Superior In vivo Efficacy of Afucosylated Trastuzumab in the Treatment of HER2-Amplified Breast Cancer

The enhancement of immune effector functions has been proposed as a potential strategy for increasing the efficacy of therapeutic antibodies. Here, we show that removing fucose from trastuzumab (Herceptin) increased its binding to Fc γ RIIIa, enhanced antibody-dependent cell-mediated cytotoxicity, and more than doubled the median progression-free survival when compared with conventional trastuzumab in treating preclinical models of HER2-amplified breast cancer. Our results show that afucosylated trastuzumab has superior efficacy in treating in vivo models of HER2-amplified breast cancer and support the development of effector function – enhanced antibodies for solid tumor therapy. Cancer Res; 70(11); 4481 – ©2010 AACR.


Introduction
Herceptin (trastuzumab) is a humanized antibody for treating HER2/ErbB2-overexpressing breast cancer. Trastuzumab was initially approved for the treatment of women with HER2-positive metastatic breast cancer in combination with standard cytotoxic chemotherapy (1) and as a monotherapy (2,3). Multiple studies have been completed in several large adjuvant or early breast cancer trials for HER2-positive breast cancer (4). A survival benefit was noted after only 2 years of follow-up, which is impressive in breast cancer (4). Multiple mechanisms of action are thought to contribute to the tumor-inhibitory effect of trastuzumab. Binding of trastuzumab to HER2 has a direct inhibitory effect on HER2-amplified tumor cells. Trastuzumab disrupts the ligand-independent HER2-HER3 interaction resulting in rapid inhibition of HER3/PI3K/AKT signaling (5). Ultimately, this leads to an increase in the CDK2 inhibitor, p27, resulting in cell cycle arrest of the cancer cells (5)(6)(7)(8). Another primary effect on tumor cells is the ability of trastuzumab to inhibit HER2 ectodomain shedding (9), which may have therapeutic significance. Trastuzumab is also known to have synergistic effects when combined with chemotherapy (10).
In addition to directly inhibiting tumor cell signaling, trastuzumab can also mediate the effector functions of immune cells through the constant region (Fc) of the antibody. As a humanized IgG 1 , it binds to Fcγ receptor III (RIII) and is a potent mediator of antibody-dependent cell-mediated cytotoxicity (ADCC). The ability of trastuzumab to mediate ADCC is strictly related to HER2 density (11). Tumor cells that overexpress HER2 are observed to have 100-fold more HER2 on their cell surface than normal adjacent breast epithelial cells (12). As a result, trastuzumab mediates ADCC very effectively against target cell lines that overexpress HER2 but displays only background activity against cells that express normal levels of HER2. Depleting trastuzumab effector functions by Fc modification or deleting the FcγR functions from the mouse results in significant reduction of trastuzumab efficacy, demonstrating the importance of immune effector functions in trastuzumab response (13). Clinical relevance for effector functions of Herceptin is supported by a report demonstrating that trastuzumab efficacy correlated with the high-affinity FcγRIII genotype (V/V; ref. 14) similar to rituximab-treated patients (15). Furthermore, increased tumor-associated natural killer (NK) cells and lytic capacity of effector cells were detected after trastuzumab treatment (16)(17)(18), and trastuzumab response correlated with high in vitro ADCC (17,18) and immune cell infiltration (17).
Together, the aforementioned studies provide a biological and clinical rationale for a strategy in which increasing trastuzumab-FcγRIIIa affinity could enhance therapeutic benefit. Currently, no published reports have shown that enhancing FcγR affinity results in enhanced in vivo efficacy for the treatment of established, nonhematologic tumors. Although promising results are described in a few published reports in which treatment commenced at the same time as tumor cell inoculation (13,19,20), these studies leave open the question of whether the observed outcomes were due to the effects on tumor implantation. The primary aim of the present study is to determine whether the efficacy of trastuzumab in suppressing the growth of established tumors is augmented by increasing FcγR binding affinity.

Cell lines
Breast

Analysis of N-linked oligosaccharides
The oligosaccharide profiles of the antibodies were analyzed by matrix-assisted laser desorption/ionization timeof-flight mass spectrometry, as previously described (22).

Fcγ receptor binding ELISA
Soluble human FcγRI, FcγRIIA, FcγRIIB, FcγRIIIA(F158), and FcγRIIIA(V158) (ref. 23), and mouse FcγRI, FcγRII, and FcγRIIIA (Genentech) were expressed in Chinese hamster ovary (CHO) cells (mFcγRI and mFcγRII in 293 cells) as recombinant fusion proteins with Gly-His6-glutathione S-transferase at the COOH terminus of the extracellular domain of the receptor α chains. Soluble mouse FcγRIV was expressed in CHO cells as a recombinant fusion protein with Gly-His8 at the COOH terminus of the extracellular domain of the receptor α chains (Genentech). MaxiSorp 384-well microwell plates (Nunc) were coated with 2 μg/mL of anti-glutathione S-transferase (clone 8E2.1.1; Genentech) or anti-His (penta his antibody, for mouse FcγRIV; Qiagen) in 50 mmol/L of carbonate buffer (pH 9.6), at 4°C overnight followed by a wash (PBS containing 0.05% polysorbate; pH 7.4). Plates were blocked for 1 hour in room temperature with PBS containing 0.5% bovine serum albumin (BSA) and washed. Fcγ receptor [0.25 μg/mL, in PBS containing 0.5% BSA, and 0.05% polysorbate 20 (pH 7.4) in assay buffer] was incubated in plates for 1 hour followed by a wash. To measure binding to the high-affinity FcγRI, IgG antibodies (0.0085-500 ng/mL in 3-fold serial dilution) in assay buffer were added to the plates. To measure binding to the low-affinity FcγRII, FcγIII, and FcγRIV, IgG antibodies were first incubated with goat F(ab′)2 anti-κ antibody (MP Biomedicals) at a 1:2 (w/w) ratio for 1 hour to form antibody complexes. Complexed IgG antibodies (0.42-25,000 ng/mL in 3-fold serial dilution) in assay buffer were added to the plates, incubated for 2 hours, and washed. Bound IgG was detected after 1 hour of incubation with peroxidase-labeled goat F (ab′)2 anti-human IgG F(ab′)2 (Jackson ImmunoResearch) in assay buffer followed by a wash, incubation with substrate (3′,5,5′-tetramethyl benzidine; Kirkegaard & Perry Laboratories), and termination with 1 mol/L of phosphoric acid. The absorbance (450 nm) at the midpoint of the standard curve (mid-OD) was calculated. The corresponding concentrations of standard and samples at this mid-OD were determined from the titration curves using a fourparameter nonlinear regression curve-fitting program (XLfit).

In vitro ADCC
In vitro ADCC assays were performed as previously described (24). In short, peripheral blood mononuclear cells (PBMC) were separated from the blood of normal volunteers using lymphocyte separation medium (MP Biomedicals). The NK cells were enriched from the PBMCs using anti-CD56 microbeads (Miltenyi Biotec). Target cells (1 × 10 4 ) were preincubated with antibodies for 30 minutes in 37°C in serumfree RPMI 1640 supplemented with 0.1% BSA before adding the effector cells in a 25:1 E/T ratio (10:1 when NKs were used as effectors). The cells were incubated for an additional 4 hours before detecting death by measuring the lactate dehydrogenase activity from the medium using Cytotoxicity Detection Kit (LDH; Roche). All measurements were done in quadruplicate. The percentage of cytotoxicity was calculated as follows: % cytotoxicity (experimental lysis − spontaneous effector lysis − spontaneous target lysis) / (maximum target lysis − spontaneous target lysis) × 100.

HER2 binding
Nunc BreakApart Immunomodule plates (Nunc) were coated with 20 ng/mL of HER2-ECD-Ig fusion protein in 50 mmol/L of HEPES (pH 8.2), and 150 mmol/L of NaCl overnight at 4°C. Nonspecific binding was blocked with 2 mg/mL of BSA, 25 mmol/L of Tris (pH 7.5), and 150 mmol/L of NaCl for 2 hours at room temperature. Wells were washed thrice with assay buffer [2 mg/mL BSA and 10 mmol/L HEPES (pH 7.2) in RPMI]. The competitive binding reaction with a dilution series of nonlabeled competitor antibodies and constant 125 I-trastuzumab was carried out for 2 hours at room temperature. The 125 I-trastuzumab bound to HER-2 was then detected by a gamma counter and the data were analyzed using the nonlinear regression method of Munson and Rodbard (25). All measurements were done in quadruplicate.

Analysis of PI3K/AKT pathway activation
Cells were rinsed with PBS and lysed with nondenaturing lysis buffer including 1% Triton (Cell Signaling Technology). Lysates were cleared of insoluble material by centrifugation. AKT phosphorylation was detected by ELISA detecting phosphorylated Ser 473 (Cell Signaling Technology).

Cell proliferation
Proliferation/viability of cells was detected using CellTiter-Glo Luminescent Cell Viability Assay (Promega). For the assay, 5 × 10 3 cells were plated on 96-well plates and incubated overnight for cell attachment before 6 days of treatment with the antibodies. All measurements were done in triplicate.

In vivo drug efficacy
To address the requirement of FcγR interaction in the trastuzumab response using T-D265A, SCID-beige mice (Charles River Labs) were used. For afucosylated trastuzumab efficacy studies, FcγRI −/− FcγRIII −/− RAG2 −/− Tg (human FcγRIIIa) mice were used. Mice xenografted with MCF7-neo/ HER2 cells were supplemented with subcutaneous estrogen pellets (0.36 mg, 60-day release; Innovative Research of America) 3 days prior to cell inoculation. Five million MCF7-neo/HER2 cells or 3 million KPL-4 cells were injected into the mammary fat pad in a 1:1 HBSS-matrigel suspension (BD Matrigel, BD Biosystems). When tumor volumes reached 100 to 300 mm 3 , mice were randomly grouped into treatment cohorts. Dosing is described in the figure legends. Tumor volumes were calculated using the formula: (mm 3 ) = (L × W 2 ) × 0.5.  Relative affinity to human FcγR

Results
Afucosylated trastuzumab has increased affinity to human FcγRIIIa which results in enhanced ADCC We produced 100% fucose-free (afucosylated trastuzumab) trastuzumab in FUT8 −/− CHO cells (30). The FUT8 gene encodes 1,6-fucosyltransferase, which catalyzes the transfer of fucose from GDP-fucose to N-acetylglucosamine. The oligosaccharide profile of the antibodies was determined by matrix-assisted laser desorption/ionization time-of-flight analysis, which confirmed the lack of fucose ( Table 1). The relative affinity of afucosylated trastuzumab was analyzed by FcγR binding ELISA. Binding to FcγRIIIa-F158 and FcγRIIIa-V158 increased by 29-fold and 6.5-fold, respectively ( Table 2). Binding to human FcγRI, FcγRIIa, and FcγRIIb was not markedly altered, indicating that afucosylated trastuzumab has selective enhanced affinity for FcγRIII (Table 2). Similarly, binding to mouse FcγRI, FcγRII, or FcγRIII was not increased and only a minor increase was detected in relative affinity to FcγRIV ( Table 2). The increases in binding affinity to mFcγRIV and huFcγRIIIa-V158 were similar (3.3-and 6.2-fold decreased K D , respectively; data not shown) when binding was measured using surface plasmon resonance-based technology (Biacore).
Afucosylated trastuzumab showed increased in vitro ADCC. The ability to mediate ADCC improved 1.9-to 7.2-fold (EC 50 ; n = 5) when F/F effectors were used and 2.1-to 7.7-fold (n = 5) when V/V donors were used ( Fig. 1A; data not shown). A representative experiment using F/F effector cells is presented in Fig. 1A. ADCC activity increase was statistically significant in 0.01 to 0.1 ng/mL antibody doses ( Fig. 1A; t test). Despite the effector cell donor-dependent variability, the EC 50 was consistently lower for afucosylated trastuzumab in each assay. ADCC activity of PBMCs was mediated by NK cells (Fig. 1A). When purified NK cells from F/F donors were used as effector cells, an 11.3-fold enhancement in ADCC was observed for afucosylated trastuzumab (Fig. 1A). Afucosylated trastuzumab did not mediate ADCC of MCF-10A cells which express low/normal levels of HER2 even at high antibody concentrations and using V/V effectors cells (Supplementary Fig. S1A). Together, these results show that afucosylated trastuzumab has increased affinity for human FcγRIIIa, which results in enhanced ADCC.

Afucosylated trastuzumab retains FcγR-independent functions of trastuzumab
In addition to the ability to mediate ADCC, trastuzumab has a direct effect on tumor cell signaling which causes cell Figure 1. In vitro characterization of afucosylated trastuzumab. A, increased affinity to human FcγRIIIa results in enhanced ADCC. The ability of trastuzumab (black symbols) and afucosylated trastuzumab (red symbols) to mediate in vitro ADCC by FcγRIIIa-F158 homozygous PBMCs (left), NK cells or NK cell-depleted PBMCs (right) was measured using assay detecting LDH released from lysed cells. HER2-amplified BT474-M1 cells were used as target cells in a 25:1 effector/target ratio (10:1 when NK cells were used as effectors). B to D, removal of fucose does not affect FcγR-independent properties of trastuzumab. B, binding to recombinant extracellular domain of human HER2 was determined by competition binding of 125 I-trastuzumab with nonlabeled trastuzumab or afucosylated trastuzumab. C, HER2-amplified SKBR-3 cells were treated for 1 h with antibodies (10 μg/mL) or Wortmannin (1 μmol/L). AKT phosphorylation was measured by ELISA detecting phosphorylated Ser 473 . Statistical significance between groups was determined using Dunnett's test. D, proliferation of SKBR-3 cells was analyzed after 6 days of treatment with trastuzumab or afucosylated trastuzumab using CellTiter-Glo Luminescent Cell Viability Assay. cycle arrest and inhibition of proliferation. To confirm that carbohydrate modification does not affect HER2 binding, we competed the binding of radioiodinated trastuzumab to HER2 ECD by adding increasing concentrations of noniodinated antibody (trastuzumab or afucosylated trastuzumab) in the reaction (Fig. 1B). The calculated affinities of antibodies were similar for trastuzumab and afucosylated trastuzumab (K D , 0.13 ± 0.04 and 0.12 ± 0.03 nmol/L, respectively), indicating that lack of fucose does not affect HER2 binding.
To explore the antisignaling properties, SKBR-3 cells were treated with trastuzumab or the afucosylated variant. Treatment with 10 μg/mL of trastuzumab or afucosylated trastuzumab caused a 69% to 66% reduction in pAKT after 60 minutes of treatment (Fig. 1C). No difference was detected in the effects of afucosylated and conventional trastuzumab treatment (P = 0.99; Dunnetts'). SKBR-3 proliferation was measured after 6 days of treatment. Both antibodies caused approximately 50% maximal reduction in proliferation (Fig. 1D). The EC 50 values were 52 ± 8 and 61 ± 9 ng/mL for trastuzumab and afucosylated trastuzumab, respectively. Together, these results confirm that the lack of fucose does not affect the FcγRindependent functions of trastuzumab. These functions include HER2 binding, the immediate inhibition of the PI3K pathway, and sustained inhibition of tumor cell proliferation.

Pharmacokinetic properties of afucosylated trastuzumab
Removal of fucose does not affect IgG binding to FcRn (23). However, carbohydrate composition or increased FcγR affinity might have an effect on target-independent clearance or drug disposition of the antibody (31). To address this question, we performed pharmacokinetic analysis of afucosylated trastuzumab. For this purpose, the systemic disposition of trastuzumab and afucosylated trastuzumab was evaluated in FcγRI −/− FcγRIII −/− RAG2 −/− Tg (human FcγRIIIa) mice following a single i.v. dose (10 mg/kg). Serum samples (n = 4/time point) were analyzed for anti-HER2 antibodies 3 minutes to 28 days after administration of the antibodies. Both molecules displayed biphasic systemic disposition, with a rapid initial phase and a prolonged terminal phase (Fig. 2). Afucosylated trastuzumab displayed increased clearance (9.0 versus 6.8 mL/d/kg), decreased terminal half-life (10.1 versus 13.1 days), and decreased AUC 0-inf (1,118 versus 1,463 d × μg/mL) relative to trastuzumab (Fig. 2). Both groups displayed a central volume of distribution values (V 1 ) that approximated mouse plasma volume. Taken together, these results show that afucosylated trastuzumab exhibited moderately altered pharmacokinetic dispositions compared with trastuzumab, slightly faster elimination from the circulation, and a modest reduction in half-life.

FcγR interaction is required for trastuzumab response in the treatment of established KPL-4 tumor xenografts
The KPL-4 breast cancer cell line harbors HER2 amplification (21) but does not respond to trastuzumab in vitro in a proliferation assay (5). A "hotspot" PIK3CA mutation (H1047R) generates constitutive PI3K/AKT signals and is the likely cause for the trastuzumab insensitivity. However, KPL-4 cells are sensitive to the FcγR-mediated trastuzumab activity similar to BT474-M1 (Supplementary Fig. S1B). Despite the inability of trastuzumab to inhibit KPL-4 cell proliferation, KPL-4 tumor xenografts respond to trastuzumab when treated with three weekly 15 mg/kg doses (Fig. 3A, left). To confirm that the in vivo response is dependent on FcγR interaction, we also treated mice with the D265A mutant of trastuzumab (T-D265A). The single alanine substitution impairs binding of trastuzumab in all human and mouse FcγR (13). Treatment of mice with T-D265A resulted in almost complete loss of response (Fig. 3A, left), confirming that trastuzumab response in the treatment of KPL-4 tumors were due to the FcγR-dependent effects of trastuzumab.

Increased in vivo efficacy of afucosylated trastuzumab
To assess the efficacy of afucosylated trastuzumab in vivo, we treated established KPL-4 xenografts grown in the mammary fat pads of SCID-beige mice with three weekly 10 mg/kg doses of trastuzumab or afucosylated trastuzumab. Although both trastuzumab and afucosylated trastuzumab inhibited the tumor growth (P < 0.001 and P = 0.003, respectively; log rank test), no difference was seen between the antibodies (P = 0.48; log rank test), indicating that the modest increase in binding affinity to FcγRIV was not sufficient to result in increased efficacy (Fig. 3A, right).
To further evaluate the activity of the afucosylated trastuzumab variant, we generated FcγRI −/− FcγRIII −/− RAG2 −/− Tg (human FcγRIIIa) mice that lack murine FcγRI (27) and FcγRIII (26) and express the human FcγRIIIa transgene (28). The transgene expression is controlled by the human FcγRIIIa promoter and is expressed in NK cells and macrophages (28). Expression of the transgene in mice was  confirmed by quantitative reverse transcription-PCR of splenocyte mRNA (Supplementary Fig. S2A). Analysis of genomic DNA showed that the transgene is the low-affinity FcγRIIIa-F158 allele (Supplementary Fig. S2B).
For the efficacy studies, KPL-4 xenografts were grown in the mammary fat pads of FcγRI −/− FcγRIII −/− RAG2 −/− Tg (human FcγRIIIa) mice until the tumor size reached an average of 125 mm 3 . At this point (day 0), mice were randomly grouped for treatment regimens (n = 9) and treated with single 1 or 3 mg/kg i.v. doses of trastuzumab, afucosylated trastuzumab, or vehicle. Afucosylated trastuzumab inhibited the growth of KPL-4 tumors more effectively than trastuzumab (Fig. 3B, left). Treatment with 3 mg/kg of trastuzumab caused a significant delay in tumor progression (P = 0.003; log rank test; Fig. 3B, right) and the efficacy of afucosylated trastuzumab was significantly improved compared with trastuzumab (P = 0.010; Fig. 3B, right). The median progression-free survival more than doubled for mice treated with afucosylated trastuzumab, compared with treatment with trastuzumab (48 versus 23.4 days, respectively; Fig. 3B, right).
We performed similar experiments using FcγRI −/− FcγRIII −/− RAG2 −/− mice without the human transgene. The KPL-4 tumors were treated with three weekly 3 mg/kg i.v. doses of the antibodies. No significant difference was observed in the response to the antibodies, demonstrating that the increase in the efficacy detected for afucosylated trastuzumab was mediated by the human FcγRIIIa transgene (Fig. 3C).
To ensure that the observed increase in efficacy was not modeldependent, the study was repeated using established MCF7neo/HER2 xenografts grown in FcγRI −/− FcγRIII −/− RAG2 −/− Tg (human FcγRIIIa) mice. Similar to KPL-4, these cells overexpress HER2 but do not respond to trastuzumab in vitro in a proliferation assay. Again, afucosylated trastuzumab was more potent than trastuzumab in treating MCF7-neo/ HER2 tumors, causing four complete responses and five partial responses (Fig. 3D). In contrast, no complete responses and only one partial response was detected in trastuzumab-treated mice. Taken together, these results show that the in vivo efficacy of afucosylated trastuzumab is superior to trastuzumab in treating two preclinical models of HER2amplified breast cancer.

Discussion
It has been hypothesized that ADCC is a general mechanism of action for many therapeutic antibodies (32). However, it remains unclear whether ADCC or the engagement of Fcγ receptors significantly contribute to trastuzumab's antitumor effects in patients with breast cancer. After the initial in vitro findings, Ravetch and colleagues showed the in vivo importance of FcγR-mediated activity of trastuzumab (13). Notably, these studies did not address the role of FcγR interaction in models with established tumors. Instead, treatments were simultaneous with cell inoculation and thus the results might portray the role of effector cell functions on tumor implantation rather than its effect on solid tumor mass. Our findings confirm the results of Clynes and colleagues and extend them to the treatment of pre-established solid tumors.
The ability to increase in vitro ADCC by antibody engineering is widely established (33,34). However, to our knowledge, it has not been previously shown that the increase in antibody-FcγR interaction leads to increased in vivo efficacy in the treatment of established nonhematologic tumors. To directly address this question, we generated afucosylated trastuzumab. The approach was selected primarily because the affinity increase is significant and selective for FcγRIII. Moreover, the variant can be readily produced and is completely devoid of fucose. Fucose in the oligosaccharide (at Asn 297 ) of IgG 1 hinders the interaction between carbohydrate of FcγRIIIa (at Asn 162 ) and regions of IgG 1 (35). Afucosylation results in increased affinity of the FcγR-IgG 1 interaction. The higher affinity observed by afucosylation is selective for human FcγRIIIa and FcγRIIIb because the other human FcγRs are not glycosylated at the corresponding position (35). Therefore, removing the fucose residue results in increased affinity selectively for FcγRIII but not for other FcγR, FcRn, or C1q (23).
We produced afucosylated trastuzumab in FUT8 −/− CHO cells (30). One alternative technology to produce afucosylated antibodies in CHO cells is to overexpress GnTIII and ManII (35). Interestingly, the clearance of afucosylated trastuzumab was slightly faster than the clearance of trastuzumab in the transgenic mouse model. Hypothetically, the faster clearance might be a result of differential biodistribution leading to the enrichment of antibodies to immune effector cell-rich organs due to increased FcγRIII affinity. Using a mouse model that expresses a human FcγRIIIa transgene in effector cells, we were able to successfully show that an increase in affinity to FcγRIII results in improved efficacy when treating solid tumors. As with any preclinical finding, confirmation of these results will require validation in human clinical trials.
The superior efficacy of afucosylated trastuzumab is not explained by an increase in antibody exposure. However, a plausible mechanism for more effective tumor growth inhibition may be due to the observation that afucosylated antibodies could escape the inhibitory effect of serum IgG. Reports by Iida and colleagues show that serum IgG markedly inhibits the efficacy of rituximab to deplete B cells (36). This inhibition has been attributed to the ability of IgG to compete for FcγRIII binding. However, due to its higher affinity for FcγRIII, afucosylated IgG might escape the inhibitory effect of serum IgG (36)(37)(38).
The differential expression of HER2 in tumors relative to normal tissue is thought to account for Herceptin's favorable safety profile. The activity of trastuzumab in ADCC assays is also specific for only high HER2-expressing target cells (11). To maintain this safety profile, it is critical that any increase in ADCC activity does not result in the cytotoxicity of cells that express normal levels of HER2. Previous reports suggest that low-fucose IgG 1 s could mediate ADCC at low antigen densities at which their corresponding high-fucose counterparts do not induce ADCC (39). However, our ADCC analysis using MCF-10A cells (commonly used as a model for nontransformed low HER2-expressing mammary epithelial cells) failed to detect any significant ADCC using afucosylated trastuzumab.
Because trastuzumab does not bind to murine ErbB2 (HER2), the mouse model used in efficacy studies is not suitable for safety studies. However, more detailed safety studies in non-human primates are required. In conclusion, our results support the development of effector function-enhanced antibodies for solid tumor therapy.

Disclosure of Potential Conflicts of Interest
All authors are employees of Genentech, Inc.