Engineered antibody Fc variant with selectively enhanced FcγRIIb binding over both FcγRIIaR131 and FcγRIIaH131

Engaging inhibitory FcγRIIb by Fc region has been recently reported to be an attractive approach for improving the efficacy of antibody therapeutics. However, the previously reported S267E/L328F variant with enhanced binding affinity to FcγRIIb, also enhances binding affinity to FcγRIIaR131 allotype to a similar degree because FcγRIIb and FcγRIIaR131 are structurally similar. In this study, we applied comprehensive mutagenesis and structure-guided design based on the crystal structure of the Fc/FcγRIIb complex to identify a novel Fc variant with selectively enhanced FcγRIIb binding over both FcγRIIaR131 and FcγRIIaH131. This novel variant has more than 200-fold stronger binding affinity to FcγRIIb than wild-type IgG1, while binding affinity to FcγRIIaR131 and FcγRIIaH131 is comparable with or lower than wild-type IgG1. This selectivity was achieved by conformational change of the CH2 domain by mutating Pro to Asp at position 238. Fc variant with increased binding to both FcγRIIb and FcγRIIa induced platelet aggregation and activation in an immune complex form in vitro while our novel variant did not. When applied to agonistic anti-CD137 IgG1 antibody, our variant greatly enhanced the agonistic activity. Thus, the selective enhancement of FcγRIIb binding achieved by our Fc variant provides a novel tool for improving the efficacy of antibody therapeutics.

Several works describe engineering the Fc region to enhance the effector function of mAbs by increasing the binding affinity for active FcgRs (FcgRIa, FcgRIIa and FcgRIIIa) with amino acid substitutions. For example, antibodies engineered to have increased binding affinity for FcgRIIIa exhibited superior in vitro ADCC activity and in vivo antitumor activity compared with wild-type mAbs (Stavenhagen et al., 2007;Zalevsky et al., 2009). In addition to protein engineering, glyco-engineered tumor-specific mAbs with afucosylated N-linked oligosaccharides at Asn297 in the Fc region showed increased binding affinity for human FcgRIIIa and mouse FcgRIV, which resulted in enhancing the therapeutic activity in mouse models (Nimmerjahn and Ravetch, 2005;Mossner et al., 2010).
In contrast to these activating FcgRs that function as immunostimulatory receptors, inhibitory FcgRIIb is reported to function as an immunomodulatory receptor (Li and Ravetch, 2011;White et al., 2011). The inhibitory receptor FcgRIIb is the only IgG Fc receptor expressed on B-cells and plays a critical role in regulating B-cell homeostasis (Heyman, 2003;Nimmerjahn and Ravetch, 2008). Immune complexes (ICs) coengage FcgRIIb and B-cell receptor (BCR) and then selectively suppress B-cells that recognize cognate antigen (Heyman, 2003). FcgRIIb also regulates other B-cell stimulators that amplify B-cell proliferation and differentiation and suppresses the expression of costimulatory molecules (Leibson, 2004;Crowley et al., 2009).
FcgRIIb also plays an inhibitory role in a homeostatic checkpoint of dendritic cells (DCs) for inducing tolerance or immunity by ICs, in contrast to FcgRIIa. While ligation of FcgRIIa led to DC maturation, targeting FcgRIIb alone did not activate DCs (Boruchov et al., 2005). FcgRIIb on DCs induced peripheral tolerance by inhibiting antigenic processing and DC activation to suppress T-cell activation (Desai et al., 2007).
Besides these inhibitory effects of FcgRIIb, several groups have recently reported that FcgRIIb enhances the agonistic activity of anti-tumor necrosis factor receptor (TNFR) superfamily antibody by working as a scaffold (White et al., 2013). Proliferation of antigen-specific T-cells induced by anti-CD40 antibodies was abrogated in FcgRIIb-deficient mice (White et al., 2011;Li and Ravetch, 2012). Anti-death receptor 5 (DR5) agonist antibody also required FcgRIIb to exert its agonistic activity and initiate apoptosis in lymphoma cells (Wilson et al., 2011). The FcgRIIb-expressing cells are considered to work as a scaffold to efficiently crosslink anti-CD40 or DR5 antibodies bound to target cells via FcgRIIb.
In order to exploit these properties, engineering Fc region to increase the binding affinity to FcgRIIb is considered to be a promising approach. Introducing S267E/L328F substitutions into the Fc region of human IgG1 increased the binding affinity to FcgRIIb 430-fold without increasing that to FcgRI, FcgRIIa H131 or FcgRIIIa (Chu et al., 2008). Anti-CD19 antibody with the Fc promoted suppression of B-cell activation and proliferation in SLE mouse model by coengaging FcgRIIb with BCR (Horton et al., 2011). Recent reports noted that anti-IgE antibody with S267E/L328F Fc variant reduced the production of IgE in vivo  and that the Fc1 domain fused with S267E/L328F Fc variant suppressed degranulation and calcium mobilization of mast cells (Cemerski et al., 2012). Moreover, agonistic antibodies against CD40 or DR5 showed more potent agonistic activity in vivo with enhanced FcgRIIb binding Ravetch, 2011, 2012). These reports clearly demonstrate that engineered Fc with enhanced binding to FcgRIIb has various therapeutic applications. However, it was reported that S267E/L328F substitutions also enhanced the binding to one of the FcgRIIa allotypes, FcgRIIa R131 , to a level similar to the binding to FcgRIIb (Smith et al., 2012). Therefore, when applying the substitutions to mAb therapeutics, the consequence of increasing the binding to FcgRIIa should be considered.
It is reported that a high incidence of thromboembolic complication was observed in patients treated with anti-CD154 or anti-VEGF antibody in clinical settings (Boumpas et al., 2003;Scappaticci et al., 2007). The ICs composing of these antibodies and the antigens activated platelets and induced thrombosis in human FcgRIIa transgenic mice by crosslinking FcgRIIa expressed on the platelets (Meyer et al., 2009;Robles-Carrillo et al., 2010). Moreover, intravenous immunoglobulins also induced FcgRIIa-mediated platelet aggregation in vitro (Pollreisz et al., 2008). These results demonstrate that even wild-type IgG1 generally has a potential risk of inducing thromboembolism through an FcgRIIa-dependent mechanism, suggesting that this risk could be increased if the binding to FcgRIIa is enhanced. Therefore, considering the therapeutic applications, antibody with selectively enhanced binding to FcgRIIb relative to both FcgRIIa R131 and FcgRIIa H131 is preferred.
In this work, we investigated antibodies with selectively enhanced binding to FcgRIIb over both FcgRIIa R131 and FcgRIIa H131 . We screened a large set of single substituted variants of human IgG1 for binding to human FcgRs and found a distinct variant that distinguishes FcgRIIb from both FcgRIIa allotypes. We solved its crystal structure in complex with human FcgRIIb and elucidated the structural recognition mechanism by which it recognizes FcgRIIb over both FcgRIIa allotypes. In order to improve its FcgRIIb binding, we conducted further optimization and identified a novel antibody variant with over 200-fold more enhanced binding to FcgRIIb without increasing the binding to either FcgRIIa allotype. Structural analysis of this variant revealed the mechanism of the improved binding to FcgRIIb. We also report that antibody that also has enhanced binding to FcgRIIa is more likely to activate platelets and induce aggregation and would have a shorter in vivo half-life compared to wild-type IgG1. We also confirmed that this variant enhanced the agonistic activity of anti-CD137 antibody in vitro. Our results indicate that the novel Fc variant with purely selectively enhanced binding to FcgRIIb is applicable to a broad range of therapeutic antibodies, since it avoids increasing the potential to activate or aggregate platelets and to have rapid clearance in vivo.

Preparation of IgG1 variants
IgG1 variants used in the experiments were expressed transiently in FreeStyle TM 293F cells transfected with plasmids encoding heavy and light chain and purified from culture supernatants using protein A. Site-directed mutagenesis of the Fc region was performed using QuikChange Site-Directed Mutagenesis Kit (Stratagene) or In-Fusion HD Cloning Kit (Clontech).

Construction, expression and purification of FcgRs
The genes encoding the extracellular region of human FcgRs were synthesized based on the sequence information obtained from the National Center for Biotechnology Information. FcgRs were fused with 6Â His-tag at their C-terminus. Vectors containing FcgRs were transfected into FreeStyle TM 293F cells. The receptors were purified from the harvested culture supernatants by using ion exchange chromatography, nickel affinity chromatography and size exclusion chromatography.
Binding and affinity analysis for human FcgRs by surface plasmon resonance (SPR) The interaction of antibody variants with FcgRs was monitored using Biacore instruments (GE Healthcare), as previously described (Richards et al., 2008). Antibody variants were captured on the CM5 sensor chip (GE Healthcare) on which antigen peptide or protein A/G (Thermo Scientific) was immobilized, followed by injection of FcgRs. The binding of each variant to each FcgR was normalized by the captured amount of antibody on the sensor chip and was expressed as a percentage of that of IgG1. Kinetic analysis was performed by global fitting of binding data with a 1 : 1 Langmuir binding model using Biacore Evaluation Software (GE Healthcare).
Crystalization, data collection and structure determination of the complex of Fc fragment of human IgG1 with P238D substitution, Fc(P238D) and FcgRIIb Crystals were obtained by 1 : 1 mixing of protein complex (10 mg/ml) with 0.1 M Bis-Tris pH6.5, 0.2 M ammonium acetate, 2.7% (w/v) D-galactose, 17% PEG3350 in sitting drop vapor diffusion setups at 208C. For data collection, crystals were flash frozen at 95 K in precipitant solution containing 22.5% ethylene glycol. Diffraction data to 2.6 Å were collected using the Photon Factory beamline BL-1A. Data were processed with Xia2 (Winter, 2010), XDS Package (Kabsch, 2010) and Scala (Evans, 2006). The structure was determined by molecular replacement with PHASER (McCoy et al., 2007) using Fc part of Fc(IgG1)/FcgRIIIa structure (PDB ID: 3SGJ) and FcgRIIb structure (PDB ID: 2FCB) as search models. The asymmetric unit was formed by a single 1 : 1 complex. A model was built with the program Coot (Emsley et al., 2010) and refined with the program REFMAC5 (Murshudov et al., 2011). Crystalization, data collection and structure determination of the complex of Fc fragment of V12 variant, Fc(V12) and FcgRIIb Crystals were obtained by 1 : 1 mixing of protein complex (10 mg/ml) with 0.1 M Bis-Tris pH6.5, 0.2 M potassium phosphate dibasic, 19% PEG3350 in hanging drop vapor diffusion setups at 208C. For data collection, crystals were flash frozen at 95 K in precipitant solution containing 20% ethylene glycol. Diffraction data to 2.86 Å were collected at the SPring-8 beamline BL-32XU. After data processing and structure determination by molecular replacement using Fc(P238D)/FcgRIIb structure, a model was built with the program Coot and refined with the program REFMAC5. The asymmetric unit was formed by a single 1 : 1 complex.
Data collection and refinement statistics of both crystals are summarized in Supplementary Table S1. All graphical presentations were prepared with PyMOL (DeLano, 2002).

Preparation of Fc fragment and FcgRIIb complexes
We cloned recombinant Fc fragments corresponding to the heavy chain residues from 216 (EU numbering) to C-terminus for the crystallization. Cys 220 was replaced with Ser so that the free cysteine would not make disulfide bonds. FcgRIIb expressed by FreeStyle TM 293F cells in the presence of the mannosidase I inhibitor, kifunensine, was purified and treated with endoglycosidase F1 fused with glutathione S-transferase for deglycosylation as described previously to minimize the heterogeneity of the oligosaccharide (Grueninger-Leitch et al., 1996). The complexes of Fc fragment and the deglycosylated FcgRIIb were prepared by mixing Fc fragment with a little excess of the deglycosylated FcgRIIb, and purified by size exclusion chromatography.

Thermal shift assay
Melting temperature (T M ) of the C H 2 region was measured as previously described (He et al., 2010). The SYPRO orange dye (Invitrogen) was added to 0.1 mg/ml protein solutions at a final dilution of 1 : 500. Fluorescence measurements were employed using a real-time polymerase chain reaction instrument, Rotor-Gene Q (QIAGEN). The temperature was increased from 30 to 998C at a heating rate of 48C/min.

Stability study of high concentration antibodies
Antibodies were concentrated to 100 mg/ml in a histidine buffer (20 mM histidine, 150 mM NaCl, pH6.0) by ultrafiltration (Millipore). The solutions were stored at 258C to assess the stability of the antibodies. High-molecular-weight (HMW) species were analyzed with TSK-GEL G3000SWXL column (TOSOH) by size exclusion high-performance liquid chromatography at initial time point and after 4 weeks. The mobile phase contained 300 mM NaCl in 50 mM phosphate at pH7.0. The percentage of HMW species peak in total protein peak area was calculated using Empower Waters software.

Binding affinity analysis of antibodies for human FcRn
Recombinant, soluble human FcRn (hFcRn) was expressed in HEK293 cells and prepared as described previously (Dall'Acqua et al., 2006). The interaction of soluble hFcRn was monitored as described previously (Igawa et al., 2010).

Platelet aggregation and activation study using washed platelets
Fresh blood from healthy volunteers whose FcgRIIa genotype is R/R or H/H was anticoagulated with 0.5 ml of 3.8% sodium citric acid -citrate in blood collection tubes. Platelet-rich plasma (PRP) was obtained by centrifuging at 200 g for 15 min and then removing the supernatant. PRP was washed in modified Tyrode buffer (137 mM NaCl, 2.7 mM KCl, 12 mM NaHCO 3 , 0.42 mM NaH 2 PO 4 , 2 mM MgCl 2 , 5 mM HEPES, 5.55 mM dextrose, 0.35% bovine serum albumin) with 1.5 U/ml apyrase and resuspended at a concentration of 3 Â 10 8 /ml in modified Tyrode buffer. Washed platelets were then incubated with preformed IC for 5 min. The Preformed IC was prepared by mixing the anti-IgE mAb having different Fc variants (200 mg/ml) with its antigen (229 mg/ml), human IgE, at a molar ratio 1 : 1. Five minutes after the incubation, 30 mM ADP was added to induce the first wave of platelet aggregation. Platelet aggregation was measured by an aggregometer (MCM Hema Tracer 712; MC Medical) at 378C with stirring at 1000 rpm.

Cells and reagents
CTLL-2 cells (mouse T lymphocyte cell line, No.RCB0637) were provided by the RIKEN BRC through the National Bio-Resource Project of the MEXT, Japan. Raji cells (human Burkitt's lymphoma cell line, ATCC No.CCL-86) were purchased from the American Type Culture Collection. Both cell lines were cultured in RPMI 1640 medium (Nacalai tesque), supplemented with 10% heat-inactivated foetal bovine serum (Bovogen). The culture medium for CTLL-2 was supplemented with 10 ng/ml recombinant mouse interleukin (IL)-2 (PeproTech). The culture medium for Raji cells was supplemented with 10 mM HEPES, 1 mM sodium pyruvate (Nacalai tesque), 4.5 g/l D-glucose, 1.5 g/l sodium bicarbonate (Sigma-Aldrich).

Flow cytometry analysis of CD137 expression
To analyze mouse CD137 expression on CTLL-2 cells, antimouse CD137 clone 1D8 variable region fused with the Fc domain of human IgG1 (Shuford et al., 1997) or human IgG1 isotype control (Alexis, Lausen) was applied. Goat F(ab 0 )2 anti-human IgG-FITC #732596 (Beckman Coulter) was used as the detection antibody. Cell Lab Quanta SC MPL system (Beckman Coulter) was used for cell acquisition and data analysis was conducted using FlowJo software (Tree Star Inc.).

Results
Screening and characterization of variants to selectively enhance FcgRIIb binding by comprehensive mutagenesis We screened a large set of over 500 variants prepared by replacing each of about 30 residues in the lower hinge and C H 2 region with every naturally occurring amino acid except for cysteine, for binding to human FcgRIIa H131 , FcgRIIa R131 and FcgRIIb in order to identify substitutions that selectively enhance FcgRIIb binding relative to both FcgRIIa allotypes. The effect of each substitution on the binding to FcgRIIa R131 , FcgRIIa H131 and FcgRIIb is compared in Fig. 1A and B. Only two substitutions, P238D and L328E, were found to enhance FcgRIIb binding while decreasing the binding to both FcgRIIa allotypes, although previously reported S267E and L328F increased both binding.
We compared the dissociation constant (K D ) of wild-type human IgG1, P238D variant and L328E variant (Supplementary Table S2) and the changes in affinity from IgG1 are shown in Fig. 2. Both P238D and L328E variants enhanced affinity to FcgRIIb, while reduced affinity to both FcgRIIa R131 and FcgRIIa H131 . P238D variant reduced the affinity for all the active FcgRs and enhanced that for FcgRIIb more significantly than L328E variant.
Crystal structure of the FcgRIIb complex with IgG1 Fc fragment with P238D Fc(P238D) bound to FcgRIIb in an asymmetric fashion between the two C H 2 domains (C H -A and C H -B) like previously reported Fc(IgG1) and FcgR complexes. In Fig. 3A, the overall structure of Fc(P238D)/FcgRIIb was compared with known Fc(IgG1)/FcgRIIa R131 (PDB ID: 3RY6) (Ramsland et al., 2011). Despite the high-sequence homology between FcgRIIb and FcgRIIa R131 (Supplementary Fig. S1), the domain arrangement of C H 2-B in Fc(P238D)-FcgRIIb interface was different from that in Fc(IgG1)/FcgRIIa R131 .
Structural comparisons of both C H 2 domains with those of Fc(IgG1) complexed with FcgRIIa R131 revealed conformational changes of loops around position 238 in both C H 2 domains ( Fig. 3B and C). Pro238 in wild-type human IgG1 formed an inside hydrophobic core with Val263, Val266, V323, Asn325 and Leu328. On the other hand, Asp238 in P238D variant shifted out of the core and it was exposed to the solvent region. Instead, Leu235 occupied the vacated space which Pro238 occupied in Fc(IgG1), forming another hydrophobic core in C H 2-A, but not in C H 2-B. As a result, there was a large cavity surrounded by Val263, Val266, Val273, Val323, Asn325 and Leu328 in C H 2-B. Asp238 in C H 2-B formed an electrostatic interaction with Lys334 in the same chain.
In the C H 2-A domain, the main chain of Gly237 formed a novel hydrogen bond with Tyr160 of FcgRIIb (Fig. 3B). A weak hydrogen bond between the main chain of Asp238 and the side chain of Thr158 of FcgRIIb was also observed. In addition, the side chain of Asp238 formed van der Waals contacts with the side chain of Thr158 in FcgRIIb. In the C H 2-B domain, the conformation of the loop around Asp270 changed from that of Fc(IgG1) complexed with FcgRIIa R131 , and Asp270 formed a salt bridge with Arg131 of FcgRIIb (Fig. 3C). Binding affinity for FcgRIIa R131 , FcgRIIa H131 and FcgRIIb was evaluated. The binding to FcgRIIa R131 , FcgRIIa H131 or FcgRIIb of template antibody was normalized to 100. The binding to FcgRIIa R131 and FcgRIIb is shown in panel A and the binding to FcgRIIa H131 and FcgRIIb is shown in panel B. Novel substitutions to selectively improve FcgRIIb binding, L328E and P238D, are indicated in blue and red, respectively. Previously reported substitutions, S267E and L328F, to improve FcgRIIb but also FcgRIIa R131 binding are indicated in green and yellow, respectively. Screening, characterization and design of variants to further enhance FcgRIIb binding by comprehensive mutagenesis using the P238D variant as a template In order to further enhance the affinity to FcgRIIb, we combined P238D with L328E or S267E/L328F, previously known variant to increase binding to FcgRIIb and FcgRIIa R131 . Unexpectedly, the additive effect by combining P238D with those substitutions was not observed. Affinities for FcgRIIb of these variants are listed in Supplementary Table S3.
Then, we screened substitutions to improve FcgRIIb binding when they were combined with P238D by comprehensive mutagenesis using the P238D variant as a template. Approximately 400 variants with an additional single substitution onto the P238D variant were generated, and the affinity for FcgRIIb was determined. Six effective substitutions (E233D, G237D, H268D, P271G, Y296D and A330R) were identified. We further combined these substitutions to create V12 variant having all the six substitutions described above. The fold increases of the variants against FcgRIIb over P238D variant are listed in Table I

Crystal structure of the FcgRIIb complex with Fc(V12)
The same conformational changes around Asp238 induced by P238D substitution in Fc(P238D)/FcgRIIb were also observed in Fc(V12)/FcgRIIb. In the C H 2-A domain, the introduced residues, Asp233, Asp237 and Arg330, were located in the interface with FcgRIIb (Fig. 5A). The weak electron density of Asp233, replaced from Glu, was observed. As Lys113 of FcgRIIb is located near Asp233, they might interact with each other. In the C H 2-A domain, Asp237, replaced from Gly, made van der Waals contacts with Trp87 and Tyr160 of FcgRIIb. In addition, the hydrogen bond distance between NH of the main chain at position 237 in Fc and OH of the Tyr160 side chain in FcgRIIb changed from 3.1 Å in Fc(P238D) (Fig. 3C) to 2.9 Å in Fc(V12) (Fig. 5A). A weak hydrogen bond between the side chain of Asp238 and the side chain of Thr158 of FcgRIIb was also observed. The electron density of the side chain of Arg330, which replaced Ala, was not clearly observed. As Glu86 of FcgRIIb is located near Arg330, a weak interaction between those residues could be present.
In the C H 2-B domain, H268D and P271G substitutions induced a further conformational change of a loop around Asp270 (Fig. 5B). The substitution from Pro to Gly at position 271 caused the main chain to flip and consequently rearranged the 266 -273 loop and Arg131 of FcgRIIb, which also formed a salt bridge with Asp270 in C H 2-B of Fc(V12) as well as in Fc(P238D). In addition, the electron density of Gly271 in C H 2-B was clearly observed, though in Pro271 in Fc(P238D) it was not. This conformational change of the loop caused a rearrangement of Arg292 in C H 2-B of Fc(V12). The other substituted residue, Asp268, formed an electrostatic interaction with the shifted Arg292.

Pharmacokinetic and biophysical property assessment of V12 variant
In order to characterize the pharmacokinetic aspect of V12 variant, we measured the affinity for hFcRn and the pharmacokinetics in hFcRn transgenic mice. The binding affinity of V12 variant for hFcRn at pH6.0 was comparable with that of wild-type IgG1 and the in vivo half-life was also comparable (Table II).
T M of the C H 2 domain was measured by thermal shift assay. The T M of V12 variant decreased by 88C relative to wild-type IgG1 (Table II). To assess the real-time stability for pharmaceutical application of V12 variant, a stability study at an antibody concentration of 100 mg/ml was performed. The formation of HMW species of V12 variant after storage for 4 weeks at 258C was comparable with that of wild-type IgG1 (Table II).

In vitro activation and aggregation of platelets by ICs consisting of IgE and anti-IgE antibody with S267E/L328F variant or V12 variant
Platelets obtained from two donors homozygous for FcgRIIa R131 genotype and incubated with IC consisting of IgE and anti-IgE S267E/L328F variant increased the expression of CD62p and PAC-1 on the platelets, but those incubated with IC consisting of IgE and anti-IgE with the V12 variant did not (Supplementary Fig. S2A and B). On the other hand, when we used platelets obtained from two donors homozygous for FcgRIIa H131 genotype, IC consisting of IgE and anti-IgE S267E/L328F variant slightly upregulated the expression of CD62p and PAC-1 on the platelets but IC consisting of anti-IgE with the V12 variant did not after the incubation compared with the control ( phosphate-buffered saline) (Supplementary Fig. S2C and D).
Next, the platelet aggregation induced by ICs was evaluated with an aggregometer. After the addition of ADP, only IC consisting of IgE and anti-IgE with S267E/L328F substitutions aggregated the platelets obtained from two donors homozygous for FcgRIIa R131 genotype, while IC consisting of IgE and anti-IgE with V12 variant did not ( Fig. 6A and B). On the other hand, IC consisting of neither variant induced the aggregation of the platelets obtained from two donors homozygous for FcgRIIa H131 genotype ( Fig. 6C and D).

Enhancement of agonistic activity of anti-CD137 antibody with enhanced FcgRIIb binding Fc
Several reports have described that agonistic anti-TNFR superfamily antibodies generally require FcgRIIb coengagment for their agonistic activity and that enhancing the binding affinity of the antibodies for FcgRIIb could increase the agonistic activity (White et al., 2011;Wilson et al., 2011;Li and Ravetch, 2012). Therefore, we tested whether V12 also has the same property using agonist antibody against CD137 (clone 1D8), which is one of the TNFR superfamily.
Mouse T lymphoma cell line CTLL-2 was used as mouse CD137-expressing cells (Fig. 7A), and human B lymphoma cell line Raji was used as human FcgRII-positive cells (Hernández et al., 2010). T-cell activating agonistic activity of anti-CD137 antibody was measured with the production of mouse IFN-g production of CTLL-2 co-cultured with Raji cells. Consistent with the previous reports, V12 variant as well as S267E/L328F variant increased IFN-g production induced by agonist anti-CD137 antibody compared with intact human IgG1 by more than 5-fold (Fig. 7B).

Discussion
Several efforts to improve FcgR binding by Fc engineering have been reported to date. While most of the engineering enhances the binding affinity for activating FcgRs (Stavenhagen et al., 2007;Zalevsky et al., 2009;Mossner et al., 2010), engineering to enhance binding affinity to inhibitory FcgR is limited (Chu et al., 2008). S267E/L328F variant, the Fc variant reported to have the highest affinity for inhibitory FcgR binding, enhanced the binding affinity for FcgRIIb by 355-fold; however, it also enhanced the binding affinity for FcgRIIa R131 to the same level as FcgRIIb (Smith et al., 2012). The high similarity between FcgRIIb and FcgRIIa R131 suggests that discriminating FcgRIIb and FcgRIIa R131 by Fc engineering is challenging (Supplementary Fig. S1).
In this work, we investigated the substitutions to distinguish FcgRIIb from FcgRIIa R131 by comprehensive mutagenesis and discovered a highly selective substitution, P238D, which provides the highest selectivity for FcgRIIb relative to all the other active FcgRs including FcgRIIa R131 . Fc variants which discriminate FcgRIIb from FcgRIIa R131 were extremely rare, probably because FcgRIIb is highly homologous to FcgRIIa R131 (Supplementary Fig. S1) and the Fc-FcgRIIb interface would be also highly homologous to the Fc-FcgRIIa R131 . To the best of our knowledge, this is the first report illustrating an Fc variant that enhances FcgRIIb binding while distinguishing FcgRIIb and FcgRIIa R131 precisely.
To elucidate the structural mechanism by which P238D variant discriminates FcgRIIb from FcgRIIa R131 , we solved the crystal structure of the complex of Fc(P238D) and FcgRIIb. Our structural analysis indicated that the high selectivity of P238D variant was achieved by a dynamic conformational change of the Fc-FcgRIIb interface that was induced by P238D substitution into wild-type IgG1. In reported Fc(IgG1) structures, Pro238 forms a hydrophobic core with its surrounding residues. Therefore, substituting this hydrophobic Pro238 to hydrophilic Asp should cause large free energy loss for desolvation of Asp to maintain the same structure. To avoid this free energy loss, Asp238 showed a large shift out of its original position to achieve access to the solvent region. Additionally, the position of Pro238 in Fc(IgG1) was occupied by Leu235 instead in C H 2-A of Fc(P238D) complexed with FcgRIIb. As the result, the large conformational change of loop233-240 attached to the hinge region was observed. This change would affect the domain arrangement between the C H 2-A and B domains because both C H 2 domains connect by S -S bonds in the hinge region and cannot move independently. In fact, the relative arrangement of C H 2-A and B domains in the Fc(P238D)/FcgRIIb complex was different from that of Fc(IgG1)/FcgRIIa R131 , despite FcgRIIb having the highest homology to FcgRIIa R131 . As the result of these dynamic conformational changes, Fc(P238D) acquired two novel interactions with FcgRIIb. One is the hydrogen bond between the main chain of Gly237 in C H 2-A and the side chain of Tyr160 in FcgRIIb. Both FcgRIIa allotype cannot make this hydrogen bond because corresponding residue of this Tyr is Phe in FcgRIIa ( Supplementary Fig. S1). So, this interaction would play a critical role for distinguishing FcgRIIb from FcgRIIa. The other one is the salt bridge between Asp270 in C H 2-A and Arg131 of FcgRIIb. This salt bridge would contribute to improve not only the binding affinity to FcgRIIb but also the selectivity over FcgRIIa H131 because this allotype has His as the corresponding residue of Arg131 in FcgRIIb. On the other hand, the S267E/L328F variant increased binding affinity to both FcgRIIb and FcgRIIa R131 to the same extent, while it did not increase binding affinity to FcgRIIa H131 or other FcgRs. From the reported complex structure of Fc(IgG1)/ FcgRIIa R131 , we elucidated that Glu267 of the Fc with S267E/ L328F substitution would form a salt bridge with Arg131 of the FcgRIIb and also of the FcgRIIa R131 . This might explain the reason for the lack of selectivity of the S267E/L328F variant to FcgRIIb over FcgRIIa R131 , both of which have Arg at position 131. Large conformational change induced by P238D would also explain the reason for lack of additive effect of combining P238D with L328E or S267E/L328F. Then, in order to further increase the binding affinity to FcgRIIb, we conducted a second comprehensive mutagenesis using the P238D variant as a template rather than combining substitution(s) effective for wild-type IgG1 template, since the interface of Fc(P238D)/FcgRIIb was considered to be significantly changed from that of Fc(IgG1)/FcgRIIb. P238D-based comprehensive and combinatorial study identified V12 variant whose affinity for FcgRIIb was significantly increased from P238D variant. Especially, P271G markedly increased affinity for FcgRIIb. The substitution of fixed proline to flexible glycine would contribute to release conformational stress when a salt bridge between Asp270 in Fc (V12) and Arg131 in FcgRIIb is formed, which is considered to contribute to the affinity improvement for FcgRIIb.
Structural analysis of P238D suggests that P238D substitution seemed to destabilize the hydrophobic core of C H 2 domain, although this could be partially compensated by Leu235 as found in C H 2-A domain. The conformational changes seem to be the cause of decreased T M for V12 variant. Although this reduced T M raised concern regarding the storage stability of Fc(V12) for pharmaceutical application, its stability (aggregation tendency at 100 mg/ml) was comparable with that of wild-type IgG1.
It is known that, for IgG1 antibody to have a long half-life in vivo, binding to FcRn at acidic pH is important. The affinity of V12 variant for hFcRn was comparable with that of wildtype IgG1 and, consistently, the in vivo half-life of V12  The group mean + SD are given for the parameter (n ¼ 3 each). a K D for FcRn was measured at pH6.0 by SPR. b Half-life of intravenously injected IgG1 and V12 variant at 1 mg/kg in hFcRn transgenic mouse.
Engineered antibody Fc variant with selectively enhanced FcgRIIb binding variant was also comparable with that of wild-type IgG1. However, it should be noted that because the binding affinity of V12 variant to murine FcgRIIb (the mouse counterpart of human FcgRIIb) was not increased, its effect on the pharmacokinetics was not addressed in this study. ICs induce platelet aggregation and activation when antibody binds to FcgRIIa expressed on the platelet surface (Boumpas et al., 2003;Scappaticci et al., 2007). The previous studies suggest that ICs consisting of IgG with enhanced binding to FcgRIIa have the potential to induce platelet aggregation and activation more intensively than IC consisting of wild-type IgG. However, the effect of engineered Fc with enhanced binding affinity to FcgRIIa on platelet activation and aggregation induced by IC has not so far been investigated. In this report, IC consisting of IgE and anti-IgE antibody with enhanced binding affinity to both FcgRIIb and FcgRIIa R131 (S267E/L328F variant) induced the activation and aggregation of platelets obtained from FcgRIIa genotype of R/R131 homozygous donors, but not from FcgRIIa genotype of H/H131 homozygous donors. On the other hand, anti-IgE antibody with selectively enhanced binding affinity to FcgRIIb did not induce the activation or aggregation of platelets from donors with either genotype. Platelets whose FcgRIIa genotype is R/H131 are considered to express both FcgRIIa allotypes on their surface, while platelets whose FcgRIIa genotype is R/R131 express only FcgRIIa R131 . Therefore, a similar tendency will be observed in the platelets from heterozygous donors, although the induction of the aggregation and activation might be milder.
It is known that FcgRIIa internalizes ICs by endocytosis and transfers them to lysosomal degradation (Zhang and Booth, 2010), while FcgRIIb recycles the internalized ICs Fig. 6. Platelet aggregation studies incubated with ICs. Platelet aggregation was evaluated after platelets were incubated with ICs consisting of IgE with anti-IgE V12 variant (blue), that of anti-IgE S267E/L328F variant (red), IgE and anti-IgE IgG1 (green) or phosphate-buffered saline (black) for 5 min after being primed with ADP. Aggregation of the platelets from two donors with FcgRIIa R/R131 homozygous genotype are shown in panels A and B, respectively. Aggregation of the platelets from two donors with FcgRIIa H/H131 homozygous genotype are shown in panels C and D, respectively.  (Bergtold et al., 2005;Mousavi et al., 2007). These reports suggest that an antibody with enhanced binding affinity to FcgRIIa would be more rapidly eliminated from plasma by FcgRIIa-mediated uptake and degradation than wild-type IgG1, while an antibody with enhanced binding only to FcgRIIb would have a longer half-life because the antibody would be recycled back to the cell surface. Indeed, an antibody with enhanced binding to both murine FcgRII and FcgRIII (counterpart of human FcgRIIa) exhibited more rapid clearance from plasma compared with an antibody with enhanced binding only to murine FcgRII (data not shown).
These facts indicate that therapeutic IgG with S267E/L328F substitution has the potential to induce platelet aggregation and activation and to be rapidly cleared from plasma in patients with FcgRIIa R131 genotype. One report described that the allelic frequency of R/R131, R/H131 and H/H131 is 31, 44 and 25%, respectively, among healthy Caucasians, and 26, 43 and 31%, respectively, among healthy African Americans (Lehrnbecher et al., 1999). Another report showed that in Japanese, Chinese and Asian Indian populations the R/R131 genotype occurs in 6, 6 and 31%, respectively (Osborne et al., 1994). Considering the proportion of populations with R/R and R/H genotypes, enhancing the binding affinity to FcgRIIa R131 would have a substantial impact.
Previous studies using S267E/L328F or S267E substitution(s) have demonstrated that enhancing affinity to FcgRIIb is a promising application for therapeutic antibodies against CD19, IgE, DR5 and CD40 Ravetch, 2011, 2012;Chu et al., 2012;Hammer, 2012). Consistent with these reports, V12 variant, as well as S267E/L328F variant, enhanced agonistic activity of antibody against CD137, one of TNFR superfamily molecules, compared with intact human IgG1. Since agonistic antibodies against TNFR superfamily are currently being explored for cancer immunotherapy, enhancement of the agonistic activity of these antibodies by selectively improving the binding affinity for FcgRIIb could be a promising approach. In addition, in Ba/F3 cells expressing constitutively active mutants of the receptor tyrosine kinase, Kit, ICs that crosslinked FcgRIIb and Kit inhibited growth factorindependent proliferation (Malbec and Daëron, 2012). In another report, ICs suppressed the TLR4-mediated response of DCs in rheumatoid arthritis patients through FcgRIIb. Each of these effects of FcgRIIb could be enhanced by applying Fc with enhanced affinity for FcgRIIb. Moreover, ICs significantly suppressed expression of CD40, CD80 and CD86 on FcgRIIb-overexpressing DCs, suggesting that in DCs, using ICs consisting of an antibody variant with selectively enhanced FcgRIIb affinity relative to FcgRIIa might polarize IC-triggered activating signals to inhibitory signals (Zhang et al., 2011).

Conclusion
In this study, we screened antibody Fc variant which selectively enhances the binding affinity to FcgRIIb over both FcgRIIa R131 and FcgRIIa H131 by comprehensive mutagenesis. We identified a distinct substitution, P238D, that could discriminate FcgRIIb from FcgRIIa R131 precisely, and crystal structural analysis revealed that this substitution substantially changed the recognition interface of Fc-FcgRIIb. We further designed an antibody variant with 200-fold higher affinity for FcgRIIb than IgG1 without increasing the affinity for other active FcgRs. The variant was comparable with IgG1 in terms of pharmacokinetics and storage stability. We also showed that an antibody with increased affinity for FcgRIIa has an increased possibility of inducing platelet activation and aggregation and of being rapidly cleared from plasma. Since previous studies and our study using agonist anti-CD137 antibody suggested that increasing the binding affinity to FcgRIIb has various therapeutic applications, our engineered Fc, which enhances binding selectivity to FcgRIIb, is expected to have a significant therapeutic potential.

Supplementary data
Supplementary data are available at PEDS online.