Discovery of potent allosteric antibodies inhibiting EGFR

ABSTRACT

In this work, we report the discovery of potent anti-epidermal growth factor receptor (EGFR) allosteric heavy-chain antibodies by combining camelid immunization and fluorescence-activated cell sorting (FACS). After immunization and yeast surface display library construction, allosteric clones were obtained by introducing the labeled EGF Fc fusion protein as an additional criterion for FACS. This sorting method enabled the identification of 11 heavy-chain antibodies that did not compete with the orthosteric ligand EGF for the binding to EGFR. These antibodies bind to a triple-negative breast cancer cell line expressing EGFR with affinities in the picomolar to nanomolar range. Those camelid-derived antibodies also exhibit interesting properties by modulating EGFR affinity for EGF. Moreover, they are also able to inhibit EGF-induced downstream signaling pathways. In particular, we identified one clone that is more potent than the approved blocking antibody cetuximab in inhibiting both PI3K/AKT and MAPK/ERK pathways. Our results suggest that allosteric antibodies may be potential new modalities for therapeutics.

Introduction

The epidermal growth factor receptor (EGFR, also described as ErbB1 or HER1) is a member of the receptor tyrosine kinase superfamily (RTK) and, with HER2 (ErbB2), HER3 (ErbB3) and HER4 (ErbB4), the ErbB subfamily.¹ EGF receptor is a transmembrane glycoprotein that adopts two main conformations: a tethered one and an extended one upon ligand binding. This binding promotes receptor dimerization and signaling.^1,2 Downstream signaling pathways, including PI3K/AKT and MAPK/ERK, trigger protein synthesis, cell proliferation and survival.^1,3,4 Overexpression of this receptor leading to signal amplification is often present in cancer and has encouraged researchers to investigate EGFR as an attractive target in oncology.^4,5 Approved anti-EGFR therapies are either small molecule tyrosine kinase inhibitors (binding to the intracellular domain)^3,6 or monoclonal antibodies (binding to the extracellular domain (ECD)).⁷ To date, the three monoclonal antibodies approved by the Food and Drug Administration (FDA), cetuximab, panitumumab and necitumumab, compete with natural ligand EGF binding and hence block ligand-induced activation and subsequently inhibit downstream signaling.^7,8 However, their actions are limited by the emergence of acquired resistance (including mutations within the binding site and mutations of downstream proteins),^9–12 and/or side effects caused by conservation of the site among receptors.⁷

Allosterism represents a paradigm shift for biologics through a unique mechanism of action by targeting non-active binding sites.¹³ Addressing allosteric sites, which refer to topographically distinct sites from the ligand binding one,^13,14 may allow identification of ligands with higher selectivity as these sites often are less conserved between protein subtypes.¹⁵ Moreover, binding of an antibody to these sites can also modulate protein affinity, activity and induce conformational changes.^13–15 Allosterism has thus gained more interest over the past decades as a new mechanism of action for therapeutic antibodies in particular for the ErbB family. For example, Herceptin® (trastuzumab) binds to an allosteric site of HER2.¹⁵ Regarding EGFR, matuzumab, a humanized antibody developed by Merck KGaA, Darmstadt Germany, binds to the opposite side of the ligand binding site on EGFR, subsequently blocking its rearrangement and dimerization.¹⁶ Clinical trials of matuzumab were discontinued in 2007 after Phase 2 studies, but they paved the way for the development of allosteric antibodies against EGFR. For example, Sym004 consists of a cocktail of two antibodies targeting two different non-overlapping epitopes promoting internalization and degradation of the receptor.¹⁷ MM-151, another combination cocktail targeting three sites of EGFR, has been shown to overcome acquired resistance of EGFR and a 65-fold decrease of ERK pathway activation compared to cetuximab.^18,19

The current research landscape in biologics, however predominantly emphasizes combinational strategies,^17–20 often overlooking the potential of allosteric antibodies as standalone treatments. This scientific gap, preventing full exploration of their therapeutic potential, exists because of the lack of development of allosteric antibodies and a comprehensive understanding of advantages and challenges associated with their modulations of EGFR structure and function. Addressing these gaps could unveil new avenues for the development of more effective allosteric antibody therapies in oncology, in particular for triple-negative breast cancer (TNBC) or colorectal cancer for which first-line treatments often encounter resistance.^7,9,10 Furthermore, single-agent tumor-specific compounds will lower de facto potential off-target effects in comparison to a combination cocktail. Our colleagues previously reported the discovery of unique ultra-long complementarity-determining region (CDR) H3 cow antibodies that target allosteric sites of EGFR and showed good inhibitory properties.²¹

Here, we report the discovery of potent anti-EGFR allosteric heavy-chain antibodies from an immune yeast surface display VHH library by combining a three-color staining method and fluorescence-activated cell sorting (FACS) to specifically identify antibodies that did not compete with EGF for EGFR binding. This approach allowed us to identify 11 candidates that were reformatted as heavy-chain antibodies. Four allosteric epitopes were characterized, one of them being the one of matuzumab. Several biochemical and cell-based assays highlighted their binding and modulation of EGFR function in comparison to the commercial blocking antibody cetuximab. In particular, our best hit, 30555_A2, showed better inhibitory properties than cetuximab.

Materials and methods

Immunization of camelids

Two llamas (Lama glama) around 5-year age were immunized using recombinant human (rh) soluble EGFR extracellular domain (ECD) (25–618, produced by Merck Healthcare KGaA) at Preclinics (Potsdam, Germany). After final bleeding, total RNA was extracted, and cDNA was synthetized.

All experimental procedures and animal care were in accordance with European Union animal welfare protection laws and regulations. All animals remain alive.

Library construction

Library generation was performed via homologous recombination as previously described.^21,22 Briefly, the destination vector (pDest) was digested with the restriction enzyme BsaI V2 hF (New England Biolabs) for 24 h at room temperature. The linearized vector was purified using Wizard® SV Gel and PCR Clean-Up System (Promega) and the concentration was measured on NanoDrop (ThermoFisher Scientific). Yeast surface display VHH library generation was performed in Saccharomyces cerevisiae EBY100 cells. Briefly, cells were resuspended in the electroporation buffer with the linearized vector and a pool of DNA-insert. Twelve electroporation reactions were performed in parallel.

Library sorting

For library screening and sorting, yeast cells were grown overnight in SD medium with drop out mix without tryptophan for 24 h at 30°C. Afterwards, cells were seeded at an optical density of 1 in SG medium with drop out mix without tryptophan for 48 h at 20°C.

rh his-tagged EGFR ECD (25–642) was produced by Merck healthcare KGaA. Recombinant human EGF Fc fusion protein was purchased from Sino Biological (10605-H01H). rh his-tagged EGFR ECD (1 µM) was incubated with rh EGF Fc (3 µM) for 1.5 h at room temperature prior to the staining procedure.

All labeling steps were performed on ice and in the dark for 30 min with 1 × 10⁷ cells in 20 µL for the controls and 5 × 10⁸ cells in 1 mL (first sorting round) or 1 × 10⁸ cells in 200 µL for the sorting (further sorting rounds). Briefly, cells were first washed with phosphate-buffered saline (PBS) followed by 30 min incubation with the pre-formed complex EGFR:EGF Fc. Cells were washed twice with PBS and then incubated for 30 min with secondary detection antibodies. VHH surface display was monitored using goat-HA Tag Alexa Fluor® 488-conjugated antibody (R&D systems, IC6875G, 1:20). Detection of antigen binding was performed using SureLight® APC Anti-6X His Tag® antibody (Abcam, ab72579, 1:20) for EGFR, R-Phycoerythrin AffiniPure Goat Anti-Human IgG, Fc Fragment specific (Jackson ImmunoResearch, 109-117-008, 1:20) for EGF Fc. After washing twice with PBS, cells were resuspended in appropriate volume for FACS sorting on a BD FACSAria™ Fusion cell sorter (BD Biosciences).

Expression and purification

After sequencing, selected hits were reformatted in-house into pTT5 plasmid as VHH-Fc fusion protein format (IgG1 backbone). DNA preparation was done using ZymoPURE II Plasmid Miniprep Kit (Zymo Research), and concentration was checked using QIA expert system (Qiagen). For expression, Expi293 and ExpiCHO cells were transiently transfected with expression vectors according to the manufacturer’s protocols (ThermoFisher Scientific). Harvest of the tubes has been done 6 (Expi293F) or 7 (ExpiCHO) days post transfection and purified using MabSelect antibody purification chromatography resin (GE Healthcare). Samples were formulated in PBS (Sigma Aldrich) and concentrations were determined using QIA expert system (Qiagen). Purity and aggregates formation were checked by size-exclusion chromatography (SEC) analysis.

Octet assays

Qualitative binding assays were performed with the antibodies using the Octet RED® biolayer interferometry (BLI) system (Sartorius) according to the manufacturer´s guidelines at 25°C with 1000 rpm agitation. For initial binding confirmation: antibodies were loaded onto AHC biosensors for 180 sec (5 µg/mL), after a 60 sec rinsing step in kinetic buffer (KB; PBS + 0.1% Tween 20 + 1% bovine serum albumin, BSA) association to rhEGFR was tested for 180 sec followed by a dissociation step in KB for 180 sec. A kinetic analysis was done for EGFR-binders, after loading of the antibody on AHC biosensors, association with varying concentrations of the EGFR (2-fold dilution from 50 nM) was measured for 300 sec followed by a dissociation step for 600 sec in kinetic buffer. Data were analyzed using Sartorius analysis software and K_Ds were determined according to a 1:1 model. Competition assays were performed by loading rhEGFR (5 µg/mL) on HIS1K biosensors for 180 sec, after a rinsing step in KB for 60 sec, the first association lasted for 360 sec followed by another association for 360 sec. EGF, matuzumab (produced by Merck Healthcare KGaA) and Erbitux® (cetuximab, produced by Merck Healthcare KGaA) were also included in this assay and, for each sample both orientations were tested to highlight overlapping epitopes. Lastly, cross reactivity binding assays were performed using rhHER2 (Sino Biological 10,004-H08H), rhHER3 (10368-RB-050), rhHER4 (1131-ER-050), recombinant cynomolgus monkey EGFR (10405-ER-050) and recombinant mouse EGFR (1280-ER-050) all from R&D Systems. Antibodies were loaded on AHC biosensors (5 µg/mL), and binding tested by association/dissociation of the cross-reactive protein (100 nM). Relevant controls were included for each experiment.

Fluorescence polarization

Recombinant human EGF (R&D Systems, 236-GMP-01 M) was fluorescently labeled using Atto 488 protein labeling kit (Sigma Aldrich) according to the manufacturer´s datasheet. Binding capacity to EGFR after labeling was confirmed by BLI experiments. Dye-to-protein ratio was determined by measuring the absorbance of the labeled EGF at 280 nm and 498 nm on the NanoDrop. Antibodies were concentrated using Amicon 10K (Sigma Aldrich). All the subsequent experiments have been performed at room temperature and fluorescence polarization (FP) was monitored regularly over 24 h using PHERAstar (BMG).

First, association of the labeled EGF on EGFR and subsequent increase of FP was tested by association of 100 nM EGF-Atto 488 with EGFR (2-fold dilution from 10 µM). An EGFR concentration resulting in 50% of the maximum FP was chosen for further assays. Secondly, a competition assay between EGF-Atto 488 and unlabeled EGF was performed. For this assay, 100 nM EGF-Atto 488, 400 nM EGFR and a 2-fold dilution of unlabeled EGF were used. Finally, antibodies were tested for their impact on the equilibrium of EGFR:EGF-Atto 488 by titrating the pre-formed complex with the samples (2-fold dilution from 10 µM).

Functional assays

EGFR-expressing breast carcinoma cell line MDA-MB-468 was obtained from ATCC and cultured in RPMI media supplemented with 10% fetal bovine serum, 1% glutamine and 1% sodium pyruvate.

Functional binding of the expressed antibodies on EGFR-expressing cells was done using an iQue3 device (Sartorius). To this end, 10⁵ cells/well were seeded and after two washing steps in PBS + 1% BSA, incubated on ice for 1 h with 100 nM of expressed antibodies. After two washing steps with PBS + 1% BSA, 250 nM of Alexa Fluor® 488 anti-human Fc (Jackson ImmunoResearch, 109-547-008) was used for staining for another 30 min on ice and in the dark. After two washing steps with PBS + 1% BSA, 20 µg/mL of propidium iodide (Invitrogen) was used to label dead cells in a total volume of 100 µL. Relevant controls were included, and gates set according to these controls.

A titration of anti-EGFR binders on MDA-MB-468 cells (EC₅₀ determination) was done according to the aforementioned protocol except for the incubation steps for which the concentration ranged from 250 nM to 0.0596 pM of the expressed antibodies (serial dilution) and 500 nM of Alexa Fluor 488 anti-human Fc detection antibody.

The impact of anti-EGFR antibodies on EGF-binding kinetics was evaluated by FACS. First, rhEGF (R&D Systems, 236-GMP-01 M) was biotinylated using EZ-Link Sulfo-NHS-Biotin (ThermoFisher Scientific). Confirmation of the biotinylation and binding capacity of the biotinylated EGF (bEGF) was confirmed by BLI and FACS experiments. Then, 10⁵ MDA-MB-468 cells were seeded per well. Cells were subsequently incubated with a ranging concentration of antibodies for 1 h on ice, washed twice with PBS + 1% BSA and incubated with 100 nM bEGF on ice for 2.5 h. Cells were washed twice before a 0.5 h-incubation step with 250 nM Streptavidin AlexaFluor® 647 conjugate (ThermoFisher Scientific); two supplemental washing steps were performed before final resuspension in PBS + 1% BSA, 20 µg/mL of propidium iodide.

MAPK/ERK signaling pathway

A cell line reporter assay using luciferase (SRE Reporter – HEK293 Cell line, BPS Bioscience) was performed to evaluate the impact of the antibodies on the MAPK/ERK signaling pathway. All assays were performed according to the manufacturer´s data sheet. First, a response assay to EGF was performed to validate its native functionality and for the determination of its EC₅₀. The second assay was the inhibition of EGF-induced reporter activity by antibodies. This assay was performed at EGF EC₈₀ to evaluate a decrease of the EGF-induced activity. The luminescence was measured using a Synergy 4 device (Biotek).

pAKT signaling pathway

The Phospho-AKT1/2/3 (Ser473) HTRF Kit (64AKSPEG, Revvity) was used to investigate the impact of our antibodies on the AKT pathway. MDA-MB-468 cells were seeded onto a sterile 96-well culture plate (3,5×10⁴ cells per well) in complete medium. After adherence, cells were serum-starved overnight. The next day, cells were pretreated with the respective antibody for 1 h in serum-free medium at 37°C. Then, cells were stimulated with 20 ng/mL of EGF, after 10 minutes medium was removed, and cells were subsequently incubated 30 min in supplemented lysis buffer at room temperature under shaking. Afterwards, lysates were analyzed using manufacturer’s protocol. HRTF measurements were done using a PHERAstar device (BMG).

Domain mapping via YSD

Domain mapping was performed via yeast surface display as described before.²³ Briefly, the ECD of EGFR was subdivided into six truncated segments consisting of residues 1–124, 1–176, 1–294, 273–621, 294–543, or 475–621, and displayed on yeast cells under the control of a galactose inducible promoter. Cells were cultured and induced in medium lacking tryptophan. Surface presentation was verified by an anti-c-myc AlexaFluor® 647 conjugated antibody (Abcam, ab223895) and antibody binding was measured via Alexa Fluor® 488 anti-human Fc (Jackson ImmunoResearch, 109-547-008). Measurement was performed using an iQue3 device (Sartorius).

Data analysis

FlowJo software was used for plotting FACS data. Other graphical and statistical analysis were done with GraphPad Prism 8 software. When one-way ANOVA comparison was performed, results values of p ≤ 0.01 were considered as statistically significant.

Results

Discovery of non-competitive binders

Aiming at harnessing allosterism as a mode of action for antibodies, the first objective was to obtain non-EGF-competitive antibodies. These should bind to EGFR at a site that is different from the EGF binding site irrespective of the presence of EGF. We selected the VHH format due to its ease of use regarding both library construction and display. Additionally, the VHH format would facilitate the rapid engineering of multispecific antibodies. To this end, camelids were immunized with EGFR ECD and a VHH yeast surface display library was constructed (Figure 1a). Then, to isolate VHHs that do not bind the EGF binding site of EGFR, we blocked its access by preincubating EGF Fc fusion protein (EGF Fc) with EGFR (ratio 3:1) for 1.5 h at room temperature. The pre-formed complex EGFR:EGF Fc was then added to the yeast surface display immune library. Yeast cells were subsequently stained with three detection antibodies: anti-HA AlexaFluor® 488 for confirmation of VHH display, anti-His-APC for detection of EGFR binding and anti-Fc-PE for detection of EGF Fc binding (Figure 1a). Cells were subsequently screened by FACS using a three-color staining strategy (Supplemental data 1). First, a two-dimensional gate was applied to yeast cells that display VHHs on their surface and bind to EGFR binding positive cells (classical gate). Then from these gated cells, the second two-dimensional gate was applied to select yeast cells binding to EGFR in the presence of EGF. In the first sorting round, 0.6% of the library were EGFR-binders while 0.54% were double positive for the complex EGFR:EGF. We were able to enrich non-EGF-blocking EGFR binders within two rounds of sorting (Figure 1b). Next, we aimed at obtaining antibodies that not only bind EGFR at an allosteric site, but also might modulate the EGFR conformation in way that eventually results in reduced EGF binding. For this purpose, we screened for binders that retain EGFR binding in presence of EGF but to a smaller extent, which could be due, besides lower affinity, to EGFR conformational modulation. Hence, we chose a sorting window that is more stringent and selected only clones from the section part of the EGFR:EGF sorting gate (i.e., low EGF binding) for which we selected around 7% of the double positive cells. Relevant controls included binding to EGF Fc only to confirm the non-enrichment of EGF binders.

Figure 1.

Discovery of allosteric antibodies. A: workflow overview. After camelid immunization with rhEGFR (1), a yeast surface display library was generated (2) and subsequently incubated with EGFR and EGF fc and their respective detection antibodies (3). Non-egf-competitive binders were selected via FACS using a three-color staining strategy (4), reformatted as heavy-chain antibodies (5) and expressed in mammalian cells (6). Antibodies were characterized using various biochemical and functional assays (7). Created with BioRender.com. B: FACS-based selection of non-egf-competitive antibodies. A three-color based sorting approach was applied, first a two-dimensional gate to identify functional VHH display in combination to EGFR binding (upper panel). Then, a second two-dimensional gate was applied to select EGFR binders that did not compete with EGF fc (lower panel). For the final round a more stringent sorting gate was designed to select antibodies that were better negatively modulating EGFR affinity for EGF fc. Applied sorting gates and corresponding cell population (as % of total cells) are shown. Plots were generated using FlowJo. C: sequence homology of egfr-binders. Green bars highlight 100% sequence identity at a particular residue, olive bars indicate an intermediate sequence conservation, and red bars represent a high sequence diversity. Alignment generated with geneious prime.

Antibodies are binding to four different epitopes

Sequencing of 192 clones revealed the isolation of 23 unique sequences in total. VHH sequences were analyzed and clustered based on their complementarity-determining region 3 (CDR3) (Figure 1c). 20 sequences were identified from the second sorting round using EGFR:EGF Fc heterodimer (5 clusters and 8 singlets) and 16 (6 clusters and 3 singlets) from the extra third round of stringent sorting. All sequences were reformatted. Variable regions were reformatted into VHH-Fc fusion format (IgG1 backbone), transiently expressed in Expi293F or ExpiCHO cells, and purified by MabSelect. Yields ranged from two to three-digit milligrams per liter with some notable differences for a few molecules (e.g., 30702_B3, 30702_C1) between Expi293F and ExpiCHO host cells (Supplemental data 2). SEC analysis showed that 19 antibodies were successfully expressed (purity >90%) (Supplemental data 2). Qualitative binding to recombinant human EGFR ECD using BLI revealed 11 rhEGFR-specific binders. Unsurprisingly, most of the non-binders were represented as singlets in the sequencing and binders were those belonging to clusters. No binding to rhEGF was shown. The competition assay indicated that none of these binders competed with EGF or cetuximab for EGFR binding, confirming the objective of the three-color staining and sorting procedure. Importantly, antibodies were able to bind EGFR in the presence and absence of EGF (data not shown). Despite harboring fundamentally different CDR3 sequences, three binders competed with matuzumab for EGFR binding (Table 1). The epitope binning assay highlighted four different non-overlapping bins, one of these bins being the one of matuzumab. We subsequently determined binding kinetics of the binders to EGFR by BLI (Table 1). Affinities ranged from sub nanomolar to two-digit nanomolar K_D values (Supplemental data 3). Unlike orthosteric sites, allosteric sites are less conserved among protein subtype than across species.¹¹ To address antibody specificity, we performed cross-reactivity binding assays to related proteins using rhHER2, rhHER3, rhHER4, cynomolgus EGFR and mouse EGFR. Interestingly, none of the antibodies bound to rhHER2, rhHER3, rhHER4. Only one molecule showed no binding to cynomolgus EGFR and mouse EGFR (30556_G9), whereas all the others bound to both orthologs (Table 1). Furthermore, because matuzumab did not bind to mouse EGFR but our three antibodies competing with matuzumab did, we assumed instead that they bind overlapping epitopes.

Table 1.

Biophysical data of anti-egfr antibodies. Affinities and binding assays were performed by BLI. Recombinant human (rh), recombinant cynomolgus (rc) and recombinant mouse (rm).

	KD (M)	Kon (1/Ms)	Koff (1/s)	Binding to rcEGFR	Binding to rmEGFR	Binding to rhHER2	Binding to rhHER3	Binding to rhHER4	Bin
30702_A11	8.51.10−10	1.50.105	1.28.10−4	yes	yes	no	no	no	1
30702_B3	1.85.10−8	1.53.105	2.83.10−3	yes	yes	no	no	no	2
30702_B4	8.27.10−10	1.58.105	1.30.10−4	yes	yes	no	no	no	1
30702_C1	1.07.10−9	1.90.105	2.03.10−4	yes	yes	no	no	no	1
30702_D10	3.06.10−9	2.04.105	6.24.10−4	yes	yes	no	no	no	1
30702_E5	1.53.10−8	7.38.104	1.13.10−3	yes	yes	no	no	no	3 (matuzumab)
30702_G5	3.03.10−9	1.72.105	5.20.10−4	yes	yes	no	no	no	1
30702_H9	2.40.10−9	1.91.105	4.58.10−4	yes	yes	no	no	no	1
30555_A2	2.25.10−9	1.88.105	4.23.10−4	yes	yes	no	no	no	3 (matuzumab)
30556_G9	5.42.10−8	2.37.105	1.29.10−2	no	no	no	no	no	4
30557_E6	5.44.10−8	3.07.105	1.67.10−2	yes	yes	no	no	no	3 (matuzumab)

Effect on equilibrium EGFR:EGF complex

Allosterism occurs when the properties of one ligand binding a target molecule are altered upon the binding of a second ligand to a topographically distinct site on the target molecule.¹⁵ To assess the impact of these newly identified antibodies on EGFR:EGF equilibrium, we developed a FP assay adapted from classical protein-protein FP assay.²⁴ EGF was labeled using Atto 488 labeling kit at a dye-to-protein ratio of approximately 1. The fact that the affinity was not impaired was confirmed by BLI experiments (data not shown). FP increase was investigated upon titration of 100 nM of Atto 488 labeled EGF with EGFR. Values were stable after one hour and for at least 24 hours (duration of the assay). For further assays we fixed the concentration of EGFR at 400 nM (50% of the maximum of FP response obtained). Competitive displacement of the signal upon the binding of an increasing concentration of unlabeled EGF was also validated (data not shown). Adaptation of this assay with antibodies highlighted three different modulations of the equilibrium (Figure 2). 30557_E6 increased FP values and therefore the affinity of EGF for EGFR (positive allosteric modulator of EGF affinity), 305556_G9 and 30702_B3 had no impact on the equilibrium whereas all the remaining antibodies decreased FP values in a comparable manner to matuzumab (negative allosteric modulators of EGF affinity) but not like cetuximab which competes with EGF for EGFR binding (bottom plateau of FP values). Interestingly, antibodies having no impact on the equilibrium were not found in the additional sorting round (left part of the double positive EGFR:EGF gate) though they were only represented as singlet in the previous round of sorting. 30557_E6 which was increasing EGF affinity was still found but as the one with the lowest affinity for rhEGFR, this can explain its presence in this gate.

Figure 2.

Impact of antibodies on EGFR:EGF equilibrium determined by fluorescence polarization. Atto-488 labeled EGF (100 nM) was incubated with EGFR (400 nM) until equilibrium. Subsequently, antibodies were added. Cetuximab was used a control as it fully competes with EGF. Fluorescence polarization values were monitored using a plate reader with adequate FP module. Graphs show means ± SD. The experiment was run twice in duplicate, data shown for one representative experiment.

Binding to triple-negative breast cancer cell line

As an initial assay, we determined the binding properties of selected VHH-Fc antibodies to EGFR-expressing cells TNBC cell line (MDA-MB-468).²⁵ All 11 antibodies were able to bind the endogenous EGFR (MDA-MB-468 cells) but not to Chinese hamster ovary cells, a negative cell line for EGFR expression, which confirmed specific binding of the antibodies (Figure 3a). We also performed a titration of the antibodies on the cells to obtain respective EC₅₀ values (Figure 3b). These values ranged from single-digit picomolar to two-digit nanomolar. Interestingly, stronger binders were those for which the bin corresponds to the one of matuzumab, unlike in the BLI kinetic experiment, proving the strong impact of avidity on high EGFR-expressing cells and conformation-dependent binding.

Figure 3.

Functional assays on triple-negative egfr-expressing breast cancer cells (MDA-MB-468). a: cellular binding of antibodies on cells detected by secondary-labeled detection antibody and recorded by FACS. The experiment was run twice in duplicate, data shown for one representative experiment. Graphs show means ± SD. b. EC₅₀ values were determined after data fitting on GraphPad prism. c: displacement of bEGF affinity by anti-egfr antibodies. The experiment was run twice in duplicate, data shown for one representative experiment. Graphs show normalized means ± SD. d IC₅₀ and EC₅₀ values were determined after data fitting on GraphPad prism.

To evaluate the impact of the antibodies on the binding affinity of EGF to EGFR, we biotinylated EGF (bEGF). Confirmation of its biotinylation and binding capacity to EGFR was confirmed by BLI and FACS experiments. An increasing bEGF concentration was titrated on MDA-MB-468 cells and EC₅₀ of bEGF was determined at 2.2 nM (data not shown). Displacement of bEGF binding affinity on EGFR-expressing cells highlighted that all antibodies but one (30702_B3) negatively impacted bEGF binding to the cells (Figure 3c). 30702_B3 had no impact on EGFR:EGF equilibrium in the FP assay. More interestingly, the two other molecules that did not negatively modulate the equilibrium in the FP assay were the only ones that did not inhibit the binding of EGF to the cells at higher concentration (30556_G9 and 30557_E6). Differences observed between FP and cell-based assays can be explained by the fact that rhEGFR and endogenous EGFR may have a slightly different conformation as highlighted by the difference between EC₅₀ values in comparison to BLI values. Nevertheless, the eight remaining antibodies that inhibited EGF binding were not as potent as cetuximab (Figure 3d). Even though three antibodies bound to an overlapping bin with matuzumab, only one showed similar characteristics to matuzumab in this assay (30557_E6) strengthening the hypothesis of overlapping epitopes.

Impact on EGFR signaling

Another aspect of allosterism is the “efficacy” of the sample on probe-induced pathways.¹⁵ Therefore, we used a reporter cell line (BPS Bioscience) to monitor the impact of our newly discovered antibodies on the activity of the MAPK/ERK signaling pathway, one of the conserved pathways among tyrosine kinases.⁴ After determination of the EC₅₀ of EGF on this reporter cell line (EC₅₀: 1.16 ng/mL), a titration of the samples at a fixed concentration of EGF (EC₈₀: 1.31 ng/mL) was performed. All antibodies negatively modulated the EGF-induced luciferase response (Figure 4a). Interestingly, antibodies that had no negative impact on the equilibrium EGFR:EGF in the FP assay or on EGF-binding on MDA-MB-468 cells were also the ones showing low inhibition of the pathway in this reporter assay. In particular, 30702_B3 reached a plateau at approximately 50% inhibition. On the contrary, one of the antibodies binding to the same bin as matuzumab was more potent to inhibit the MAPK/ERK pathway than the control antibody cetuximab: IC₅₀ was 99.4 pM for cetuximab and 38.2 pM for 30555_A2 (Figure 4b). Matuzumab-bin antibodies 30702_E5 and 30557_E6 inhibitory concentrations were comparable to matuzumab, in the two-digit nanomolar range. These values are in accordance with the fact that matuzumab can inhibit PI3K/AKT pathway, but to a lesser extent the MAPK/ERK one,²⁶ suggesting again overlapping epitopes.

Figure 4.

Inhibition of egf-induced MAPK/ERK and PI3K/AKT pathways by anti-egfr antibodies. a: inhibition of MAPK/ERK pathway using BPS bioscience reporter cells. Cells were seeded, after adherence and overnight serum starvation, incubated with the respective antibodies for 24 h. Cells were stimulated with EGF at its EC₈₀ for 6 h before reading of luminescence via a plate reader. Graphs show normalized means ± SD. The experiment was run three times in duplicate, data shown for one representative experiment (left panel). b: IC₅₀ values were obtained after data fitting on GraphPad software (right panel). c: inhibition of PI3K/AKT pathway using HTRF kit from revvity to monitor AKT phosphorylation. After seeding of MDA-MB-468 cells, adherence and overnight serum-starvation, cells were treated with 250 nM of the respective antibodies for 1 h, subsequently cells were activated with 20 ng/mL of EGF for 10 min. Homogeneous time-resolved fluorescence ratio (HTRF ratio) was calculated according to the manufacturer’s instructions as follows: acceptor (665 nm)/donor (620 nm) × 10,000. Graphs show means ± SD. ****p-value < 0.0001. The experiment was run three times in triplicate.

Subsequently, the impact of our antibodies on the PI3K/AKT pathway was evaluated using the Phospho-AKT1/2/3 (Ser473) HTRF Kit (Revvity) (Figure 4c). As expected, cetuximab and matuzumab effectively reduced AKT phosphorylation (p < 0.0001). Interestingly, 30555_A2 was the only molecule inhibiting EGF-induced AKT phosphorylation at a concentration of 250 nM. This antibody also showed better inhibition than the control antibody cetuximab as for the ERK pathway. All the other antibodies did not inhibit significantly AKT phosphorylation.

Domain mapping via YSD

To scrutinize domain coverage, we performed YSD-based domain mapping as described previously²³ (supplemental data 4). Briefly, the different domains of EGFR were displayed on the surface of yeast individually and antibody binding to a specific domain was tested by FACS. It was impossible to identify the binding domain for samples binding to the first bin. Hence, we assumed these antibodies bind to an epitope that is not displayed properly on yeast surface or an epitope spanning two domains. 30702_B3 and 30556_G9 bins have been associated with binding to domain II of EGFR (they bind exclusively to 1–294 variant). Notably, these samples did not show any sequence similarity or competition indicating the targeting of distinct epitopes. These two samples had the least effects on EGF affinity and efficacy, and this might be explained by the fact that they are binding to domain II of EGFR stabilizing EGF-bound extended conformation and may not completely block the dimerization of the receptor. Even though three antibodies overlap with the binding site of matuzumab, binding to domain III was confirmed for only one molecule, 30555_A2 binds to 273–621 and 294–543 regions as matuzumab. This strongly suggested that those antibodies must have overlapping binding epitopes with matuzumab rather than strictly the same one.

Discussion

The ErbB family and particularly EGFR was identified early on as an attractive target in oncology because of its tumor cell overexpression and pivotal role for tumor cells.^4,5 Although many efforts in the EGFR-targeting therapy area have been made, most of the patients do not respond to currently approved modalities and the responders frequently develop resistance.^9–12 Therefore, to target this receptor, novel therapeutic approaches are needed.¹⁰ Aiming to identify novel potent anti-EGFR antibodies with a distinct mechanism of action than currently approved molecules,⁷ we screened a yeast surface display VHH library derived from camelid immunized with the rhEGFR.

As reported here, we generated a panel of antibodies targeting allosteric sites on different domains of EGFR by introducing the labeled EGF in early screening to block access to the ligand binding site. While Traxlmayr and colleagues recently reported the discovery of an engineered domain binding specifically to the ligand-bound conformation of EGFR suitable for chimeric antigen receptor (CAR) T cells engineering,²⁷ we have not identified such a specific EGFR binder. However, we believe that our allosteric antibodies that bind strongly to EGFR in a ligand-independent manner will not further enhance the on target/off tumor effects.

We show that these allosteric antibodies have an influence on both EGFR:EGF equilibrium (allosteric modulators of EGF affinity) and subsequent EGF-induced signaling pathway activations (allosteric modulators of EGF efficacy). Moreover, we report that antibodies that were less potent for negatively modulating EGFR:EGF equilibrium both in biochemical and cell-based assays also showed low pathway inhibition.

Remarkably, we were able to identify one clone (30555_A2) that was more potent in inhibiting both PI3K/AKT and MAPK/ERK pathways in a TNBC cell line than the approved EGF-competitive antibody cetuximab. Targeting EGFR through an allosteric site can therefore significantly improve its inhibition. We demonstrate here that allosteric antibodies show similar or even better inhibitory properties than their competitive counterparts and should also be valued during the discovery process.

Intriguingly, these findings indicate that antibodies targeting the dimerization domain of EGFR (30702_B3 and 30556_G9) do not result in the most potent inhibition of the signaling. These antibodies might rather stabilize the EGF-bound EGFR conformation. In addition, while matuzumab is known for sterically obstructing the EGFR rearrangement essential for dimerization,¹⁶ it is plausible that 30555_A2 exerts additional conformational effects, potentially accounting for its heightened potency.

Given the fact that we identified antibodies with broad epitope coverage on EGFR and that current trends in antibody discovery are directed toward engineering of bispecific antibodies,^28–30 an interesting concept would be to take advantage of the high affinity binders and generate natural-killer cell engagers to maximize tumor recognition,^31–33 An emerging approach would be to engineer biparatopic antibodies targeting two distinct epitopes of EGFR, which could also increase anti-tumor activity promoted by receptor cross-linking.³⁴ This type of construct has already been described for EGFR^23,35,36 and also for HER2,^37,38 including Jazz´s investigational Zanidatamab.³⁸

The discovery of potent allosteric antibodies targeting EGFR presents substantial advantages over EGF-competitive compounds. Their noncompetitive nature confers high selectivity with no off-target binding profiles highlighted. Furthermore, along with advancements in precision medicine, particularly in the identification of mutation patterns, these could lead to individualized targeting applications. Allosteric antibodies would thus be increasingly valued for targeted therapies in the future.

Funding Statement

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement [No 956314 (ALLODD)].

Abbreviations

BLI

biolayer interferometry

BSA

bovine serum albumin

CDR

complementarity-determining region

ECD

extracellular domain

EGFR

epidermal growth factor receptor

FACS

fluorescence-activated cell sorting

FDA

food and drug administration

FP

fluorescence polarization

KB

kinetic buffer

PBS

phosphate-buffered saline

rh

recombinant human

RTK

receptor tyrosine kinase

SEC

size-exclusion chromatography

TNBC

triple-negative breast cancer

VHH

variable domain from heavy-chain-only antibodies

Disclosure statement

Léxane Fournier, Lukas Pekar, Birgitta Leuthner, Lars Toleikis and Stefan Becker are employees of Merck Healthcare KGaA. Harald Kolmar states no conflict of interest.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19420862.2024.2406548