An Antibody That Locks HER3 in the Inactive Conformation Inhibits Tumor Growth Driven by HER2 or Neuregulin

Discovery of an anti-HER3 antibody that inhibits both ligand-dependent and -independent HER3 activity in vitro

Previous short-hairpin RNA (shRNA) data have shown that loss of HER3 expression in HER2-amplified models is sufficient to inhibit cell proliferation (14). Treatment of the HER3 shRNA-sensitive cell line SKBR-3 with commercially available HER3-targeted antibodies further validates the ability to modulate cell growth in this setting from the extracellular compartment. AF234 (a HER3-specific goat polyclonal) induced significant inhibition of growth, whereas MAB3481, a ligand-blocking mouse monoclonal, had little to no effect (Fig. 1A). The greater activity observed with AF234 was correlated with its more robust downregulation of HER3 and prolonged inhibition of pAKT (Fig. 1B). Thus, antibodies that simply block ligand binding to HER3 and fail to inhibit AKT phosphorylation are unlikely to be active in this setting.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Identification of LJM716, a HER3-binding monoclonal antibody capable of blocking ligand-dependent and –independent HER3 signaling and cell proliferation. A, SKBR-3 were treated with increasing concentrations of IgG, MAB348, or AF234 for five days followed by cell viability assessment using the CTG assay. B, SKBR-3 cells were treated with 10 μg/mL IgG, MAB3481, and AF234 for one or 24 hours. Cell lysates were harvested and immunoblotted with antibodies directed against the indicated proteins. C, HER3-targeted antibodies were profiled for their ability to inhibit HER3 tyrosine phosphorylation and cell growth in HER2-amplified SKBR-3 cells. Maximal growth inhibition was calculated relative to that achieved with AF234. pHER3 was measured via MSD assay and Family 15 members are highlighted in red. D, MCF7 cells were treated with HER3-targeted antibodies before stimulation with NRG1 (50 ng/mL) and their impact on HER3 phosphorylation and proliferation determined. Maximal growth inhibition was calculated relative to that achieved with AF234. pHER3 was measured via MSD assay and Family 15 members are highlighted in red. E, SKBR-3 cells were treated with LJM716 (red) or IgG (black) for one hour and the level of pHER3 quantified via MSD assay. F, MCF7 cells were treated with LJM716 (red) or IgG (black) for 30 minutes before NRG1 stimulation for 10 minutes. pHER3 levels were measured by MSD assay. G, SKBR-3 cells grown in full serum were treated with LJM716 (red) or IgG (black) for five days and cell viability determined by CTG. H, serum-starved MCF7 cells were treated with LJM716 (red) or IgG (black), stimulated with NRG1 (50 ng/mL), and cell viability determined after five days using CTG. Cell proliferation values relative to unstimulated cells are plotted.

To identify a HER3 therapeutic antibody capable of inhibiting both ligand-dependent and ligand-independent HER3 signaling, we isolated HER3-targeted Fabs using MorphoSys HuCAL phage-display technology (24) against a number of recombinant HER3 antigens. Corresponding full IgGs were screened for their ability to neutralize HER3 tyrosine phosphorylation and cell proliferation driven by either HER2 (SKBR3 cells) or NRG1 (MCF7 cells). From all confirmed hits, only four antibodies were capable of effectively inhibiting HER2-driven HER3 phosphorylation and proliferation in SKBR-3 cells (Fig. 1C). In contrast, a large number of antibodies efficiently inhibited HER3 phosphorylation and cell proliferation in NRG1-stimulated MCF7 cells (Fig. 1D) due to their ability to prevent ligand binding (data not shown). Three antibodies were active in both settings (Fig. 1C and D), suggesting they may be uniquely capable of inhibiting multiple modes of HER3 activation. The sequence of these antibodies differed only in residues contained within hCDR2 and, thus, these antibodies were highly related and are referred to herein as Family 15.

We further characterized LJM716, an antibody derived from Family 15. LJM716 binds to human, mouse, rat, and cynomolgus monkey HER3 with high affinity as determined by solution equilibrium titration (Table 2) and flow cytometry (Supplementary Table S2). LJM716 potently inhibited phosphorylation of HER3 driven by either HER2 overexpression or NRG1 stimulation (Fig. 1E and F, Supplementary Table S3, Supplementary Fig. S2A) in SKBR-3 and MCF7 cells. The inhibitory activity of LJM716 similarly led to inhibition of PI3K pathway signaling including downregulation of AKT phosphorylation (pAKT S⁴⁷³; Supplementary Fig. S2B and 2C and Table S3) and inhibition of cell proliferation (Fig. 1G and H, Supplementary Table S4) in both SKBR-3- and NRG1-stimulated MCF7 cells. Despite their high degree of sequence homology, no binding of LJM716 to other ErbB family members was observed (Supplementary Fig. S3A). Together, these data suggest that LJM716 is a specific and potent inhibitor of ligand-dependent and -independent HER3 signaling and growth.

LJM716 displays single-agent activity in HER2- and NRG1-driven in vivo models

Having established that LJM716 inhibits multiple modes of HER3 activation in vitro, we next asked whether this activity was maintained in vivo. SCID mice bearing HER2-amplified BT474 breast cancer xenografts were treated with a single-dose of 20 mg/kg LJM716, resulting in maximal reduction of pHER3 by 52% and pAKT by 84% over a 24-hour period when compared with an equivalent dose of an isotype control antibody (Fig. 2A). Similarly, in SCID mice bearing ligand-expressing BxPC-3 pancreatic xenografts, intravenous dosing of 20 mg/kg LJM716 resulted in 86% maximal inhibition of pHER3 and 74% inhibition of pAKT compared with isotype-matched treated controls (Fig. 2B). These pharmacodynamic studies show that LJM716 is capable of inhibiting both HER2- and ligand-mediated HER3 signaling in vivo as evidenced by decreased pHER3/pAKT in both xenograft models.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

LJM716 inhibits HER3 signaling in vivo and is efficacious in multiple ligand-dependent and independent xenograft tumor models. A, BT474 xenografts were dosed intravenously with a single 20 mg/kg dose of LJM716, and tumors were harvested at 0, 4, or 24 hours followed by analysis of lysates for pHER3 and pAKT levels by MSD assay. B, BxPC-3 xenografts were dosed intravenously with a single 20 mg/kg dose of LJM716, and tumors were harvested at 0, 4, or 72 hours followed by analysis of lysates for pHER3 and pAKT levels by MSD assay. C, BT474 xenografts were dosed intravenously every other day with 20 mg/kg LJM716 (red) or IgG (black). D, HBCx-13A primary human xenografts were dosed intravenously every other day with 20 mg/kg LJM716 (red) or IgG (black). E, BxPC-3 xenografts were dosed intravenously every other day with 20 mg/kg LJM716 (red) or IgG (black). F, FaDu xenografts were dosed intravenously every other day with 20 mg/kg LJM716 (red) or IgG (black). Two-tailed non-paired t test was conducted for A+B. Data in C–F are presented as mean tumor volume ±SEM. All δ volumes subjected to One Way ANOVA and Tukey post hoc analysis. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

LJM716 displayed significant antitumor activity in BT474 (17% T/C) and HBCx-13A (7.7% T/C) HER2-amplified xenografts (Fig. 2C and D). Treatment of SCID mice bearing NRG1-expressing xenografts with 20 mg/kg LJM716 induced tumor regression in FaDu (SCCHN), tumor stasis in Hara (lung), and significant (T/C < 25%) tumor growth inhibition in BxPC-3 and T3M4 (pancreatic) xenografts (Fig. 2E and F, Supplementary Table S5). LJM716 is, therefore, efficacious in HER2-amplified and NRG1-expressing tumor models in vivo.

LJM716 traps HER3 in the inactive conformation

In an effort to understand the mechanism by which LJM716 inhibits HER3, the epitope was identified by determining the three-dimensional (3D) structure of the Fab fragment of MOR09825 bound to the HER3 extracellular domain. MOR09825, the parent Fab molecule of LJM716, has identical CDR regions and an identical biologic profile to LJM716 (data not shown). The 3.4Å crystal structure reveals that MOR09825 binds to HER3 in the tethered (inactive) ErbB conformation and recognizes a nonlinear epitope formed by two distinct domains D2 and D4 (Fig. 3A). The epitope is centered over the D2/D4 interface with a similar occluded surface area for each domain (D2-706 Ų; D4-546 Ų; Fig. 3B). This unique binding mode has not been previously observed with other ErbB-targeted antibodies and can only occur when domains 2 and 4 are juxtaposed in the tethered (inactive) HER3 conformation (27–29). The structure also reveals that the antibody heavy chain and light chain contribute approximately equally to the recognition of HER3 with the paratope comprising all three heavy chain CDRs and two light chain CDRs (Supplementary Fig. S3B).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

LJM716 locks HER3 in the inactive conformation and does not block ligand binding. A, crystal structure of the HER3 extracellular domain in complex with MOR09825. Fab is in gold and the HER3 domains are individually colored: D1 (gray), D2 (purple), D3 (white), and D4 (blue). HER3 is in the inactive (tethered) conformation and the ligand binding site located in D1 and D3 is not occluded. B, HER3 residues comprising the MOR09825 epitope (yellow ball-and-sticks) are clustered in D2 and D4. MOR09825 is removed from the view to aid visualization. C, close-up of MOR09825 interactions with K267 and L268 located near the interface of the HER3 dimerization loop (D2) and auto-inhibitory domain (D4). MOR09825 light-chain residues are highlighted in gold and MOR09825 heavy-chain residues in orange. D, binding of LJM716 to HER3 mutants K267A, L268A, and K267A/L268A expressed as recombinant proteins and determined by biochemical ELISA. E, interaction analyses conducted by capturing biotinylated NRG1 on the surface of a Biacore sensor chip. Preformed HER3/antibody complexes at the indicated concentrations were injected over reference and active surfaces and their interactions with NRG1 observed. F, MCF7 cells were incubated with the indicated concentrations of antibodies before the addition of 10 nM NRG1-β1 EGF domain. Cell-surface-bound NRG1 was quantified by flow cytometry and is plotted as mean channel fluorescence on the y-axis.

The crystallographically defined epitope is corroborated by biochemical characterization of the anti-HER3/HER3 interaction. ELISA characterization of MOR09825 and other Family 15 members binding to recombinantly expressed HER3 domains and ECD shows that binding is only detected for the full ECD construct. No binding is observed for constructs containing isolated domains or domain pairs, D1-D2, D2, D3–4, or D4 (Supplementary Fig. S3C). In addition, inspection of the crystal structure indicated that HER3 residues Lys267 and Leu268 located within the D2 β-hairpin dimerization arm are involved in numerous interactions with both the heavy and light chains of the antibody (Fig. 3C), suggesting they may be fundamentally important for binding. Mutation of Lys267 and/or Leu268 to alanine abolishes LJM716 binding (Fig. 3D), whereas binding of a D3-targeted antibody is largely unaffected by these mutations (Supplementary Fig. S3D). These data show that residues within the β-hairpin dimerization arm are an integral part of the LJM716 epitope. We conclude that binding of LJM716 locks HER3 in the inactive conformation, thus preventing the large-scale structural rearrangements required for transition of HER3 to the extended conformation characteristic of activated ErbB receptors (30–32).

LJM716 does not prevent neuregulin binding to HER3

The crystal structure indicated that the ligand-binding site of HER3, which has been mapped to D1 and D3 by analogy to EGFR (31, 32), was not occluded by Fab binding (Fig. 3A and B). We explored the impact of LJM716 upon the HER3–NRG1 interaction. LJM716–HER3 complexes were preformed and passed over a surface plasmon resonance biosensor chip previously coated with NRG1. LJM716/HER3 complexes were capable of binding immobilized NRG1 whereas a complex comprised of HER3 and a ligand-blocking HER3 antibody (MOR09624) prevented binding to NRG1 (Fig. 3E). Furthermore, the presence of LJM716 had no significant impact on the binding affinity of HER3 for NRG1 (Supplementary Fig. S4), indicating that LJM716 does not influence NRG1 binding in biochemical assays. The affinity of NRG1 was consistent with that previously determined using a mutant form of HER3 that is in the locked conformation (33). To confirm these results in a cell-based system, we tested whether MCF7 cells pre-bound with LJM716 were capable of binding NRG1. Using flow cytometry we found that pre-binding of MCF7 with LJM716 had no impact on subsequent binding by NRG1 (Fig. 3F). These findings strongly suggest that the LJM716 epitope is not essential for ligand binding and that LJM716 and NRG1 can bind concurrently.

LJM716 effectively combines with HER2 and EGFR inhibitors

The inability of trastuzumab to fully inhibit HER2/3 complexes is thought to limit its therapeutic benefit, owing to incomplete PI3K pathway suppression (2, 3). When we tested the ability of trastuzumab to inhibit pHER3 in a panel of HER2-amplified cell line models grown in full serum, robust (>50%) pHER3 inhibition was observed in only 3/19 cell lines (Fig. 4A). In contrast, LJM716 treatment was more broadly active, lowering the IC₅₀ of pHER3 inhibition (e.g., MDA-MB-453) as well as enhancing maximal inhibition (e.g., NCI-N87), so that 14/18 cell lines showed robust (>50%) pHER3 inhibition (Fig. 4A and B). LJM716 induced inhibition of pAKT in a subset of cell lines (Fig. 4A), all of which contained high levels of HER3 phosphorylation, and in these cell lines, LJM716 was capable of inducing growth inhibition (Fig. 4C, Supplementary Fig. S1A and E). Trastuzumab similarly induced inhibition of pAKT (Fig. 4A) and proliferation (Fig. 4C) in an overlapping subset of cell lines despite its inability to effectively inhibit HER3. These data may indicate that HER2 can activate PI3K signaling in a HER3-dependent and -independent manner, and that modulation of pAKT may be a key predictor of single-agent activity for both LJM716 and trastuzumab.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

LJM716 displays in vitro and in vivo combination activity with trastuzumab and cetuximab. A, HER2-amplified cell lines grown in full-serum were treated with 10 nmol/L LJM716 or trastuzumab for one hour and the impact on both pHER3 (Y¹¹⁹⁷) and pAKT (S⁴⁷³) determined by MSD assay. Percentage inhibition relative to control IgG-treated cells is visualized in the form of a heat map colored from blue (0% inhibition) to red (100% inhibition). Cell lines harboring hotspot PI3K mutations are highlighted in red. B, the HER2-amplified cell lines UACC812, MDA-MB-453, NCI-N87, and SKBR-3, grown in full serum, were treated with increasing doses of LJM716 or trastuzumab for one hour and the impact on pHER3 (Y¹¹⁹⁷) was determined by MSD assay. C, HER2-amplified cell lines were dosed with LJM716 (33 nmol/L), trastuzumab (33 nmol/L), or LJM716 plus trastuzumab. Cells were grown for five days and cell viability determined by CTG. Percentage of inhibition relative to untreated cells is visualized in the form of a heat map. D, BT474 tumor xenografts were grown in NSG mice and treated with IgG (20 mg/kg, q2d), LJM716 (20 mg/kg, q2d), trastuzumab (10 mg/kg, q2w), or LJM716/trastuzumab. Data are presented as mean tumor volume ±SEM. *, P < 0.05 by ANOVA post hoc Holm–Sidak test. E, SCCHN cell lines were treated with LJM716 (11 nmol/L), cetuximab (11 nmol/L), or LJM716/cetuximab. Cells were grown for five days in full serum, and cell viability determined by CTG. Percentage inhibition relative to untreated cells is visualized in the form of a heat map. F, FaDu tumor-bearing mice were dosed for 14 days with IgG (20 mg/kg, q2d), LJM716 (20 mg/kg, q2d), cetuximab (20 mg/kg, q2w), or LJM716/cetuximab. The regrowth of tumors was specifically monitored following cessation of dosing on day 28. *, P < 0.05 by Kruskal–Wallis ANOVA on ranks post hoc Dunn test. q2d, dosed every other day; q2w, dosed every second week.

To investigate the consequence of HER2 and HER3 dual inhibition we determined the growth inhibitory effect of simultaneously combining LJM716 with trastuzumab. In vitro, the addition of LJM716 to trastuzumab both increased the degree of growth inhibition obtained in sensitive cell lines (e.g., SKBR-3) and also induced a response in cell lines refractory to either agent alone (e.g., KPL4; Fig. 4C, Supplementary Fig. S7). The efficacy of LJM716/trastuzumab combination was tested in vivo using BT474 xenografts grown in NSG mice. BT474 xenografts grown in nude mice are exquisitely sensitive to trastuzumab, due in part to its potent ability to induce antibody-dependent cell-mediated cytotoxicity (ADCC; ref. 34). Consequently, NSG mice were used since they lack functional NK cells in addition to B and T cells, thus potentially mimicking the impaired ability of trastuzumab to induce ADCC in those patients possessing low-affinity Fcγ receptor polymorphisms (35). In NSG mice, both LJM716 and trastuzumab slowed tumor growth when dosed as single agents (Fig. 4D). The combination of both agents was well tolerated and showed superior activity compared with either agent alone, inducing prolonged tumor stasis (Fig. 4D, Supplementary S5A).

Because LJM716 also effectively inhibits NRG1-dependent HER3 activation, we assessed its activity in models of SCCHN, a lineage that has recently been shown to be driven by a NRG1/HER3 autocrine loop (22). LJM716 displayed single-agent growth inhibitory activity in several of these cell lines (Fig. 4E). Interestingly, the combination of LJM716 with the EGFR-targeted antibody cetuximab, which is approved for the treatment of metastatic SCCHN, was more potent in vitro compared with either single agent (Fig. 4E). In vivo the LJM716–cetuximab combination was well-tolerated and capable of completely eradicating FaDu xenograft tumors (Fig. 4F, Supplementary S5B) and, in this regard, was superior to either agent alone. These data indicate that LJM716 can inhibit multiple modes of HER3 activation and the combination results in enhanced activity with both HER2- and EGFR-targeted agents in indications driven by either HER2 or NRG1.

The HER2-targeted antibody pertuzumab, which blocks dimerization of HER2 with other ligand-activated ErbB members, was recently approved for treatment of metastatic breast cancer. We sought to compare the combination activity of trastuzumab–pertuzumab (T+P) with that of trastuzumab–LJM716 (T+L) in a set of HER2-amplified breast cancer cell lines. Although T+P and T+L combinations were both highly potent at inhibiting colony, as well as 3D spheroid formation in the trastuzumab-sensitive cell line BT474, the LJM716-containing combination was significantly more active in trastuzumab-insensitive models like HR6 (36) and MDA-MB-453 cells (Fig. 5A and B, Supplementary Fig. S6). Increased activity of LJM716 containing treatments was accompanied by enhanced inhibition of HER3 and AKT (Fig. 5C). In trastuzumab-sensitive BT474 xenografts in vivo, T+P and T+L combinations are equally active at inducing tumor regressions (Fig. 5D, Supplementary Fig. S5C). When monitoring overall survival following a 35-day dosing period, both combinations equally prolong survival compared with either T or P alone (Fig. 5E). These studies show that addition of LJM716 to trastuzumab is at least as active as the T+P combination in trastuzumab-sensitive settings, but enables enhanced inhibition of oncogenic signaling and proliferation in trastuzumab-resistant settings.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

The addition of LJM716 to trastuzumab enables more potent HER3–PI3K pathway signaling and growth inhibition than trastuzumab–pertuzumab combination, particularly in trastuzumab-insensitive models. A, cells were plated at 10,000 to 50,000 cells per well in six-well plates and treated in triplicate with dimethyl sulfoxide (DMSO), 10 μg/mL LJM716, pertuzumab, and/or trastuzumab. Media containing antibodies was replenished every three to four days. Cells were stained with crystal violet when control-treated cells were confluent, ranging from 14 to 21 days. Representative images and quantification of integrated intensity (% control) are shown. *, P < 0.05, t test. B, cells were seeded in Matrigel and allowed to grow in the absence or presence of 10 μg/mL LJM716, pertuzumab, and/or trastuzumab as indicated. Medium was subsequently changed every three days. Images shown were recorded 15 to 19 days after cell seeding. Acini burden was quantified using the GelCount system. Each bar graph represents the mean + SEM. of triplicate samples. *, P < 0.05, t test. C, BT474 and MDA453 cells were treated with 10 μg/mL LJM 716, 10 μg/mL pertuzumab, and/or 10 μg/mL trastuzumab for one or 24 hours. Whole-cell lysates were prepared and separated in a 7% SDS gel followed by immunoblot analysis with antibodies directed against the indicated proteins. D, BT474 xenografts were treated with either IgG (20 mg/kg), trastuzumab (20 mg/kg), pertuzumab (20 mg/kg), LJM716 (20 mg/kg, q2d), trastuzumab+pertuzumab, or trastuzumab+LJM716 for 35 days. Data are presented as mean tumor volume ±SEM. E, Kaplan–Meier survival analysis following the end of 35 days of treatment. Mice were monitored for tumor regrowth and sacrificed when tumor burden was larger than 2,000 mm³. q2d, dosed every other day; q2w, dosed every second week.