Enhanced Natural Killer Cell Binding and Activation by Low-Fucose IgG1 Antibody Results in Potent Antibody-Dependent Cellular Cytotoxicity Induction at Lower Antigen Density

Antibodies of the human IgG1 isotype are commonly used for therapeutic applications as they can mediate multiple effector functions including antibody-dependent cellular cytotoxicity (ADCC), complement-dependent cytotoxicity (CDC), and direct apoptosis induction (1–3). ADCC, a lytic attack on antibody-targeted cells, is triggered following binding of leukocyte receptors (FcγR) to the antibody Fc region. Several mouse and clinical studies indicate that ADCC is an important therapeutic mechanism of clinically effective antibodies (4–7). FcγRIIIa is the predominant cγR of natural killer (NK) cells responsible for ADCC activation. The FcγRIIIa gene (FCGR3A) displays an allelic polymorphism that generates receptors containing either a phenylalanine (F) or a valine (V) at a position critical in mediating ADCC, amino acid position 158. This variation results in human IgG1 antibodies binding with higher affinity to the NK cells of homozygous FCGR3A-158V donors than those of homozygous FCGR3A-158F donors, and seems to result in more effective NK cell activation (8, 9). Importantly, several reports have recently shown that FCGR3A genotype influences the clinical efficacy of human IgG1-type anti-CD20 antibody rituximab (3, 10, 11) with the clinical response of patients bearing FcγRIIIa-158F being significantly inferior to the patients with the FcγRIIIa-158V receptors (5–7). These reports underscore the importance of ADCC in clinical outcomes.

ADCC activity is influenced by the structure of complex-type oligosaccharides linked to CH2 domain of the antibody Fc region. The content of galactose (12, 13), bisecting N-acetylglucosamine (14, 15), and fucose (16, 17) in the antibody oligosaccharide have each been reported to effect ADCC. In previous studies, we have shown that fucose is the most critical antibody oligosaccharide component and that the removal of fucose from IgG1 oligosaccharides results in a very significant enhancement of both ADCC in vitro (∼100 fold) and antitumor activity in vivo (17, 18). However, many therapeutic antibodies currently approved or under clinical development are produced using Chinese hamster ovary cells that express high level of α1,6-fucosyltransferase and consequently produce low amounts antibody lacking fucose (17). Therefore, we generated a fucosyltransferase knockout Chinese hamster ovary cell line that can stably produce nonfucosylated antibodies with enhanced ADCC (19) that behaves in other respects indistinguishable from the parental line.

Although therapeutic antibodies are demonstrating increasing success in the clinic, especially in the field of cancer treatment (1, 2), the full potential may be limited by the low and variable antigen expression on the target cells. This is due to three factors that can limit the efficacy of therapeutic antibodies. First, there are often individual patients or clinical subtypes of the cancer that are unresponsive to the antibody therapy due to limited or heterogeneous antigen expression (3, 20). Second, residual tumor cells after antibody therapy may be selected to express less antigen than pretreated tumor cells (21, 22), become resistant to additional treatment, and, thus, may lead to poor prognosis. Third, antigen may be thought unsuitable as antibody targets due to low antigen expression. Consequently, it is important to improve antibody efficacy for tumor cells with lower antigen expression level.

Previous studies have shown that ADCC depends on antigen expression levels on target cells (23). Hence, we asked if low-fucose IgG1 with enhanced ADCC might overcome the problem of low antigen density on target cells. In this study, we quantitatively analyzed the effect of antigen levels on the ability of low-fucose IgG1 to induce ADCC on target cells to determine the potential therapeutic advantage of low-fucose IgG1 for future clinical applications. Additionally, we focused on NK cell binding to low-fucose IgG1 and the resultant cellular activation to understand the mechanism of potent ADCC induction, especially at low antigen density.

Blood Donors. Blood donors were randomly selected from healthy volunteers registered at Tokyo Research Laboratories, Kyowa Hakko Kogyo, Co., Ltd. All donors gave written informed consent before analyses.

Cell Lines. Mouse T-cell lymphoma cell line EL4, human B lymphoma cell line CA46, CCRF-SB (ATCC CCL-120), ST486, Raji, and Daudi were purchased from the American Type Culture Collection (Rockville, MD). A human B lymphoma cell line P32/ISH (JCRB0095) was purchased from Health Science Research Resources Bank (Osaka, Japan).

Preparation of Target Cell Lines. A human CC chemokine receptor 4 (CCR4) expression plasmid CAG-pcDNA-CCR4 bas been previously described (18). An expression plasmid encoding human CD20, designated pKANTEXCD20, was constructed by inserting CD20 gene, cloned from a human leukocyte cDNA library (BD Biosciences Clontech, Palo Alto, CA) by PCR into mammalian cell expression vector pKANTEX93 (24). EL4 cells were transfected with CAG-pcDNA-CCR4 or pKANTEXCD20 by electroporation and grown in the presence of 0.5 mg/mL G418 sulfate to obtain G418-resistant clones. For some clones with higher CD20 expression, gene amplification in the presence of methotorexate (4-amino-10-methylpteroylglutamic acid, MTX) was done. Multiple clones with differential protein expression levels were screened by nonquantative flow cytometry as described below.

Antigen Expression Analysis by Flow Cytometry. A nonquantitative flow cytometry was done for the screening of transfected target cell clones. Biotinylated KM2760 was prepared using EZ-Link Sulfo-NHS-LC-Biotin (Pierce, Rockford, IL) as described by the manufacturer. Approximately 1 × 10⁶ cells were stained for 1 hour on ice with 3 μg/mL of biotin-labeled KM2760 for CCR4⁺ clones or a 10-fold dilution of FITC-conjugated anti-CD20 monoclonal antibody (Beckman Coulter, Tokyo, Japan) for CD20⁺ clones. For CCR4-expressing clones, cells were washed and then stained with phycoerythrin-conjugated streptavidin (Becton Dickinson Japan, Tokyo, Japan) as the secondary reagent. The stained cells were analyzed on an EPICS XL-MCL flow cytometer (Beckman Coulter).

To determine the absolute number of the antibody binding sites per cell, a quantitative flow cytometry analysis (25) was done using DAKO QIFIKIT (DakoCytomation, Kyoto, Japan). Briefly, 1 × 10⁶ cells were stained for 1 hour on ice with saturating concentration of KM2160 (60 μg/mL) for CCR4⁺ clones or mouse anti-CD20 monoclonal antibody (clone 2H7, BD Biosciences PharMingen, San Diego, CA; 40 μg/mL) for CD20⁺ clones. Cells were then washed and then stained with FITC-conjugated anti-mouse IgG (DakoCytomation) for 1 hour on ice. Standard beads coated with known amount of mouse IgG molecules were also stained with FITC-conjugated anti-mouse IgG. The stained samples were analyzed using a flow cytometer, and the numbers of binding sites per cell were calculated by comparing the mean fluorescent intensity value of the stained cells to a standard curve obtained by regression analysis of the mean fluorescent intensity values of standard beads.

Cytotoxicity Assay. ADCC was measured using a standard 4-hour ⁵¹Cr release assay as previously described (17). CDC was measured by a nonradioactive method. Target cells (1 × 10⁶ cells/50 μL medium), varying concentrations of antibodies (in 50 μL medium), and human complement serum (50 μL, ×2 diluted with medium; Sigma, St. Louis, MO) were distributed into 96-well flat-bottomed plates. All cells and reagents were diluted with RPMI 1640 (Life Technologies, New York, NY) containing 10% heat-inactivated fetal bovine serum. After incubation at 37°C for 2 hours, aliquots of the cell proliferation reagent WST-1 (Roche Diagnostic GmbH, Penzberg, Germany) were added to each well (15 μL) and the plates were incubated for a further 4 hours to allow the formazan dye production by the metabolically active cells. The percent cytotoxicity was calculated from the absorbance at 450 nm minus the reference absorbance at 650 nm (A₄₅₀ − A₆₅₀) of each well according to the formula:

where E is the A₄₅₀ − A₆₅₀ of experimental well, S is that in the absence of monoclonal antibody (cells were incubated with medium and complement alone), and M is that of medium and complement alone.

Isolation of Natural Killer Cells. Peripheral blood mononuclear cells (PBMC) were prepared from peripheral blood by density gradient centrifugation using Lymphoprep (AXIS SHIELD, Dundee, United Kingdom). PBMC were then subjected to negative magnetic sorting to obtain NK cell fraction, by removing CD3, CD14, CD19, CD36, and IgE-positive cells using MACS NK cell isolation kit and MidiMACS (Miltenyi Biotec, Bergisch Gladbach, Germany). The phenotype of the isolated NK cell fraction was confirmed as >95% CD56⁺CD3⁻ before any experiments.

IgG1 Binding Analysis to Natural Killer Cells. Isolated NK cells (1 × 10⁶) were incubated on ice for 1 hour with 1% bovine serum albumin (BSA)/PBS containing 10 μg/mL IgG1. After incubation with IgG1, cells were washed twice with PBS and analyzed after staining with phycoerythrin-conjugated anti-human IgG Fab'₂ (DakoCytomation) on a flow cytometer. In some experiments, IgG1 was incubated with 20 μg/mL BSA-conjugated CCR4 peptide before the incubation with NK cells for the IgG1 ligation. The BSA-conjugated CCR4 peptide was prepared by conjugating the partial CCR4 peptide including the binding site of KM2760 and KM3060 (corresponding to amino acid residues 2 to 29 of human CCR4; ref. 18) to the amino groups of BSA using 4-(N-maleimidomethyl)-cyclohexane-1-carboxylate N-hydroxysuccinimido ester (Sigma) as previously described (26).

Analysis of Natural Killer Cell Activation. Isolated NK cells (1 × 10⁵/100 μL medium/well) and the equal number of target cells in 100 μL medium were dispensed in round-bottom 96-well plate. IgG1 (final concentration, 1 μg/mL) or medium alone were added to each well and incubated at 37°C. Cells were harvested periodically (4, 24, and 72 hours) and double-stained with phycoerythrin-conjugated anti-CD56 monoclonal antibody (Beckman Coulter) and FITC-conjugated anti-CD69 monoclonal antibody (BD Biosciences PharMingen) on ice for 1 hour. After washing, NK cells, which can be gated out by their relative smaller values of forward scatter and side scatter compared with target cells, were measured of their CD56/CD69 expression on a flow cytometer.

Antibody-Dependent Cellular Cytotoxicity of Anti-CD20 IgG1 Fucose Variants against B Lymphoma Cell Lines with Differential CD20 Expression Levels. KM3065 (17) is a low-fucose variant of anti-CD20 chimeric IgG1 rituximab, which is widely used for the treatment of B-cell disorders. KM3065 and rituximab have identical amino acid sequences and thus show identical CD20 binding activities, whereas they differ in the percentage of antibody lacking fucose-containing carbohydrates linked to Asn²⁹⁷ in CH2 domain of heavy chain (Table 1). Due to its low fucose contents and consequent strong FcγRIIIa binding, KM3065 has been shown to exhibit enhanced ADCC against CD20⁺ Raji cells and WIL2-S cells (17, 27). However, how the ADCC enhancement by defucosylation was influenced by the intrinsic properties of target cells remains to be verified. To investigate the relationships between the enhancement of ADCC by defucosylation and the antigen expression level on target cells, we investigated ADCC of the two anti-CD20 IgG1s upon six B lymphoma cell lines with differential CD20 expression levels as shown in Fig. 1. CD20 binding sites on each cell lines were determined by quantitative flow cytometry method (25) with values ranging within an order of magnitude (7.8 × 10⁴ to 5.9 × 10⁵). Using human PBMCs with effector cells, KM3065 showed enhanced ADCC upon all the target cell lines compared with rituximab. Although the correlation between the ADCC and the numbers of CD20 binding sites was not statistically significant, ADCC values mediated by both the two IgG1s tended to be higher for target cell lines with higher CD20 expression. The result suggests that target cell lines from different origins are not suitable for the quantitative analysis of ADCC depending on antigen expression because their own cellular backgrounds, other than the antigen amount, determines their inherent sensitivities against ADCC.

Preparation of Transfectant Target Cells with Varying Expression Levels of Exogenous Antigen. To establish a simpler system for the investigation of the influence of the antigen amount on tumor cells independently from different cellular backgrounds, we next constructed panels of transfectant tumor cell lines with wide ranges of antigen expression levels. We used two antigen-specific systems for this aim, CCR4 and CD20. In addition to anti-CD20 IgG1s described above, two fucose variants of chimeric anti-CCR4 IgG1s have also been generated: the conventional highly fucosylated anti-CCR4 IgG1, KM3060, and its low-fucose counterpart, KM2760 (Table 1). They share identical amino acid sequences and CCR4 binding activities, whereas KM2760 has shown enhanced ADCC against CCR4⁺ T leukemia cells (18).

Panels of target cells with a range of expression levels of the chosen antigens were constructed by transfecting CCR4 or CD20 genes into murine thymoma EL4 cells. EL4 cells were chosen as the host cell to generate target cells for the quantitative evaluation of ADCC because of their resistance to antibody-independent cytotoxic activity of human NK cells that could obscure antibody-dependent cytotoxicity. Following single cell cloning, eight CCR4-positive clones (designated CCR4/EL4-A to CCR4/EL4-H) and seven CD20-positive clones (CD20/EL4-A to CD20/EL4-G) were screened by nonquantitative flow cytometry as illustrated in Fig. 2. To generate high expressing CD20⁺ clones, gene amplification in the presence of MTX was used: CD20/EL4-D and CD20/EL4-F were produced using 200 nmol/L MTX selection and CD20/EL4-G was produced using 1,000 nmol/L MTX.

The numbers of antibody binding sites per cell on each established EL4 clones were then determined using quantitative flow cytometry method and ranged from 1.3 × 10³ to 5.4 × 10⁴ for CCR4 and from 1.2 × 10⁴ to 5.8 × 10⁵ for CD20 (Table 2). All clones were confirmed to have similar cell diameters by flow cytometry (data not shown), implying that antigen densities on the surface of each target cell were expected to be proportional to the numbers of antibody binding sites shown in Table 2. Although the parent murine EL4 cell would not be expected to express human antigens and no visible staining were observed in nonquantitative flow cytometry (Fig. 2), the methodology used calculated ∼600 sites per cell for each of the antigens. This was probably because of the marginal nonspecific or cross-reactive staining in the quantitative flow cytometry analysis. It should be noted that the CCR4 expression levels of all the human T-cell leukemia cell lines that we have previously described (18) ranged between those of CCR4/EL4 clone A and clone G used in these studies, as observed in nonquantitative flow cytometry (data not shown). CD20 expression on clinical B lymphoma cells has been reported to be ∼10⁵, although it depends on the clinical subtype of the lymphoma to some extent (3, 28–30). Taken together, the expression levels of the target cell clones established in this study were considered to be representative of the in vivo expression levels of patient's tumor cells.

Antibody-Dependent Cellular Cytotoxicity of IgG1 Fucose Variants against Experimental Target Cells with Various Antigen Expression Levels. We next measured the ADCC of both IgG1 fucose variants upon the experimental target cell lines described above. As shown in Fig. 3, using PBMC from peripheral blood of two healthy donors (donor A and B) as effector cells, the low-fucose KM2760 showed higher ADCC activity than that of the high-fucose KM3060 against each of the CCR4-transfected target cells. Analysis of the ADCC activity of the two IgG1s were each found to fit to a sigmoid-shaped curve with different maximal levels of cytotoxicity for the different IgG1s calculated by the four-parameter regression equations (Table 3). In addition, the minimum amount of antigen that was required to induce detectable level of ADCC of KM2760 was less than that of KM3060. For example, with donor A effector cells, KM2760 exhibited ADCC against all CCR4-expressing clones (≥1.3 × 10³ CCR4 binding sites), whereas KM3060 required at least 1.5 × 10⁴ CCR4 binding sites in the presence of 3 μg/mL IgG1s (Fig. 3, top right). Further, we calculated the number of CCR4 binding sites required to achieve the half-maximal cytotoxicitiy of KM3060 for both KM2760 and KM3060 (Table 3). For example, with PBMC from donor A and 3 μg/mL antibody concentration, KM3060 required 2.1 × 10⁴ CCR4 binding sites on target cells for its half-maximal cytotoxicity (14.4%), whereas KM2760 required only 9.6-fold less antigen to achieve the same level of cytotoxicity, indicating KM2760 required only approximately one tenth of CCR4 expression than KM3060 in this experimental condition. Similar quantitative relationships in the ADCC of the two IgG1s were observed in all cases in Fig. 3 except for a little smaller difference in efficacies of the two IgG1s (5.6-fold) with donor A PBMC and 0.1 μg/mL antibody concentration.

A similar advantage of low-fucose IgG1 in the lysis of low-antigen-expressing cells and the maximal cytotoxicity was also observed for CD20 system using PBMC from two other donors (Fig. 4), indicating that low-fucose IgG1 can reduce the antigen number required for ADCC in CD20 system in addition to increasing the maximal ADCC achievable. However, the shifts in required antigen amount was somewhat less compared with CCR4 system (the low-fucose KM3065 required ∼3-fold less CD20 amount than rituximab in each experimental condition; Table 4). This may be due to the relatively higher ADCC mediated by the high-fucose IgG1 (rituximab). We also investigated another important Fc-mediated function of antibody, CDC, of the two anti-CD20 IgG1s on the CD20-transferred target clones (Fig. 5). In contrast to ADCC, the cytotoxicities of the two IgG1s were not significantly different. The anti-CCR4 IgG1s did not exhibit any measurable CDC activity against any of the CCR4-transfected target cells irrespective of their fucose contents (data not shown). These differing results between CCR4 and CD20 might be due to many factors including the expression levels of the antigens, the inherent capacity of the antigens to mediate CDC, or the antigen-binding affinity of individual antibody clone.

Natural Killer Cell Binding of IgG1 Fucose Variants. The mechanism responsible for ADCC enhancement by fucose removal from IgG1 is an increase of IgG1 binding to FcγRIIIa (16, 27). This enhancement also results in effective ADCC at a lower antigen density. We have previously shown the improved FcγRIIIa affinity of KM3065 compared with rituximab by ELISA method (31) and determined the kinetic and thermodynamic parameters of the interaction between FcγRIIIa and the two anti-CD20 IgG1s (27). The enhanced FcγRIIIa binding by fucose removal was also confirmed for anti-CCR4 IgG1s (data not shown).

Furthermore, the binding of IgG1 fucose variants to purified NK cells prepared from blood of one healthy donor was measured by flow cytometry (Fig. 6). The low-fucose anti-CCR4 and anti-CD20 IgG1s each showed slightly higher binding to NK cells than their highly fucosylated counterparts, which was consistent with the result of other analyses (16). Next, we examined whether IgG1 binding to NK cells was enhanced when anti-CCR4 IgG1 was ligated with clustered antigen (BSA-conjugated CCR4 peptide) to imitate the physiologic condition in which FcγRIIIa molecules on NK cells are cross-linked by the complex of antibody and target cells. Under these conditions, binding to NK cells was further enhanced for each antibody; however, the increase was greater for the low-fucose KM2760.

Antigen-Specific Natural Killer Cell Activation. To investigate whether increased binding of low-fucose IgG1 to FcγRIIIa and NK cells is a key step for NK cell activation during ADCC, we next investigated expression pattern of the activation marker CD69 on CD56⁺ NK cells in the presence of both anti-CD20 IgG1 and four of the target cell clones with differing CD20 expression levels (EL4, CD20/EL4-C, CD20/EL4-E, and CD20/EL4-G; the numbers of CD20 binding sites per cell of each clone were shown in Table 2). The control experiment was first done by stimulating purified NK cells in the presence of phorbol 12-myristate 13-acetate and ionomycin for 24 hours to confirm the up-regulation of CD69 in whole population of NK cells (Fig. 7A).

When mixed with the EL4-derived target cells, NK cells could be clearly identified and removed from the flow cytometeric analysis due to their relative small diameter reflected in forward scatter (Fig. 7B); accordingly, CD56/CD69 expression levels in NK cells mixed with target cells for 4 and 24 hours were analyzed in the absence or presence of various antibodies (KM3065, rituximab, or anti-CCR4 KM2760 as an irrelevant control antibody; Fig. 7C). In the absence of antibody, no CD69 up-regulation was observed either at 4 or 24 hours (Fig. 7C,, medium column), which was consistent with the resistance of the EL4 clones to antibody-independent cytotoxicity observed in the ADCC assays (Figs. 3 and 4).

In the presence of CD20-positive target cells at 4 hours, the two anti-CD20 IgG1s each increased the proportion of activated NK cells in a manner dependent on CD20 expression level. The low-fucose anti-CD20 antibody, KM3065, consistently induced higher numbers of activated NK cells than rituximab: 36% to 45% activation of NK cells for KM3065 versus 8% to 22% for rituximab. At 24 hours, KM3065 activity was increased, but no longer obviously antigen-density dependent, with activation of majority of NK cells (64-71%) irrespective of CD20 expression levels on target cells. With rituximab at 24 hours, the activation was also significantly increased from 4-hour incubation (24-48%), although still lower than that with KM3065, and increased at higher antigen density. NK cell activation by both IgG1s were considered maximal at this time point because the CD69-positive NK cell proportions in the presence of anti-CD20 IgG1s at 24 hours was unchanged when the incubation time was extended to 48 hours (data not shown). Taken together, these results provide evidence of a possible mechanism by which low-fucose IgG1 can increase ADCC on target cells with low antigen expression, namely activation of more NK cells by low fucose antibodies than conventional high-fucose IgG1.

When mixed with CD20-negative EL4 cells and anti-CD20 IgG1, or when mixed with CD20-positive cells and irrelevant IgG1 KM2760, NK cells did not exhibit up-regulated CD69 expression at any time during the experimental period. These results indicate the strict dependency of low-fucose IgG1 in mediating ADCC on the presence of antigen, despite its antigen-independent binding capacity to NK cells as shown in Fig. 6.

There are two distinct subsets of human NK cells identified by cell surface density of CD56: a small population of CD56^bright (<10%) and the remainder CD56^dim (32). To identify which population was involved in anti-CD20 IgG1-mediated NK cell activation, the flow cytometer histograms obtained above were reanalyzed according to CD56 expression on NK cells. As a result, no apparent increase in CD69 expression in CD56^bright subset was observed in all the samples shown in Fig. 7C, suggesting that the NK cell subset responsible for ADCC mediated by both the IgG1 fucose variants was CD56^dim. An example is shown in Fig. 7D, 24-hour incubation in the presence of CD20/EL4-E.

One of the major obstacles to optimizing the efficacy of therapeutic antibodies is low and heterogeneous antigen expression that (a) allows cells to evade the effective treatment, (b) induces the selection of low antigen cell, and (c) makes some antigenic targets resistant to antibody therapeutics. For example, the most successful anticancer antibody, rituximab, shows an inferior clinical response rate for lymphoma subtypes that have a relatively lower expression of CD20, such as chronic lymphocytic leukemia or small lymphocytic lymphoma, compared with highly responsive follicular lymphoma with higher CD20 expression, although it should be noted that other factors such as the differential expression of complement-inhibitory proteins (CD46, CD55, CD59) among these clinical subtypes might also affect the rituximab responsiveness (3). It has been shown that antigen expression is a critical factor of the efficacy of the anti-HER2 IgG1 trastuzumab (1, 2, 4, 33), for which treatment is restricted to HER2-overexpressing 20% to 30% breast cancer patients as determined by immunohistochemistry or fluorescence in situ hybridization before therapy (34). Trastuzumab-mediated ADCC, considered a critical therapeutic mechanism of the antibody, is dependent on the levels of HER2 on target cells: The antibody can induce ADCC only against tumor cells with ≥10⁵ HER2 molecules per cell (35, 36). In addition, residual tumor cells with lower antigen expression could escape from antibody therapy, which might lead to relapse and resultant poor prognosis. As the problem of the low or heterogeneous antigen expression on target cells is relevant for many other therapeutic antibodies, enhanced ADCC due to the augmentation of low-fucose IgG1 binding to FcγRIIIa is a promising way to address this problem.

Previous studies have only shown the enhanced ADCC of low-fucose IgG1 upon target cell lines with a fixed and relatively high density of antigen (16–18). How the antigen expression of target cells modulates the advantage by fucose reduction on ADCC remained unanswered. To investigate whether low-fucose IgG1 could exhibit potent ADCC on target cells with less antigen amount, we constructed panels of target cells with a range of known antigen expression levels. Because the obtained clones were uniform in their cell diameters and susceptibilities to antigen-independent lysis by NK cells, the use of these experimental target cells enabled the measurement of the dependence of ADCC activity on antigen amount, and hence density, per target cell. Interestingly, antigen amount required for ADCC induction for low-fucose IgG1 was lower than that for high-fucose IgG1, suggesting that improvement of IgG1 binding to FcγRIIIa on effector cells by fucose depletion can reduce antigen amount necessary for ADCC induction. Because the antigen-binding activities do not vary among fucose variant IgG1s and consequently the IgG1 amount bound on the surface of target cells were unaffected by fucose content in IgG1 as confirmed by flow cytometry (data not shown), the potent ADCC induction by low-fucose IgG1 might be due to the efficient activation of effector cells via more effective interactions with FcγRIIIa even in the presence of low IgG1 density on target cells.

Although we did not investigate FCGR3A-158 genotype of the PBMC donors in the current studies, FCGR3A-158V/F polymorphism is also likely to affect the antigen density required for ADCC because our data imply that IgG1-FcγRIIIa affinity affects the antigen density threshold required for ADCC induction. However, fucose depletion is expected to lower the required antigen density for any PBMC donors because it improves IgG1-FcγRIIIa binding independently of the FCGR3A genotypes (31), although the difference in required antigen densities between IgG1 fucose variants may vary with the FCGR3A genotype.

Although reduction of fucose showed the benefits of ADCC induction in low antigen density for both CCR4 and CD20 specificities, the CCR4 system displayed a more marked difference between ADCC of fucose variants than CD20 system. Possible explanations for this difference include the following: The higher amount of CD20 on the experimental target cells (approximately one order of magnitude higher than CCR4; Table 2) might be enough for the ADCC induction of high-fucose IgG1. Alternatively, the inherent structures of each antigen molecules, especially the mobility and the configuration of the molecules in lipid bilayer on cell surface, might affect the efficacy of high-fucose IgG1. In contrast to CCR4, which is a seven-transmembrane-spanning G-coupled receptor (37) and is presumed to have poor mobility in lipid bilayer, CD20 is highly mobile and naturally forms multimer (38–40). In addition, most of anti-CD20 antibodies including rituximab can redistribute CD20 into lipid rafts, which determines the capacity to induce strong CDC (41). Although the relation of ADCC and the redistribution into lipid rafts has been poorly defined, high density of IgG1 bound on CD20 multimer or localized CD20 in lipid rafts may induce ADCC even by high-fucose IgG1. The hypothesis that clustering IgG1 could enhance ADCC is supported by the reports of dimerized antibodies having potent ADCC (42, 43).

As expected from the identical binding capacities of fucose variant IgG1s to C1q (16), KM3065 and rituximab did not vary in their CDC activities against the target cell clones with varying antigen expression. Although the anti-CD20 IgG1s exerted CDC even upon low or intermediate CD20⁺ clones, whose antigen expression levels overlapped with those of higher CCR4 expressing clones, anti-CCR4 IgG1s did not mediate detectable CDC activity against all CCR4⁺ clones. This discrepancy may again support the hypothesis that CCR4 and CD20 might differ in their inherent properties of the molecule, such as clustering capacity in lipid bilayer, which is critical for CDC susceptibility (41) and possibly affect the susceptibility to high-fucose IgG1-mediated ADCC shown in Fig. 4. Because the low-fucose anti-CCR4 IgG1 greatly augmented the ADCC of high-fucose anti-CCR4 IgG1, the inability of anti-CCR4 antibody to mediate CDC also implies the therapeutic benefit of low-fucose IgG1 that it would be applicable for wider range of molecular targets where CDC cannot be expected due to inherent molecular properties of the antigen.

We further investigated the mechanism of enhanced ADCC by low-fucose IgG1 in terms of NK cell activation. Consistent with the ADCC analysis, KM3065 induced more CD69-positive activated NK cells than rituximab even in the presence of lower antigen density, suggesting that low-fucose IgG1 can efficiently recruit and activate effector cells via the increased binding to FcγRIIIa. The activated NK subset was predominantly CD56^dim, a not unexpected finding because CD56^bright NK cells express low level of FcγRIIIa and are presumed to function as cytokine responder/producer cells rather than cytotoxic killer cells (30, 44–46). Further, NK cells were shown to be activated only by IgG1 bound on target cells despite low-fucose IgG1 itself could bind slightly to NK cells in the absence of antigen (Fig. 6). This is a very important finding as it suggests that enhanced FcγRIIIa/NK interactions should not induce nonspecific side effects. The mechanism of the tumor-specific activation of NK cells via low-fucose IgG1 remains unknown; however, it may be that intracellular signals through FcγRIIIa are not triggered until the density of receptor-bound IgG1 is elevated above a minimal level by antigen molecules present on target cells. Enhanced binding of low-fucose IgG1 to FcγRIIIa might lower the threshold of IgG1 density on target cells necessary for NK cell activation and consequently results in ADCC induction against targets with fewer antigen molecules. This hypothesis is supported by the results herein that low-fucose IgG1 binding was greatly enhanced by preligating the antibody with antigen to mimic cell-surface antigen-antibody clusters increase NK cell binding at concentration of highly fucosylated IgG1 did not bind (Fig. 6).

In conclusion, our data showed that low-fucose IgG1 shows potent ADCC upon target cells with lower antigen density compared with conventional antibodies, through the effective and antigen-specific activation of NK cells due to augmented binding to FcγRIIIa. These features of low-fucose IgG1 could be therapeutically beneficial because clinical tumors often display the heterogeneity in their antigen expression levels and in addition the number of effector cells accessible to in vivo tumor would be fewer compared with in vitro experimental condition.

We thank Dr. Philip Wallace for helpful suggestions and critical reading of the manuscript.