Fc‐engineered EGF‐R antibodies mediate improved antibody‐dependent cellular cytotoxicity (ADCC) against KRAS‐mutated tumor cells

Oncogenic mutations of the KRAS gene have emerged as a common mechanism of resistance against epidermal growth factor receptor (EGF‐R)‐directed tumor therapy. Mutated KRAS leads to ligand‐independent activation of signaling pathways downstream of EGF‐R. Thereby, direct effector mechanisms of EGF‐R antibodies, such as blockade of ligand binding and inhibition of signaling, are bypassed. Thus, a humanized variant of the approved EGF‐R antibody Cetuximab inhibited growth of wild‐type KRAS‐expressing A431 cells, but did not inhibit KRAS‐mutated A549 tumor cells. We then investigated whether killing of tumor cells harboring mutated KRAS can be improved by enhancing antibody‐dependent cellular cytotoxicity (ADCC). Protein‐ and glyco‐engineering of antibodies’ Fc region are established technologies to enhance ADCC by increasing antibodies’ affinity to activating Fcγ receptors. Thus, EGF‐R antibody variants with increased affinity for the natural killer (NK) cell‐expressed FcγRIIIa (CD16) were generated and analyzed. These variants triggered significantly enhanced mononuclear cell (MNC)‐mediated killing of KRAS‐mutated tumor cells compared to wild‐type antibodies. Additionally, cells transfected with mutated KRAS were killed as effectively by ADCC as vector‐transfected control cells. Together, these data demonstrate that KRAS mutations are not sufficient to render tumor cells resistant to ADCC. Consequently Fc‐engineered EGF‐R antibodies may prove effective against KRAS‐mutated tumors, which are not susceptible to signaling inhibition by EGF‐R antibodies.

A fter approximately 30 years of translational research, the epidermal growth factor receptor (EGF-R), ErbB1, has emerged as a validated target antigen for molecular therapies. (1) Today, both EGF-R-directed tyrosine kinase inhibitors (TKI) and EGF-R antibodies have obtained approval for tumor therapy. (2,3) As expected from clinical experience with TKI against other target antigens, (4) mutations in the EGF-R kinase domain critically affect sensitivity and resistance against EGF-R directed TKI. (5) These EGF-R mutations proved less relevant for tumor cell killing by EGF-R antibodies in vitro (6) and in animal models, (7) and early clinical data support these observations. (8) On the other hand, resistance against EGF-R antibodies in colorectal cancer (CRC) patients has been associated with activating mutations of KRAS (9)(10)(11)(12) and other mediators of downstream signaling such as v-raf murine sarcoma viral oncogene homolog B1 (B-RAF), (13) phosphoinositide-3-kinase (PI3K), phosphatase and tensin homolog (PTEN), and others. (14) This clinical resistance of KRAS-mutated CRC led to the development of guidelines that recommend to limit the application of EGF-R-directed antibodies to CRC patients bearing wild-type KRAS tumors. (15) KRAS belongs to the family of RAS proto-oncogenes that activate intracellular signaling downstream of receptor kinases. (16) Thereby, oncogenic point mutations in the KRAS gene promote cellular growth and apoptosis resistance -leading to more aggressive tumor phenotypes with increased resistance to chemo-and radiotherapy. Oncogenic RAS proteins display impaired intrinsic GTPase activity preventing their inactivation by GTPase activating proteins (GAPs). (17) This makes the development of small molecule inhibitors inherently difficult, (18) although promising approaches are emerging. (19) Furthermore, inhibitors of RAS processing, such as farnesyltransferase-inhibitors, often do not inhibit KRAS4b -the most common RAS isoform in solid tumors. (20) Therefore, substances blocking molecules downstream of KRAS, such as PI3K and MEK are actively investigated, (21) but other strategies to overcome RASmediated resistance to tumor therapy are required.
EGF-R antibodies can recruit a broad panel of tumor cell killing mechanisms. (22) These mechanisms may be divided into those mediated by the F(ab¢) region -called direct mechanismsand indirect mechanisms that are triggered by antibodies' constant Fc region. (23) For EGF-R antibodies, direct mechanisms such as blockade of ligand binding, inhibition of signaling, and receptor down-regulation are considered to be particularly important. (2,3) However, in vitro EGF-R antibodies can also effectively recruit indirect mechanisms such as complement-dependent cytotoxicity (CDC) (24) and antibody-dependent cellular cytotoxicity (ADCC). (25) In mice, the efficacy of ErbB2 (HER-2 ⁄ neu)directed antibodies has been demonstrated to critically depend on the expression of activating Fcc receptors, (26) and -against other target antigens -on antibodies' relative binding affinities to activation compared to inhibitory Fc receptors (A:I ratio). (27) Importantly, ADCC operated at lower antibody concentrations compared to direct effector mechanisms. (28) The contribution of ADCC to ErbB antibodies' efficacy in patients is supported by studies demonstrating associations between therapeutic efficacy and the expression of certain Fcc receptor allotypes, (29)(30)(31) which display enhanced antibody binding and increased ADCC activity compared to their less active alloforms. (32,33) Together, these observations stimulate research into antibodies with enhanced affinity to activating Fcc receptors. (34,35) So far, two approaches of engineering antibodies' Fc regions have proved to be particularly powerful for enhancing Fcc receptor binding and for improving Fcc receptor-mediated effector functions: reducing the amount of fucose in the CH2 attached glycosylation (36,37) (glyco-engineering), and site-directed mutagenesis of amino acids in the hinge or CH2 regions of antibodies (protein-engineering). (38) Here, we demonstrate that EGF-R antibodies of IgG1 isotype can kill KRAS-mutated tumor cells via ADCC. Tumor cell killing was significantly enhanced when Fc-engineered EGF-R antibodies were compared to their wild-type counterpart. These results support approaches to enhance the ADCC activity of EGF-R antibodies, particularly in patients with KRAS-mutated tumors, who otherwise have a low probability of responding to currently available EGF-R-directed therapies. (15) Material and Methods Study population and consent. Experiments reported here were approved by the Ethics Committee of the Christian-Albrechts-University, Kiel, Germany, in accordance with the Declaration of Helsinki. Blood donors were randomly selected from healthy volunteers, who gave written informed consent before analyses. Distribution of FccRIIIa genotypes among analyzed individuals was 25% (V ⁄ V), 33.3% (F ⁄ F), and 41.7% (V ⁄ F).

Construction of a KRAS expression vector and stable
transfection of A431 cells. Full-length wild-type KRAS4b was amplified by PCR from cDNA of H1975 cells. The G12V mutation was inserted by PCR, and the resulting PCR product was cloned between the HindIII and NotI sites of the expression vector pSec (Invitrogen). An N-terminal myc-tag was added into the NheI and HindIII restriction sites with the following oligonucleotides 5¢-CTAGCATGGAGCAGAAGCTGATCAGCGAG-GAGGACCTGATGACTGAATATA-3¢ and 5¢-AGCTTATAT-TCAGTCACAGGTCCTCCTCGCTGATCAGCTTCTGCTCC-ATG-3¢. To control vector expression, an internal ribosomal entry site (IRES) and green fluorescence protein (GFP) were subcloned from the vector pIRES-ZsGreen (Clonetech, St-Germain-en-Laye, France) into the pSec KRAS4b G12V vector or into the pSec vector -serving as the control vector. The resulting vectors were transfected into A431 with lipofectamine2000 (Invitrogen) following the manufacturer's protocol. Transfectants were sorted for GFP expression with a FACSAria cytofluorometer (BD Biosciences, Erembodegem, Belgium).
Immunoprecipitation. Cells were seeded in 10-cm dishes at 5 · 10 6 cells ⁄ dish and grown overnight. The next day, cells were washed with ice-cold PBS and pelleted by centrifugation. Cells were lysed according to the manufacturer's instructions using native lysis buffer (NLB; Cell Signaling Technology, Danvers, MA, USA). Five-hundred lg of native protein were incubated with 10 lL of Raf-1 RBD agarose beads (Millipore, Temecula, CA, USA) for 1 h at 4°C to precipitate activated RAS GTP . Precipitates were washed three times with NLB, denatured by boiling in 40 lL 2 · Laemmli buffer, and separated by SDS-PAGE. After transfer onto PVDF membranes, samples were probed for precipitated RAS using rabbit anti-RAS or rabbit anti-myc antibody (both Cell Signaling Technology). Whole protein extracts were separated by SDS-PAGE and immunoblotted against total RAS to control equal protein loading.
Determination of viable cell mass (MTS assay). The CellTiter 96 non-radioactive cell proliferation assay was performed with A431 and A549 cells according to manufacturer's instructions (Promega, Madison, WI, USA). Cells were seeded at 5000 cells ⁄ well in 96-well flat-bottom tissue culture plates (Becton Dickinson, Meylan Cedex, France) and grown in complete growth medium in a final volume of 100 lL. Cells were treated for 72 h with antibodies before the absorbance of formazan was measured at 490 nm on a Sunrise ELISA reader (Tecan Group, Männedorf, Switzerland).
Isolation of human effector cells. Human effector cells such as peripheral mononuclear cells (MNCs) or polymorphonuclear cells (PMNs) were isolated from peripheral blood from healthy volunteers as previously described. (39) Antibody-dependent cellular cytotoxicity (ADCC) assays. ADCC assays were performed as described in (39) but without stimulation of PMNs with granulocyte-macrophage colony-stimulating factor. For whole blood assays, unseparated human blood was freshly drawn from healthy volunteers, anticoagulated with 500 U ⁄ mL heparin and directly used as effector source in ADCC assays. Percentage of cellular cytotoxicity was calculated using the following formula: % specific lysis = (experimental cpm ) basal cpm) ⁄ (maximal cpm -basal cpm) · 100. For calculation of relative tumor cell lysis (%), specific lysis triggered by 10 lg ⁄ mL of S239D ⁄ I332E was equated with 100% lysis and set into relation with all other lysis rates.
EGF-R antibodies. The approved EGF-R antibody Cetuximab (chimeric IgG1; Erbitux) was bought from Merck (Darmstadt, Germany). A humanized IgG1 version of Cetuximab was generated, and produced in HEK293E cells (wildtype). (40) Protein-or glyco-engineered antibody variants of this wild-type antibody were constructed, expressed, and purified as described, (38,41) except that expression was carried out in HEK293E cells using the pTT5 vector system (NRC-BRI, Montreal, Quebec, Canada). (38) An human IgG1 antibody against keyhole limpet hemocyanin (HuMab-KLH), which was kindly provided by Genmab, served as irrelevant control antibody.
Determination of Fc receptor binding affinities. Human Fc receptors with 6 · His tags were expressed in HEK293T cells and purified using nickel affinity chromatography. Binding affinities of EGF-R antibodies to FccRs were determined by Biacore as described previously. (41) Flow cytometric analyses. For indirect immunofluorescence, cells were incubated with antibodies at various concentrations in PBS supplemented with 0.5% bovine serum albumin (Sigma-Aldrich) and 0.1% sodium-azide (PBS buffer) for 30 min on ice. After washing, cells were stained with FITC-conjugated F(ab¢) 2 fragments rabbit antihuman IgG (Dako, Glostrup, Denmark), respectively. Quantitative surface EGF-R expression was determined using murine EGF-R and control antibodies, and the QIFIKIT (Dako), according to the manufacturer's instructions. Immunofluorescence was analyzed on a flow cytometer (Epics Profile; Beckman Coulter, Fullerton, CA, USA).
Data processing and statistical analyses. Data are displayed graphically and were statistically analyzed using GraphPad Prism 4.0. Curves were fitted using a nonlinear regression model with a sigmoidal dose response (variable slope). Statistical significance was determined by the Student's t-test (paired, two-tailed) and the respective results are displayed as mean ± SEM. P-values were calculated and null hypothesis was rejected when P £ 0.05.

Results
Characterization of used cell lines. Cell lines were selected according to their KRAS status from the Cancer Genome Project online database at the Sanger Institute and from literature. (42) While the A431 cell line harbors wild-type KRAS, (43) A549, H2030, HCT-116, and SW-480 cells carry activating mutations of the KRAS gene. As measured by using the QIFIKIT, all of these cell lines differ in the expression of EGF-R molecules per cell, with A431 cells showing the highest and HCT-116 cells the lowest EGF-R expression. The characteristics of the aforementioned cell lines are summarized in Table 1. To analyze whether point mutations of the KRAS gene, described for A549, H2030, HCT-116, and SW-480 cells, influence the activation status of RAS, all five cell lines were assayed by immunoprecipitation against activated RAS GTP . As expected, enhanced activation of RAS was detected for the cell lines harboring oncogenic KRAS (A549, H2030, HCT-116, and SW480 cells) when compared to A431 cells (Fig. 1).
Humanized EGF-R antibody did not inhibit proliferation, but mediated ADCC against KRAS-mutated cells. In order to investigate effector mechanisms of humanized Cetuximab (wild-type) against cell lines carrying wild-type or oncogenic KRAS, growth inhibition and ADCC assays were performed with A431 and A549 cells, respectively. While the EGF-R antibody induced significant growth inhibition of A431 cells when compared to the non-binding control antibody, it failed to affect A549 cells. Vital cell masses at the highest antibody concentration (200 lg ⁄ mL) of wild-type antibody ranged from 61.6% ± 6.4% for A431 cells to 103.7% ± 1.9% for A549 cells after 72 h of incubation, respectively ( Fig. 2A,C). In a next set of experiments, ADCC mediated by MNC effector cells was analyzed. Interestingly, wild-type antibody induced significant ADCC activity (P £ 0.05) against A431 cells (50.3% ± 12.6%) and A549 cells (36.2% ± 7.9%) when compared to control antibody (Fig. 2B,D).
Construction and characterization of Fc-engineered EGF-R antibody variants. As the KRAS-mutated A549 cell line was susceptible to ADCC, it was investigated whether Fc engineering may improve ADCC against EGF-R-expressing tumor cells. Binding of the wild-type antibody to FccRIII (CD16) was enhanced by protein engineering of its CH2 domain (I332E and S239D ⁄ I332E) as described previously. (38,41) Furthermore, the wild-type antibody was expressed in Lec-13 cells to generate a non-fucosylated variant. Binding to activating human Fcc receptors (FccRIIa and FccRIIIa) and their alloforms (FccRIIa-131H ⁄ R and FccRIIIa-158V ⁄ F) was analyzed by Biacore. Results are summarized in Table 2, and were as expected from previous studies. Binding of the variants to human FccRI and FccRIIb has been described in a recent study. (41) SDS-PAGE and immunoblotting experiments with HRP-conjugated antihuman-IgG antibodies and a HRP-conjugated Aleuria Aurantia lectin specific for fucose linked (a-1,6) to N-acetylglucosamine or (a-1,3) to N-acetyllactosamine-related structures revealed that the Lec-13 expressed antibody did not contain detectable levels of fucose (Fig. S1). In Figure 3(A) a model structure of a human IgG1 antibody comprising protein-engineered variants is represented.
Wild-type and protein-engineered antibody variants were then analyzed with regard to their F(ab¢)-mediated effector functions. As expected, all three antibodies demonstrated very similar binding characteristics to EGF-R-expressing A431 cells, resulting in EC 50 values ranging from 5.2 lg ⁄ mL (S239D ⁄ I332E; 95% confidence interval [CI], 4.1-6.5 lg ⁄ mL) to 6.9 lg ⁄ mL (I332E; 95% CI, 5.4-8.9 lg ⁄ mL) (Fig. 3B) -indicating that protein-engineering did not affect antibody affinity to its antigen. EGF-R affinity of the antibodies was also indistinguishable when tested by Biacore (data not shown). To assess Fc-mediated effector functions, ADCC assays were performed with A431 target cells, and isolated effector cell populations or human whole blood samples.  (Fig. 3C). However, when ADCC assays were performed with isolated PMN effector cells, ADCC was less effective with the I332E variant compared to wild-type antibody, and abolished with the S239D ⁄ I332E variant (Fig. 3D). To analyze the cytolytic capacity of variant antibodies under more physiological conditions, unseparated human whole blood samples were assayed in ADCC experiments. As shown in Fc-engineered EGF-R antibody variants did not affect direct effector mechanisms. Antibody variants were compared with regard to their capacity to inhibit growth of A431 and A549 cells. In contrast to A431 cells where vital cell masses of 59.8% ± 6.6% with S239D ⁄ I332E or 64.3% ± 5.7% with I332E variants and 61.7% ± 6.4% with wild-type at the highest antibody concentration (200 lg ⁄ mL) were detected (Fig. S2A), no growth inhibition was induced in A549 cells (Fig. S2B). These results were significantly (P < 0.05) different from that received in experiments with a control antibody, where the detected vital cell mass was 86.5% ± 1.9% at highest concentration. Thus, for both antibody variants, no alterations concerning their capacity  to directly inhibit cell proliferation could be detected -demonstrating F(ab¢)-mediated effector functions not to be affected by Fc protein-engineering.
To assess whether glyco-engineering also results in enhanced ADCC against KRAS-mutated tumor cells, similar experiments were performed comparing a non-fucosylated and the wild-type antibody (Fig. 5). Interestingly, all four cell lines proved susceptible to ADCC by the glyco-engineered EGF-R antibody. The wild-type antibody demonstrated significantly lower ADCC activity with all cell lines tested. Moreover, MNC mediated ADCC against KRASmutated tumor cells with both protein-and glycoengineered variants is not only distinguished by lower EC 50 concentrations compared to wild-type, but also by remarkably increased killing at saturating antibody concentrations.
In conclusion, data presented in Figures 5 and 6 demonstrated that KRAS-mutated cells were susceptible to ADCC, and that ADCC activity could be enhanced by protein-or glyco-engineering of EGF-R antibodies.   (Fig. 7A). Next, ADCC experiments with MNC effector cells were performed with A431 KRAS4b G12V cells and A431 cells transfected with the control vector. As presented in Figure 7(B), no differences in Cetuximab-triggered cytotoxicity were observed between both transfectants -demonstrating that oncogenic KRAS did not affect MNC-mediated cytotoxicity.

Discussion
In the present study, a humanized variant of Cetuximab did not inhibit the growth of KRAS-mutated A549 cells, but triggered limited ADCC activity against these cells. Interestingly, estab-lished approaches to enhance ADCC activity of antibodies by protein-or glyco-engineering significantly improved killing against a panel of KRAS-mutated tumor cells by MNC effector cells. This difference between wild-type and Fc-engineered antibodies is well explained by the higher affinity of the latter for FccRIIIa -the predominant Fcc receptor on human natural killer (NK) cells (38) that constitute the main effector cell population in MNC. Interestingly, protein-engineered antibodies with increased binding affinity for FccRIIIa (I332E and S239D ⁄ I332E) demonstrated lower ADCC activity with PMN effector cells than the respective wild-type antibody (Fig. 3D). We have previously reported similar results for glyco-engineered variants of the EGF-R antibody Zalutumumab where low-fucosylated compared to completely-fucosylated Zalutumumab demonstrated higher binding to FccRIIIa, improved ADCC by MNC effector cells, but reduced tumor cell killing by PMN. (39) Similar results were observed in the present study with a glyco-variant of the humanized Cetuximab antibody (data not shown). These unexpected results with PMN effector cells may be explained by the homology between FccRIIIa and the PMNexpressed FccRIIIb. Thus, improving the affinity for FccRIIIa simultaneously increases binding to FccRIIIb, (39) which is not a cytotoxic trigger molecule on PMN. (44) Glyco-engineering predominantly modulates binding affinity to FccRIIIa, while protein-engineering can alter binding characteristics to FccRIIIa and FccRIIa, as well as FccRI and FccRIIb. (41) Whether these differential FccR binding profiles of protein-and glyco-engineered antibodies will impact their therapeutic effects in vivo will require further studies. Notably, Fc-engineered EGF-R antibodies proved more effective than the wild-type antibody in ADCC against both KRASunmutated and -mutated cell lines (Figs 3-5). The negative impact of KRAS mutations on the clinical efficacy of EGF-R antibodies in CRC has now been confirmed in numerous clinical trials (reviewed in reference (15) ), while more preliminary studies in NSCLC patients suggested that the KRAS status may not be predictive for Cetuximab's efficacy in these patients. (8,45) Presently, it is not clear whether these differences are explained by statistical considerations (lower response rates and lower incidence of KRAS mutations in NSCLC vs CRC) or whether they reflect biological differences between both tumor types. There is experimental evidence to suggest that the relative contribution of antibodies' effector mechanisms depends on antibody concentrations (28) and potentially on the tumor location. (46) Further studies are required to address these interesting issues.
Results from in vitro studies with tumor cell lines are often confounded by multiple genetic abnormalities that are commonly observed in these cells. These often poorly defined alterations impair the interpretation of results with respect to specific mutations. Thus, results presented in and that mutated KRAS is not sufficient to render tumor cells resistant to ADCC, but they do not allow the direct assessment of the impact of KRAS mutations on the susceptibility of tumor cells against ADCC. To address this important issue, wild-type KRAS-expressing A431 cells were transfected with oncogenic KRAS4b G12V or a control vector. Interestingly, both cell lines were similarly susceptible to MNC-mediated ADCC by EGF-R antibodies. Furthermore, HCT-116 cells additionally carry an activating PI3K mutation (http://www.sanger.ac.uk), which is discussed as another mechanism of resistance against EGF-R antibodies. (47) Our preliminary results suggest that also this mechanism of resistance could be overcome by Fc-engineered EGF-R antibodies. Potentially, enhancing Fc-mediated effector functions of EGF-R antibodies -such as ADCC and CDCmay prove as a more general approach to overcome resistance against EGF-R-directed therapy. (48) An important question is whether our in vitro results have relevance for the understanding of the mechanisms of action for EGF-R antibodies in vivo, and whether they impact potential approaches to enhance the efficacy of EGF-R directed therapy. Recent recommendations suggested to exclude patients with KRAS-mutated CRC from EGF-R directed therapies, as they showed low response rates in clinical trials. (10)(11)(12) These clinical observations supported the concept that inhibition of EGF-Rmediated signaling might be the predominant mechanism of action for EGF-R antibodies. Accordingly, activating mutations of downstream signaling molecules -like KRAS or v-raf murine sarcoma viral oncogene homolog B1 (B-RAF) -would render tumor cells resistant to EGF-R inhibitors. (13,14) Further support for the contribution of Fc-mediated effector mechanisms for Cetuximab's clinical efficacy was derived from statistical correlations between Fcc receptor alloforms and clinical outcomes. (29) These studies suggested that Fc-mediated effector functions were relevant for Cetuximab's clinical efficacy in CRC patients -as previously demonstrated for other therapeutically approved antibodies. (31,49,50) Importantly, individual patients with favorable Fcc receptor alloforms (FccRIIa-131H and ⁄ or FccRIIIa-158V) responded to Cetuximab therapy -even when their tumors harbored oncogenic KRAS mutations. (29) As previously reported by others, (38) Fc engineering improved ADCC by FccRIIIa-158F ⁄ F, -158F ⁄ V, and 158V ⁄ V donors.
Presently, there is a wide gap between results from in vitro models addressing mechanisms of sensitivity and resistance against EGF-R-directed antibody therapy and clinical data that validate them. Unfortunately, studies in small animals are of limited value for many of these specific questions due to differences between the murine and the human Fcc receptor systems.
In studies investigating the impact of Fc engineering on antibody efficacy, those differences in species characteristics additionally impede the transfer of in vitro results to in vivo relevance. (51) On the other hand, protein-and glyco-engineered antibodies against other target antigens (e.g. CD30, CD20) have already entered clinical trials, while trials with glyco-engineered EGF-R antibodies are expected to start soon. Another approach to clinically assess the contribution of ADCC versus other antibody effector mechanisms would include studies that prospectively investigate the impact of Fcc receptor polymorphisms on therapeutic outcomes, for example in FccRIIIa-158V ⁄ V donors, as is currently on-going for Trastuzumab (a humanized monoclonal IgG1 antibody directed against human epidermal growth factor receptor 2, ErbB2).
In conclusion, KRAS-mutated tumor cells can be effectively killed by ADCC, indicating that mutated KRAS is not sufficient to confer resistance to antibody-mediated effector cell killing. Consequently, approaches to enhance ADCC may have the potential to increase the clinical activity of EGF-R antibodiesparticularly in patients whose tumors harbor activating KRAS mutations.