Inhibition of Tumor Cell Growth and Cancer Stem Cell Expansion by a Bispecific Antibody Targeting EGFR and HER3

The EGFR is a receptor tyrosine kinase (RTK) involved in the regulation of cell proliferation and differentiation. Overexpression of the EGFR is frequently observed in various cancer entities. Activation of the EGFR stimulates survival, proliferation, and motility of tumor cells and is involved in tumor initiation and progression (1–3). In the past 15 years, antagonistic antibodies such as cetuximab and panitumumab and several small molecules (e.g., gefitinib, erlotinib) targeting EGFR have been developed and approved for the treatment of colorectal cancer, head and neck squamous cell carcinoma (HNSCC), and non–small cell lung cancer (NSCLC; refs. 4, 5).

EGFR belongs to the ErbB/HER family of RTKs, additionally comprising HER2, HER3, and HER4. Binding of specific ligands such as EGF, TGFα, or amphiregulin to the extracellular domain of the receptor induces a conformational change, enabling receptor homo- or heterodimerization, typically with other HER family members. This leads to activation of the intracellular kinase domain and trans-phosphorylation of specific tyrosine residues within the cytoplasmic tails of the receptors. These phosphotyrosines generate binding sites for effector proteins, triggering the activation of downstream signaling pathways (2, 3).

Despite expressing EGFR, certain tumors are intrinsically resistant to EGFR-blocking treatments. The triple-negative subgroup of breast cancers (TNBC), for example, is characterized by the frequent overexpression of EGFR; however, cetuximab has shown limited clinical benefit in this disease setting (6, 7). Furthermore, even those tumors that initially respond to EGFR blockade eventually escape inhibition by activating compensatory signaling mechanisms. Alternatively, subclones in the tumor population carrying genetic alterations that confer a selective growth advantage to EGFR blockade can undergo expansion (8–10). One such resistance mechanism is the upregulation of HER3, or its ligand heregulin/neuregulin-1 (HRG/NRG1; refs. 11–13). Expression of HRG in the mammary gland induces adenocarcinomas in transgenic mouse models (14) and is sufficient for the metastatic spreading of breast cancer cells (15). In patients with colorectal cancer, elevated circulating HRG levels were found to be associated with de novo and acquired resistance to cetuximab therapy (16). Similarly, the comparison of paired samples of patients with TNBC treated with cetuximab or panitumumab revealed that in 25 of 42 patients, the abundance of HER3 posttreatment was increased (13).

Although HER3 has an impaired kinase domain, by heterodimerizing with a signaling-competent receptor, HER3 becomes phosphorylated and can then serve as a signaling platform. The presence of several consensus sites for the p85 subunit of PI3K mediates the potent induction of PI3K–Akt survival signaling by phosphorylated HER3, explaining its contribution to drug resistance (17). Because HER3 kinase activity is very low, HER3 cannot efficiently be inhibited by small molecules. Thus, functional blockade is rather achieved by antibodies directed toward the extracellular domain of the receptor, preventing ligand binding or dimerization (18, 19). We recently developed a HER3-targeting antibody, IgG 3-43, which potently inhibited the proliferation of gastric, lung, breast, and HNSCC cell lines in vitro and tumor growth of xenografted HNSCC cells in vivo (20). IgG 3-43 competed with the binding of HRG to HER3-expressing cells, efficiently inhibited phosphorylation of HER3 as well as downstream signaling, and induced receptor internalization and degradation. In addition, in a colorectal cancer model with oncogenic Ras-induced HRG production, IgG 3-43 suppressed proliferation and restored polarized cyst formation in 3D cultures (21).

Considering the frequent activation of HER3 signaling as a resistance mechanism to EGFR-targeted therapy, combination therapies targeting both receptors are more likely to yield a durable response. This can be achieved by cotreating with two different therapeutic antibodies. Alternatively, bispecific antibodies can be used, simultaneously targeting the two receptors. In this study, we have developed a novel bispecific tetravalent antibody (scDb-Fc) simultaneously targeting EGFR and HER3 by integrating the antigen-binding sites of a humanized version of cetuximab (IgG hu225) and IgG 3-43 (20, 22). We show that the scDb-Fc preserves the properties of the parental antibodies and potently blocks proliferation of the HNSCC cell line FaDu in vitro and in vivo. We further demonstrate in TNBC cell lines that scDb-Fc decreased the cancer stem cell (CSC) population in vitro and in vivo and delayed tumor regrowth of orthotopically implanted TNBC cells in vivo. Compared with the parental antibodies alone or the combination of both, the new bispecific scDb-Fc antibody has superior inhibitory activity warranting its further preclinical development.

Horseradish peroxidase (HRP)- and phycoerythrin (PE)-conjugated anti-human Fc antibodies were purchased form Sigma-Aldrich (A0170; P9170). Anti-His-HRP (HIS-6 His-Probe-HRP, sc-8036) was purchased from Santa Cruz Biotechnology. Antibodies for immunoblotting were purchased from Cell Signaling Technology [phospho-EGF receptor (Tyr1068) (D7A5) XP rabbit mAb #3777; phospho-HER2/ErbB2 (Tyr1221/1222) (6B12) rabbit mAb #2243; phospho-HER3/ErbB3 (Tyr1289) (21D3) rabbit mAb #4791; Akt (pan) (40D4) mouse mAb #2920; phospho-Akt (Ser473) (D9E) XP rabbit mAb #4060; p44/42 MAPK (Erk1/2) (3A7) Mouse mAb #9107; phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204) antibody #9101; Cell Signaling Technology], from Santa Cruz Biotechnology [EGFR (sc-03) rabbit mAb #sc-03, Santa Cruz Biotechnology], from Thermo Fisher Scientific [HER-2/c-erbB-2/neu Ab-17, mouse mAb # MS-730-P-A; ErbB3, clone: 2F12, Invitrogen mouse mAb #MA5-12675), from Sigma-Aldrich [α-tubulin mouse mAb #T6793], and Dianova [goat IgG anti-mouse IgG (H+L)-HRPO, MinX Hu,Bo,Ho #115-035-062].

MDA-MB-468 (CLS, CVCL_0419), HCC1806 (ATCC, CRL-2335), and MCF-7 (IKP Stuttgart) cells were cultured in RPMI1640 (Thermo Fisher Scientific, 11875093) and FaDu (ATCC HTB-43) cells were cultured in DMEM (Thermo Fisher Scientific, 41965039), supplemented with 10% FBS (PAA Laboratories) at 37°C in a humidified chamber with 5% CO₂. Cell lines were authenticated by SNP profiling [Multiplexion GmbH; FaDu (2017), MDA-MB-468 (2019), HCC1806 (2017) and MCF-7 (2016)] and regularly tested for Mycoplasma contamination (Lonza, LT07-318).

The variable domains of IgG hu225 and IgG 3-43 were combined into a scDb-Fc fusion protein (scDb hu225×3-43-Fc) and cloned into a variant of the pSecTagA vector (Invitrogen, Thermo Fisher Scientific, V90020; refs. 20, 23) HEK293-E cells were transiently transfected with the plasmids using polyethylenimine (PEI, Merck, 408727) and 24 hours after transfection, tryptone N1 (TN1, Organo Technie) was added to initiate the production. After cultivation at 37°C for 4 days, proteins were purified from the supernatant using Protein A (Protein A Sepharose 4 Fast Flow, Pharmacia Biotech, 17-0974-03) affinity chromatography. Protein preparations were dialyzed against PBS at 4°C overnight and stored at −20°C.

Integrity and purity of the antibodies was analyzed by SDS-PAGE and staining with Coomassie-Brilliant Blue G-250 (Thermo Fisher Scientific, 20279) and by size-exclusion chromatography (SEC) performed by high-performance liquid chromatography (Waters 2695 HPLC) in combination with a TSKgel SuperSW mAb HR column (Sigma Aldrich, 822854) at a flow rate of 0.5 mL/minute. Na₂HPO₄/NaH₂PO₄ (0.1 mol/L), Na₂SO₄ (0.1 mol/L), pH 6.7 was used as mobile phase. The following standard proteins were used: thyroglobulin (669 kDa, R_S 8.5 nm), apoferritin (443 kDa, R_S 6.1 nm), β-amylase (200 kDa, R_S 5.4 nm), BSA (66 kDa, R_S 3.55 nm), and carbonic anhydrase (29 kDa, R_S 2.35 nm).

Analysis of real-time binding kinetics were performed as previously described using the A-100 C-Fast or Cell-200 C-Fast biosensors (Attana; ref. 24). scDb hu225×3-43-Fc was immobilized to the Attana LNB Carboxyl Sensor Chip (3623–3103, Attana). Extracellular domains of EGFR or HER3 were used as binding analytes. For the analysis of IgG hu225, the extracellular domain III of EGFR was immobilized, and the antibody used as an analyte. Sensor chip regeneration for the scDb-Fc was accomplished by injecting once 20 mmol/L glycine (pH 3.5) and once 158 nmol/L NaOH for 12 seconds. For IgG hu225 analysis, the sensor was washed twice with 20 mmol/L glycine, pH 3.5. The analysis was performed at 37°C.

ELISA-based binding analysis was performed as described previously (25). For flow cytometry–based binding studies, 1×10⁵ FaDu, MCF-7 or MDA-MB-468 cells were seeded per well and incubated with serial dilutions of the proteins in PBA [PBS, 2% (v/v) FBS, 0.02% (w/v) sodium azide] at 4°C for 1 hour. Bound antibodies were detected using a PE-labeled anti-human Fc antibody (Merck, P9170). Flow cytometry was performed using a MACSQuant Analyzer 10 or MACSQuant VYB (both Miltenyi Biotec) and data were analyzed using FlowJo (Tree Star). Relative mean fluorescence intensities (rel. MFI) were calculated as follows: rel. MFI = [MFI_sample − (MFI_{detection system} − MFI_cells)]/MFI_cells.

RIPA lysis buffer (150 mmol/L NaCl, 1% Nonident P-40, 0.5% sodium deoxycholate, 0.1% SDS, 25 mmol/L Tris pH 7.4) supplemented with cOmplete protease inhibitors (Merck, 4693116001) and PhosSTOP (Merck, 4906845001) was used to lyse cells. Lysates were centrifuged and equal amounts of protein were loaded on NuPage Novex 4–12% Bis-Tris gels (Thermo Fisher Scientific, NP0336). The iBlot system was used for blotting (Thermo Fisher Scientific, IB301002). The Roche blocking reagent (Merck, 11096176001), 0.5% (v/v) in PBS containing 0.05% (v/v) Tween-20 (PBST), was used for membrane blocking (1 hour at room temperature). Membranes were incubated with primary antibodies diluted in blocking buffer at 4°C overnight. Membranes were washed three times for 5 minutes with PBST. HRP-conjugated secondary antibodies diluted in blocking buffer were incubated at room temperature for 1 hour. Membranes were washed three times for 5 minutes with PBST before incubation with SuperSignal chemiluminescent substrates (Thermo Fisher Scientific, 34075, 34578), followed by detection using a FUSION SOLO [Vilber Lourmat] device.

HCC1806 cells (2×10³ cells per well) were seeded into 96-well plates (Greiner Bio-One, 655083). The next day, cells were left untreated or treated with IgG hu225, IgG 3-43, and scDb hu225×3-43-Fc, respectively. After 4 days, the medium was discarded and 50 μL of prewarmed detection reagent [1:2, medium:CellTiter-Glo (Promega, G9242)] was added. The cells were incubated for 10 minutes at room temperature and luminescence was measured at 560 nm and an integration-time of 1 second, using the Spark microplate reader (Tecan).

For two-dimensional (2D) cell proliferation assays, FaDu and MDA-MB-468 cells (2×10³ cells per well) were seeded into 96-well plates. For the three-dimensional (3D) cell proliferation assays, 96-well plates (Greiner Bio-One, 655083) were first coated with a mixture of Matrigel and collagen (1:2; Corning, 354230; Advanced BioMatrix, 5015-20ML), followed by seeding of the cells in medium containing 2% Matrigel. The next day, the medium was exchanged for starvation medium containing only 0.2% FBS. After 24 hours, cells were pretreated with the different antibodies (50 nmol/L) for 60 minutes prior to ligand stimulation with 30 ng/mL recombinant human heregulin β-1 (HRG; PeproTech, 100-03). After 4 days (3D) or 5 days (2D), the medium was discarded and 50 μL of prewarmed detection reagent [1:2, medium:CellTiter-Glo (2D assay: Promega, G9242; 3D assay: Promega, G9682)] was added. The cell lysates were incubated for 10 minutes at room temperature and luminescence was measured at 560 nm and an integration time of 1 second, using the Spark microplate reader (Tecan).

FaDu cells (2.5×10⁴ per 96-well) were seeded into a 96-well plate (Essen BioScience, catalog no. 4379). The next day, medium was exchanged for starvation medium containing 0.2% FBS. After 24 hours, a scratch was introduced into the confluent cell monolayer using the 96-pin IncuCyte WoundMaker Tool (Essen BioScience, catalog no. 4563) followed by two washes with 100 μL medium. Next, antibodies (50 nmol/L of IgG hu225, IgG 3-43, and scDb hu225×3-43-Fc) and ligand [30 ng/mL HRG (PeproTech, 100-03)] diluted in 100 μL medium were added. Analysis of migration into the wound was performed using the IncuCyte S3 live-cell analysis system (Essen BioScience).

For primary oncosphere formation assays, MDA-MB-468 and HCC1806 cells were trypsinized, singularized using a 27G needle and seeded onto Poly(2-hydroxyethyl methacrylate)-(pHEMA)-treated (Merck, P3932) 12-well plates (3×10³ cells per well) in sphere formation medium [DMEM/F12 GlutaMAX (Thermo Fisher Scientific, 31331028); 20 ng/mL HRG (Peprotech, 100-03)]. HCC1806 cells required the addition of 1× B27 (Thermo Fisher Scientific, 17504044) to the sphere formation medium. The antibodies were immediately added. After 5 days, spheres were imaged and the sphere area was analyzed using ImageJ.

For extreme limiting dilution assays (ELDA), cells were seeded (2 × 10⁵ cells per 10-cm plate) and grown as primary oncospheres as described above. After 5 days, primary oncospheres were harvested and singularized. One, 10, or 100 cells were seeded onto pHEMA-treated 96-well plates, respectively. Cells were cultured for 10 days in 300 μL sphere formation medium, supplemented with 1× B27 (20 technical replicates per condition). Wells positive for spheres were counted and the relative estimated stem cell frequency was determined by ELDA software (http://bioinf.wehi.edu.au/software/elda/index.html), provided by the Walter and Eliza Hall Institute (26).

The ALDEFLUOR Assay was purchased from StemCell Technologies (01700). A total of 1×10⁵ singularized HCC1806 or MDA-MB-468 cells were diluted in 100 μL analysis reagent (100 μL Assay Buffer + 1.5 μL activated ALDEFLUOR). Subsequently, 50 μL cell suspension were transferred to a new 96-well and mixed with 1.5 μL DEAB Reagent (control). After incubation at 37°C for 30 minutes, samples and respective controls were analyzed using the MACSQuant Analyzer 10 (Miltenyi Biotec). FlowJo (Tree Star) was used for flow cytometry data evaluation. DEAB control served as an internal control and was used to define the ALDH-positive population.

For relative surface receptor expression, cells were trypsinized and singularized. A total of 1×10⁵ cells per sample were incubated with antibodies against EGFR or HER3. For quantitative surface receptor expression, the QIFIKIT was used according to manufacturer's instructions (Agilent Dako, K007811-8). Flow cytometry was performed using a MACSQuant Analyzer 10 (Miltenyi Biotec). FlowJo (Tree Star) was used for flow cytometry data evaluation. Relative mean fluorescence intensities (rel. MFI) were calculated as follows: rel. MFI = [MFI_sample − (MFI_{detection system} − MFI_cells)]/MFI_cells.

All animal studies were approved by state authorities and performed in accordance to federal guidelines. Proteins (25 μg diluted in 100 μL PBS) were injected into the tail vein of CD-1 mice (Charles River, three animals per molecule). After 3 minutes, 1, 6, 24, 72, and 169 hour blood samples were taken, and immediately incubated on ice. Serum samples were centrifuged (16,000×g, 4°C, 20 minutes), and stored at −20°C until analysis. Antibody serum concentrations were determined by ELISA using EGFR-His, and HER3-His as antigen, and HRP-conjugated anti-human Fc antibody for detection.

All animal studies were approved by state authorities (reference number 35-9185.81/0456 & 35-9185.81/0463) and performed in accordance to federal guidelines. Eight-week-old female SCID beige mice (Charles River, CB17.Cg-Prkdc^scidLyst^bg-J/Crl) were anesthetized with isoflurane during the injection of 5×10⁶ FaDu or MDA-MB-468 cells. FaDu cells (100 μL PBS) were subcutaneously injected in the left and right flanks, whereas MDA-MB-468 cells (100 μL PBS) were orthotopically injected into the right and left 4th mammary fat pads. Tumor growth was monitored with a caliper and tumor volume was calculated as follows: tumor volume = (a×b²)/2 (a, longitudinal diameter of tumor; b, transverse diameter of tumor). Mice were randomly assigned to control and 4 treatment groups after the tumors had reached 100 mm³ (n = 7 mice per group). Mice received six intravenous antibody injections [FaDu xenograft model: 300 μg antibody per injection in 100 μL PBS; MDA-MB-468: 300 μg antibody in 100 μL PBS for the first injection, the dose was then adjusted to 200 μg (IgG hu225) and 240 μg (IgG 3-43 or scDb-Fc), respectively] twice weekly for three consecutive weeks. MDA-MB-468 tumors were further dissociated for quantitative surface receptor expression analysis, oncosphere formation assays, and ALDEFLUOR analysis. Tumor dissociation was performed according to manufacturer's instructions (Miltenyi Biotec, 130-096-730).

All values are presented as mean ± SD. Significance between multiple groups was determined by one-way ANOVA and Tukey test for multiple comparison. Significance between two groups was determined by t test. For statistical analyses of in vivo experiments, the 95% confidential interval (95% CI) was used. Total group effects were analyzed by two-way ANOVA and results at a specific time point were compared via one-way ANOVA followed by Tukey multiple comparison test (posttest). Data was analyzed, using GraphPad Prism 7. P values: P ≤ 0.05 (*), P ≤ 0.01 (**), P ≤ 0.001 (***), P ≤ 0.0001 (****).

A tetravalent bispecific antibody (scDb hu225×3-43-Fc; Fig. 1A) was generated by combining the antigen-binding site of the anti-HER3 antibody 3-43 with the antigen-binding site of humanized cetuximab (hu225) into a bispecific single-chain diabody (scDb hu225×3-43), which was further fused to the hinge region of a human Fcγ1 chain (Fig. 1A). The scDb moiety was arranged in the VH-VL orientation [VH(3-43)-linker1-VL(hu225)-linker2-VH(hu225)-linker3-VL(3-43); see Supplementary Fig. S1 for details]. We previously found that assembly of the antigen-binding sites within the scDb moieties might be affected by the connecting linkers (27). Hence, two scDb derivatives were generated to analyze the influence of the middle linker 2 on scDb assembly. Thus, in scDb-1 we used a linker 2 with a length of 15 residues, and in scDb-2 we used a length of 20 residues, with linker 1 and 3 comprising five residues (GGGGS). These scDb moieties were then fused via a GGSGG linker to the hinge region of a human γ1 Fc chain. This scDb-Fc format combines the variable domains of two antibodies as well as the Fc region within a single polypeptide chain, facilitating the production of a tetravalent bispecific antibody (28). The scDb and scDb-Fc molecules were produced in transiently transfected HEK293-6E suspension cells and purified from the cell culture supernatant by IMAC (scDb) or protein A affinity chromatography (scDb-Fc), respectively.

Both scDb molecules showed a single band in SDS-PAGE and eluted in SEC with a major peak with apparent molecular masses of 41 kDa (scDb-1) and 43.2 kDa (scDb-2), respectively (Supplementary Fig. S1). Obviously, scDb-1 possessing a shorter linker 2 assembled into a more compact molecule. Of note, for both scDbs also a small fraction with an apparent molecular mass of 81 kDa was observed in SEC, most likely corresponding to noncovalently linked dimers (27).

SDS-PAGE analysis of the two scDb-Fc molecules revealed a single band under nonreducing conditions (apparent molecular masses of 193 kDa for scDb-Fc-1 and 204 kDa for scDb-Fc-2, respectively) as well as under reducing conditions (apparent molecular masses of 85 kDa for scDb-Fc-1 and scDb-Fc-2, respectively; Supplementary Fig. S1). In SEC, a major peak corresponding to an apparent molecular mass of 152 kDa and a Stoke's radius (R_S) of 4.7 nm was observed for scDb-1-Fc, while scDb-2-Fc eluted with a major peak with an apparent molecular mass of 178 kDa (R_S of 5.1 nm). These findings confirm correct covalent linkage of the two scDb-Fc chains and assembly into homodimeric molecules. In comparison, purified IgG hu225 and IgG 3-43 showed a single peak corresponding to an apparent molecular mass of 163 kDa and an R_S of 4.8 nm (IgG hu225) and 194 kDa and 5.2 nm (IgG 3-43), respectively (Fig. 1C). In summary, both scDb configurations can be used to produce scDb-Fc fusion proteins with a size similar to IgG molecules. Production yields were similar for scDb-1-Fc and scDb-2-Fc (approx. 40–50 mg/L supernatant). All further studies were performed with scDb-1-Fc (Supplementary Fig. S2), which was further purified by an additional preparative size-exclusion chromatography (SEC) step to remove minor multimers and aggregates (Fig. 1C).

The purified scDb-Fc was capable of binding to the recombinant extracellular region of EGFR and HER3 as demonstrated by ELISA, with EC₅₀ values similar to those of the parental antibodies (Fig. 1D). In addition, the scDb-Fc was able to simultaneously bind to immobilized EGFR-Fc and soluble HER3-His in a sandwich ELISA (Fig. 1E). Further analysis of the real-time binding of scDb-Fc to EGFR or HER3 using a quartz crystal microbalance (QCM) revealed a K_D value of 21 nmol/L for EGFR and a K_D value of 19 nmol/L for HER3, respectively (Fig. 1G). For IgG hu225, a K_D value of 8 nmol/L for EGFR was determined (Supplementary Fig. S3). In a previous study, we determined a K_D value of 11 nmol/L for binding of IgG 3-43 to HER3 (20). In summary, the bispecific antibody retained the binding properties of the parental antibodies, thus, all antigen-binding sites are accessible for antigen binding and the scDb-Fc was able to simultaneously bind both antigens with affinities similar to those of the parental antibodies.

Flow cytometry analysis demonstrated strong binding of the scDb-Fc to the EGFR- and HER3-expressing HNSCC cell line FaDu (EC₅₀: 205 ± 96 pmol/L), and the breast cancer cell lines MCF-7 (76 ± 25 pmol/L) and MDA-MB-468 (596 ± 116 pmol/L). EC₅₀ values and maximum binding of scDb-Fc were similar to that of IgG hu225 on FaDu and MDA-MB-468 cells expressing high numbers of EGFR and lower numbers of HER3 (approximately 145,000 EGFR and approximately 3,000 HER3 molecules for FaDu (20) and > 723,000 EGFR and approximately 6,000 HER3 molecules for MDA-MB-468), while on MCF-7 cells, expressing low numbers of EGFR and higher numbers of HER3 (approximately 1,000 EGFR and 17,000 HER3 molecules; ref. 20), scDb-Fc showed binding similar to that of IgG 3-43. Thus, scDb-Fc combined the antigen-binding specificity of the parental antibodies, as shown by ELISA, and retained the strong target cell–binding properties of the parental antibodies.

Next, we analyzed the inhibitory activity of scDb-Fc with respect to receptor phosphorylation (Fig. 2B; Supplementary Figs. S4 and S5). Thus, serum-starved MCF-7 celIs were preincubated with the antibodies and then stimulated with EGF or HRG, or remained unstimulated. Immunoblot analysis of the cell lysates revealed that EGF potently induced EGFR phosphorylation (Y1068) while HRG induced HER3 (Y1289) and EGFR phosphorylation. In EGF-stimulated cells, the pEGFR signals were reduced by the preincubation of cells with IgG hu225 and scDb-Fc as well as the combination of the parental antibodies, while IgG 3-43 had no effects (Fig. 2B, quantified data: Supplementary Fig. S4). HRG-induced EGFR and HER3 phosphorylation were almost completely suppressed by IgG 3-43, the combination of the parental antibodies, and the scDb-Fc, while IgG hu225 did not affect the phosphorylation of EGFR and HER3 (Fig. 2B, quantified data: Supplementary Fig. S4). Furthermore, in unstimulated and EGF-stimulated cells, scDb-Fc, IgG 3-43 or the combination of IgG hu225 and IgG 3-43 reduced the levels of HER3 (Fig. 2B), indicative of antibody-mediated HER3 internalization and degradation. Degradation of HER3 was also triggered by HRG stimulation, and this was not further enhanced by incubation with the antibodies. Inhibition of EGF- or HRG-induced receptor phosphorylation was also observed for FaDu cells (Supplementary Fig. S5). Of note, titration of the antibodies on FaDu cells revealed approximately 18-fold higher inhibition of HRG-induced HER3 phosphorylation by scDb-Fc in comparison with IgG 3-43 and 10-fold higher inhibition when compared with the combination of IgG 3-43 and cetuximab, while a reduced IC₅₀ value for scDb-Fc compared with cetuximab or the combination of cetuximab and IgG 3-43 was observed for EGF-induced EGFR phosphorylation. No differences in inhibition of HRG-induced HER3 phosphorylation between scDb-Fc and the parental antibodies were observed on MCF-7 cells, which express low EGFR and moderate HER3 levels (20), indicating that the HER3-binding sites in the bispecific antibody benefit from EGFR cotargeting especially on cells exhibiting high EGFR and low HER3 levels. Taken together, scDb-Fc efficiently blocks ligand-induced EGFR and HER3 phosphorylation and triggers HER3 degradation, thus maintaining the inhibitory properties of the parental antibodies.

The proliferation of FaDu cells is known to be driven by an autocrine signaling loop involving HRG (29). We therefore measured the growth inhibition of FaDu cells by scDb-Fc in the absence of exogenous HER ligands. Whereas single EGFR and HER3 inhibition suppressed cell proliferation by approximately 30% (IgG hu225) and 20% (IgG 3-43), respectively, the combination of IgG hu225 and IgG 3-43 or scDb-Fc reduced proliferation by approximately 50% (Fig. 3A). Titration of the antibodies revealed a superior activity of scDb-Fc when compared with the combination of the parental antibodies under unstimulated (IC₅₀: 10 pmol/L vs. 71 pmol/L) and HRG-stimulated (IC₅₀: 48 pmol/L vs. 462 pmol/L) conditions (Fig. 3B).

Western blot analysis of FaDu cells incubated with the antibodies at a concentration of 50 nmol/L revealed HER3 downregulation by IgG 3-43, the combination of the parental antibodies and scDb-Fc, respectively, which was associated with reduced pAkt(T308) signals (Fig. 3C; Supplementary Fig. S6). Of note, strong inhibition of Akt phosphorylation was observed when EGFR and HER3 were blocked simultaneously, either by combining the parental antibodies or by scDb-Fc. Furthermore, Erk phosphorylation [pErk1/2(T202/Y204)] was reduced by IgG hu225, while no effects were observed for IgG 3-43. Although scDb-Fc–treated FaDu cells displayed an increased pEGFR signal compared with the control and the single antibody treatments, this did not translate into activation of Erk or Akt (Fig. 3C; Supplementary Fig. S6). We further analyzed how simultaneous EGFR and HER3 blockade affect the motility of FaDu cells in a scratch wound assay. Despite producing HRG in an autocrine manner, motility of FaDu cells was stimulated by addition of exogenous HRG [relative wound density (%) at 24 hours: without HRG: 34.34 ± 3.47, with HRG: 47.28± 1.51]. Compared with the single treatments, both, the scDb-Fc and the combination of IgG hu225 and IgG 3-43 significantly inhibited cell motility in the absence or presence of HRG (Fig. 3D).

Inhibition of cell proliferation was also observed when FaDu cells were embedded into a 3D matrix, more closely recapitulating the conditions found in vivo (30). Under these conditions, IgG hu225 strongly inhibited proliferation by about 60%, which was similar to the effects of scDb-Fc (∼68%). IgG 3-43 and the combination of the parental antibodies also led to an inhibition of proliferation, however, to a lesser extent (∼34 and ∼50%; Fig. 3E).

Next, we analyzed the antitumor activity of scDb-Fc and its parental antibodies in a subcutaneous FaDu xenograft model, injecting 300 μg protein twice weekly for three weeks. The scDb-Fc, IgG hu225, and IgG hu225 + IgG 3-43 treatments induced almost complete tumor remission that was stable until the end of treatment, whereas IgG 3-43 caused only partial tumor regression, followed by regrowth after termination of treatment (Fig. 3F). When comparing serum concentrations after the first and last injection, an accumulation of the different antibodies was observed in all cases (approximately 3-fold). We therefore analyzed the pharmacokinetic properties of scDb-Fc after a single intravenous (i.v.) injection into CD-1 mice. ScDb-Fc revealed a profile comparable to IgG 3-43 (Fig. 3G). The terminal half-life was approximately 65 hours using EGFR-His for detection of functional antibodies in serum and 49 hours using HER3-His, respectively. Half-life and drug exposure (AUC) of scDb-Fc were similar to that of IgG 3-43 (Fig. 3G; Supplementary Table S1). In summary, the scDb-Fc exhibited IgG-like pharmacokinetic properties and combined the antitumor activities of the parental antibodies.

TNBC cells frequently overexpress EGFR, whereas HER3 expression is typically rather low. However, HRG can drive TNBC cell proliferation and HER3 is known to be upregulated as a survival signal in TNBC cells (12, 13). Hence, we investigated whether dual EGFR and HER3 inhibition suppresses proliferation of the TNBC cell lines MDA-MB-468 and HCC1806.

In the absence of external ligands and in low serum, IgG hu225 was sufficient to block the basal proliferation of MDA-MB-468 cells. In the presence of HRG, however, the inhibitory activity of IgG hu225 was impaired and scDb-Fc treatment inhibited proliferation more potently than the treatment with IgG hu225 or IgG 3-43 alone (71.1% vs. 94.3% vs. 108.4%; w/o antibody control: 132.0%; Fig. 4A). The proliferation of HCC1806 cells was also markedly inhibited by scDb-Fc treatment while treatment with IgG hu225 or IgG 3-43 showed no or weaker effects (IgG hu225: 94.1 ± 7.2% vs. IgG 3-43: 101.2 ± 4.0% vs. scDb-Fc: 83.3% ± 5.4%; Supplementary Fig. S7A).

Sphere formation in serum-free suspension culture is characteristic of cells with stem cell–like and tumor-initiating properties (31). HRG was reported to drive oncosphere formation of breast cancer cells (32–34). We therefore investigated whether scDb-Fc inhibited cancer stem cell survival and expansion. In serum-free medium supplemented with HRG, both IgG hu225 and scDb-Fc significantly reduced the sphere formation efficiency (SFE) of MDA-MB-468 cells, with scDb-Fc yielding the strongest sphere formation suppression (2.9 ± 0.5 vs. 5.7 ± 1.5 and 11.5 ± 0.8 for IgG hu225 and IgG 3-43, respectively; without antibody control: 16.8 ± 6; Fig. 4B). In addition, all three treatments, scDb-Fc, IgG hu225, and IgG 3-43, respectively, modestly reduced the size of the spheres (0.0034 mm² vs. 0.0042 mm² vs. 0.0043 mm²; without antibody control: 0.0049 mm²; Fig. 4C). To confirm that the number of cells with stem cell–like properties was reduced by scDb-Fc, we measured the activity of ALDH, an enzyme expressed in stem cells (Fig. 4D; Supplementary Fig. S7D). Indeed, treatment of MDA-MB-468 cells with scDb-Fc for 5 days reduced the ALDH^high stem cell population by almost 50% as measured by ALDEFLUOR assay (Fig. 4D). Similar results were obtained for HCC1806 cells, in which scDb-Fc also reduced the SFE in oncosphere assays and the ALDH^high population (Supplementary Fig. S7B–S7D). We further compared the expression of EGFR and HER3 between MDA-MB-468 cells grown in monolayer and oncosphere culture. Both EGFR and HER3 were significantly upregulated in oncospheres (approximately 2-fold), explaining the increased activity of scDb-Fc in this assay (Fig. 4E). This prompted us to determine the stem cell frequency after scDb-Fc treatment in an extreme limiting dilution assay (ELDA). To do so, primary oncosphere cultures were first treated with scDb-Fc. After 5 days, the oncospheres were collected, reseeded into the ELDA (1, 10, or 100 cells per well) and cultivated for 10 additional days (Fig. 4F). Remarkably, in both MDA-MB-468 and HCC1806 cell lines, pretreatment with scDb-Fc significantly reduced the outgrowth of secondary oncospheres and the estimated stem cell frequency was reduced by approximately 74% and 54%, respectively (Fig. 4G, and H; Supplementary Fig. S7E and S7F).

Finally, we analyzed the in vivo antitumor activity of scDb-Fc in comparison with its parental antibodies in an orthotopic xenograft model using MDA-MB-468 cells. Compared with the PBS control, treatment with IgG hu225, IgG 3-43 and the combination of both antibodies delayed the growth of the tumors; however, tumor growth was halted only by the treatment with scDb-Fc, where growth inhibition persisted until animals were sacrificed (Fig. 5A and B). The tumors were then collected and dissociated, and the tumor cells analyzed in ALDH and oncosphere formation assays. Both, the scDb-Fc and IgG 3-43–treated tumors contained lower numbers of ALDH^high cells, compared with PBS-treated tumors (Fig. 5C). Furthermore, although all treatment groups showed a reduced SFE compared with the PBS control group, cells derived from scDb-Fc-treated tumors had the lowest SFE (SFE of PBS: 13.5 ± 10.7, IgG hu225: 6.63 ± 2.3, IgG 3-43: 7.5 ± 3.7, IgG hu225 + IgG 3-43: 6.1 ± 1.5, scDb hu225×3-43-Fc: 3.5 ± 0.6), of note, sphere size was not affected by the treatment with the antibodies, indicating that the number of cells with stem cell–like properties was reduced by the antibody treatments in vivo, agreeing with the results obtained in vitro (Fig. 5D and E).

Here we report the development of a novel tetravalent bispecific antibody that dually targets EGFR and HER3. We provide evidence that the scDb-Fc–based antibody fully preserves the properties of the IgG hu225 and IgG 3-43 parental antibodies with respect to antigen binding as well as inhibition of ligand-induced HER receptor phosphorylation, downstream signaling and receptor internalization and degradation. In FaDu cells, which are driven by autocrine HRG, the antiproliferative activity of scDb-Fc was greater than that of the combination of the parental antibodies. Furthermore, compared with the single parental antibodies the scDb-Fc and the combination of both parental antibodies significantly reduced cell migration, which is a prerequisite for dissemination and the formation of metastasis. However, when grown in 3D culture, scDb-Fc and the combination of the parental antibodies were similarly effective as IgG hu225, indicating dominant EGFR signaling under these conditions. This translated into comparable antitumor activities in a FaDu xenograft mouse model in which strong and long-lasting growth inhibition was observed for the bispecific antibody, the combination of both antibodies, but also for IgG hu225 alone.

We reasoned that the bispecific antibody would perform especially well in a setting that depends on survival signaling mediated by HER3. Interestingly, both EGFR and HER3 were upregulated when MDA-MB-468 cells were grown in serum-free suspension culture supplemented with HRG. Thus, under these conditions cells are expected to be more prone to inhibition by dual targeting of EGFR and HER3. Indeed, the bispecific antibody efficiently reduced SFE of MDA-MB-468 and HCC1806 cell lines, which was accompanied by a reduced number of cells with stem cell–like properties as determined by ALDH measurements and a strongly reduced stem cell frequency, as determined by ELDA. These in vitro findings further translated into clear antitumor effects in an orthotopic MDA-MB-468 tumor model. Analysis of the tumors at the end of treatment revealed lowest SFE for those tumors treated with the bispecific antibody. Thus, the in vitro results regarding SFE were also observed in vivo and might contribute to increased antitumor activity of the bispecific antibody.

Several anti-HER3 antibodies have been tested in clinical trials either as single agents or in combination with EGFR antibodies, small-molecule inhibitors of signaling pathways and/or chemotherapy in different tumor entities, including HNSCC, CRC, NSCLC, and TNBC (18, 19). Surprisingly, results from preclinical studies could not be confirmed in these clinical trials and most studies did not show significant clinical benefits. This has been attributed to the complexity of receptor and ligand expression and downstream signaling (19). As yet, only one bispecific antibody, duligotuzumab, for dual targeting of EGFR and HER3 has been studied in clinical trials. A first phase II study in patients with recurrent/metastatic HNSCC during or following chemotherapy versus cetuximab did not improve patient outcomes in comparison with cetuximab (35). This result was confirmed in another phase II trial where duligotuzumab did not appear to improve the outcomes in patients with wild-type RAS metastatic colorectal cancer compared with single-agent cetuximab (36). One should note that duligotuzumab is a two-in-one bispecific antibody, thus, although possessing four potential binding sites, it only can bind in a bivalent one plus one mode, that is, either mono- or bispecifically. There is strong evidence that parameters such as epitope specificity, affinity, geometry, and valency influence neutralizing and therapeutic activity of monospecific and bispecific antibodies (37, 38). This is confirmed by data obtained for a tetravalent, bispecific DVD-Ig directed against EGFR and HER3 (39). In vitro, the DVD-Ig induced stronger antiproliferative effects on cell lines expressing high levels of EGFR and HER3, including FaDu, than a bivalent, bispecific IgG included in the study (39). This was attributed to avidity effects of the tetravalent bispecific antibody, which also applies to our tetravalent, bispecific scDb-Fc fusion protein, for example, as shown by the increased inhibition of proliferation of FaDu cells compared with the combination of both parental antibodies.

Although clinical results for duligotuzumab in comparison with cetuximab treatment did not establish superior effects of the bispecific antibody, it was argued that dual blockade of EGFR and HER3 might be beneficial upon careful patient selection, that is, those who have developed EGFR resistance upon treatment with EGFR blocking antibodies (40). Of note, a recent "window-of-opportunity" study of anti-HER3 antibody CDX-3379 (KTN3379) in patients with HNSCC demonstrated inhibition of HER3 phosphorylation as well as measurable tumor regression in a substantial number of patients (41). This led to the initiation of a phase II study to evaluate CDX-3379 in combination with cetuximab in advanced HNSCC (NCT03254927) supporting further testing of dual EGFR×HER3 targeting strategies.

Dual targeting of EGFR and HER3 with a bispecific antibody might also affect safety by combining the side effects stemming from the single treatments. Of note, patients treated with duligotuzumab alone or in combination with chemotherapy suffered less from skin rash compared with cetuximab or cetuximab in combination with chemotherapy. However, in the same clinical trials, duligotuzumab led to more gastrointestinal toxicity, infections, and diarrhea (35, 36). The combination of cetuximab with the HER3-targeting antibody seribantumab revealed toxicities comparable with cetuximab alone, indicating that the grades of toxicity depend on the specific combination of binding moieties (42). Without doubt, further studies are required to analyze the safety profile of our dual-targeting tetravalent scDb-Fc molecule.

Only a few anti-HER3 antibodies, for example, seribantumab have been clinically studied in TNBC (43) and none of these antibodies have been combined with EGFR targeting. However, an enhanced response to PI3K pathway inhibitors in TNBC was described for duligotuzumab in preclinical models (13). EGFR inhibition causes resistance formation via HER3 upregulation in TNBC, and HER3 upregulation is highly associated with worse overall survival (44). Furthermore, HRG was shown to enhance the proliferation and self-renewal of breast tumor–initiating cells (32, 34). Such TNBC stem cells are accountable for decreased patient survival rates and are associated with therapy resistance, tumor recurrence, and metastasis formation (45). Thus, strategies targeting not only actively dividing cells but also cancer stem cells, as achieved by our bispecific antibody, are expected to be more effective than conventional antitumor therapies.

In summary, our new tetravalent scDb-Fc–based antibody with dual EGFR and HER3 specificity demonstrated antitumor activity in two different tumor models and potently blocked TNBC stem cell survival and expansion in vitro and in vivo, warranting further studies of this bispecific antibody in TNBC but also other tumor entities.

M.A. Olayioye is a consultant at SunRock Biopharma and has ownership interest (including patents) in HER3 antibody. R.E. Kontermann is a consultant at Sunrock Biopharma, reports receiving a commercial research grant from Sunrock Biopharma, and has ownership interest (including patents) in HER3 antibody. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Rau, W.S. Lieb, M.A. Olayioye, R.E. Kontermann

Development of methodology: A. Rau, W.S. Lieb, J. Honer

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Rau, W.S. Lieb, D. Birnstock, F. Richter

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Rau, W.S. Lieb, O. Seifert, D. Birnstock, F. Richter, R.E. Kontermann

Writing, review, and/or revision of the manuscript: A. Rau, W.S. Lieb, O. Seifert, J. Honer, M.A. Olayioye, R.E. Kontermann

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Rau, W.S. Lieb, O. Seifert, N. Aschmoneit, M.A. Olayioye

Study supervision: M.A. Olayioye, R.E. Kontermann

This project was supported by the German Cancer Aid (DKH 70112564) and through a research grant from SunRock Biopharma (Spain). We would like to thank Nadine Heidel and Doris Göttsch for excellent technical assistance.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.