Structural basis for improved efficacy of therapeutic antibodies on defucosylation of their Fc glycans

Removal of the fucose residue from the N-glycans of the Fc portion of immunoglobulin G (IgG) results in a dramatic enhancement of antibody-dependent cellular cytotoxicity (ADCC) through improved affinity for Fcγ receptor IIIa (FcγRIIIa). Here, we present the 2.2-Å structure of the complex formed between nonfucosylated IgG1-Fc and a soluble form of FcγRIIIa (sFcγRIIIa) with two N-glycosylation sites. The crystal structure shows that one of the two N-glycans of sFcγRIIIa mediates the interaction with nonfucosylated Fc, thereby stabilizing the complex. However, fucosylation of the Fc N-glycans inhibits this interaction, because of steric hindrance, and furthermore, negatively affects the dynamics of the receptor binding site. Our results offer a structural basis for improvement in ADCC of therapeutic antibodies by defucosylation.


Introduction
The therapeutic applications of antibodies pioneered by von Behring & Kitasato (1890) at the end of the 19th century were revolutionized by the advent of monoclonal antibody technology, and subsequently, improved by genetic engineering approaches such as humanization (Mayforth 1993). To date, over 20 recombinant monoclonal antibodies have been licensed as drugs against various cancers and chronic diseases. Furthermore, antibody-based therapeutics currently account for most recombinant proteins in clinical use, and over 130 human monoclonal antibodies entered clinical trials between 2001 and 2008 (Pavlou & Belsey 2005;Reichert et al. 2005). Although the demand for industrial production of recombinant monoclonal antibodies is increasing, and the number of approved antibodies for therapeutic uses will increase during the next few years, one drawback of antibody medicine is the high cost of production. Production costs impose an economic limit on the therapeutic benefit of these approaches. Recently, a key breakthrough in the development of therapeutic antibodies has been achieved that may dramatically enhance effector functions through a cellular engineering technique for modifying the sugar chains displayed on the Fc region of immunoglobulin G (IgG) (Umaña et al. 1999;Kaneko et al. 2006;Jefferis 2009;Kubota et al. 2009;Yamane-Ohnuki & Satoh 2009).
The Fc region of IgG possesses a conserved glycosylation site at Asn-297 in each of the C H 2 domains. The N-linked oligosaccharides expressed at this site are the biantennary complex type with microheterogeneities, resulting from the presence or absence of the nonreducing terminal residues, and they have a significant effect on the effector functions of IgG ( Fig. 1A) (Burton & Woof 1992;Jefferis et al. 1998;Yamaguchi et al. 2006Yamaguchi et al. , 2007. With regard to the clinical applications of glycoengineered antibodies, the removal of the core fucose residue from the N-glycans of IgG-Fc results in dramatic enhancement (>50-fold) of antibody-dependent cellular cytotoxicity (ADCC) through improved IgG binding to Fcc receptor IIIa (FccIIIa) (Shields et al. 2002;Shinkawa et al. 2003;Taniguchi et al. 2006;Kubota et al. 2009;Yamane-Ohnuki & Satoh 2009). Indeed, a phase I trial of a defucosylated humanized anti-CC chemokine receptor 4 (CCR4) antibody, KW-0761, in relapsed CCR4-positive adult T-cell leukemia-lymphoma or other peripheral T-cell lymphomas revealed that the antibody had antilymphoma activity even at a low dose of 0.01 mg ⁄ kg (Niwa et al. 2004). Our crystallographic data on uncomplexed IgG1-Fc fragments indicated that the overall structures of the fucosylated and nonfucosylated human IgG1-Fc glycoforms are quite similar except for the hydration mode around Tyr-296 (Matsumiya et al. 2007). The crystal structures of the Fc-FccRIIIb complexes have been reported for fucosylated IgG1-Fc and the soluble form of FccRIIIb (sFccRIIIb), which was expressed in Escherichia coli, and therefore not glycosylated (Sondermann et al. 2000;Radaev et al. 2001). The extracellular domains of FccRIIIa possess five N-glycosylation sites at positions 38, 45, 74, 162, and 169 (Ravetch & Perussia 1989). Mutational deglycosylation analyses showed that the N-glycan expressed at Asn-162 has negative and positive effects on the binding affinities of the fucosylated and nonfucosylated IgG glycoforms, respectively (Ferrara et al. 2006;Shibata-Koyama et al. 2009). To gain detailed knowledge of the mechanisms underlying the improvement in ADCC on defucosylation, it is necessary to obtain detailed structural information on the interaction between nonfucosylated IgG and sFccRIIIa possessing this N-glycan.

Results
Overall structure of Fc-sFccRIIIa complex We solved the crystal structure of the nonfucosylated glycoform of human IgG1-Fc complexed with human sFccRIIIa glycosylated only at Asn-45 and Asn-162 ( Fig. 1B). N-glycosylations at these two sites have been reported to be the minimal requirement for the expression of sFccRIIIa without proteolytic degradation in mammalian cells (Shibata-Koyama et al. 2009). The Fc fragment was cleaved from the nonfucosylated IgG1 produced in the Fut8 ) ⁄ ) cell line Ms704 (Matsumiya et al. 2007) by papain digestion, whereas the bis-glycosylated sFccRIIIa was expressed in the CHO ⁄ DG44 cell line with the Asn-to-Gln mutation of the remaining glycosylation sites (Shibata-Koyama et al. 2009) and treated with sialidase, resulting in homogeneous glycosylation with a biantennary fucosylated complex-type oligosaccharide (Fig. 1A). The crystal structure of nonfucosylated Fc complexed with the bis-glycosylated sFccRIIIa was determined by a molecular replacement technique with the crystal structure of a human fucosylated IgG1-Fc fragment complexed with nonglycosylated FccRIIIb (PDB ID code 1E4K) (Sondermann et al. 2000) and refined to 2.2-Å resolution ( Table 1).

Fc-sFccRIIIa interface
The present crystal structure exhibits a contact surface area of 1218 Å 2 , which is larger than those observed in the previously reported crystal structures of the complexes formed between fucosylated Fc and nonglycosylated sFccRIIIb (876 Å 2 in 1E4K, 887 Å 2 in 1T83, and 800 Å 2 in 1T89). In the present crystal structure, the surface areas occupied by the carbohydrate moieties and the protein portion of the Fc are 261 and 957 Å 2 , respectively. The sFccRIIIa-binding sites consist of the lower hinge, B strand, loop B ⁄ C, loop D ⁄ E, and loop F ⁄ G of Fc chain A as well as the lower hinge, B strand, and loop F ⁄ G of Fc chain B (Table 2).
On the receptor side, the Fc contact surface is composed of the loops located in the D2 domain. Although the D1 domain of sFccRIIIa makes little direct contacts with Fc, the carbohydrate chain linked to Asn-45 is oriented toward the C H 2 domain of Fc chain B (Table 2). Interestingly, the N-glycan at Asn-162 interacts with the Fc N-glycans, particularly those of chain A, through hydrogen bonds and van der Waals contacts (Fig. 2). The carbohydrate-carbohydrate interaction occupies approximately 12% (145 Å 2 ) of the total interface area.
The present crystal structure also shows that the aromatic ring of Tyr-296 of Fc chain A is flipped out and forms a hydrogen bond and van der Waals contacts with Man-4 of the Asn-162 N-glycan as well as Lys-128 of sFccRIIIa (Fig. 3), thereby stabilizing the complex. Surface plasmon resonance (SPR) data consistently indicated that alanine substitution of Tyr-296 resulted in impaired affinity for sFccRIIIa in both fucosylated and nonfucosylated IgG1 glycoforms ( Fig. S2 in Supporting Information). In the previously reported crystal structures of fucosylated Fc and nonglycosylated sFccRIIIb (i.e., 1E4K, 1T83, and 1T89), the D ⁄ E loop of Fc chain A, which contains this tyrosine residue, exhibits two alternate conformations. In one crystal structure (1T83), which bears the closest resemblance in overall structure to the present structure, the Tyr-296 aromatic ring is also flipped out, interacting with Lys-128 of sFccRIIIb. In the remaining crystal structures (1E4K and 1T89), this tyrosine ring makes intramolecular contacts with the core fucose residue and turns away from sFccRIII. These data suggest that Fc and sFccRIII have two binding modes with high and low affinities, depending on the orientation of the aromatic ring of Tyr-296 (Fig. 3).

Involvement of carbohydrate moieties in Fc-sFccRIIIa interactions
To assess the importance of the observed carbohydrate-carbohydrate interactions, we investigated the possible effects of enzymatic trimming of the N-glycans of sFccRIIIa on its binding to fucosylated and nonfucosylated IgG glycoproteins by SPR measurements (Fig. 4). The Fc binding affinity slightly increased on the removal of the outer lactosamine branches of the sFccRIIIa N-glycans, although Glc-NAc-5 of the Asn-162 N-glycan positively interacts with Man-4¢ and Arg-301 of Fc chain A, at least in the case of its nonfucosylated glycoform (Fig. 2B,C). This affinity enhancement can be ascribed to the removal of the outer carbohydrate branches of the N-glycan at Asn-45, which are in close spatial proximity to the C H 2 domain of Fc chain B and may cause steric hindrance, although their electron densities are not interpretable (Fig. 5A). Mutational deglycosylation at Asn-45 was reported to enhance the affinity for Fc (Shibata-Koyama et al. 2009). The removal of GlcNAc-2 and the trimannosyl parts of the sFccRIIIa N-glycans (i.e., Man-3, Man-4, and Man-4¢) have more pronounced positive and negative effects on the interactions with fucosylated and nonfucosylated Fc glycoforms, respectively (Fig. 4). These data are consistent with previously reported mutational deglycosylation data (Ferrara et al. 2006;Shibata-Koyama et al. 2009), as well as the present crystal structure in which these sugar residues show productive contacts with Tyr-296 and the nonfucosylated GlcNAc-1 residue of Fc chain A. Furthermore, our structural model indicates that fucosylation of GlcNAc-1 in chain A causes steric hindrance with GlcNAc-2 and Man-3 of the sFccRIIIa-Asn-162 N-glycan (Fig. 5B), which results in impaired interaction between these two glycoproteins. This explains why affinity enhancement on the extensive trimming of the sFccRIIIa N-glycans was selectively observed in the fucosylated glycoform of Fc.  , GN1  L157  L235  V158  L235, G236  G159  L235, G236  K161  G236, G237, P238, S239,  I332  GN1  GN1  GN2  GN1  M3  Y296  M4  M4¢  M4¢  Y296  GN5 R301, GN2, M4¢ GN5¢ Y296 Fuc GN5 The underlined residues are involved in hydrogen-bonding interactions.

Discussion
Although it is widely recognized that glycosylation of immune receptors can influence their affinities for the cognate ligands (Standley & Baudry 2000), not much is known about the specific roles of the individual glycans from a structural aspect. The present crystallographic data demonstrated that one of the two N-glycans (Asn162 N-glycan) of sFccRIIIa mediates the interaction with nonfucosylated Fc, thereby stabilizing the complex. To the best of our knowledge, this is the first atomic description of the carbohydratecarbohydrate interactions that mediate the formation of complexes between glycoproteins. Fucosylation of the Fc N-glycans inhibits these positive interactions, because of steric hindrance, thereby impairing IgG binding to FccRIIIa and the consequent ADCC activity. On the other hand, the N-glycan displayed at Asn-45 of sFccRIII negatively affects its binding to Fc (Shibata-Koyama et al. 2009), most probably because of steric clash with the C H 2 domain (Fig. 5A). Moreover, the present crystal structure, in comparison with those previously reported for the low-affinity complexes, indicates that Fc and sFccRIII have two binding modes depending on the orientation of the aromatic ring of Tyr-296 of Fc chain A (Fig. 3). Our previous NMR data demonstrated that Tyr-296 of the nonfucosylated Fc glycoform exhibits conformational multiplicity in its uncomplexed state (Matsumiya et al. 2007), suggesting that conformational selection is governed by the presence or absence of the fucose residue of the Fc N-glycan. Fucose depletion increases the incidence of the active conformation of Tyr-296, and thereby accelerates the formation of the high-affinity complex. This interpretation is supported by our previous data indicating that an increase in affinity by defucosylation was primarily ascribed to an enhanced association rate (Okazaki et al. 2004). Consistently, our SPR data show that sFccRIIIa has higher affinity for the nonfucosylated Fc than for the fucosylated Fc, even after extensive trimming of its N-glycans (Fig. 4).
These data are consistent with previously reported thermodynamic data indicating that affinity enhancement on defucosylation is characterized by favorable DH, but opposed by unfavorable DS (Okazaki et al. 2004). The favorable DH can be ascribed, at least partially, to the productive contacts caused by the flipping out of the Tyr-296 ring and accommodation The present crystallographic data provide a structural basis for the improvement in ADCC on defucosylation of IgG-Fc through the enhancement of its affinity for FccRIIIa, thus offering new clues for designing and engineering antibody medicines with improved efficacy. From a more general viewpoint, this study shows that the oligosaccharides displayed on proteins can modulate complex formation, positively and negatively, not only through intermolecular carbohydrateprotein and carbohydrate-carbohydrate interactions but also by influencing protein dynamics coupled with the selection of protein-protein interaction modes.

Preparation of IgG-Fc
The CHO ⁄ DG44 cell line (Urlaub et al. 1986) was kindly provided by Dr Lawrence Chasin (Columbia University, NY). The fucosylated and nonfucosylated forms (designated KM3060 and KM3416, respectively) of an anti-CCR4 chimeric antibody with human IgG1 ⁄ j constant regions along with their Y296A mutants, constructed using the Quik-Change Ò Multi Site-Directed Mutagenesis kit (Stratagene), were expressed by the CHO ⁄ DG44 cell line and the FUT8 ) ⁄ ) cell line Ms704, respectively, as previously described (Matsumiya et al. 2007). The nonfucosylated and fucosylated IgG1 glycoproteins were expressed and purified as previously described (Yamaguchi et al. 2006;Matsumiya et al. 2007). The Fc fragments were prepared by papain digestion as previously described (Yamaguchi et al. 2006), and the purity of the isolated Fc fragment was examined by SDS-PAGE.

Preparation of sFccRIIIa
The human sFccRIIIa mutant with two N-glycosylation sites at Asn-45 and Asn-162 was expressed and purified as previously described (Shibata-Koyama et al. 2009). In brief, the extracellular region of human FccRIIIa with an N-terminal hexahistidine tag and glutamine substitutions at Asn-38, model of sFccRIIIa bound to fucosylated Fc, in which the tyrosine ring does not make contact with the receptor, but makes intramolecular contact with the core fucose residue. The model is based on the crystal structure of the complex between fucosylated Fc and nonglycosylated sFccRIIIb (1E4K) with substitution of the receptor molecule by bis-glycosylated sFccRIIIb in the present crystal structure by superposing its D2 domains. In this model, the core fucose residue was attached to GlcNAc-1 by the a1-6 glycosidic linkage in the N-glycans in Fc chain A by superimposing the crystal structure of the fucosylated IgG1-Fc (3AVE). Chains A and B of Fc are cyan and pink, respectively, and sFccRIIIa is yellow.
Asn-74, and Asn-169 was expressed by the CHO ⁄ DG44 cell line as a recombinant protein modified exclusively with sialylated biantennary N-glycans (Shibata-Koyama et al. 2009). A series of sFccRIIIa glycoforms were prepared according to the protocols described below.
In each case, the reaction mixture was neutralized with 1.5 M Tris-HCl (pH 8.0) and then subjected to Ni-NTA chromatography to purify sFccRIIIa. sFccRIIIa was incubated under the same conditions in the absence of the corresponding enzyme(s) to prepare mock-treated controls.

Glycosylation profiling
N-glycosylation profiling of Fc and sFccRIIIa was carried out by the HPLC mapping method on the basis of elution profiles of pyridylamino derivatives of their N-linked oligosaccharides on Shim-pack HRC-octadecyl silica columns (Shimadzu), as previously described (Takahashi et al. 2002;Yamaguchi et al. 2006).

Crystallization, data collection, and structure determination
The nonfucosylated Fc fragment and desialylated sFccRIIIa mutant were mixed at a molar ratio of 1 : 1 and then applied to a gel filtration column (Superose 12; GE Healthcare) equilibrated with 20 mM Tris-HCl buffer, pH 7.5, containing 100 mM NaCl. Fractions containing the Fc-sFccRIIIa complex were concentrated to a total protein concentration of 20 mg ⁄ mL and used for crystallization.
Crystals were grown by the sitting-drop vapor-diffusion method at 20°C by mixing the Fc-sFccRIIIa complex with a reservoir [0.1 M MES, pH 6.5, 12% (w ⁄ v) PEG 20000]. Crystals were soaked in a cryoprotectant solution [0.1 M MES, pH 6.5, 20% (w ⁄ v) PEG 20000, 15% (v ⁄ v) glycerol] and flash-frozen. Diffraction data were collected at 100 K using a wavelength of 0.9 Å on a beamline BL44XU (SPring-8). Data processing and reduction were carried out using the HKL-2000 software package (Otwinowski & Minor 1997). The crystals belong to the space group P4 1 2 1 2, with cell dimensions a = b = 77.3 Å , c = 350.3 Å at 2.2-Å resolution. The value of the Matthews coefficient is 3.34 Å 3 ⁄ Da for one molecule, corresponding to a solvent content of 63.2%. Data collection, phasing, and refinement statistics are summarized in Table 1. The structure of the complex between nonfucosylated Fc and bis-glycosylated sFccRIIIa was determined by molecular replacement using the structures of the complex between fucosylated Fc and nonglycosylated sFccRIIIa (Sondermann et al. 2000) (PDB ID code 1E4K) using MOLREP (CCP4, 1994;Vagin & Teplyakov 1997). The model was furthermore built using the program COOT (Emsley & Cowtan 2004) and then improved by several cycles of manual rebuilding and refinement using the program REFMAC5 (Murshudov et al. 1997). The final model contains the Fc residues 229-443 (chain A) and 229-444 (chain B), and the sFccRIIIa residues 5-30 and 41-174. Phasing and refinement statistics are summarized in Table 1. There are no residues in disallowed regions of the Ramachandran plot. Buried surface area and surface complementarity were calculated using AREAIMOL (Lee & Richards 1971). SFCHECK and PROCHECK (CCP4, 1994) were used for structure validation. Molecular graphics were prepared using PyMOL (DeLano 2002). Atomic coordinates and structure factors have been deposited in the PDB under accession code 3AY4.

Surface plasmon resonance measurements
Interactions of the sFccRIIIa glycoforms with fucosylated and nonfucosylated IgG glycoproteins were analyzed by SPR using the T100 biosensor system (GE Healthcare). Mouse anti-tetra-His IgG antibody (Qiagen) was immobilized on CM5 biosensor chips by an amine coupling method according to the manufacturer's instructions. The individual glycoforms of the hexa-His-tagged sFccRIIIa glycoproteins were captured by the immobilized anti-tetra-His antibodies at a flow rate of 5 lL ⁄ min using HBS-EP+ buffer (10 mM HEPES, 150 mM NaCl, 3 mM EDTA, and 0.05% v ⁄ v surfactant P20, pH 7.4) at 25°C. Assays were performed using nonfucosylated and fucosylated IgG glycoproteins at seven concentrations (ranging from 4 to 267 nM) in a mobile phase at a flow rate of 30 lL ⁄ min using the HBS-EP+ buffer at 25°C. The dissociation constant (K D ) was calculated by steady-state affinity analysis using Biacore T100 evaluation software version 2.0.1 (GE Healthcare). To repeat experiments, sFccRIIIa and human IgG1 were removed from the sensor tips by injecting 10 mM glycine-HCl, pH 1.5, at a flow rate of 60 lL ⁄ min for 1 min. K D values are the mean ± SD of three independent experiments.
Figure S1 Superposition of the present crystal structure of the complex between IgG-Fc (cyan and pink) and sFccRIIIa (yellow) and the structures previously reported for the IgG-Fc-sFccRIIIb complexes (gray).

Figure S2
Contribution of Tyr-296 of human IgG1 to its interactions with sFccRIIIa.
Additional Supporting Information may be found in the online version of this article.
Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.