Rational identification of an optimal antibody mixture for targeting the epidermal growth factor receptor

Abstract

The epidermal growth factor receptor (EGFR) is frequently dysregulated in human malignancies and a validated target for cancer therapy. Two monoclonal anti-EGFR antibodies (cetuximab and panitumumab) are approved for clinical use. However, the percentage of patients responding to treatment is low and many patients experiencing an initial response eventually relapse. Thus, the need for more efficacious treatments remains. Previous studies have reported that mixtures of antibodies targeting multiple distinct epitopes are more effective than single mAbs at inhibiting growth of human cancer cells in vitro and in vivo. The current work describes the rational approach that led to discovery and selection of a novel anti-EGFR antibody mixture Sym004, which is currently in Phase 2 clinical testing. Twenty-four selected anti-EGFR antibodies were systematically tested in dual and triple mixtures for their ability to inhibit cancer cells in vitro and tumor growth in vivo. The results show that targeting EGFR dependent cancer cells with mixtures of antibodies is superior at inhibiting their growth both in vitro and in vivo. In particular, antibody mixtures targeting non-overlapping epitopes on domain III are efficient and indeed Sym004 is composed of two monoclonal antibodies targeting this domain. The superior growth inhibitory activity of mixtures correlated with their ability to induce efficient EGFR degradation.

Introduction

The epidermal growth factor receptor (EGFR/ErbB1) belongs to the ErbB family of receptor tyrosine kinases, which also includes HER-2/ErbB2, HER-3/ErbB3 and HER-4/ErbB4.¹ EGFR is known to bind at least six ligands, including epidermal growth factor (EGF) and transforming growth factor-α (TGFα).² Unliganded EGFR is thought to rest in a tethered inactive form where a putative “dimerization arm” localized in domain II (dII) is sequestered by interactions with domain IV (dIV). For ligand binding to occur, domain 1 (dI) and domain III (dIII) must undergo conformational changes that bring them close together for formation of a ligand binding pocket. These ligand-mediated structural changes induce an untethered active receptor conformation where the dimerization domain is exposed and capable of promoting receptor dimerization with other monomers, shifting the monomer ↔ dimer equilibrium in favor of the dimeric receptor state.^3–5 Receptor homo- or hetero-dimerization with EGFR or other members of the ErbB family results in activation of intracellular tyrosine kinase domains, leading to autophosphorylation that controls signaling cascades important for cell growth, proliferation, survival and motility.^6,7 Owing to involvement in these cellular processes, EGFR signaling is frequently unbalanced in human malignancies, either due to increased ligand production, receptor overexpression, receptor mutations or cross-talk with other receptor systems.^8–11 Changes in EGFR status have been linked to the development and maintenance of a malignant phenotype and correlated to poor clinical prognosis.¹² For this reason EGFR is an attractive target for anti-cancer therapy.¹³ Two monoclonal antibodies (mAbs), cetuximab (Erbitux^®) and panitumumab (Vectibix^®), are approved for the treatment of colorectal cancers^14,15 and head and neck cancers (cetuximab only¹⁴).

One of the main mechanisms of tumor growth inhibition by mAbs is thought to be disruption of ligand induced EGFR tyrosine kinase activation. The mAbs do this by sterically blocking access of the ligand to EGFR dIII.¹⁶ Despite the clinical success of cetuximab and panitumumab, recent in vitro results show that cancer cells that were initially sensitive to treatment escape by switching on alternative receptor tyrosine kinases (HER2, MET or IGF1R). However, importantly the cells remain dependent on the presence of EGFR.^17–20

Previous studies have described how mixtures of antibodies targeting multiple distinct epitopes on HER-2 are more effective than single mAbs at inhibiting growth of HER-2 dependent cells in vitro and tumor growth in various animal models.^21–24 Furthermore, in vitro studies have demonstrated that mixtures of antibodies with non-overlapping epitopes are more efficient at inducing EGFR downregulation compared with the individual antibodies.^22,25 A mixture of the mAbs 806 and 528 have also been shown to induce degradation of the Δ2–7 EGFR in a xenograft model of the U87-MG-Δ2–7 EGFR cell line.²⁶

Finally, antibody effector functions like antibody-dependent cellular cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC) could contribute to the activity of tumor growth inhibitory anti-EGFR mAbs by inducing tumor cell killing. While single anti-EGFR mAbs have been demonstrated to activate ADCC through binding to FcγIIIa receptor, only antibody mixtures have been shown to activate CDC in vitro.^27,28

Based on these reports, we hypothesized that a systematic discovery and selection approach might lead to improved antibody mixtures for targeting EGFR. The result was the identification of Sym004, a drug candidate composed of an equimolar mixture of the two chimeric mAbs, 992 and 1024, recognizing non-overlapping epitopes on EGFR dIII. Sym004 was superior to reference mAbs and the individual antibodies 992 and 1024 at inhibiting cancer cell growth in vitro and tumor growth in an animal xenograft model.

Results

Generation and initial selection of chimeric anti-human EGFR mAbs.

Mice were immunized with human EGFR antigen in various presentations to generate a comprehensive and diverse panel of murine antibodies recognizing multiple epitopes on human EGFR. Cognate V_H-V_L gene pairs were prepared by PCR from single-cell sorted splenic plasma cells and expressed as chimeric IgG1 molecules using the murine Symplex™ technology. Antibody supernatants were screened for binding to an EGFR expressing cell line and recombinant sEGFR. The V_H and V_K genes of all positive clones were DNA sequenced and subjected to ClustalW analysis to investigate the clonal relationship between antibodies (Fig. 1). The analysis revealed a total of 88 unique antibodies estimated to originate from 48 distinct V(D)J rearrangements or so-called clusters.

Figure 1.

Phylogenetic analysis of the clonal and functional diversity of anti-EGFR antibodies. Cladogram showing the clonal and functional relationship between unique V gene rearrangements (clusters). The 24 selected antibodies that were analyzed in detail in this paper are highlighted in red, and genetically related antibody families or clusters originating from same V gene rearrangements are highlighted in color bars. Growth Inhibition HN5: Percentage growth inhibitory effect of antibody supernatants when tested by WST-1 assay on HN5 cells and compared with a negative control antibody. Negative values (green bars) indicate growth stimulatory effects.

All 88 mAbs were tested for their ability to inhibit proliferation of the sensitive head and neck cancer cell line HN5 using a standard viability assay (Fig. 1). Antibodies were identified that had either no effects, inhibitory effects or stimulatory effects on proliferation of HN5 cells. Twenty-four antibodies from various clusters were selected for further functional evaluation.

Ranking of in vitro efficacies of individual mAbs and mixtures.

The HN5 cell line is very sensitive to inhibition by anti-EGFR mAbs; therefore, the human cancer cell line A431NS, which is partially resistant to mAbs, was selected for the functional analysis of antibody mixtures in order to allow mixtures with superior activity to be identified (Fig. 2A). The 24 selected antibodies were tested as individual antibodies and in all possible mixtures of two for their ability to inhibit proliferation of the A431NS cell line at a total antibody concentration of 2 µg/ml. The functional activity of the mixtures was ranked according to their maximum level of growth inhibition. In total, more than 400 mixtures were evaluated in this way (data not shown). Many antibody mixtures showed enhanced efficacy compared with the individual antibodies; a set of 10 antibodies (992, 1024, 1030, 1260, 1261, 1277, 1284, 1320, 1347 and 1565) that were frequent in mixtures with high activity were selected for further analyses. The individual antibodies had a diverse range of inhibitory activity (Fig. 2B). The 10 antibodies were then tested in all possible mixtures of two for their ability to inhibit proliferation of the HN5 cell line and in all possible mixtures of two and three for their ability to inhibit proliferation of the A431NS cell line using the WST-1 viability assay. mAbs and mixtures were ranked according to their level of inhibition (Fig. 2C and Sup. Fig. S1). Eight mixtures of two antibodies and 38 mixtures of three antibodies had enhanced efficacy compared with the individual antibodies, and were statistically superior to the reference cetuximab in the A431NS cell line (Table 1). No noticeable increase in the level of inhibition of the A431NS cell was found by adding a third antibody to the mixtures of two antibodies.

Figure 2.

Functional evaluation of antibody mixtures. (A) Dose-response curves for inhibition of growth of the two cancer cell lines A431NS and HN5 by cetuximab. Cells were treated with cetuximab for 96 hours. The metabolic activity was measured by WST-1 addition and calculated as percentage of untreated control shown as mean ± SD. (B) Dose-response curves for inhibition of growth of the cancer cell line HN5 by 10 selected anti-EGFR antibodies and cetuximab after 96 hours of treatment with metabolic activity calculated as percentage of untreated control. (C) HN5 and A431NS cells treated with 2 µg/ml of the indicated antibody mixtures or cetuximab for 96 hours. The metabolic activity was measured by WST-1 addition and calculated as percentage of untreated control.

Table 1.

Characteristics of individual anti-EGFR antibodies and mixtures

%MAC: Percentage of metabolically active cells (MAC) compared with untreated control at 2 µg/ml of total antibody. SEM: Standard error of mean. Significance: degree of statistical significance in %MAC between the mixtures and cetuximab. *p < 0.05, **p < 0.01, ***p < 0.005 and ****p < 0.001. SI: Synergy index.

Epitope binning and mapping of selected anti-human EGFR mAbs.

To better understand the molecular mechanism causing the superior anti-proliferative effect of antibody mixtures, the epitopes of each of the 10 IgGs were characterized by cross-blocking experiments performed in pairs using Surface Plasmon Resonance. Initially, domain specificity was determined using a panel of reference antibodies with known domain specificity²⁹ (data not shown). Subsequently, the 10 selected mAbs, cetuximab and panitumumab were analyzed by cross-competition (Fig. 3A). The analysis indicated four separate epitope bins: One was located on dI and dII and at least three separate bins were located on dIII (III/A, III/B and III/C). The reference antibodies cetuximab and panitumumab were found to bind overlapping epitopes and both belonged to epitope bin III/B. The antibodies 992 and 1024 shared the ability to cross-compete with mAbs from two separate bins (III/A + III/B and III/B + III/C respectively) indicating distinct and non-overlapping epitopes for both mAbs. Furthermore, it was repeatedly observed (n = 3 experiments) that 992 and 1024 binding was 50% enhanced (negative inhibition values) when they were tested in combination, indicating cooperative binding.³⁰ Fine specificity was addressed by binding to variants of human sEGFR where surface exposed residues spanning from the end of dII (position 289) to the middle of dIV (position 517) were successively replaced with the murine sequence. Again, the antibody panel could be divided into four unique epitope bins (Figs. 3B and S2), confirming the cross-competition studies. Antibodies categorized into epitope bin III/A were very sensitive to the H359R mutation, while antibodies classified as III/C were very sensitive to the combined substitution of S460P and G461N. Cetuximab, panitumumab, and 992 (III/B) were all susceptible to the I467M and S468N substitutions in agreement with these positions making contact with cetuximab in the X-ray crystal structure.³¹ Only 992 was found to be sensitive to the K443R substitution when tested on sEGFR and cell surface expressed full length EGFR (Figs. 3B and S3), distinguishing the 992 epitope as different from the epitopes of cetuximab and panitumumab. Pairing epitope data with the functional data clearly showed that mixtures with the highest levels of growth inhibition contained non-competing antibodies that bind domain III (Table 1). Mixtures of 992 with 1320, 1024 or 1030 were the most efficacious. mAbs belonging to epitope bin I/II did not significantly enhance the inhibitory effect of bin III antibodies. One important exception was in mixtures with 992, which seemed to be able to act synergistically with all mAbs having non-overlapping epitopes. No functional antibodies targeting domain IV were identified.

Figure 3.

Epitope characterization of the selected set of 10 anti-EGFR antibodies. (A) Epitope binning by surface plasmon resonance competition assay. The anti-EGFR antibodies identified were binned according to EGFR domain by competition with murine reference antibodies with known EGFR specificity (data not shown) and according to cross-blocking. At least 50% inhibition (red) or 50% enhancement (green) were set as arbitrary cut-off values for antibody pairs binding overlapping epitopes or displaying binding co-operativity. (B) Mutational analysis of antibodies binding to EGFR variants detected by surface plasmon resonance analysis. The effect of binding to EGFR mutants is calculated as percentage maximum response level relative to wild type. Red color indicates less than 50% binding response compared with wild type. (C) Schematic diagram of the EGFR with the approximate location of the three domain III epitope bins shown.

Cross-reactivity to cynomolgus EGFR.

Cross-reactivity to orthologue cynomolgus EGFR is a requirement for adequate toxicological evaluation of a potential lead candidate during preclinical development. Thus, the binding of all ten antibodies to recombinant extracellular domains of human EGFR and cynomolgus EGFR was tested by enzyme-linked immunosorbent assay (ELISA). Apparent affinities of soluble purified IgG1 (EC₅₀) were in the picomolar (pM) range, indicating high binding affinity (Table S1). Furthermore, it was evident that 1320 did not cross-react with cynomolgus EGFR orthologue, and hence it was undesirable to have 1320 in a drug candidate mixture. Apparent affinities to EGFR overexpressed on A431NS cells were all in the pM range (Fig. S4) and the measured EC₅₀ values of cetuximab and panitumumab were in very good agreement with previously published avidities of 87 pM and 83 pM respectively measured by surface plasmon resonance.²⁷

Antibody induced degradation of EGFR.

The relationship between antibody-induced EGFR degradation and growth inhibition for each of the 58 mAbs and mixtures of two antibodies was investigated in the cell lines HN5 and A431NS (Fig. 4A–D). It is evident from the results that mixtures of antibodies are superior to mAbs at inducing EGFR degradation in both cell lines. A clear correlation between response and the level of EGFR degradation was found for both mAbs and mixtures in the HN5 cell line, but only for mixtures in the A431NS cell line (Fig. 4A and B). The levels of EGFR degradation induced by selected mAbs and mixtures were validated by immunoblot analyses (Fig. 4C and D). All the investigated mixtures of antibodies induced higher levels of EGFR degradation compared with the individual antibodies although minor differences between the mixtures were evident. The low level of EGFR degradation induced by mAbs in the A431NS cell lines may, at least partly, explain its resistance to mAb targeting.

Figure 4.

Correlation between level of growth inhibition and level of EGFR degradation. Plots derived from dot blot analysis showing the correlation between level of viable cells and level of EGFR degradation for mAbs and mixtures in (A) the A431NS cell line and (B) the HN5 cell line. (C) Immunoblots of cell lysates of HN5 or A431NS cells treated with mAbs or mixtures for 48 hours. (D) Odyssey quantification of band intensities from the immunoblots. EGFR levels were calculated as percentage of untreated control by using MFI values normalized to β-actin levels.

Functional comparison of selected mixtures.

The most efficacious mixtures of antibodies with unique epitope bin combinations were selected for further functional evaluation. Mixtures containing 1320 were excluded because the antibody failed to bind to cynomolgus EGFR. Dose-response analysis was performed for the A431NS and HN5 cell lines to determine potency (IC₅₀) and efficacy (maximum level of inhibition) of the mixtures. Representative dose-response curves are shown in Figure 5A and B. In the A431NS cell lines, all mixtures reached the same level of inhibition and all mixtures were superior to cetuximab. The 992 + 1347 mixture was five times less potent than the mixtures of domain III antibodies. No additional benefit was derived by adding a third antibody to the bi-clonal mixtures. In the HN5 cell line, all mixtures and cetuximab reached the same level of inhibition and with approximately similar potencies.

Figure 5.

Functional evaluation of lead candidate mixtures in vitro and in vivo. Dose-response curves for inhibition of growth of the cell line A431NS (A) and HN5 (B) by the indicated antibodies and antibody mixtures. Cells were treated with antibody for 96 hours. The metabolic activity was measured by WST-1 addition and calculated as percentage of untreated control shown as mean ± SD. (C) The growth inhibitory effects of two lead candidate mixtures and cetuximab in A431NS human tumor xenografts. (D) The growth inhibitory effects of Sym004, the two individual antibodies 992 and 1024, and cetuximab in A431NS human tumor xenografts.

In vivo efficacy of selected antibody mixtures.

Mice bearing A431NS tumors were treated with negative control antibody, 992 + 1024, 992 + 1024 + 1030 or cetuximab for 25 consecutive days with twice weekly injections of 50 mg/kg of total antibody (Fig. 5C). Cetuximab partially suppressed tumor growth, while the group of mice treated with antibody mixtures had tumor regression and sustained tumor suppression. Tumors were slower at responding to treatment with the 992 + 1024 + 1030 mixture compared with the 992 + 1024 mixture and tumor re-growth was observed in one mouse upon end of treatment. No recurring tumors were detected at day 90 in the 992 + 1024-treated mice, demonstrating a long-lasting antitumor effect of this antibody mixture.

To confirm the synergy of the 992 + 1024 mixture observed in vitro, A431NS tumors were treated with the individual antibodies 992 and 1024 or the mixture of these in vivo (Fig. 5D). Only the 992 + 1024 mixture was able to induce complete tumor regression and tumor suppression of all treated tumors, while the individual antibodies, 992, 1024 and cetuximab only delayed tumor growth.

Discussion

Conventional, antibody-based drugs targeting EGFR are mAbs that block EGFR signaling by interfering with ligand binding. However, Friedman and colleagues showed that mixtures of antibodies with non-overlapping epitopes were able to inhibit EGFR and HER2 driven tumor growth more efficiently than single mAbs.²² These results suggest that rational approaches involving early testing of antibody mixtures can be applied to drug discovery, leading to identification of optimal mixtures of antibodies with different and more efficacious mechanisms of action than mAbs.

To test this hypothesis, a novel comprehensive panel of chimeric antibodies specific for human EGFR was generated. High throughput screening of antibody mixtures in vitro allowed comparison of hundreds of mixtures and rapid identification of highly efficacious antibody mixtures. Synergistic effects were frequently observed, but only with mixtures of antibodies that bind to non-overlapping epitopes on EGFR. Hence, the underlying molecular mechanism is likely to involve receptor cross-linking and extensive lattice formation on the cell surface, which in turn trigger widespread EGFR internalization and degradation as described by Ben-Kasus et al. and Pedersen et al.^23,32 These results are supported by the strong correlation between the level of growth inhibition and the level of EGFR degradation in two human cancer cell lines found in this work.

However, a high level of growth inhibition and synergy was not observed for all mixtures of antibodies binding non-overlapping epitopes, suggesting that the spatial and steric constraints involved in antibody binding are equally important for efficient receptor cross-linking and degradation. The most potent and efficacious antibody mixtures contained antibodies binding to uniquely positioned non-overlapping epitopes on dIII, which may be ideal for receptor cross-linking and building of higher order aggregates on the cell surface.

The antibody 992 was unique because of its ability to mediate synergistic growth inhibition with all mAbs having non-overlapping epitopes. One likely explanation is that in complex with 992, EGFR adopts a conformation highly suitable for receptor clustering on the cell surface. In support, 992 is the only antibody that is sensitive to the mutation K443R present on the lateral face of the solvent exposed dIII of EGFR, located above the relatively narrow and extended domain IV and on the opposite side of the more bulky domain I and II (Fig. S2). Assuming that the 992 epitope is located around this mutation, the second unligated Fab portion of 992 will thus extend away from EGFR almost parallel to the cell surface, making it readily available for cross-linking with other EGF receptors. Because dI and dII are able to change between an auto-inhibited tethered conformation and an unfolded extended conformation where they are rotated to bring dI within a close distance of dIII,⁵ such structural changes could impose steric hindrance and interfere with receptor cross-linking mediated by dI/dII and dIII/A antibodies. Furthermore, a large N-linked carbohydrate is found on residue N328 in very close proximity to the presumed binding site for dIII/A antibodies, which together with N-linked glycans on dI³³ could limit the flexibility of dI/II and dIII/A antibodies. Finally, antibodies binding to epitopes on dI, dII and dIV are all predicted to be closer to the cell membrane than dIII antibodies and hence sterically more constrained and less able to cross-link receptors, which may explain the non-synergistic activity of such antibodies in combinations. Three mixtures consisting of two antibodies and two mixtures consisting of three antibodies, all containing antibodies binding non-overlapping epitopes and with a high level of growth inhibition in both cancer cell lines, were selected for more detailed comparisons. Dose-response curves showed that all mixtures were highly efficacious, and that all mixtures were statistically superior to cetuximab at inhibiting growth of the A431NS cell line and as efficient in the HN5 cell line. The mixture 992 + 1347 had a lower potency than the other mixtures in the A431NS cell line and 992 + 1030 had a slightly lower potency in the HN5 cell line. A mixture of two antibodies, 992 + 1024, and a mixture of three antibodies, 992 + 1024 + 1030, were selected for final evaluation and comparison in an in vivo model based on the cell line A431NS. Both mixtures contained antibodies binding to non-overlapping epitopes on domain III of the EGFR. Both mixtures induced tumor regression, whereas the reference cetuximab was only able to delay tumor growth. The mixture of 992 + 1024 had a faster inhibitory effect on tumor growth than the mixture containing three antibodies and better tumor control, as tumor re-growth was observed in one mouse at day 75 in the 992 + 1024 + 1030 treatment group. This outcome was unexpected because no significant difference was evident in vitro, which indicates that other factors are contributing to growth inhibition in vivo. One important factor could be the dose of individual antibodies. In the mixture containing two antibodies, each antibody was administered at a dose of 25 mg/kg (50 mg/kg total) compared to 16.6 mg/kg of each antibody in the mixture containing three antibodies (50 mg/kg total). Because concentration drives penetration of the antibodies into the tumor and the individual serum antibody concentration was lower in the mixture containing three antibodies, this lower concentration may explain the delay in growth inhibition and, ultimately, the tumor control observed for the 992 + 1024 + 1030 treatment group compared with the results of 992 + 1024 treatment.

Based on these results the mixture of 992 + 1024 was selected as the optimal mixture for targeting EGFR. Synergy of the two antibodies was confirmed in vivo by treatment of A431NS tumors with the individual antibodies 992 and 1024 and the mixture. Only the mixture was able to control A431NS tumor growth. The individual antibodies were as good as cetuximab at inhibiting growth of these tumors, which was surprising because in vitro results showed cetuximab to be more efficacious than either antibody. However, all three mAbs induced a similar level of EGFR degradation in the A431NS cell line, which could explain the finding and supports the hypothesis that the level of EGFR degradation is the best predictor of response to anti-EGFR antibody targeting.

Antibody effector functions like ADCC and CDC may contribute to the inhibition of tumor growth of anti-EGFR mAbs by inducing tumor cell killing. In vitro experiments with Sym004 showed that the antibody mixture was able to induce ADCC at levels similar to cetuximab and, in contrast to cetuximab, also high levels of CDC (Sup. Fig. S5). These results are supported by studies demonstrating that anti-EGFR mAbs can induce ADCC while only antibody mixtures are able to induce CDC in vitro.^27,28 To further delineate the role of complement activation and ADCC, we performed a time-course in vitro experiment (Sup. Fig. S6). This experiment clearly showed that the level of CDC induced by Sym004 was rapidly declining over time as a result of Sym004-induced EGFR internalization. Levels of ADCC were unaffected by EGFR internalization, which suggests that CDC, in contrast to ADCC, is much more dependent on high antibody density at the cancer cell surface. However, ADCC may have a limited effect on established tumors in mouse xenograft models as recently shown by Overdijk et al.³⁴ In light of this observation and the experimental data described above, we conclude that the superior growth inhibitory effect of Sym004 compared to individual mAbs and cetuximab observed in the A431NS xenograft model is likely mainly driven by prevention of EGFR signaling due to receptor internalization and degradation, and not as a result of ADCC or CDC. Interestingly, Sym004 also induced superior apoptosis compared with cetuximab in vitro (Sup. Fig. S7), indicating that activation of cell apoptosis secondary to EGF receptor internalization and degradation is part of the mechanism of action of Sym004.

In summary, based on a systematic evaluation of the in vitro efficacy of single, dual and triple antibody mixtures, it was possible to identify a very efficacious and potent antibody mixture composed of the antibodies 992 and 1024. The therapeutic potential of this mixture, called Sym004, was demonstrated by the superior inhibition of growth of cancer cells in vitro and tumor growth in vivo compared with the reference cetuximab. The clinical potential of Sym004 is currently being evaluated.

Materials and Methods

Antigens.

The extracellular domain of EGFRvIII35 was cloned together with full length extracellular human and cynomolgus EGFR (sEGFR, amino acids 1–620). Antigens were expressed in Human Embryonic Kidney (HEK) 293 cells in fusion with an N-terminal polyhistidine purification tag (6xHIS-tag). The following proteins were commercially available: Recombinant human sEGFR, R&D Systems (cat. 1095-ER); recombinant human sEGFR-Fc, R&D Systems (cat. 344-ER); and full length EGFR (amino acids 1-1,186), Sigma-Aldrich (cat. E3641).

Cell lines.

A431NS, ATCC (cat. CRL-2592), and SKBR-3, ATCC (cat. HTB-30). HN5 ³⁶ was kindly provided by the Department of Radiation Biology, Copenhagen University Hospital, Denmark.

Antibodies.

Anti-human EGFR antibodies from Thermo Scientific: 199.12 (cat. MS-396-P0), EGFR.1 (cat. MS-311-P0), 111.6 (cat. MS-378-P0), 225 (cat. MS-269-P0), 528 (cat. MS-268-P0). ICR10, AbCam (cat. ab231). Therapeutic EGFR mAbs cetuximab (Erbitux^®, Merck), panitumumab (Vectibix^®, Amgen). Negative control anti-respiratory syncytial virus palivizumab (Synagis^®, Abbott). Sheep anti-human EGFR polyclonal antibody, Fitzgerald (cat. 20-Es04), Rabbit anti-human beta actin, Cell Signaling Technology (cat. 4970), Secondary anti-bodies from LI-COR Biosciences: Donkey anti-Goat IRDye 800 CW (cat. 926-32214), Donkey anti-Rabbit IRDye 800 CW (cat. 926-32213), Goat anti-Human 800 CW (cat. 926-32232).

Immunizations and generation of antibody repertoires.

BALB/c mice, strain A, females, 8–10 weeks old were immunized with either sEGFR, sEGFR combined with sEGFRvIII, or alternating HN5 cells and sEGFR and full length EGFR. Spleens were recovered and macerated through a 74 µm cell strainer, Corning (cat. 136350-3479) to generate single cell suspensions. Single plasma B cells were isolated by fluorescence activated cell sorting (FACS) and used as genetic sources for the murine Symplex™ technology^37,38 using primers specific for mouse IgG heavy chain and kappa light chain (unpublished data). Symplex™ is a PCR-based method for cloning of antibodies from single sorted plasma cells. Due to the simultaneous amplification and linkage of V_H and V_K chains in a single PCR step, the natural heavy and light chain pairing is maintained for all antibodies. Subsequently, murine V_H-V_K gene pairs were joined with human CH and CK constant region genes, and cloned in libraries as chimeric full-length IgGs. Antibody supernatants were made by transient expression in Chinese hamster ovarian cells and screened for antibodies binding to SKBR-3 cells using FMAT assay (AB8200 Cellular Detection System, Applied Biosystems) and recombinant sEGFR by ELISA. Double positive hits were sequenced and grouped according to their V gene rearrangements and pattern of somatic mutations, which were determined by alignment to germline sequences from the ImMunoGeneTics sequence directory. Cladograms displaying the relationship between unique V gene rearrangements (antibody clusters) were generated using ClustalW. Antibodies were said to belong to the same cluster if they shared V_H, J_H, V_K and J_K family segments, heavy chain CDR3 lengths, and a consensus sequence in CDR3 for both heavy and light chains.

Viability assay.

A standard WST-1 viability assay, Roche Applied Science (cat. 05015944001) was used to measure the growth inhibitory activity (G.I.) of the mAbs and mixtures and performed as described in Pedersen et al.³² All viability assays were performed in media containing 0.5% FBS and cells were incubated with antibodies for 4 days. Metabolic activity was calculated as percentage of untreated control. All antibodies were stored individually at −80°C or at 4°C for shorter periods. Antibody mixtures were generated prior to performing experiments and mixed in ratios of 1:1 (w/w) or 1:1:1 (w/w) and immediately thereafter added to experimental wells or injected into mice. Statistical significance was calculated using the Student's t test. IC₅₀ values were determined using GraphPad Prism by fitting the dose-response curves to the equation Y = Bottom + (Top − Bottom)/(1 + 10^{((LogIC₅₀ - X)*HillSlope)}) using the least squares method. Synergy index (SI) was calculated as SI = ((% MACmab1 + % MACmab2)/2)/% MACmab1mab2, with % MAC referring to the percentage of metabolically active cells compared with the untreated control.

Surface plasmon resonance-based competitive binding assay.

Antibody competitions were performed in pairs by surface plasmon resonance analysis using a Biacore 2000 instrument from GE Healthcare. Antigen and antibodies were diluted in HBS-EP buffer, GE Healthcare (cat. BR-1000-88), and injected with a flow rate of 5 µl/minute. A CM5 sensor chip, GE Healthcare (cat. BR-1005-30) was conjugated with 10,000 Resonance units (RU) of a tetra His antibody, Qiagen (cat. 34670) according to the manufacturer's instructions. Next 6xHIS sEGFR fusion protein was captured on the anti-HIS surface and the antibody combinations were evaluated in competitive binding experiments at a concentration of 40 µg/ml by recording the maximum response levels. The chip surface was regenerated with 10 mM glycine-HCl, pH 2.

Cloning and expression of EGFR point-mutants.

To minimize significant changes in the overall EGFR structure, 20 human to mouse amino acid substitutions were created at surface exposed residues spanning the end of dII (position 289) to the middle of dIV (position 517). Site-directed mutagenesis was performed by synthesis of the whole DNA sequence encoding sEGFR at Geneart (Regensburg, Germany). DNA fragments were subcloned into expression vector in fusion with an N-terminal polyhistidine purification tag (6xHIS-tag) and protein was expressed in FreeStyle™ 293 Expression System, Invitrogen (cat. K9000-01). Antibody binding to EGFR mutants or wild type EGFR was determined by surface plasmon resonance analysis using same assay conditions as for epitope binning.

Dot blot analyses.

For analysis of EGFR levels, cancer cells were seeded in 96-well plates and treated with 10 µg/ml of antibody for 24 h. Cells were then washed in 1 x PBS and whole cell lysates prepared using RIPA lysis buffer, ThermoFisher Scientific (cat. 89901). One µl of lysate was transferred to nitrocellulose membranes and probed for levels of EGFR and β-actin using sheep anti-EGFR and rabbit anti-beta actin primary antibodies. Dots were visualized by addition of 800 CW fluorescently labeled secondary antibodies and imaged and quantified using the Odyssey^® System from LI-COR Biosciences. The levels of EGFR were calculated as percentage of untreated control using MFI values normalized to β-actin levels. Pearson correlation coefficient r and two-tailed p value were calculated using GraphPad Prism.

Immuno blot analyses.

EGFR levels were determined by preparing whole-cell lysates from cell lines and analyzed by immunoblot as described previously in reference ³⁹. One µg of protein was loaded on the gel and the membranes probed for levels of EGFR and β-actin using sheep anti-EGFR and rabbit anti-beta actin primary antibodies. Band intensities were quantified using the Odyssey^® System, LI-COR Biosciences. The levels of EGFR were calculated as percentage of untreated control using MFI values normalized to β-actin levels.

Mouse xenograft in vivo studies.

10⁶ A431NS cells suspended in 0.1 ml of PBS were inoculated subcutaneously into the right flanks of 6–8 week old BALB/c nu/nu mice. Mice were monitored daily and tumors measured two to three times weekly using calipers. Tumor volume was calculated using the formula: ½ × L × W² (L = length; W = width). Mice were randomized to treatment when the mean tumor volume reached a pre-determined size. Antibodies were administered at 50 mg/kg twice weekly by intraperitoneal injection. Data are expressed as the mean ± standard error of mean. Statistical testing was performed using Student's t test. All in vivo studies were performed in accordance with the Danish law on animal experimentation and approved by the Animal Ethical Committee, Denmark.

Abbreviations

ADCC

antibody-dependent cellular cytotoxicity

CDC

complement-dependent cytoxicity

EGFR

epidermal growth factor receptor

sEGFR

extracellular EGFR

V_H

antibody heavy chain variable region-encoding gene

V_L

light chain variable region-encoding gene

mAb

monoclonal antibody

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.