Triple Targeting of HER Receptors Overcomes Heregulin-mediated Resistance to EGFR Blockade in Colorectal Cancer

Colorectal cancer is the third most common cancer worldwide. Survival rates are especially low in patients with advanced colorectal cancer. In addition to chemotherapy, monoclonal antibodies (mAbs) targeting the EGFR have been approved for the treatment of metastatic KRAS wild-type (WT) colorectal cancer, either in combination with chemotherapy or as a maintenance therapy in chemorefractory tumors (1). The mAbs cetuximab and panitumumab compete with ligand binding to the EGFR, thereby preventing receptor dimerization, phosphorylation, and the activation of downstream signaling pathways. In patients with colorectal cancer with mutational activation of the small GTPase KRAS, EGFR-targeting mAbs were found to be ineffective. The presence of KRAS mutations is thus an exclusion criteria for anti-EGFR therapy (2, 3). However, even in patients with KRAS WT disease, the clinical benefit of cetuximab and panitumumab is limited because of frequent intrinsic resistance of the tumors or the development of acquired resistance in response to targeted treatment (4). To achieve lasting clinical response, it is essential to elucidate the molecular mechanisms underlying drug resistance to support the development of more effective tailored therapies.

The EGFR belongs to the ErbB/HER family of receptor tyrosine kinases (RTKs) further comprising HER2/ErbB2, HER3/ErbB3, and HER4/ErbB4. The receptors are activated by a variety of different peptide ligands with EGF, TGFα and amphiregulin binding to EGFR, betacellulin and epiregulin binding to EGFR and HER4, whereas the heregulins (HRGs; also known as neuregulins) bind HER3 and HER4 (5, 6). In addition to homodimers, ligand binding promotes the formation of receptor heterodimers with HER2 being the preferred dimerization partner owing to its constitutively open conformation (7). Whereas in patients with colorectal cancer amphiregulin and epiregulin are predictive markers for the response to anti-EGFR therapy (8, 9), the autocrine and paracrine production of HRG has been associated with resistance to EGFR-blocking therapies (10). Similarly, amplification or transcriptional upregulation of HER2 has also been linked to anti-EGFR therapy resistance (11, 12). In breast and gastric cancers, where HER2 amplification is particularly prevalent, HER3 was shown to be required for HER2-driven transformation and HER3 upregulation has been associated with resistance to HER2-blocking therapies (13–15). High HER3 expression was also observed in primary and metastatic colorectal cancer tissues (16–18). The cooperative signaling of the different HER receptors and their broad contribution to tumor progression and resistance development has thus prompted the development of horizontal targeting approaches involving more than one HER receptor.

Because the tyrosine kinase activity of HER3 is impaired, conventional targeting approaches involving ATP-competitive small-molecule inhibitors are not effective. Functional blockade is rather achieved by antibodies directed toward the extracellular domain of the receptor, preventing ligand binding or dimerization, or locking the receptor in an inactive state (19). We recently developed a HER3-targeting antibody, IgG 3-43, which potently inhibited the proliferation of gastric, lung, breast, and head and neck cancer cell lines in vitro and tumor growth of xenografted head and neck cancer cells in vivo (20). IgG 3-43 competed with the binding of HRG to HER3-expressing cells, 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 three-dimensional (3D) cultures (18). By integrating the antigen-binding sites of IgG 3-43 and IgG hu225, a humanized version of cetuximab, we generated scDb hu225×3-43-Fc, a bispecific and tetravalent antibody targeting EGFR and HER3 (21). This bispecific antibody fully retained the receptor binding and neutralizing properties of the parental antibodies and gave rise to long-lasting growth suppression in a subcutaneous head and neck xenograft tumor model (21).

The limitations of current anti-EGFR therapies motivated our analysis of scDb-Fc as a treatment option for colorectal cancer. Here we show that the presence of HRG abrogates the inhibitory activity of EGFR blockade in two cetuximab-sensitive colorectal cancer cell line models. A combinatorial HER-targeting approach involving scDb-Fc and the HER2-blocking antibody trastuzumab prevented HRG-induced HER receptor phosphorylation, downstream signaling, and proliferation. Importantly, this was recapitulated in primary patient-derived organoid (PDO) cultures, which are known to faithfully predict clinical responses to targeted treatments. Our results thus provide strong support for a pan-HER receptor blocking approach to combat anti-EGFR therapy resistance of KRAS WT colorectal cancer tumors mediated by the upregulation of HRG and/or HER2/HER3 signaling.

For flow cytometry, PE-conjugated anti-human Fc antibody was purchased from Dianova [goat IgG anti-Human IgG (Fc)-RPE, 109-115-098 (RRID:AB_2337675)], FITC-conjugated anti-mouse immunoglobulins antibody was purchased from Agilent [polyclonal goat anti-mouse immunoglobulins/FITC, goat F(ab')2, F047902-2, Agilent], and anti-human EGFR antibody (AY13), anti-human HER2 antibody (24D2), anti-human HER3 (1B4C3) were purchased from BioLegend [purified anti-human EGFR Antibody, No. 352902 (RRID:AB_10916396); purified anti-human CD340 (erbB2/HER-2) Antibody, No. 324402 (RRID:AB_756118); purified anti-human erbB3/HER-3 Antibody, No. 324702 (RRID:AB_756156), BioLegend, Fell, Germany]. Antibodies for immunoblotting were purchased from Cell Signaling Technology [phospho-EGF Receptor (Tyr1068) (D7A5) XP Rabbit mAb No. 3777 (RRID:AB_2096270); phospho-HER2/ErbB2 (Tyr1221/1222) (6B12) rabbit mAb No. 2243 (RRID:AB_490899); phospho-HER3/ErbB3 (Tyr1289) (21D3) rabbit mAb No. 4791 (RRID:AB_2099709); Akt (pan) (40D4) mouse mAb No. 2920 (RRID:AB_1147620); phospho-Akt (Thr308) (D25E6) XP Rabbit mAb No. 13038 (RRID:AB_2629447); p44/42 MAPK (Erk1/2) (3A7) mouse mAb No. 9107 (RRID:AB_10695739); phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204) antibody No. 9101 (RRID:AB_331646); anti-rabbit IgG, horseradish peroxidase–linked antibody No. 7074 (RRID:AB_2099233; Cell Signaling Technology Europe B.V], from Santa Cruz Biotechnology [EGFR (sc-03) rabbit mAb No. sc-03 (RRID:AB_631420), Santa Cruz Biotechnology], from Thermo Fisher Scientific [HER-2/c-erbB-2/neu Ab-17, mouse mAb No. MS-730-P-A (RRID:AB_141769); ErbB3, clone: 2F12, Invitrogen Mouse mAb No. MA5-12675 (RRID:AB_11001597)], from Sigma-Aldrich [α-Tubulin mouse mAb No. T6793 (RRID:AB_477585)] and Dianova [goat IgG anti-mouse IgG (H+L)-HRPO, MinX Hu, Bo,Ho No. 115-035-062 (RRID:AB_2338504)]. For mass cytometry, the following anti-human antibodies were purchased preconjugated from Fluidigm: Anti-pAkt (S473) (D9E)-152Sm – No. 3152005A (RRID:AB_2811246), Anti-pS6 (S235/S236) (N7-548)-175Lu – No. 3175009A (RRID:AB_2811251), Anti-pERK1/2 (T202/Y204) (D13.14.4E)-171Yb – No. 3171010A (RRID:AB_2811250), Anti-Human Ki-67 (B56)-168Er – No. 3168007B (RRID:AB_2800467), Anti-Cleaved Caspase 3 (D3E9)-142Nd – No. 3142004A (RRID:AB_2847863), Anti-Human cleaved PARP (F21-852)-143Nd – No. 3143011A, Anti-Human CD44 (BJ18)-166Er – No. 3166001B (RRID:AB_2744692), Anti-Human CD24 (ML5)-169Tm – No. 3169004B (RRID:AB_2688021). Purified anti-Rb (S807/S811) was purchased from BD [No. 558389 (RRID:AB_647295)] and conjugated to 163Dy using the Maxpar X8 Antibody Labeling Kit from Fluidigm (No. 201163A).

DiFi and LIM1215 cells were cultured in RPMI1640 (Thermo Fisher Scientific) supplemented with 10% FBS (PAA Laboratories) at 37°C in a humidified chamber with 5% CO₂. LIM1215 cells were purchased from ECACC in 2018 (RRID:CVCL_2574) and reauthenticated by SNP profiling (Multiplexion GmbH) and tested negative for Mycoplasma contamination (Lonza, LT07-318) in 2020. DiFi cells were obtained from Dr. Julia Schueler (Oncotest) in 2012 (RRID:CVCL_6895), and reauthenticated by short tandem repeat profiling (Multiplexion GmbH) and tested negative for Mycoplasma contamination in 2020. After thawing, cells were kept in culture for a maximum period of 12 weeks.

PDO cultures (PDO1: IKP261-016; PDO2: IKP261-010; PDO3: IKP261-004; PDO4: IKP261-011) were generated and maintained as in refs. 22, 23. In short, human colorectal cancer tissue was minced and partially digested with 1 mg/mL collagenase type I (Sigma-Aldrich) and 100 µg/mL DNAse I (AppliChem), and cell clusters were embedded in matrigel (Matrigel, Corning). Matrigel was overlayed with Advanced DMEM/F12 medium supplemented with 10 mmol/L HEPES, 1.25 µg/mL amphotericin (Merck), penicillin/streptomycin, GlutaMAX, 1× N2 supplement, 1× B27 supplement (Thermo Fisher Scientific), 1 mmol/L N-acetylcysteine (Sigma-Aldrich), 50 ng/mL EGF, 20 ng/mL bFGF (Peprotech), and 10 µmol/L Y27632 (Absource Diagnostics). For maintenance culture, the ROCK inhibitor Y27632, which reduces initial cell death of epithelial cells in suspension (22), was omitted and EGF and bFGF were exchanged for recombinant human heregulin β-1 (HRG; 50 ng/mL, Peprotech) for at least three passages to model continuous presence of HRG as a drug resistance mechanism. For panel sequencing for frequent oncogenic driver mutations, organoid DNA was isolated and sequenced using the colorectal cancer panel (24) and an IonTorrent sequencer (Thermo Fisher Scientific) according to the manufacturer's instructions. Experiments were approved by the Ethics Committee of the Medical Faculty of the University of Tübingen (Tübingen, Germany).

IgG hu225, IgG 3-43, and scDb hu225×3-43-Fc were produced and purified as described previously (20, 21). Cells were incubated with serial dilutions of the antibodies diluted in PBA [PBS, 2% volume for volume (v/v) FBS, 0.02% sodium azide] at 4°C for 1 hour. Antibodies were detected by flow cytometry (MACSQuant Analyzer 10, Miltenyi Biotec) using PE-labeled anti-human FC antibody (Dianova). Relative mean fluorescence intensities (relative mean fluorescence intensity) were calculated as follows: rel. MFI = [MFI_sample − (MFI_{detection system} − MFI_cells)]/MFI_cells.

Viability was determined by CellTiter-Glo assay as described previously (21). For two-dimensional (2D) assays, DiFi and LIM1215 cells were seeded into 96-well plates (2 × 10³ cells/well). After 24 hours, the medium was exchanged for starvation medium containing 0.2% FBS. Another 24 hours later, cells were pretreated with the different antibodies (50 nmol/L) for 1 hour prior to addition of 30 ng/mL HRG. After 5 days, the medium was aspirated and 50 µL of prewarmed detection reagent (1:2, RPMI1640:CellTiter-Glo, Promega) was added. Cell lysates were incubated for 12 minutes at room temperature, and luminescence was measured using the Spark microplate reader (Tecan). For assays in 3D, 96-well plates (Greiner Bio-One) were precoated with a matrigel:collagen mixture (1:2; PureCol-S Advanced BioMatrix), DiFi and LIMI1215 cells were seeded and then overlaid with medium containing 2% matrigel. In the case of PDO cultures, singularized cell clusters were seeded onto the matrigel:collagen coated plates (1–4 × 10³ cell clusters/well) and then overlaid with PDO medium containing 2% matrigel. 3D cultures were incubated with antibodies and stimulated with HRG as described above. After 4 days (DiFi and LIM1215 cells) or 3 days (PDO cultures), viability was analyzed using CellTiter-Glo 3D reagent (Promega).

Cell-cycle analysis was performed using Click-iT EdU Flow cytometry assay kit Alexa Fluor 647 (Invitrogen) in combination with FxCycle Violet stain (Invitrogen) according to the manufacturer's instructions. In brief, 75,000 cells were seeded in 24-well plates and on the next day, starved overnight in medium containing 0.2% FBS. Cells were then pretreated with the different antibodies (50 nmol/L) for 1 hour before adding 30 ng/mL HRG. Twenty-four hours later, cells were labeled with 10 µmol/L of EdU for 2 hours prior to harvesting by trypsinization and staining. Flow cytometry was performed using a MACSQuant Analyzer 10 (Miltenyi Biotec).

Analysis of apoptosis was performed using Annexin V-GFP produced in-house and propidium iodide (PI; Sigma). A total of 10,000 cells were seeded in 96-well plates and on the next day, cells were starved overnight in medium containing 0.2% FCS. Cells were then pre-treated with the different antibodies (50 nmol/L) for 1 hour prior to addition of 30 ng/mL HRG. After 5 days, the cell culture supernatant was collected and the adherent cells harvested by trypsinization followed by staining of cells with Annexin V-GFP and 50 µg/mL PI in binding buffer (10 mmol/L HEPES, 140 mmol/L NaCl, 2.5 mmol/L CaCl₂, pH 7.4) for 10 minutes. Flow cytometry was performed using a MACSQuant VYB (Miltenyi Biotec).

DiFi and LIM1215 cells were seeded onto poly(2-hydroxyethyl methacrylate; pHEMA)-treated (Merck) 12-well plates (3 × 10³ cells/well) in sphere formation medium [DMEM/F12 GlutaMAX (Thermo Fisher Scientific); 20 ng/mL HRG, 1× B27 (Thermo Fisher Scientific)] and immediately treated with antibodies. After 5 days, spheres were imaged and counted, and the sphere area was analyzed using ImageJ software.

Cells were lysed using 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, cOmplete protease inhibitors (Merck) and PhosSTOP (Merck)]. Lysates were loaded on NuPage Novex 4% to 12% Bis-Tris gels (Thermo Fisher Scientific) and transferred to membranes using the iBlot system (Thermo Fisher Scientific). Membranes were blocked for 1 hour at room temperature with Roche blocking reagent (Merck), 0.5% (v/v) in PBS containing 0.05% (v/v) Tween-20 (PBST) and then incubated with primary or secondary antibodies diluted in blocking solution. Between incubations with primary and secondary antibodies, as well as before detection, membranes were washed with PBST. For detection, membranes were incubated with SuperSignal chemiluminescent substrates (Thermo Fisher Scientific), followed by imaging using a FUSION SOLO (Vilber Lourmat) device. Signals were quantified using ImageJ (RRID:SCR_003070). For relative protein levels, signals were normalized to the α-tubulin signals.

Preparation of CyTOF samples was performed (25) using their already established antibody panel. Data were acquired on a Helios CyTOF system and further processed using R in conjunction with various packages of the Tidyverse family (26). For deconvolution of barcoded samples, the CATALYST package (ref. 27; RRID:SCR_017127) was used. All data were normalized for unwanted variation, possibly related to the physical size and therefore total protein content of cells, using information from the Cell-ID palladium barcode, iridium, total-ERK, and pan-Akt channels, by regressing these channels’ levels out of the other channels. For this, multiple linear regression models were built for each channel separately for 2D and 3D samples. The intercept of the model depended on the sample and whether the individual cell belonged to the apoptotic subpopulation.

All values are presented as mean ± SD. Significance between multiple groups was determined by one-way ANOVA and Tukey posttest for multiple comparison. Significance between two groups was determined by t test. Data were analyzed, using GraphPad Prism 7 (RRID:SCR_002798). P values: ^*, P ≤ 0.05; ^**, P ≤ 0.01; ^***, P ≤ 0.001; ^****, P ≤ 0.0001; n.s., not significant.

Data were generated by the authors and are available on request.

The colorectal cancer cell lines DiFi and LIM1215 are established cetuximab-sensitive cellular models, with EGFR inhibition efficiently blocking growth both in vitro and in xenograft in vivo models (28). This is reflected by the potent suppression of the viability of DiFi (∼75%) and LIM1215 cells (∼50%) by the humanized cetuximab antibody IgG hu225 (29), as measured by CellTiterGlo assays performed in standard 2D and in 3D matrigel cultures (Fig. 1A–D). In the presence of exogenous HRG, however, IgG hu225 was far less effective at suppressing the proliferation of DiFi cells and completely lost its inhibitory activity in LIM1215 cells (Fig. 1A–D). HRG binds to HER3 and HER4 receptors, triggering heterodimerization mainly with HER2, but also with EGFR. We recently developed an antagonistic HER3-blocking antibody, IgG 3-43, and a bispecific antibody scDb hu225×3-43-Fc, which combines the antigen-binding domains of IgG hu225 and IgG 3-43, thereby blocking EGFR and HER3 receptors simultaneously (20, 21). Whereas the combined inhibition of EGFR and HER3 by IgG hu225 plus IgG 3-43 or the bispecific scDb-Fc antibody suppressed the viability of DiFi cells in the presence of HRG, none of the treatments interfered with HRG-driven proliferation in LIM1215 cells, despite antibody binding in the picomolar range and comparable binding of the bispecific scDb-Fc and the parental IgG hu225 antibody (Supplementary Fig. S1A and S1B).

Although LIM1215 cells express a higher number of HER3 receptors on their surface when comparing to DiFi cells and a similar number of HER2 receptors, the EGFR levels on LIM1215 cells are almost 30-fold lower than those on DiFi cells (∼22,000 vs. 700,000; Supplementary Fig. S1C). This suggests that HRG engages more HER3:HER2 than HER3:EGFR heterodimers in the LIM1215 cell line. We therefore tested whether HER2 targeting by trastuzumab might synergize with EGFR and HER3 inhibition to suppress HRG-driven proliferation of LIM1215 cells. In agreement with the hypothesis that HER2 signaling plays a minor role in DiFi cells, trastuzumab did not further enhance the inhibition observed upon EGFR and HER3 targeting in these cells (Fig. 1E). Interestingly, in the case of LIM1215 cells, the combination of scDb-Fc and trastuzumab, but not the combination of the three parental antibodies (IgG hu225, trastuzumab, and IgG 3-43), significantly inhibited HRG-induced proliferation (66.8% vs. 94.0%; w/o antibody control: 100%; Fig. 1F). Similar inhibition of cell viability was observed when pertuzumab was combined with scDb-Fc (Supplementary Fig. S1D). Taken together, these data show that exogenous HRG overrides EGFR inhibition in two cetuximab-sensitive colorectal cancer cell lines and, under these conditions, inhibition of proliferation of LIM1215 cells in particular is achieved most effectively by scDb-Fc in combination with HER2-blocking antibodies.

We next sought to obtain insight into how triple-HER targeting (scDb-Fc plus trastuzumab) interferes with HRG signaling in comparison with single (IgG hu225) and dual HER-targeting (scDb-Fc). To this end, serum-starved DiFi and LIM1215 cells were preincubated with the different antibodies for 1 hour, followed by HRG stimulation and harvesting of cells 24 hours later. Multiplexed samples were then subjected to CyTOF analysis to determine the activation states of key intracellular signaling pathways upon HER targeting (Fig. 2A). In DiFi cells, dual targeting of EGFR and HER3 potently suppressed the PI3K and MAPK pathways, as seen by the downregulation of pAkt/pS6 and pERK signals, respectively (Fig. 2A). Analysis of pRb and Ki67 as markers of proliferation and cleaved caspase-3 and PARP as markers of apoptosis revealed that dual HER-targeting inhibited proliferation and induced apoptosis, with trastuzumab providing marginal additional benefit (Fig. 2A), agreeing with our previous data. Intriguingly, dual and triple-HER-targeting increased the stem cell markers CD24 and CD44 (Fig. 2A), which might be explained by the previously observed connection between MAPK inactivation and the induction of stemness signatures in colorectal cancer (30–32). Dual inhibition of EGFR and HER3 by scDb-Fc was sufficient to almost completely block EdU incorporation in response to HRG stimulation, validating the potent inhibition of S-phase entry and cell-cycle progression (cells in S-phase: HRG = 54% vs. HRG + scDb-Fc = 5.8%; Fig. 2B). Similarly, Annexin-V staining of DiFi cells grown in the presence of HRG for 5 days confirmed that dual EGFR and HER3 blockade by scDb-Fc potently triggered apoptosis and was more effective than the combination of the parental antibodies (Annexin-V⁺/PI⁻ cells: HRG = 1.3%, HRG + scDb-Fc = 9.2%, HRG + IgG hu225 + IgG 3–43 = 4.9%; Fig. 2C).

Interestingly, CyTOF analysis of LIM1215 cells revealed that single EGFR and dual EGFR/HER3 targeting increased pAkt and pERK signals, respectively, suggesting the formation of alternative receptor heterodimers and/or activation of intracellular signaling feedback in response to HER receptor blockade. However, by additionally targeting HER2, pERK was partially and pAkt fully repressed below the control setting (Fig. 2A). Slightly increased apoptosis was observed particularly when HER2 was inhibited by trastuzumab (Supplementary Fig. S2B). The increased pRb and Ki67 positivity upon triple-HER targeting suggested that cell proliferation was inhibited in a distinct manner compared to DiFi cells, most likely by blocking cells in S-phase. Serum starvation neither altered the percentage of LIM1215 cells in S-phase nor triggered apoptosis (Supplementary Fig. S2A and S2B), which might be explained by autocrine TGFα produced by these cells (33). Of note, prolonged HRG treatment led to the accumulation of cells in S- and G₂–M-phases of the cell cycle (Supplementary Fig. S2C). The blocking antibodies seem to further slow down the cell cycle, leading to the prolongation of S-phase associated with pAkt and pERK suppression despite increased pRb and Ki67 signals. Thus, HRG does not appear to be a potent driver of LIM1215 proliferation but rather exerts a protective effect mainly when EGFR is inhibited (Supplementary Fig. S2; see Fig. 1).

To understand in more detail how LIM1215 cells respond to HER receptor inhibition, we preincubated serum-starved cells with the different antibodies for 1 hour before HRG stimulation for 15 minutes. Immunoblotting of cell lysates using phosphospecific antibodies revealed that HRG-induced phosphorylation of EGFR (Y1068), HER2 (Y1221/1222), and HER3 (Y1289) were strongly and significantly suppressed whenever HER3 was targeted, either by IgG 3-43 or scDb-Fc, whereas targeting of EGFR (IgG hu225) or HER2 (trastuzumab) was not sufficient to block receptor phosphorylation (Fig. 3A). HRG-induced downstream ERK (T202/204) phosphorylation was inhibited strongest and significantly by the simultaneous targeting of EGFR, HER2, and HER3 (Fig. 3A). However, only scDb-Fc plus trastuzumab partially suppressed Akt (T308) phosphorylation, although this difference was not statistically significant, whereas the combination of the parental antibodies failed to do so.

To analyze whether the suppression of HRG signaling was maintained by scDb-Fc plus trastuzumab beyond 15 minutes, we analyzed the phosphorylation status of the HER receptors and ERK and Akt in HRG-stimulated LIM1215 cells for up to 24 hours (Fig. 3B). While compensatory EGFR phosphorylation was observed at 6 and 24 hours, in addition to fully suppressing HER3 and ERK phosphorylation, triple-HER-targeting blunted HRG-induced Akt activation, agreeing with the results obtained by CyTOF analysis.

HRG was shown to drive growth and stemness of breast (34, 35) and colorectal cancer cells in serum-free suspension culture (36). The formation of oncospheres is characteristic of cancer cells with stem cell–like properties, with the PI3K pathway being particularly important for mediating survival under these conditions (37). We therefore investigated how single-, double-, and triple-antibody–mediated HER blockade affected oncosphere forming efficiency (SFE) of DiFi and LIM1215 cells. In DiFi cells, dual HER targeting by scDb-Fc was sufficient to impair SFE when compared with the untreated control, consistent with HER2 being less important in these cells. Nevertheless, addition of trastuzumab yielded even stronger and significant SFE suppression (1.4 ± 1.2 vs. 16.8 ± 3.7; Fig. 4A, top). Of note, oncosphere size was reduced whenever EGFR was targeted, indicating that EGFR signaling is required for the proliferation of DiFi cells under these conditions (Fig. 4A, bottom). In LIM1215 cells, the combination of scDb-Fc and trastuzumab was the only antibody combination that strongly and significantly inhibited SFE when comparing with the untreated control (SFE 1.7 ± 0.7 vs. 28.0 ± 14.3; Fig. 4B, top). The mean area of the oncospheres was not significantly affected by any of the treatments, indicating that triple-receptor blockade does not simply suppress general proliferation but rather cancer stem cell survival and oncosphere formation under these conditions (Fig. 4B, bottom). Taken together, triple-HER targeting by scDb-Fc and trastuzumab potently inhibited HRG-induced expansion of oncospheres in suspension cultures.

We finally sought to validate our triple targeting approach in primary PDO cultures established from resected KRAS WT colorectal adenocarcinoma tissues. Hematoxylin and eosin staining confirmed that the architecture of the organoid cultures resembled that of the primary tumor epithelium (Supplementary Fig. S3A). We next tested whether HRG could substitute for the EGF and bFGF growth factors present in the standard serum-free culture medium. HRG fully supported the survival and growth of the cultures without any marked changes in proliferation rates or morphology (Supplementary Fig. S3B). In full agreement with the response of the established colorectal cancer cell lines, treatment of the PDO1 culture with the different HER-targeting antibodies revealed strongest suppression of proliferation by the simultaneous targeting of EGFR, HER2, and HER3. This was confirmed in three additional KRAS WT PDO cultures, with IgG hu225 having little or no effect, while dual targeting of EGFR and HER3 led to stronger inhibition of proliferation. In all cases and regardless of the genetic profile of the PDO culture (Supplementary Fig. S3C) or the HER receptor expression levels (Supplementary Fig. S3D), the biological activity of scDb-Fc was slightly higher than that of the parental antibodies, and this was also seen when trastuzumab was added [scDb-Fc + trastuzumab vs. mix of parental antibodies: 41.7 ± 4.0% vs. 46.6 ± 7.6% (PDO1); 32.4 ± 3.0% vs. 39.0 ± 4.8% (PDO2); 12.2 ± 5.2% vs. 19.0 ± 7.2% (PDO3); 50.9 ± 24.5% vs. 65.0 ± 18.4% (PDO4)] (Fig. 5A–D). Note that overall HER receptor transcript levels were lowest in PDO4 (Supplementary Fig. S3D), correlating with weaker response to the HER-blocking antibodies.

Consistent with the results obtained in the cell lines, CyTOF analyses of PDO1 and PDO2 cultures confirmed the potent suppression of Akt and S6 activities by triple-HER targeting with scDb-Fc plus trastuzumab (Fig. 5E). Compared with single or dual targeting, triple targeting was more effective at reducing the cell cycle and proliferation markers pRb and Ki67 and triggering apoptosis, as seen by the elevated cleaved caspase 3 and PARP signals, particularly in PDO2 lacking mutations in p53 (Fig. 5E). In PDO1 in which pERK suppression was stronger, CD24 levels increased upon HER targeting, similar to our observations in the DiFi cell line (Fig. 2A). Comparison of PDO subpopulations with high versus low pAkt levels revealed pAkt positively correlating with pRb and Ki67, and in the PDO2 culture in particular, inversely correlating with cleaved caspase 3 (Fig. 5F), underscoring the relevance of the PI3K/Akt pathway to HRG-mediated survival and proliferation. Collectively, our data demonstrate that our newly developed bispecific EGFR×HER3-targeting scDb-Fc antibody together with trastuzumab overrides HRG signaling whereby the expansion of KRAS WT primary colorectal cancer cultures is broadly inhibited.

Although anti-EGFR therapy provides clinical benefit in a subset of patients with advanced colorectal cancer, responses are seldom durable. Although the expansion of mutant clones can lead to relapse, tumor cells can also adapt to the treatment despite lacking genetic alterations known to cause resistance to EGFR blockade. The latter can be mediated by both paracrine and autocrine mechanisms that stabilize drug tolerant states (38). Here, we provide evidence that pan-HER inhibition using a newly developed bispecific antibody targeting EGFR and HER3 in combination with trastuzumab effectively suppresses KRAS WT cell survival and proliferation driven by HRG, a prominent resistance factor in colorectal cancer (10). When treating metastatic patient-derived colorectal cancer xenografts with cetuximab alone, drug tolerance of residual disease was shown to be associated with the upregulation of ligands such as betacellulin and the activation of HER2 and HER3 receptors (31). As HER3-containing receptor heterodimers are known to potently activate the PI3K/Akt survival pathway, the simultaneous blocking of the EGFR and PI3K pathways led to more sustained treatment responses (31). This is in line with our results in which the suppression of Akt activation and downstream S6 phosphorylation in colorectal cancer cell lines and primary PDOs correlated with enhanced therapeutic activity, as seen by the inhibition of proliferation and colonosphere expansion, and the induction of apoptosis.

In the viability assays with LIM1215 cells in particular, the treatment settings comprising scDb-Fc were more effective than those containing the corresponding parental antibodies, despite binding to the same epitopes. Superior biological activity was also reflected in an orthotopic triple-negative breast cancer model, in which in vivo tumor growth was inhibited strongest by scDb-Fc when comparing with the parental antibodies IgG hu225, IgG 3-43, or the combination of IgG hu225 and IgG 3-43 (21). Beyond epitope specificity, parameters such as geometry and valency influence the biological activity of monospecific and bispecific antibodies. The potency of scDb-Fc might be explained by avidity effects arising from the tetravalent bispecific format and might also be associated with efficient internalization and degradation of its target receptors. By targeting two tumor-associated antigens, bispecific antibodies can give rise to improved tumor selectivity while minimizing side effects in normal tissues (39). However, because the grade of toxicity depends on the precise combination of binding moieties, determination of the safety profile of scDb-Fc in combination with trastzumab will require performing more advanced (pre)clinical studies in vivo.

Previously, duligotuzumab, a bispecific antibody binding to EGFR and HER3 was evaluated in combination with FOLFIRI as second-line therapy in patients with RAS WT metastatic colorectal cancer (40). No benefit for the dual inhibition of EGFR and HER3 over EGFR alone and no relationship between progression-free survival or overall response rate and HRG expression was observed, suggesting the activation of alternative drug evasion strategies. Therefore, considering the extensive signaling crosstalk and redundancy within the HER receptor family, pan-HER targeting approaches involving more than two receptors might hold more promise. HER2 has been recognized to play a more important role in primary and metastatic colorectal cancer than initially anticipated. Accordingly, multispecific antibody cocktails such as Sym013, a mixture of six antibodies targeting EGFR, HER2 and HER3, or H2EH3, a three-in-one aptamer-siRNA chimera targeting these receptors, showed stronger tumor growth inhibition in vivo (41, 42). Similarly, a mixture of three antibodies targeting EGFR, HER2, and HER3 (3×mAbs) was shown to overcome resistance of lung cancer to treatment with the third-generation EGFR kinase inhibitor osimertinib, resulting from a compensatory loop leading to increased expression of HER2 and HER3 (43). An increasing number of clinical trials investigate multispecific therapies based on antibody cocktails, or antibodies in combination with kinase inhibitors (44, 45). However, some clinical trials revealed overlapping toxicities, for example, when combining cetuximab with pertuzumab in patients with colorectal cancer (46), or no or only modest improvements in the case of cetuximab in combination with the pan-HER inhibitor afatinib in patients with non–small cell lung cancer (47, 48). Because afatinib does not affect HER3, incorporating HER3-specific inhibitory antibodies should be considered as an alternative approach. However, biomarkers associated with HER3 pathway activation that predict clinical response to pan-HER blockade have yet to be identified.

A recent report also established a connection between HER2 signaling and cancer stem cell survival in colorectal cancer (49). CD44v6-positive colorectal cancer stem cells were shown to express high levels of HER2, which was associated with constitutive PI3K/Akt activation and cetuximab resistance. Inhibition of HER2, MEK, and PI3K pathways promoted cell death and regression of tumor xenografts regardless of their mutational status (49). We show here that the addition of trastuzumab potentiated the inhibitory activity of scDb-Fc on colonosphere formation and expansion of DiFi and LIM1215 cells in addition to suppressing the growth of primary colorectal cancer PDOs in the presence of HRG. HRG not only supports cellular diversity and intestinal epithelial stem cell maturation during normal homeostasis (50, 51), but it is also strongly upregulated by mesenchymal niche cells during tissue regeneration to promote the proliferation of intestinal stem cells (52). The importance of HRG for stem cell function is corroborated by the inducible loss of HRG in adult mice producing a significant reduction in the proliferation of stem and progenitor cells (52). Importantly, the activity of HRG is not limited to the intestinal epithelium. We previously demonstrated that scDb-Fc inhibited HRG-induced oncosphere formation of triple-negative breast cancer cell lines and reduced the ALDH^high cancer stem cell population in vitro and in vivo (21).

The phenotypic differences between the DiFi and LIM1215 cell lines, which are both established cetuximab-sensitive colorectal cancer models, mirror the heterogeneous responses seen in patients with colorectal cancer upon anti-EGFR therapy. In colorectal cancer, MAPK inhibition was shown to induce stem cell plasticity via Wnt pathway activation and give rise to cetuximab-resistant residual disease characterized by a Paneth cell-like phenotype (30, 31). Thus, drug combinations that prevent such transdifferentiation as an escape mechanism are expected to elicit a more effective and durable antitumor response. Given the general importance of RTK signaling and HER receptors in particular for intestinal development and homeostasis, this may well be achieved by horizontal RTK blocking strategies. Our data also show that biological activities of therapeutic antibodies can greatly vary depending on the format. The better understanding of structure-function relationships in combination with computational engineering approaches may thus facilitate the rational design of more efficacious protein therapeutics in the future.

T. Sell reports grants from BMBF Germany during the conduct of the study. M. Schwab reports grants from Robert Bosch Stiftung during the conduct of the study; grants from Green Cross WellBeing Co. Ltd., Gilead Sciences Inc., and Robert Bosch GmbH; other support from Agena Bioscience GmbH and Corat Therapeutics GmbH outside the submitted work; and is editor-in-chief of Pharmacogenetics and Genomics and Drug Research, section editor of Genome Medicine, and receives honoraria for oral presentations at academically organized congresses and meetings. R.E. Kontermann reports grants from German Cancer Aid and grants and personal fees from SunRock Biopharma during the conduct of the study and personal fees from Roche, Immatics, Oncomatryx, and SunRock Biopharma outside the submitted work; in addition, R.E. Kontermann has a patent for HER3 antibody pending and licensed to SunRock Biopharma. M.A. Olayioye reports grants from German Cancer Aid during the conduct of the study; in addition, M.A. Olayioye has a patent for HER3 antibody pending and licensed to SunRock Biopharma. No disclosures were reported by the other authors.

A. Rau: Investigation, visualization, methodology, writing–original draft. N. Janssen: Investigation, visualization, methodology, writing–review and editing. L. Kühl: Investigation, visualization, writing–review and editing. T. Sell: Visualization, methodology. S. Kalmykova: Visualization, methodology. T.E. Mürdter: Supervision, writing–review and editing. M. Dahlke: Resources. C. Sers: Resources. M. Morkel: Conceptualization, supervision, methodology, writing–review and editing. M. Schwab: Supervision, funding acquisition, writing–review and editing. R.E. Kontermann: Conceptualization, supervision, funding acquisition, writing–review and editing. M.A. Olayioye: Conceptualization, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

This study was supported by a German Cancer Aid grant (DKH 70112564) to M.A. Olayioye, R.E. Kontermann, and M. Schwab and through a research grant from SunRock Biopharma (Spain), the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany's Excellence Strategy - EXC 2180 – 390900677; M. Schwab), and the Robert Bosch Stiftung, Stuttgart (M. Schwab, T.E. Mürdter, N. Janssen).

We thank Nadine Heidel and Sabine Münkel (University of Stuttgart) as well as Kathleen Siegel (Dr. Margarete Fischer-Bosch Institute for Clinical Pharmacology) for excellent technical assistance. We further acknowledge excellent service by the BIH Cytometry core facility.

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.