Exploratory clinical trials using therapeutic anti-HER3 antibodies strongly suggest that neuregulin (NRG1; HER3 ligand) expression at tumor sites is a predictive biomarker of anti-HER3 antibody efficacy in cancer. We hypothesized that in NRG1-expressing tumors, where the ligand is present before antibody treatment, anti-HER3 antibodies that do not compete with NRG1 for receptor binding have a higher receptor-neutralizing action than antibodies competing with the ligand for binding to HER3. Using time-resolved–fluorescence energy transfer (TR-FRET), we demonstrated that in the presence of recombinant NRG1, binding of 9F7-F11 (a nonligand-competing anti-HER3 antibody) to HER3 is increased, whereas that of ligand-competing anti-HER3 antibodies (H4B-121, U3-1287, Ab#6, Mab205.10.2, and MOR09825) is decreased. Moreover, 9F7-F11 showed higher efficacy than antibodies that compete with the ligand for binding to HER3. Specifically, 9F7-F11 inhibition of cell proliferation and of HER3/AKT/ERK1/2 phosphorylation as well as 9F7-F11–dependent cell-mediated cytotoxicity were higher in cancer cells preincubated with recombinant NRG1 compared with cells directly exposed to the anti-HER3 antibody. This translated in vivo into enhanced growth inhibition of NRG1-expressing BxPC3 pancreatic, A549 lung, and HCC-1806 breast cell tumor xenografts in mice treated with 9F7-F11 compared with H4B-121. Conversely, both antibodies had similar antitumor effect in NRG1-negative HPAC pancreatic carcinoma cells. In conclusion, the allosteric modulator 9F7-F11 shows increased anticancer effectiveness in the presence of NRG1 and thus represents a novel treatment strategy for NRG1-addicted tumors. Mol Cancer Ther; 16(7); 1312–23. ©2017 AACR.
Retrospective analyses of clinical trials correlated the expression level of the HER3 ligand neuregulin (NRG1) with the efficacy of anti-HER3 antibodies in solid tumors (1–7). NRG1 tumor expression is not only predictive of the response to HER3 inhibitors (1, 2, 5, 6, 8–10), but could also represent a prognostic marker of cancer recurrence, as demonstrated in head and neck squamous cell carcinoma (10). HER3 expression has been associated with worse prognosis in solid tumors (11), but no clear correlation has been found between HER3 and NRG1 expression in cancer. High NRG1 expression in cancer cells (12) defines a population of tumors that may be dependent or addicted to ligand-activated signaling via HER3 and/or HER4. This autocrine loop that involves high NRG1 expression and leads to HER3 activation has been described in head and neck (10, 13, 14), breast (15–17), lung (18, 19), and ovarian cancer (20). Activation of this loop induces resistance to EGFR- or HER2-targeted therapies (21, 22) and to chemotherapy (23) that can be overcome by treatment with agents against HER3, as demonstrated in ovarian (20), colorectal (24, 25), breast (26), and lung cancer (27, 28). Moreover, NRG1 gene fusions have been identified as oncogenic drivers of subtypes of lung, breast, and ovarian cancer (29–33) that could be indirectly targeted by HER3 inhibitors (34). Another potential role of NRG1 as a therapeutic predictive biomarker is linked to its expression in the tumor microenvironment (12). Indeed, NRG1 is secreted by cancer-associated fibroblasts and mesenchymal stem cells and subsequently, in a paracrine manner, can activate HER3 signaling, thus promoting resistance to kinase inhibitors, in melanoma (35, 36), colorectal (37), gastric (38), and pancreatic cancer (39).
Most of the current efforts focus on the development of anti-HER3 agents that directly interfere with or allosterically block NRG1 binding site (40). However, these molecules have failed in phase III clinical trials (NCT02134015), possibly because in NRG1-positive tumors, the ligand is already bound to HER3 before the antibody treatment. Therefore, we hypothesized that an allosteric, nonligand competing anti-HER3 antibody could be more effective than ligand-competing antibodies because (i) it will not need to displace NRG1 from HER3 to be effective, and (ii) it will be more active when the ligand is already expressed in tumors. Allosteric small molecules with this particular profile have been already developed to manipulate G-protein–coupled receptors (GPCR; refs. 41, 42).
We thus generated the non-NRG1 competing allosteric anti-HER3 antibody 9F7-F11 by immunizing mice with fibroblasts that express HER2/HER3 and that were prestimulated with NRG1 to favor HER3 active conformation. 9F7-F11 binds specifically to HER3- or HER2/HER3-transfected fibroblasts, but not to EGFR-, HER2-, or EGFR/HER4-transfected fibroblasts (43). This antibody blocks the PI3K/AKT pathway (43, 44) and induces HER3 downregulation (45), leading to in vivo tumor regression. We now wanted to determine whether 9F7-F11 acts as a nonligand-competing allosteric modulator and modifies NRG1 activity. By using time-resolved–fluorescence energy transfer (TR-FRET) we showed that 9F7-F11 binding to HER3 is enhanced by NRG1. Reciprocally, 9F7-F11 increased NRG1 binding to HER3. This translated into a better efficacy of 9F7-F11 in inhibiting NRG1-mediated cell proliferation and signaling and in promoting ADCC in tumor cells compared with NRG1-competing antibodies. Finally, as a positive modulator of NRG1 binding and negative modulator of NRG1 biological effects, the allosteric 9F7-F11 antibody reduced tumor growth of NRG1-expressing pancreatic, lung, and breast cancer cell xenografts more potently than a ligand-competing anti-HER3 antibody.
Materials and Methods
The BxPC3 and HPAC (pancreas), HCC-1806 and MDA-MB-453 (breast), and A549 (lung) human cancer cell lines were obtained from the ATCC. All cell lines were free of mycoplasma contamination, determined by using the MycoAlert Detection Kit (Lonza), and were authenticated by short tandem repeat profiling using the Promega PowerPlex 21 System.
Recombinant proteins and antibodies
Recombinant human HER3 extracellular domain (ECD) and human CD16a (FcγRIIIA) were purchased from R&D Systems. Human recombinant NRG1-β1 ECD (all experiments) and NRG1-β3 EGF domain (Fig. 1 only) were provided by R&D Systems and Millipore, respectively. The fully human H4B-121 (ligand-competing) and the mouse monoclonal 9F7-F11 (nonligand competing) anti-HER3 antibodies (developed in our laboratory) were obtained as described previously (43). The control antibody Px is an IgG1 mAb purified from the mouse myeloma cell line MOPC21. The anti-HER3 antibody MAB3481 was purchased from R&D Systems, whereas Ab#6 (described in patent WO2008/100624; parental molecule of the MM-121 antibody; ref. 46), U3-1287 (described in patent WO2007/077028; parental molecule of patritumab; ref. 26), MOR09825 (described in patent WO2012/022814; parental molecule of the LJM716 antibody; ref. 47), Mab205.10.2 (described in patent WO2012/022814; parental molecule of the RG7116 antibody; ref. 48), and the chimeric antibody 9F7-F11 (ch9F7-F11) were produced as recombinant antibodies in HEK293-F cells (IgG1 format) using the FreeStyle MAX Expression System (Invitrogen) according to the manufacturer's protocol. The low-fucose antibodies H4B-121-Emb and ch9F7-F11-Emb were produced in the YB2/0 cell line according to the Emabling Technology developed by LFB Biomanufacturing. For Western blotting, rabbit mAbs against total and phosphorylated HER3 (at Tyr1289, Tyr1197, or Tyr1222), total and phosphorylated AKT (pAKT; Ser473), total and phosphorylated ERK1/2 (Thr202/Tyr204), β-actin, and β-tubulin were from Cell Signaling Technology.
SNAP-tagged HER3 expression in HEK293 cells and Lumi4-Tb labeling
The Tag-lite platform (Cisbio Bioassays), which combines homogenous time-resolved fluorescence (HTRF) with the SNAP-tag Technology, was used to study anti-HER3 antibodies and NRG1 binding to HER3. A Tag-lite plasmid that encodes HER3 fused to SNAP-tag (Cisbio) was transiently expressed in HEK293 cells. At 80% confluency, the cell medium was removed and replaced by 12 mL of fresh cell culture medium. The transfection mixture [20 μL of 1 μg/mL SNAP-tagged HER3 plasmid (Cisbio), 60 μL of Lipofectamine 2000, and 8 mL of Opti-MEM medium (Life Technologies Inc.)] was preincubated at room temperature for 20 minutes before adding to the cells. Cells were then incubated for 24 hours before labeling with 200 nmol/L SNAP-Lumi4-Terbium (Tb) substrate (Cisbio) in Tag-lite medium (Cisbio) at 37°C for 1 hour. After four washings, Lumi4-Tb-SNAP-tagged HER3-expressing cells were resuspended in Tag-lite medium at a suitable density to perform TR-FRET experiments.
One hundred microliters of 1 mg/mL anti-HER3 antibodies in pH 8 phosphate buffer were labeled with the acceptor dye using the d2 Labeling Kit (Cisbio). Antibody–d2 conjugates at the optimal fluorophore/antibody ratio of 2.5 were purified on NAP5 columns (GE Healthcare) in 50 mmol/L phosphate buffer, pH 7. For competition experiments with unlabeled NRG1, 10-fold serial dilutions (from 4 × 10−12 to 4 × 10−5 mol/L) of NRG1-β1 ECD or NRG1-β3 EGF domain were prepared in Tag-lite medium. Lumi4-Tb-SNAP tagged HER3-expressing cells were seeded in white 384-well plates at a density of 20,000 cells/10 μL/well, followed by addition of serial dilutions of competitor NRG1 (5 μL/well) and 0.5 nmol/L of anti-HER3 antibody–d2 conjugates (5 μL/well) diluted in Tag-lite medium. For the competition experiments with unlabeled anti-HER3 antibodies, 5-fold serial dilutions (from 5 × 10−12 to 8 × 10−7 mol/L) of anti-HER3 antibodies, as competitors, were coincubated with Lumi4-Tb-SNAP tagged HER3-expressing cells and 12.5 nmol/L NRG1–d2 conjugate in 384-well plates. After incubation at room temperature for 5 hours 30 minutes, the TR-FRET signal (665 nm/620 nm emission ratio) was measured on a Pherastar FS reader in time-resolved fluorescence mode and normalized to 100% binding. Negative control wells contained only cells and Tag-lite buffer, whereas 100% binding was obtained by incubating antibody–d2 conjugates without NRG1, or NRG1-d2 conjugates without antibodies. The positive control for competition consisted in coincubating serial dilutions of unlabeled NRG1 with 12.5 nmol/L NRG1–d2 conjugate (Cisbio). All experiments were done in triplicate.
HER3 and CD16a (FcγRIIIA) ELISA binding assays
Flat-bottom 96-well Maxisorp plates (Nunc) were coated with 50 ng/well of recombinant human HER3 ECD, or 200 ng/well of recombinant CD16a at 4°C for 18 hours, and then blocked with 2% BSA in PBS. After washings in PBS/0.1% Tween 20 (PBS-T), anti-HER3 antibodies were added at 37°C for 1 hour. After washes with PBS-T, antibody binding to HER3 was detected by incubation with a horseradish-conjugated goat F(ab')2 antibody against human F(ab')2 (Jackson Immunoresearch) at 37°C for 2 hours. For the CD16a ELISA assay, anti-HER3 antibodies were preincubated with the horseradish conjugate (Jackson Immunoresearch) at room temperature for 1 hour, before adding to the CD16a-coated plates at 37°C for 1 hour. After three washes with PBS-T, the TMB substrate (3,3,5,5 tetramethylbenzidine; Sigma) was used for detecting peroxidase activity before the addition of 1 mol/L H2SO4 to stop the reaction. Absorbance was measured at 450 nm.
MTS cell viability assay
Cancer cells (5,000) were dispensed in each well of sterile 96-well flat-bottom plates, the day before starvation in RPMI complete medium/1% FCS for 18 hours. Ten-fold dilutions of anti-HER3 antibodies were then added for 5 days, with or without 3 × 10−9 mol/L NRG1-β1 ECD. Cell viability was then measured using the CellTiter 96 Aqueous One Solution Cell Proliferation Assay (Promega). Colorimetry was measured at 490 nm absorbance. All experiments were done in triplicate.
MDA-MB-453 breast carcinoma cells (2 × 104; target; T) were added to each well of sterile flat-bottom 96-well plates. One day later, 10-fold serial dilutions of anti-HER3 antibodies and control Px antibody were added, with or without 3 × 10−9 mol/L NRG1-β1 ECD, 30 minutes before the addition of peripheral blood mononuclear cells (PBMC) as effector cells (E). PBMCs from healthy donors (Etablissement Français du Sang) were prepared by density gradient centrifugation (GE Healthcare) according to the manufacturer's instructions, and 3 × 105 cells were added to each well (E:T ratio, 15:1). ADCC was assessed using the Cytotox 96 Non-radioactive Cytotoxicity Assay (Promega) that measures the release of lactate dehydrogenase (LDH) from damaged cells. After 20 hours of incubation, measurement was performed according to the manufacturer's instructions. The percentage of specific lysis of each sample was determined using the following formula: percentage specific lysis = (sample LDH release – E-cell spontaneous LDH release – T-cell spontaneous LDH release)/(T-cell maximum LDH release – T-cell spontaneous LDH release) × 100.
HER3/AKT signaling analysis by Western blotting
A total of 2 × 106 BxPC3 cells were grown in 10-cm culture plates at 37°C for 24 hours. After serum starvation in RPMI/1% FCS for 18 hours and washing, cells were prestimulated with various concentrations of NRG1-β1 ECD (3 × 10−12 to 3 × 10−9 mol/L) for 5 minutes, before incubation with 330 nmol/L of anti-HER3 antibodies at 37°C for 25 minutes. Alternatively, BxPC3 cells, prestimulated with 1 × 10−9 mol/L NRG1-β1 ECD for 5 minutes, were incubated with 10-fold serial dilutions (from 330 to 0.33 nmol/L) of anti-HER3 antibodies for 25 minutes. Cells were then washed, scraped, and lysed with buffer containing 20 mmol/L Tris-HCl pH 7.5, 150 mmol/L NaCl, 1.5 mmol/L MgCl2, 1 mmol/L EDTA, 1% Triton, 10% glycerol, 0.1 mmol/L phenylmethylsulfonyl fluoride, 100 mmol/L sodium fluoride, 1 mmol/L sodium orthovanadate (Sigma), and one tablet of complete protease inhibitor mixture (Roche Diagnostics). After 30 minutes, the insoluble fraction was cleared by centrifugation and protein concentration in cell lysates was determined with the Bradford assay. After SDS-PAGE electrophoresis under reducing conditions, proteins were transferred to polyvinylidene difluoride membranes (Millipore) that were then saturated in TNT buffer (25 mmol/L Tris pH 7.4, 150 mmol/L NaCl, 0.1% Tween) containing 5% nonfat dry milk at 25°C for 1 hour. Membranes were incubated with primary antibodies against HER3 or AKT and their phosphorylated forms, diluted in TNT/5% BSA buffer, at 4°C for 18 hours. After five washes in TNT buffer, the appropriate peroxidase-conjugated secondary antibodies (Sigma) were added in TNT buffer/5% nonfat dry milk at 25°C for 1 hour. After five washes in TNT buffer, blots were visualized using a chemiluminescent substrate (Western lightning Plus-ECL, Perkin Elmer). Positive bands were pixel-quantified using the Image J software and 100% phosphorylation was quantified in NRG1-stimulated cells without antibody.
pAKT and ERK quantification by TR-FRET
pAKT and ERK1/2 (pERK1/2) levels were quantified using the HTRF pAKT Ser473 and pERK Thr202/Tyr204 Kits (Cisbio Bioassay). A total of 5 × 104 BxPC3 cells/well were seeded in sterile 96-well flat-bottom plates and cultured overnight before starvation in RPMI/1% FCS for 18 hours. After removing the medium, cells were prestimulated with low (1 × 10−9 mol/L) or high (1 × 10−7 mol/L) concentrations of NRG1-β1 ECD for 5 minutes, before adding 10- or 2-fold dilutions of anti-HER3 antibodies, starting from 330 nmol/L, for another 25 minutes at 37°C. NRG1 and the antibodies were removed by extensive washing and cells were lysed in the supplemented lysis buffer (Cisbio Bioassay). Plates were incubated at room temperature with shaking for 30 minutes to lyse cells. Lysates were transferred to white 384-well plates. Anti-pERK1/2-cryptate/anti-pERK1/2-d2 or anti-pAKT-cryptate/anti-pAKT-d2 antibody pairs were added to each well and left in the dark at room temperature for 4 hours. The TR-FRET signal (665 nm/620 nm emission ratio) was measured on a Pherastar FS reader and normalized to 100% binding. Negative control wells contained unstimulated/nontreated cells and labeled antibodies, whereas 100% binding was obtained by stimulating BxPC3 cells with NRG1 without antibody treatment. Experiments were done in triplicate.
Quantitative PCR analysis
Total RNA was isolated from HCC-1806, BxPC3, HPAC, and A549 cells using the RNeasy Mini Kit (Zymo Research). RNA was quantified by UV spectroscopy. Total RNA (1 μg) was reverse-transcribed using the M-MLV Kit (Invitrogen). Real-time quantitative PCR was performed using a LightCycler 480 instrument (Roche Diagnostics), according to the manufacturer's instructions. The amplification specificity was checked by melting curve analysis. Real-time PCR values were determined by reference to a standard curve that was generated by real-time PCR amplification of a serially diluted cDNA sample using primers specific for NRG1α, NRG1β, EGF, and hypoxanthine phosphoribosyl transferase (HPRT). Data were normalized to the reference gene HPRT.
Tumor xenografts and treatment
All in vivo experiments were performed in compliance with the French regulations and ethical guidelines for experimental animal studies in an accredited establishment (agreement no. C34-172-27). BxPC3 (3 × 106), HCC-1806 (1 × 106), A549 (4 × 106), and HPAC (3.5 × 106) cancer cells were subcutaneously injected in the right flank of 6-week-old female athymic mice (Harlan Labs). Tumor-bearing mice were randomized to different treatment groups (at least six animals/group) when tumors reached a volume of 100 mm3. Animals were treated by intraperitoneal injection of 15 mg/kg of ch9F7-F11-Emb or H4B-121-Emb twice a week for 4 weeks. Tumor volumes were calculated by using the formula: D1 x D2 x D3/2. For survival analysis, mice were sacrificed when tumors reached a volume of 1,500 or 2,000 mm3.
NRG1 is a positive allosteric modulator of 9F7-F11 binding to HER3
We previously showed that 9F7-F11 does not compete with NRG1 for binding to HER3 in SKBR3 cells (43). Now we tested whether NRG1 influenced the binding of 9F7-F11 or of other antibodies to HER3 using a cell-based HTRF assay. First, we confirmed by ELISA that ch9F7-F11-Emb and H4B-121-Emb, U3-1287, Ab#6, Mab205.10.2, and MOR09825 bound equally to HER3 in a dose-dependent manner, with similar half-maximal concentrations (EC50; from 0.11 to 0.44nmol/L; Fig. 1A). However, the HTRF assay showed that binding of d2-labeled H4B-121-Emb, U3-1287, Ab#6, Mab205.10.2, and MOR09825 to HER3 in SNAP-Tag lumi4-Tb HER3-expressing HEK293 cells was abrogated in a dose-dependent manner upon coincubation with increasing concentrations of NRG1. Only binding of d2-labeled ch9F7-F11-Emb to HER3 was increased by 13-fold upon coincubation with NRG1 (Fig. 1B). Similarly, binding of 9F7-F11 to HER3 was enhanced 15-fold by coincubation with NRG1, while binding of H4B-121 was completely abolished (Supplementary Fig. S1). As a positive control, binding of d2-labeled NRG1 to HER3 was inhibited by free NRG1 (Supplementary Fig. S1). These results suggest that 9F7-F11, differently from other anti-HER3 antibodies, binds to a unique allosteric epitope outside the NRG1 binding site, and demonstrate that NRG1 is a positive allosteric modulator of 9F7-F11 binding to HER3.
Reciprocally, binding of d2-labeled NRG1 to SNAP-Tag lumi4-Tb HER3–expressing HEK293 cells was inhibited by coincubation of all the anti-HER3 antibodies tested, except ch9F7-F11-Emb (Fig. 1C) and 9F7-F11 (Supplementary Fig. S2) that increased by 1.5-fold NRG1 binding to HER3. The control antibody Px did not affect NRG1 binding to HER3 (Fig. 1C). Taken together, these results indicate that NRG1 improves 9F7-F11 binding to HER3, whereas it inhibits binding of ligand-competing anti-HER3 antibodies.
9F7-F11 inhibits cancer cell viability, and this effect is enhanced by NRG1
Coincubation with 3 nmol/L NRG1 enhanced the dose-dependent inhibitory effect of 9F7-F11 (nonligand competing anti-HER3 antibody) on BxPC3 and HPAC pancreatic cell viability compared with control cells (no NRG1), whereas no improvement was observed for the H4B-121 antibody. The control antibody Px, did not have any effect on cell viability both with or without NRG1 (Fig. 2).
NRG1 enhances 9F7-F11–mediated ADCC and reduces H4B-121–mediated ADCC
ADCC is one of the main mechanisms of action of therapeutic antibodies and is increased when using antibodies with low-fucose Fc. Accordingly, low-fucose ch9F7-F11-Emb and H4B-121-Emb showed a stronger dose-dependent binding to the CD16a receptor (FcγRIIIA) by ELISA than ch9F7-F11 and H4B-121 (Fig. 3A). This translated into higher and dose-dependent ADCC of MDA-MB-453 breast cancer cells by PBMCs in the presence of ch9F7-F11-Emb and H4B-121-Emb than with ch9F7-F11 and H4B-121 (Fig. 3B, top). The control antibody Px did not induce ADCC. Addition of 3 nmol/L NRG1 intensified ADCC mediated by native or low-fucose ch9F7-F11 (Fig. 3B; compare bottom and top). Conversely, NRG1 almost fully inhibited ADCC induced by H4B-121 and dramatically reduced ADCC mediated by H4B-121-Emb (Fig. 3B).
9F7-F11 is a negative allosteric modulator of NRG1-mediated cell signaling
Most anti-HER3 antibodies hinder HER3-triggered cell signaling in cancer cells, mainly by blocking the PI3K/AKT and ERK pathways. Therefore, the effect of 330 nmol/L ch9F7-F11-Emb or H4B-121-Emb in BxPC3 pancreatic cancer cells preincubated with increasing concentrations of NRG1 (from 3 × 10−12 to 3 × 10−9 mol/L) was evaluated by Western blotting. NRG1 induced, in a dose-dependent manner, HER3 phosphorylation at Tyr1289, Tyr1197, and Tyr1222 (p85-binding sites), and consequently AKT phosphorylation on Ser473 (Fig. 4). The maximal phosphorylation was observed with 1 nmol/L NRG1. ch9F7-F11-Emb completely abrogated ligand-mediated HER3 and AKT phosphorylation, independently of the NRG1 concentration. Conversely, H4B-121-Emb partially reduced HER3 and AKT phosphorylation induced by low (0.003–0.03 nmol/L), but not high (1–3 nmol/L) NRG1 concentrations. Similarly, HER3 (Tyr1289, Tyr1197, and Tyr1222) and AKT (Ser473) phosphorylation induced by prestimulation with 1 nmol/L NRG1 were inhibited, in a dose-dependent manner, by incubation with increasing (0.33–330 nmol/L) concentrations of ch9F7-F11-Emb in BxPC3 cells (EC50: 3.3 nmol/L of antibody; Supplementary Fig. S3). In the same conditions, H4B-121-Emb did not affect phosphorylation of HER3 on Tyr1289 and of AKT on Ser476 (Supplementary Fig. S3). Taken together, these results again suggest that 9F7-F11 is a negative-allosteric modulator of NRG1-mediated cell signaling and that it is more efficient than the NRG1-competing antibody H4B-121. Moreover, 9F7-F11, but not H4B-121, effect on cell signaling was independent of NRG1 concentration.
Then, TR-FRET was used to quantify and compare the effect of ch9F7-F11-Emb and of other anti-HER3 antibodies (U3-1287, Ab#6, Mab205.10.2, and MOR09825) on ligand-induced cell signaling. Like H4B-121-Emb, these antibodies block NRG1 binding, except MOR09825 (47). TR-FRET analysis confirmed that ch9F7-F11-Emb strongly reduced NRG1-induced AKT phosphorylation on Ser473, independently of NRG1 concentration (1 or 100 nmol/L), with an EC50 of about 0.33 nmol/L (Fig. 5A). This was confirmed using 9F7-F11 or ch9F7-F11 (Supplementary Fig. S4). All the other anti-HER3 antibodies also inhibited AKT phosphorylation (EC50 between 33 and 330 nmol/L) after stimulation with 1 nmol/L NRG1, although less efficiently than ch9F7-F11-Emb. Moreover, their inhibitory effect was completely abrogated, except for MOR09825, when cells were prestimulated with 100 nmol/L NRG1. Similarly, NRG1-mediated ERK1/2 phosphorylation on Thr202/Tyr204 was most efficiently inhibited by ch9F7-F11-Emb (EC50 of about 33 nmol/L compared with EC50 up to 330 nmol/L for the other antibodies). Such inhibition was blocked when cells were prestimulated with 100 nmol/L NRG1, except when using ch9F7-F11-Emb and MOR09825 (Fig. 5B). The TR-FRET results on the antibody-mediated inhibition of NRG1-induced cell signaling confirm those obtained by Western blotting, and strengthen the unique pharmacologic profile of ch9F7-F11-Emb as a negative-allosteric modulator of NRG1 activity, independently of the ligand concentration.
The nonligand-competing 9F7-F11 reduces tumor growth of NRG1-positive tumors more efficiently than the ligand-competing H4B-121 antibody
We finally asked whether 9F7-F11 profile (negative-allosteric modulator of NRG1 activity favored by ligand permeation) could translate into a better antitumor efficacy in vivo. Athymic mice were xenografted with NRG1-expressing tumor cells (pancreatic BxPC3, lung A549, and TNBC HCC-1806) or with NRG1-negative HPAC pancreatic tumor cells. NRG1α, NRG1β, and EGF mRNA expression in these cells were quantified by qPCR analysis (Supplementary Fig. S5). NRG1 did not affect in vitro cell proliferation of HCC-1806, A549, and HPAC cells, but increased viability of BxPC3 cells (Supplementary Fig. S6). ch9F7-F11-Emb treatment reduced growth of NRG1-expressing BxPC3, HCC-1806, and A549 tumor cell xenografts more potently than H4B-121-Emb (Fig. 6A–C, left). Interestingly, this effect was observed in cancer cells that express HER3 at very low level (e.g., A549 cells) and translated into a longer survival (Fig. 6A–C, right). Specifically, the mean tumor growth inhibition in ch9F7-F11-Emb and H4-B121-Emb–treated mice was 95% versus 75% at day 49 post-BxPC3 cell graft, 70% versus 48% at day 36 post-HCC-1806 cell graft, and 75% versus 20% at day 72 post-A549 cell graft, respectively. In contrast, in mice bearing NRG1-negative HPAC tumor cell xenografts, treatment with ch9F7-F11-Emb or H4-B121-Emb led to the same mean tumor growth inhibition (40%) at day 40 post-graft (Fig. 6D, left) and to a similar survival rate (Fig. 6D, right). Altogether, these in vivo results demonstrate that the non-NRG1 competing allosteric modulator 9F7-F11 inhibits tumor growth of NRG1-positive pancreatic, lung, and breast cancers more efficiently than the NRG1-competing orthosteric H4B-121 antibody, in agreement with the effects observed in vitro on NRG1-mediated cell viability, signaling, and ADCC.
Here, we show that the dual-allosteric modulator 9F7-F11 is well designed for the treatment of NRG1-positive tumors because stimulation with high NRG1 concentrations promotes its binding to HER3 and its biological effects in cancer cells. 9F7-F11′s unique pharmacologic profile has never been observed in all the previously described anti-HER3 antibodies that mostly act by blocking NRG1 binding and the binding of which is reciprocally inhibited by NRG1 (40). The nonligand-competing 9F7-F11 antibody profile is better fitted for targeting NRG1-positive tumors than the U3-1287, Ab#6, and Mab205.10.2 antibodies that directly block ligand binding (26, 46, 48), or the MOR09825 or KTN3379 antibodies that preferentially bind to HER3 in the inactive configuration (47, 49). Indeed, in NRG1-positive tumors, the ligand might be already bound to HER3 before antibody treatment, leading in vivo to a prevalence of HER3 receptors in the active conformation. Moreover, in HER2-amplified tumors, where HER3-HER2 heterodimers can be formed independently of NRG1 activation, HER3 targeted therapy is less effective, as confirmed recently in a phase II clinical trial in ovarian cancer (2). 9F7-F11 shows a unique HER3 binding profile because it is promoted by NRG1 presence. Garner and colleagues (47) showed that the anti-HER3 antibody LJM716 targets a conformational epitope located within domains 2 and 4 of HER3 inactive configuration and does not prevent NRG1 binding to HER3. Such binding characteristics have been demonstrated using preformed HER3/LJM716 complexes to avoid subsequent binding of various concentrations of NRG1 to the receptor (47). In our coincubation experiments (Fig. 1C), less than 1 nmol/L of free MOR09825 (the parental antibody of LJM716) could displace 50% of the binding of 20 nmol/L d2-conjugated NRG1 to lumi4-Tb-HER3-expressing HEK293 cells. Reciprocally, higher NRG1 concentrations (around 100 nmol/L) were necessary to efficiently shift the equilibrium of 0.5 nmol/L MOR09825 complexed with HER3 toward the extended active conformation (Fig. 1B). Thus, in our experimental conditions, MOR09825 acts as a ligand-blocking antibody with a slow off-rate and its HER3 binding affinity is greater than that of NRG1. We can speculate that the high-affinity barrier imposed by preformed LJM716/HER3 complexes cannot be overcome by subsequent addition of NRG1 that shows lower affinity to HER3 compared with the antibody. In contrast, when 1 or 100 nmol/L NRG1 is added before antibody treatment, only the highest MOR09825 concentration could displace the ligand from HER3 and inhibit ligand-mediated AKT and ERK1/2 phosphorylation. In this case, 9F7-F11 was more efficient than MOR09825 at inhibiting signaling, independently of the ligand concentration. The other ligand-competitive antibodies (U3-1287, Ab#6, Mab205.10.2, and H4B-121) did not work at high NRG1 concentration.
Nevertheless, the definition of “physiologic” NRG1 concentration within tumors (as opposed to NRG1 addiction) is still an unresolved issue. An NRG1 detection/quantification assay could be of great value for selecting patients who could benefit of anti-HER3 treatment. NRG1 expression in head and neck, esophageal, lung, and cervical cancers could be higher than in other tumor types (4, 10, 13, 50); moreover, acquired NRG1 expression could also drive resistance to targeted therapies (5, 25, 28). Normal NRG1 plasma level (28, 51–54) and NRG1 concentration in body fluids in various diseases (51, 52, 54–56), including colorectal (57), lung (7, 25, 28, 58), and ovarian (58) cancers, are in the picomolar–nanomolar range and higher level in cancers are considered to be a predictive biomarker for HER3-targeted therapy (5). Currently, there is no standard measurement to define NRG1 level (low or high) in tumors in situ. MacBeath and colleagues (3) defines high NRG1 as >5 by RT-qPCR or ≥1+ by RNA-in situ hybridization. We can postulate that NRG1 concentration within tumors is higher than in body fluids and becomes an obstacle for the efficiency of ligand-blocking anti-HER3 antibodies. This could explain the disappointing results of phase III clinical trials on NRG1-competitive HER3 antibodies (NCT02134015). Therefore, NRG1 should be quantified in tumor tissue biopsies just before treatment decision-making. The optimal method of detection (RNA-in situ hybridization or RT-PCR), threshold for NRG1 positivity (continuous variable), and the reliability of results obtained with primary archived tissues compared with fresh tumor biopsies are still debated.
Moreover, we think that besides NRG1-overexpressing tumors, 9F7-F11 could be effective also in other cancer subtypes that harbor NRG1 fusion genes (29–33). These fusion genes lead to NRG1 expression at the tumor cell surface and to NRG1 binding to HER3 in an autocrine or juxtacrine manner. In these tumors, active HER3 receptors are continuously permeated by NRG1 fusion proteins, and 9F7-F11, which binds to ligand-targeted HER3 with greater affinity than to HER3-inactive monomers, could be very efficient. This antibody could also be useful for tumors that are activated through paracrine ligand–receptor stimulation from the cell microenvironment (35–39).
Receptors from the HER family are flexible molecules which can be ortho- or allosterically activated, depending on the ligand type and its binding site. HER behavior can closely resemble that of GPCR activation and trans-conformation (41). Therefore, the nature of the allosteric interaction (e.g., 9F7-F11 binding to HER3) could be closely dependent on the orthosteric ligand (the probe), in accordance with the concept of probe-dependence and biased signaling developed for GPCR (42). Here, we demonstrate by TR-FRET analysis that 9F7-F11 binding to HER3 is promoted by the presence of the orthosteric ligand NRG1 and, reciprocally, that 9F7-F11 binding to HER3 increases NRG1 binding. The resulting output is the inhibition of NRG1-mediated signaling and, more largely, the intensification of the in vitro and in vivo antibody-mediated biological effects when NRG1 permeates tumor cells. This dual-allosteric profile (i.e., positive cooperativity for NRG1 binding and simultaneously negative modulation of NRG1-mediated effects) has been already demonstrated for Org 27569, an allosteric modulator of cannabinoid receptor 1 (59). Tailoring allosteric antibodies against the HER family of receptors by using “oriented” immunization strategies has been already investigated with the anti-EGFR antibody mAb806 (60) that recognizes a conformationally exposed epitope in tumors that overexpress wild-type EGFR or mutant EGFRvIII. However, to our knowledge, this is the first report showing that receptor targeting by an antibody (9F7-F11) is increased in the presence of the orthosteric ligand. The mechanism of action of other drugs might bring some clues. Org 27569 stabilizes cannabinoid receptor 1 structure in a conformation at the transmembrane level that is different from the one observed during “classical” GPCR activation (59). The anti-EGFR antibody mAb806 targets an epitope on the D2 ECD that is masked in the inactive monomer and in ligand-activated dimers, but exposed when EGFR is overexpressed, or upon EGFR ECD truncation (61) or glycosylation changes (62). Interestingly, it has been suggested that HER3 glycosylation changes are crucial for NRG1-induced dimer formation, cell signaling, and tumor progression (63, 64). The 9F7-F11 antibody binds to an epitope located in the D1 domain of HER3 (43). Despite the different binding profile, 9F7-F11 specificity also suggests a conformation-sensitive epitope that is partially masked in the inactive monomer and better exposed in the NRG1-liganded HER3. Finally, NRG1 activates HER3, but also HER4, to transduce efficient signaling, mainly through the AKT and ERK pathways. It has been shown that the NRG1/HER4-inducible Hippo pathway promotes YAP-driven oncogenic mechanisms through the binding of the PPxY motif in HER4 to the WW domains of YAP (65). HER3 also harbors a PPxY domain that could be involved in 9F7-F11–induced HER3 degradation (45). Thus, HER3 may act as an alternative activator of the Hippo–YAP pathway that could, therefore, be indirectly blocked by the anti-HER3 antibody 9F7-F11.
In summary, we demonstrated that the binding to HER3 and biological effects on tumor cells of the novel, very potent dual-allosteric anti-HER3 antibody 9F7-F11 are paradoxically facilitated by the natural ligand NRG1. Therefore, by hijacking NRG1 addiction of cancer cells to promote its inhibitory effects on NRG1-mediated tumor growth and resistance, 9F7-F11 displays a unique potential for targeted treatment of NRG1-positive cancers.
Disclosure of Potential Conflicts of Interest
C. Larbouret has ownership interest (including patents) in WO2015/067986 and WO2012/156532. A. Pèlegrin has ownership interest (including patents) in WO2012/156532 patent “Anti-human HER3 and uses thereof” and WO2015/067986 patent “Neuregulin allosteric anti-HER3 antibody." T. Chardès has ownership interest (including patents) in WO2012/156532 and WO2015/067986. No potential conflicts of interest were disclosed by the other authors.
Conception and design: C. Le Clorennec, H. Bazin, O. Dubreuil, C. Larbouret, J.-M. Barret, T. Chardès
Development of methodology: H. Bazin, O. Dubreuil, C. Larbouret, C. Ogier, Y. Lazrek, V. Garambois, G. Mathis, T. Chardès
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Le Clorennec, H. Bazin, C. Larbouret, C. Ogier, V. Garambois, T. Chardès
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): O. Dubreuil, J.-M. Barret, A. Pèlegrin, T. Chardès
Writing, review, and/or revision of the manuscript: C. Le Clorennec, H. Bazin, O. Dubreuil, M.-A. Poul, J.-M. Barret, J.-F. Prost, A. Pèlegrin, T. Chardès
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): O. Dubreuil, C. Larbouret, G. Mathis, T. Chardès
Study supervision: J.-M. Barret, P. Mondon, J.-F. Prost, A. Pèlegrin, T. Chardès
We thank G. Heintz and S. Bousquié (IRCM) for cell culture and antibody production. The animal facility staff at the IRCM is greatly acknowledged. We also thank S. Kadi (Cisbio) for performing antibody labeling and TagLite assays.
This work was supported by the program “Investissement d'Avenir” (grant agreement: Labex MabImprove, ANR-10-LABX-53-01; to A. Pèlegrin) and by the grant AAP13 “Fonds Unique Interministériel” FUI UmAbHER3 F120402M (to T. Chardès).