized structure and overall conformation of the Fc-fragment dimers irrespective of the presence of fucose. A new interaction pattern between the two oligosaccharides belonging to the two opposite C_H2 domains was observed.

Previously it was shown that Fc:Fc interactions play a role in the formation of ICs. This work provides evidence of a specific Fc:Fc interaction with participation of the C_H2–C_H3 interface. Such types of interaction may be involved in aggregation of intact immunoglobulins in linear assemblies and so could contribute to the formation of ICs.

Glossary

Abbreviation:

IC

immune complex