Structure of an isolated unglycosylated antibody C(H)2 domain.

The C(H)2 (C(H)3 for IgM and IgE) domain of an antibody plays an important role in mediating effector functions and preserving antibody stability. It is the only domain in human immunoglobulins (Igs) which is involved in weak interchain protein-protein interactions with another C(H)2 domain solely through sugar moieties. The N-linked glycosylation at Asn297 is conserved in mammalian IgGs as well as in homologous regions of other antibody isotypes. To examine the structural details of the C(H)2 domain in the absence of glycosylation and other antibody domains, the crystal structure of an isolated unglycosylated antibody gamma1 C(H)2 domain was determined at 1.7 A resolution and compared with corresponding C(H)2 structures from intact Fc, IgG and Fc receptor complexes. Furthermore, the oligomeric state of the protein in solution was studied using size-exclusion chromatography. The results suggested that the unglycosylated human antibody C(H)2 domain is a monomer and that its structure is similar to that found in the intact Fc, IgG and Fc receptor complex structures. However, certain structural variations were observed in the Fc receptor-binding sites. Owing to its small size, stability and non-immunogenic Ig template, the C(H)2-domain structure could be useful for the development by protein design of antibody domains exerting effector functions and/or antigen specificity and as a robust scaffold in protein-engineering applications.

Introduction

Antibodies, also called immunoglobulins (Igs), comprise two identical light chains and heavy chains linked by disulfide bonds. The light chain contains a variable (VL) domain and a constant (CL) domain. The heavy chain has four to five domains, depending on the isotype, including a variable (VH) domain and several constant (CH) domains: three CH domains in IgG, IgA and IgD and four CH domains in IgM and IgE. The antigen-binding fragment (Fab) consists of the light chain (VL and CL) and the first two domains of the heavy chain (VH and CH1) and is specifically involved in antigen binding. The Ig Fc (fragment crystallizable) portion consists of two constant domains, namely CH2 and CH3, from each heavy chain and binds to effector molecules in order to elicit host responses. CH2 is the only antibody domain that exhibits very weak carbohydrate-mediated interchain protein–protein inter­actions, in contrast to the extensive interactions that occur between VH–VL, CL–CH1 and CH3–CH3 domains as seen in intact antibody structures. The crystal structures of intact IgG (Larson et al., 1991 ▶; Harris et al., 1992 ▶, 1998 ▶; Saphire et al., 2001 ▶) and Fc (Krapp et al., 2003 ▶) reveal that the CH2 domain interacts with the other CH2 domain through sugar moieties only. The N-linked glycosylation at Asn297 of the CH2 domain is conserved in all mammalian IgG molecules and the homologous regions of IgM, IgD and IgE. From a functional point of view, the CH2 domain contains large portions of the entire binding sites for complement and Fc receptors that are critical for the effector function of antibodies (Vidarsson & van de Winkel, 1998 ▶; Woof & Burton, 2004 ▶) as well as for binding to the neonatal Fc receptor (Martin et al., 2001 ▶), which is important for the preservation of antibody stability. The isolation and characterization of a CH2 domain from myeloma IgG was reported quite early on as this domain is involved in complement activity (Seon & Pressman, 1975 ▶). Previously, a series of amino-acid substitutions in the CH2 domains of various IgG subclasses was carried out to enhance the differential affinity for the Fc receptor (Canfield & Morrison, 1991 ▶). In the crystal structures of IgG1 Fc–FcγIII receptor complexes, Fc receptor (FcγR) exclusively interacts through the CH2 domains of IgG1 Fc, in which the binding site includes the Asn297 residue of the CH2 domain (Sondermann et al., 2000 ▶; Radaev et al., 2001 ▶). Several recent studies have also been focused on the characterization of the antibody constant domains, CH2 and CH3, in order to understand the folding and stability mechanisms of these domains (Demarest et al., 2004 ▶; Feige et al., 2004 ▶; McAuley et al., 2008 ▶). Of these, kinetic studies exploring the folding mechanism of the CH2 domain suggested that an unglycosylated murine CH2 domain is a monomer with relatively low stability (Feige et al., 2004 ▶). Although several crystal structures are available of intact IgG (Larson et al., 1991 ▶; Harris et al., 1992 ▶, 1998 ▶; Saphire et al., 2001 ▶), Fc (Krapp et al., 2003 ▶; Matsumiya et al., 2007 ▶), Fab fragments (Stanfield et al., 2006 ▶) and variable antibody domains VH or VL (Dottorini et al., 2004 ▶; Park et al., 2008 ▶), there are no structures of the CH2 domain. Such a structure would allow us to determine the structural details in the absence of glycosylation and other antibody domains. Here, we present the crystal structure of an isolated un­glycosylated CH2 antibody domain. The isolated CH2 domain is a monomer and is similar to the CH2-domain structures embedded in the intact Fc, IgG and Fc receptor complex despite the absence of extensive interactions with sugar moieties and its isolation from other antibody domains. However, certain differences do exist which could be important in the use of such domains as scaffolds for high-affinity binders.

High-affinity binders based on variable domains (also termed domain antibodies; dAbs; Holt et al., 2003 ▶) have attracted much attention in recent years owing to their small size and stability, which are suitable for targeting the sterically confined binding sites on antigens or other protein surfaces. Since the unglycosylated antibody CH2 domain appears to be a stable monomer and to be structurally independent, it might be useful as an alternative to domain antibodies for the generation of high-affinity binders. Therefore, the structure may be useful for optimization of the antibody CH2-domain stability, creation of alternative antibody scaffolds based on the CH2 domain (nanoantibodies) and as a small structural scaffold in protein design. Also, the use of CH2 as a scaffold is much easier and cost-effective if expressed in bacteria that do not support glycosylation.

Materials and methods

Cloning, protein expression and purification

The gene encoding the CH2 antibody domain was obtained by PCR amplification using the cDNA of the MAK33 γ1 heavy chain as a template. Protein preparation and purification were carried out as previously described (Zhang et al., 2004 ▶).

Size-exclusion chromatography

Purified CH2 was loaded onto a Superdex75 10/300 GL column that had been pre-equilibrated with phosphate-buffered saline (PBS). The protein was eluted with PBS at 0.5 ml min−1. The Superdex75 column was calibrated with protein molecular-weight standards of 669.0, 440.0, 232.0, 158.0, 67.0, 44.0, 25.0 and 13.7 kDa.

Crystallization, data collection and structure determination

High-throughput screening of crystallization conditions was carried out with a Hydra II Plus crystallization robot (Matrix Technologies, Hudson, New Hampshire, USA) using the Crystal Screen kit from Hampton Research (Laguna Niguel, California, USA). Thin plate-shaped crystals were grown from a crystallization condition consisting of 30% PEG 1500 with equal volumes of protein and well solutions. The crystals were cryoprotected with the well solution after 25% glycerol had been introduced. The diffraction data were collected on the SER-CAT 22-ID beamline of the Advanced Photon Source (APS), Argonne National Laboratory. Data were processed and scaled with the HKL-2000 program suite (Otwinowski & Minor, 1997 ▶). Data-collection statistics are given in Table 1 ▶. The structure of the CH2 antibody domain was solved by molecular replacement with the CCP4 version of AMoRe (Navaza, 2001 ▶) using the CH2-domain structure extracted from the intact antibody IgG b12 structure (Saphire et al., 2001 ▶; PDB code 1hzh) as a search model. The initial model obtained from molecular replacement was iteratively refined using CNS (Brünger et al., 1997 ▶) and rebuilt with Coot (Emsley & Cowtan, 2004 ▶) and O (Jones et al., 1991 ▶). Water molecules were added automatically using CNS followed by visual inspection and refinement. The refinement statistics are presented in Table 1 ▶. Figures were prepared with PyMOL (DeLano, 2002 ▶).

Table 1

X-ray data-collection and refinement statistics for the antibody CH2 domain

Data collection
Wavelength (Å)	1.0
Space group	P21
Unit-cell parameters (Å, °)	a = 36.14, b = 40.68, c = 39.13, β = 106.7
Resolution range (Å)	26.36–1.75 (1.82–1.75)
Observations	42457
Unique reflections	10493
Redundancy	4.0 (3.0)
Completeness (%)	94.4 (72.0)
I/σ(I)	27.2 (7.0)
Rmerge† (%)	0.040 (0.129)
Refinement statistics
R factor‡ (%)	20.1 (23.6)
Rfree‡ (%)	22.7 (31.1)
No. of atoms: protein/water	866/111
R.m.s.d. bond distances (Å)	0.006
R.m.s.d. bond angles (°)	1.4
Wilson B value (Å2)	17.0
Average B values (Å2)
Protein atoms	20.8
Water O atoms	28.7
Ramachandran plot
Most favored ϕ and ψ angles (%)	95.7
Additional allowed ϕ and ψ angles (%)	4.3

R merge = .

R factor and R free = , where R free was calculated over 5% of the amplitudes chosen at random and not used in the refinement.

Results and discussion

Structure of the isolated unglycosylated CH2 domain

We used recombinant DNA techniques to express the CH2 domain of an antibody (IgG) in Escherichia coli and purified the protein, which resulted in the production of the isolated unglycosylated CH2 domain with a molecular weight of 12 kDa. Crystals appeared in 30% PEG 1500 within a week and grew as large plates that were suitable for X-ray diffraction. The crystal structure of the CH2 domain was determined at 1.7 Å resolution by molecular replacement using the glycosylated CH2 domain from the structure of an intact antibody IgG b12 determined previously at 2.7 Å resolution (Saphire et al., 2001 ▶). A summary of X-ray data-collection and refinement statistics is presented in Table 1 ▶. The asymmetric unit contained one CH2 domain and the final model included a total of 107 amino-acid residues, with more than 95% of residues in the most favored region of the Ramachandran plot (Ramachandran & Sasisekharan, 1968 ▶). A ribbon diagram of the unglycosyl­ated CH2 antibody domain is shown in Fig. 1 ▶(a), with a color coding corresponding to the B-factor values: blue for lower (∼11 Å2), green for medium (∼23 Å2) and red for higher (∼48 Å2) values. The overall structure is similar to the intact glycosylated CH2 domain as found in the Fc and IgG antibody structures (Figs. 1 ▶ b and 1 ▶ c), displaying a stable immunoglobulin fold with minor differences in the loop regions, the termini and the orientations of side chains of the binding site or surface-exposed residues. No significant intermolecular interactions that warrant the consideration of oligomerization are ob­served in the crystal lattice, which is in agreement with our size-exclusion chromatography data suggesting that the protein exists as a monomer in solution (Fig. 2 ▶). Our current data combined with the previous results from thermodynamic studies on the CH2 domain of an IgG antibody (Feige et al., 2004 ▶) have confirmed that the isolated un­glycosylated CH2 domain is a stable mono­mer in the absence of glycosyl­ation and other antibody domains.

Figure 1

Structure of the CH2 antibody domain and structural comparison with the corresponding region in the Fc and IgG structures. (a) Ribbon diagram of the isolated unglycosylated CH2 domain from IgG γ1 is shown with a gradient ramp of colors according to the temperature factors (B factors): blue for lower (∼11 Å2), green for medium (∼23 Å2) and red for higher (∼48 Å2) values. The N- and C-­termini as well as strands A–G are marked. (b) The isolated CH2 domain (green) was superimposed with a least-squares algorithm using the Cα traces of the CH2 domains of fucosylated (blue; PDB code 2dtq) and nonfucosylated (purple; PDB code 2dts) Fc structures. (c) Superposition of the isolated CH2 structure (green) with that of CH2 portions of an intact IgG (PDB code 1hzh) using the Cα-trace alignment. The heavy and light chains of IgG are shown in red and blue, respectively. The carbohydrate moieties between the CH2 domains of the Fc and IgG structures in (b) and (c) are omitted for clarity.

Figure 2

A sample of purified CH2 was analyzed on Superdex75 10/300 GL column calibrated with molecular-weight standards. The arrows indicate the positions where the 43.0, 25.0 and 13.7 kDa molecular-weight standards eluted.

Structural comparisons of the CH2 domain with intact Fc, IgG and Fc receptor complex structures

To analyze the conformational features of the isolated CH2 domain, we compared it with recent crystal structures of human IgG Fc fragments with and without a fucose residue attached to the sugar moieties at Asn297 (Matsumiya et al., 2007 ▶; PDB codes 2dts and 2dtq), intact IgG b12 (Saphire et al., 2001 ▶; PDB code 1hzh) and Fc receptor complexes (Sondermann et al., 2000 ▶; Radaev et al., 2001 ▶; PDB codes 1e4k, 1t83 and 1t89). The two CH2 domains in these antibody structures interact with each other through sugar moieties. Superposition of the unglycosylated CH2 domain on the corresponding CH2 domains of Fc and IgG yielded root-mean-square deviations (r.m.s.d.s) of 0.5 and 0.6 Å, respectively (Figs. 1 ▶ b and 1 ▶ c). This clearly indicates that glycan removal as well as isolation of the domain does not affect the structural integrity of the monomeric CH2 antibody domain. Furthermore, we superimposed the CH2 antibody domain with the CH2 regions of FcγRIII–Fc complex structures available in three different crystal forms and found that the r.m.s.d.s ranged from 0.5 to 1.0 Å (Fig. 3 ▶ a). In all three reported FcγRIII–Fc complexes, the D/E loop (residues 296–299 between β-strands D and E) of the CH2 domain in Fc directly makes critical intermolecular inter­actions with the FcγRIII receptor. However, in the ortho­rhombic form only (PDB code 1t83), Tyr296 of the D/E loop in the CH2 protrudes out at the tip and makes contacts with Lys128 and Asp129 of FcγRIII (in magenta; Fig. 3 ▶ b). In the two hexagonal forms of the FcγRIII–Fc complexes, Tyr296 in the D/E loop has a different conformation (blue and orange in Fig. 3 ▶ b), which suggests a requirement for conformational flexibility of Tyr296 for the binding of CH2 to FcγRIII. When we overlaid the isolated CH2 domain on these complex structures (in green; Fig. 3 ▶ b), we observed that Tyr296 of the isolated CH2 domain exhibited an upright conformation with the hydroxyphenyl side chain pointing out of the D/E loop at the tip as found in the Fc receptor complex structure of the orthorhombic form. The other CH2 domain of Fc region in the complex also makes interactions with the FcγRIII receptor through its F/G loop (residues 325–331 between β-strands F and G), where Pro329 of the CH2 domain is sandwiched by Trp90 and Trp113 of the receptor, which is also observed in the orthorhombic structure (Fig. 3 ▶ c). This Fc receptor-binding site on the CH2 domain is structurally well conserved, in contrast to the other binding site where Tyr296 of the D/E loop exhibits significant conformational flexibility. From these structural analyses, we found that the isolated unglycosylated CH2 domain has a similar conformation to that embedded in the intact Fc and IgG structures. However, a significant variation was noted in the Fc receptor-binding sites when compared with the FcγRIII–Fc complexes, particularly at residue Tyr296 of the CH2 domain.

Figure 3

Stereoviews showing structural comparisons between the isolated CH2 domain and similar CH2 regions in Fc receptor complexes. (a) Cα-trace superposition of the isolated CH2 domain (green) and the CH2 domains of FcγRIII–Fc complex structures in various crystal forms: hexagonal forms (PDB codes 1e4k and 1t89, in blue and orange, respectively) and an orthorhombic form (PDB code 1t83, magenta). The FcγRIII receptor molecules in the complexes are shown in red. Arrows point to the Fc receptor-binding sites of the CH2 domains. The carbohydrate moieties between the CH2 domains of the FcγRIII–Fc complexes are omitted for clarity. (b) A close-up view of one of the binding sites from the D/E loop of the CH2 domain, highlighting the orientation of the Tyr296 residue. (c) A close-up view of another binding site from the F/G loop of the CH2 domain is shown. Amino-acid side chains are labeled according to the orthorhombic structure (PDB code 1t83).

Isolated antibody fragments can be used as scaffolds for binders. The smallest functional antigen-binding fragment of an antibody, a variable domain, either VH or VL, has been used successfully. From the structural point of view, the major difference between the variable and constant domains is the connecting loops between the β-strands. The complementarity-determining region (CDR) loops in the variable domains that make contacts with the antigen are longer than those found in the constant domains which interact with effector molecules. Using the structural details together with in vitro phage-display selection (Weiss & Penner, 2008 ▶; Dimitrov & Marks, 2008 ▶) and computational protein-loop design (Hu et al., 2007 ▶), the CH2 domain could be engineered to have predetermined specificities for various antigens and proteins. In general, the Ig fold is shared by many evolution­arily unrelated or distantly related proteins (Halaby et al., 1999 ▶). A DALI database search (Holm & Sander, 1998 ▶) using the CH2-domain fragment resulted in more than 470 different protein structures with a Z score greater than 2, mainly of immune-system and cell-adhesion molecules. Therefore, the scope for protein design using the antibody CH2-domain template may have wider applications in addition to thera­peutic high-affinity binders and stable structural scaffolds.

PDB reference: antibody C