Acquired Resistance to a MET Antibody in Vivo Can Be Overcome by the MET Antibody Mixture Sym015

Author Manuscript Published OnlineFirst on March 15, 2018; DOI: 10.1158/1535-7163.MCT-17-0787 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Acquired resistance to a MET antibody in vivo can be overcome by the MET antibody mixture Sym015

Authors:

Sofie Ellebæk Pollmann1, Valerie S. Calvert2, Shruti Rao3, Simina M. Boca3, Subha Madhavan3, Ivan D. Horak1, Andreas Kjær4, Emanuel F. Petricoin2, Michael Kragh1, and Thomas Tuxen Poulsen1

Authors’ affiliations: 1Symphogen A/S, Ballerup, Denmark 2Center for Applied Proteomics and Molecular Medicine, George Mason University, Virginia, USA 3Innovation Center for Biomedical Informatics, Georgetown University, Washington DC, USA 4Dept. of Clinical Physiology, Nuclear Medicine & PET and Cluster for Molecular Imaging, Dept. of Biomedical Sciences, Rigshospitalet and University of Copenhagen, Denmark

Running title: In vivo acquired resistance to a MET-targeting antibody.

Keywords: Antibody, MET, Resistance, in vivo, RPPA

Corresponding author: Sofie Ellebæk Pollmann Symphogen A/S Pederstrupvej 93 DK-2750 Ballerup, Denmark E-mail: [email protected] Tel: +4545265050

Disclosure of potential conflicts of interest: Sofie Ellebæk Pollmann, Ivan D. Horak, Michael Kragh, and Thomas Tuxen Poulsen are all employed by Symphogen A/S. Subha Madhavan is a member of the scientific advisory board of and a consultant for Perthera. Emanuel F. Petricoin is chief scientific officer, co-founder, member of board of directors of Perthera, of which he has ownership interests. He is also a member of the scientific advisory board of/consultant for Perthera and Avant Diagnotics.

Downloaded from mct.aacrjournals.org on September 30, 2021. © 2018 American Association for Cancer Research. Author Manuscript Published OnlineFirst on March 15, 2018; DOI: 10.1158/1535-7163.MCT-17-0787 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Disclosure of financial support: Subha Madhavan, Simina M. Boca, and Emanuel F. Petricoin has received commercial research funding from Symphogen A/S.

Word count: 5107

Number of figures: 5

Abstract

Failure of clinical trials due to development of resistance to MET-targeting therapeutic agents is an emerging problem. Mechanisms of acquired resistance to MET tyrosine kinase inhibitors are well described, whereas characterization of mechanisms of resistance toward MET-targeting antibodies is limited. This study investigated mechanisms underlying in vivo resistance to two antibody therapeutics currently in clinical development: an analogue of the MET-targeting antibody emibetuzumab and Sym015, a mixture of two antibodies targeting non-overlapping epitopes of MET. Upon long-term in vivo treatment of a MET-amplified gastric cancer xenograft model (SNU- 5), emibetuzumab-resistant, but not Sym015-resistant, tumors emerged. Resistant tumors were isolated and used to establish resistant cell lines. Characterization of both tumors and cell lines using extensive protein and signaling pathway activation mapping along with next-generation sequencing revealed two distinct resistance profiles, one involving PTEN loss and the other involving activation of the PI3Kinase pathway, likely via MYC and ERBB3 copy number gains. PTEN loss left one model unaffected by PI3Kinase/AKT-targeting but sensitive to mTOR-targeting, while the PI3Kinase pathway-activated model was partly sensitive to targeting of multiple PI3Kinase pathway proteins. Importantly, both resistant models were sensitive to treatment with Sym015 in vivo due to antibody-dependent cellular cytotoxicity-mediated tumor growth inhibition, MET degradation, and signaling inhibition. Taken together, our data provide key insights into potential mechanisms of resistance to a single MET-targeting antibody, demonstrate superiority of Sym015 in preventing acquired resistance, and confirm Sym015 anti-tumor activity in tumors resistant to a single MET antibody.

Introduction

MET is a receptor tyrosine kinase (RTK), activated by hepatocyte growth factor (HGF), which induces various downstream signaling pathways, leading to cell proliferation, survival, migration, and angiogenesis [1]. Aberrant activation of MET, along with amplification and activating mutations of the MET gene, are found in many malignancies, including non-small cell lung cancer (NSCLC) and gastric cancer. This has led to intensive study of MET as a potential therapeutic target [2]. MET consists of an extracellular ligand-binding domain and an intracellular kinase domain. The extracellular part is made up of a seven-bladed β-propeller semaphorin unit (SEMA), a cysteine-rich plexin-semaphorin-integrin unit, and four immunoglobulin (Ig)-like units. The intracellular domain is made up of a juxtamembrane part, a tyrosine kinase, and a carboxy-terminal tail [1]. Binding of HGF to the extracellular part of MET induces receptor dimerization, resulting in phosphorylation of tyrosine residues in the cytoplasmic domain. Some of these tyrosine residues can then serve as docking sites for signaling molecules, initiating a cascade of signaling activation through, for example, the Ras/ERK/MAPK and PI3Kinase/AKT pathways [1,2]. Furthermore, MET cross-talks with other RTKs, including human epidermal growth factor receptor (HER)-family receptors, RET, RON, AXL, and vascular endothelial growth factor receptor 2 (VEGFR2), by transphosphorylation of the kinases or indirectly through downstream signaling molecules [2–6].

Gastric cancer is the fourth leading cause of cancer death worldwide1 and the effectiveness of chemotherapy, the current standard treatment for patients with metastatic or nonresectable gastric cancer, is very limited [7]. Therefore, multiple targeted treatments for gastric cancer have been developed, including the HER2-targeting monoclonal antibody trastuzumab and the VEGFR2- targeting monoclonal antibody ramucirumab in patients whose tumors overexpress these receptors [7–9]. High MET activation, arising from amplification or mutation of MET, is also observed in a subset of gastric cancers correlating with poor prognosis [7]. MET also plays a role in the development of resistance to epidermal growth factor receptor (EGFR) inhibitors, where MET amplification and increased HGF expression have been observed [10,11]. Together, these observations suggest that MET is a relevant target for therapeutic agents such as tyrosine kinase inhibitors (TKIs) and antibodies.

1 World Health Organization. WHO Fact sheet on Cancer 2017. http://www.who.int/mediacentre/factsheets/fs297/en/

Many MET-targeting TKIs are in clinical development, including PHA-665752, crizotinib, AMG337, and foretinib. These inhibitors have been shown to block MET phosphorylation, inhibit in vitro proliferation, and in vivo tumor growth of numerous MET-amplified cancer models [5,12– 14]. Furthermore, MET-targeting therapeutic antibodies have been developed. These include the monoclonal antibodies ABT-700 and DN30, the one-armed onartuzumab (MetMAb), the humanized IgG4 antibody emibetuzumab, and more recently the antibody mixture Sym015, comprising two antibodies targeting non-overlapping MET epitopes. Their diverse mechanisms of action include ligand blockade, internalization/degradation of MET, and activation of secondary effector functions such as antibody-dependent cellular cytotoxicity (ADCC). All these antibodies demonstrated promising preclinical results and some have shown activity in early clinical settings [15–22]. Despite huge efforts devoted to development of MET-targeted agents, disappointing results were obtained in later clinical trials, including failure of onartuzumab in Phase 3 and limited activity of emibetuzumab in both NSCLC and gastric cancer patients in Phase 2 [23–27], warranting more in-dept investigation of mechanisms of resistance to MET-targeted agents.

Mechanisms of resistance to MET-targeting TKIs have been investigated in vitro and include MET- resistance mutations, leading to a conformational change that blocks TKI binding, autocrine activation by HGF, bypass activation via HER-family signaling, MET and KRAS gene amplification, MYC upregulation, and constitutive activation of downstream pathways [28–33]. Acquired resistance to MET-targeting TKIs has also been investigated in vivo, revealing activating mutations in MET and overexpression of PI3K p110α as two resistance mechanisms [28,34]. In contrast, resistance to treatment with therapeutic anti-MET antibodies has received less attention, showing increased MET gene copy number as a mechanism of resistance to a monovalent Fab fragment of DN30 in preclinical studies [35].

In this study, we induced resistance by long-term exposure of a MET-dependent and MET- amplified xenograft tumor model to an analogue of emibetuzumab and to the antibody mixture Sym015. While no Sym015-resistant tumor models were established, we successfully generated, isolated, and characterized emibetuzumab-resistant tumors and derived cell lines, which were analyzed for mechanisms of resistance to MET-targeting antibody therapeutics. Our results show that resistance to a single MET-targeting antibody may be overcome by inhibition of PI3Kinase pathway signaling and/or re-treatment with the MET-targeting antibody mixture Sym015.

Materials and Methods

Cell lines, in vitro proliferation and ADCC

The SNU-5 cell line obtained from ATCC (CRL-5973, February 2011) was propagated for a limited number of passages in IMDM+20% fetal bovine serum (FBS)+antibiotics (Life Technologies) at

37°C, 5% CO2. Emibetuzumab analogue (50 µg/mL) was added to the resistant SNU-5 cell lines upon passaging. All cell lines were routinely tested for mycoplasma infection using MycoAlert (Lonza) or PCR (IDEXX Bioresearch, July 2016). Cell line authenticity and species determination were documented for all cell lines by short tandem repeat DNA typing (LGC Standards, July 2016 and DSMZ, February 2017). Cells plated in medium containing 2% FBS, were subjected to treatment and metabolic activity was measured after 96h using WST-1 proliferation reagent (Roche). The ADCC assay was described elsewhere [20].

Therapeutic antibodies and TKIs

An in-house-produced analogue of emibetuzumab (LY-2875358), an antibody-binding non-human antigen (used as a negative control), Sym015, and Sym015 LALA (a mixture of effector-silent Fc- mutated antibodies with two Leu to Ala mutations) were generated as described elsewhere [20]. PI- 103 and sirolimus were from Selleckchem.

Cell and tumor lysates

Cells were plated and serum-starved in 2% FBS for 2-3 days. 24 h before lysis, cells were treated with emibetuzumab, Sym015, PI-103, sirolimus, DMSO, or negative control antibody. Cells were harvested, pelleted by centrifugation for 5 min at 300×g, washed with ice-cold PBS and resuspended in RIPA Buffer (Sigma-Aldrich) containing protease inhibitor cocktail (Thermo Scientific) and phosphatase inhibitor cocktail set II (Calbiochem). Tumor pieces were lysed in ice- cold RIPA Buffer containing inhibitors and homogenized on a gentleMACS™ Octo Dissociator (Miltenyi Biotec), prior to sonication. Lysates were cleared by centrifugation at 4°C prior to concentration measurement using a Pierce BCA Protein Assay (Thermo Scientific).

Simple Western analyses

Simple Western analyses were performed on a Sally Sue instrument (ProteinSimple), with standard settings. Rabbit primary antibodies against MET, phosphorylated (p)MET(Y1234/1235) (D26), pan-actin, pERK1/2(T202/Y204) (D13.14.4E), ERK1/2, VEGFR2 (D5B1), AXL (C2B12), pS6

ribosomal protein(S235/236) (D57.2.2E), S6 ribosomal protein (5G10), Ras, pPI3Kinase p85(Y458), pmTOR(S2448) (D9C2), mTOR (7C10), MYC (D84C12), PTEN (138G6), pPTEN(S380), and HER3 (D22C5), all from Cell Signaling Technology, and mouse anti-PI3Kinase (Clone4) from BD Biosciences, were diluted 1:50. Signal intensities were quantified using Compass software version 3.0. Statistical significance was calculated using Student’s t-test.

Tumor xenograft models

107 cells were inoculated subcutaneously into the flank of 6- to 10-week-old female ADCC-capable BALB/c nu/nu mice (Janvier labs). Tumors were measured 3 times/week using caliper, and tumor volumes were calculated using the formula: 0.5 × length × (width)2. At an average tumor size of 120 mm3, mice were randomized for treatment. At a tumor volume of 600-850 mm3 the mice were euthanized by cervical dislocation for tumor isolation. Treatment of resistant cell line xenografts was initiated on the day of cell inoculation. All in vivo studies were performed in accordance with applicable laws and regulations relating to the care and use of laboratory animals.

Tumor preparation

Upon collection, a piece of tumor was submerged completely in optimum cutting temperature (OCT) compound, placed on dry ice, and frozen at -80°C for laser capture microdissection (LCM). Another small piece of each tumor was frozen in RNAlater (Thermo Scientific) for next-generation sequencing (NGS) and a third piece was snap frozen in liquid nitrogen and stored at -80°C for lysate preparation. Finally, a fourth piece was placed in growth medium for immediate cell line generation as follows: Tumors were minced with razorblades prior to tissue degradation in collagenase type III and TrypLE Select, filtering through a 70 μm strainer, and plating for passaging.

Live cell imaging

Cells were plated in medium containing 2% FBS with/without emibetuzumab and images were recorded every 4 h using IncuCyte ZOOM (Essen Bioscience) using IncuCyte ZOOM 2015A software for data analysis. Data were normalized to initial confluency level. One-way ANOVA was used to compare treatment efficacy between groups.

LCM and reverse phase protein array (RPPA) analysis

8 μm cryostat sections were cut from each OCT-frozen tumor at -20°C and transferred to microscopy slides, fixed in 70% ethanol, stained in hematoxylin, and dehydrated in increasing concentrations of ethanol (70%-95%-100%) and 100% xylene, prior to LCM using CapSure Macro Caps and an ArcturusXT instrument (Thermo Fisher) using infrared laser-based direct cellular capture.

Lysates for RPPA analysis were prepared using T-PER lysis buffer (Thermo Scientific). Lysates were diluted to 0.25 mg/mL and boiled prior to printing in technical triplicates on nitrocellulose- coated glass slides using an Aushon 2470 arrayer (Aushon Biosystems). Analysis of 175 proteins/phosphoproteins (Suppl. Table S1) was carried out as described elsewhere [36,37].

Only proteins/phosphoproteins with intensities above detection limit in at least 95% of the samples (162 proteins) were analyzed. Intensity values were log2-transformed in order for the data to be more suitable to standard statistical analysis techniques. The limma package in R [38] was used for comparing profiles of each emibetuzumab-resistant cell line with SNU-5 upon treatment with emibetuzumab (50 µg/mL). This method uses a moderated Student’s t-test, which borrows strength between proteins when estimating variances, using an Empirical Bayes approach [39]. Volcano plots were used to present the relationship between significance and fold change for each protein/phosphoprotein detection in the resistant cell lines compared with SNU-5.

Network analysis

Systems biology analyses were conducted on outliers in each volcano plot using Pathway Studio (Elsevier). Networks were generated to determine the direct interactions between proteins/genes that were significantly altered. Each network interaction is supported by at least one published reference from the ResNet Mammal database (version 11.2.5.6).

NGS analysis

NGS using ClearSeq Comprehensive Cancer panel (Agilent Technologies) was carried out at Eurofins Genomics. Genomic DNA was isolated from 5×106 cells or 25 mg tumor tissue, previously frozen in RNAlater, using a QIAamp DNA Mini Kit (Qiagen) and subjected to library preparation and sequencing with 2×100 bp paired-end reads using SureSelect on Illumina HiSeq 2500 sequencer. Investigation of sequence homology to human/mouse genome (Suppl. Fig. S1) and data cleaning discarding unmapped reads and reads aligning better to mouse than human genome,

was performed. Analysis of single-nucleotide polymorphisms and copy number variation (CNV) was performed using SureCall software version 3.5.1.46 (Agilent).

Results

MET-dependent SNU-5 cells develop resistance to emibetuzumab, but not Sym015, upon long-term treatment in vivo

SNU-5 is a highly MET-dependent MET-amplified gastric cancer cell line, which has previously been used to generate cell lines resistant to MET-targeting therapeutics [12,13,15,34]. To confirm the MET dependency of SNU-5 in vitro, metabolic activity of SNU-5 after treatment with emibetuzumab or Sym015 was measured. SNU-5 cells were effectively inhibited by treatment with emibetuzumab and Sym015 (Fig. 1A), with emibetuzumab being more potent at lower antibody concentrations, confirming that SNU-5 depends on MET signaling to grow and proliferate. This translated into decreases in MET and pMET protein levels (Fig. 1B). Next, SNU-5 xenografts were established in vivo to generate resistant tumors by long-term continuous treatment with emibetuzumab and Sym015. All 16 mice treated with emibetuzumab initially responded with substantial tumor shrinkage (Fig. 1C). Four mice had complete tumor regression at the end of the study, but after more than 100 days of continuous treatment some tumors started to regrow. At outgrowth, tumors 7333, 7360, and 4626 were isolated for characterization. In contrast, no resistant tumors regrew upon continuous treatment with Sym015, which resulted in complete tumor regression in all 16 mice for the full duration of the study (Fig. 1D). The levels of MET, pMET, and pERK1/2 (T202/Y204) were lower in the isolated emibetuzumab-resistant tumors than in vehicle control-treated tumors (Fig. 1E).

In vivo-derived emibetuzumab-resistant cell lines retain resistance in vitro

Cell lines were successfully generated from two of the isolated emibetuzumab-resistant tumors (7333 and 4626), which were both shown to match the ATCC SNU-5 cell line profile and to be free from mouse cells (Suppl. Fig. S1). First, resistance of the cell lines to emibetuzumab was investigated. Both cell lines retained resistance to emibetuzumab in vitro compared to SNU-5 (compare graphs in Fig. 1A and 2A). Next, live cell imaging was used to study cell proliferation in real time with or without emibetuzumab treatment. 7333 and 4626 grew at a similar rate regardless of emibetuzumab treatment, although slower than untreated SNU-5 cells, whereas SNU-5 treated with emibetuzumab grew significantly slower than the emibetuzumab-resistant cell lines (Fig. 2B), indicating that the resistant cell lines are not addicted to emibetuzumab. In vivo growth of the emibetuzumab-resistant cell lines was also investigated, by re-inoculating the resistant cell lines on mice. Both models maintained resistance to emibetuzumab in vivo, as both were established and

grew continuously during emibetuzumab treatment, while inoculation of SNU-5 cells failed to establish tumors in emibetuzumab treated mice (Suppl. Fig. S2). The in vivo growth rates of emibetuzumab-treated 7333 and untreated SNU-5 tumors were similar, whereas the growth rate and time for tumor establishment of 4626 tumors was reduced (Fig. 2C). Furthermore, to verify that emibetuzumab still affect MET activation in the in vivo-generated cell lines, MET, pMET, and pERK1/2 levels were quantified. Decreased levels of MET and pMET in 7333 cells (Fig. 2D) correlated with decreased tumor protein levels (Fig. 1E). In contrast, the pMET level was maintained high in 4626 cells compared with emibetuzumab-treated SNU-5 cells, whereas the level was decreased in the 4626 tumor compared with control-treated SNU-5 tumors (Fig. 1E and Fig. 2D). Also, no difference in MET levels were observed with or without emibetuzumab treatment in 7333 and 4626 (Suppl. Fig. S3). Finally, pERK levels were not significantly altered in the resistant cell lines.

Protein and signaling pathway activation mapping reveals two different resistance profiles

Activation of alternative signaling pathways has previously been identified as a mechanism of resistance to MET-targeted therapeutics [28,29,32]. Therefore, the activated protein signaling architectures of the two emibetuzumab-resistant tumors and derived cell lines were investigated and compared to profiles of control-treated SNU-5 tumors and emibetuzumab-treated SNU-5 cells, respectively, using RPPA-based network analysis. The volcano plots show fold changes of 162 protein/phosphoproteins levels detected in each of the two emibetuzumab-resistant cell lines compared with SNU-5 cells (Fig. 3A). In 7333 cells, phosphorylation levels of a large group of proteins, including many RTKs, were decreased compared with the levels in SNU-5 cells (Fig. 3B and Suppl. Table S2). In contrast, fewer protein signals were increased in 7333, including total Ras, phosphorylated S6 ribosomal protein (S6RP), and pAKT (Fig. 3A and B). The resistance profile of 4626 cells showed significantly increased levels of many proteins and phosphoproteins, including phosphorylated VEGFR2 and various other RTKs, along with elevated phosphorylation of downstream signaling mediators, such as AKT, S6RP, 4E-BP1, and p70 S6 Kinase (Fig. 3A-B and Suppl. Table S3). Selected RPPA findings in the cell lines were validated by Simple Western analysis of whole cell lysates, which demonstrated that VEGFR2, AXL, and pS6RP levels were upregulated in both resistant cell lines and Ras protein level was increased in 7333 compared with SNU-5 (Fig. 3C).

In order to verify the RPPA cell line results and to confirm that the cell lines could be used as an in vitro tool for investigating acquired resistance in vivo, RPPA was performed on LCM samples of the tumors used for cell line generation. Comparison of protein and phosphoprotein levels in control-treated SNU-5 vs. resistant tumors showed upregulation of many of the same markers in the tumors and cell lines, including AXL, VEGFR2, pAKT, and pS6RP (Fig. 3D). Importantly, the same pattern of downregulated phospho-RTK levels in 7333 and increased phospho-RTK levels in 4626 was observed in the tumors (Fig. 3D).

PI3Kinase pathway inhibition partly overcomes emibetuzumab resistance

Together, the results presented above suggest that 4626 may have switched from primarily MET- driven signaling to signaling through multiple RTKs, such as the HER-family, VEGFR2, and AXL, and the downstream PI3kinase pathway, whereas 7333 appear to have switched to PI3Kinase pathway activation independent of RTK signaling. Therefore, we tested a number of relevant therapeutic inhibitors and antibodies in vitro to try to resensitize the emibetuzumab-resistant cells. However, despite increased activation of various RTKs in 4626 cells (Suppl. Table S3), the metabolic activity of 4626 cells was not inhibited compared with SNU-5 or 7333 cells in response to AXL, VEGFR2, and RET inhibition in vitro (Suppl. Table S4). Likewise, despite the upregulated total AXL and VEGFR2 levels in 7333 (Fig. 3), 7333 cells were not sensitive to AXL or VEGFR2 inhibition using various TKIs (Suppl. Table S4). Therefore, broader-acting inhibitors targeting downstream signaling proteins in the PI3Kinase pathway were tested. Inhibition of PIK3CA/mTOR and pan-AKT with PI-103 and MK-2206, respectively, reduced the metabolic activity of 4626 cells, whereas neither 7333 nor SNU-5 was inhibited to the same extent (Fig. 4A and Suppl. Table S4). Furthermore, both 7333 and 4626 were significantly more sensitive to mTOR inhibition by sirolimus than SNU-5 cells (Fig. 4A). These changes in metabolic activity is further supported by decreases in PI3Kinase and mTOR activation upon treatment with PI-103, and decreases in mTOR and S6RP activation upon treatment with sirolimus, in both 7333 and 4626 cells (Fig. 4B). Also, the combination of PI-103 and AXL, VEGFR2, RET, or HER3-targeting was tested in 4626, but no added inhibitory effects were observed (Suppl. Table S4). These results indicate that 4626 has increased dependency on PI3Kinase pathway signaling, whereas 7333 has increased dependency on mTOR.

Mutations and gene amplification are known reported mechanisms of resistance to MET-targeting therapeutics [30,34,35,40]. Therefore, mutational profiling and CNV determination were performed

by NGS to compare resistant cell lines and tumors with SNU-5 cells and control-treated SNU-5 tumors, respectively. This revealed a deletion of the tumor suppressor PTEN (-0.74 log2fold) in 7333 cells (Fig. 4C) correlating with downregulated PTEN and pPTEN protein levels in both cells and tumor (Suppl. Table S2, Fig. 3D and Fig. 4D). Furthermore, gain of the proto-oncogene MYC

(1.21 log2fold) and PVT1 (1.07 log2fold), which encodes a long non-coding RNA that stabilizes and potentiates the activity of MYC [41], was observed in both 4626 cells and tumor (Fig. 4C and Suppl. Fig. S4). This correlates with increased MYC levels in both 4626 cells and tumor compared to SNU-5 cells and control-treated tumors (Suppl. Table S3, Fig. 3D, and Fig. 4D). Finally, ERBB3 gene copy number gain (0.77 log2fold) was also observed in both 4626 cells and tumor (Fig. 4C and Suppl. Fig. S4), correlating with increased HER3 protein levels in both cells and tumor (Suppl. Fig. S5 and Fig. 4D). Proposed models for mechanisms of resistance, based on the in vitro characterization of 7333 and 4626 using RPPA and NGS analyses, are presented in Figure 4E.

Sym015 overcomes emibetuzumab resistance in vivo mainly due to ADCC

Due to the somewhat limited effect of PI3Kinase pathway inhibition (Fig. 4), a different strategy to overcome emibetuzumab-resistance was considered. It was speculated that the maintained high level of pMET in 4626 (Fig. 2E), which was not due to autocrine production of HGF (Suppl. Table S5) or genetic alterations, as no amplification or mutation of MET was detected, could indicate that this model retained sensitivity to MET-inhibtion by other more effective means. Because SNU-5 was sensitive to MET-targeting by Sym015 and failed to develop resistance to the antibody mixture in vivo (Fig. 1), we treated the emibetuzumab-resistant cell lines with Sym015. Notably, Sym015 did not affect the metabolic activity of 7333 and 4626 in vitro (Suppl. Fig. S6) nor did the combination of emibetuzumab or Sym015 with PI-103 treatment increase their sensitivity compared with PI-103 alone (Suppl. Table S4). Due to the superiority of Sym015 over emibetuzumab in vivo (Fig. 1C-D), and previous reports that Sym015 induces secondary effector functions [19,20], it was speculated that this might also contribute to the effect of Sym015 in resistant settings. The ability of Sym015 to induce ADCC was therefore investigated. In contrast to emibetuzumab, Sym015 induced ADCC in SNU-5 cells and the two emibetuzumab-resistant cell lines (Fig. 5A). Next, we investigated the effect of Sym015 in vivo in the two emibetuzumab-resistant models and compared it with the effect of Sym015 LALA, a Fc-muted variant of Sym015 not capable of inducing ADCC. Both 7333 and 4626 were sensitive to treatment with Sym015 in vivo, and were less sensitive or resistant to Sym015 LALA (Fig. 5B). This is in contrast to the effect in SNU-5, where Sym015 and

Sym015 LALA were equally effective (Suppl Fig. S7), likely due to the strong inherent MET dependency of SNU-5. Furthermore, compared with emibetuzumab, both Sym015 and Sym015 LALA resulted in decreased MET, pMET, and pERK1/2 levels in 4626 tumors, whereas the effect was less pronounced in 7333 tumors upon Sym015 or Sym015 LALA treatment (Fig. 5C). Together, these findings indicate that acquired emibetuzumab resistance may be overcome by treatment with a MET-targeting antibody mixture such as Sym015 likely due to triggering of secondary effector functions.

Discussion

In recent years, many MET-targeting agents have entered clinical development and a number of studies have demonstrated efficacy of MET-blockade in clinical settings [2,18,42]. Emibetuzumab and Sym015 are antibody therapeutics in clinical development that have demonstrated promising results in early clinical studies and preclinical studies, respectively [18–20]. In this study, we generated resistant models from a MET-amplified and MET dependent cell line after long-term continuous exposure to an analogue of emibetuzumab in vivo. The two derived resistant cell lines were characterized through phosphoproteomic and genomic profiling revealing distinct resistance and growth profiles, illustrating the potential diversity in resistance mechanisms evolving from a single MET-dependent model in response to long-term treatment with an anti-MET antibody. Despite diverse mechanisms of resistances, these results indicate that antibody-induced ADCC can overcome resistance in vivo, and that ADCC might play a role in treating patients with acquired resistance to single MET-targeting antibodies with an antibody mixture such as Sym015.

Analysis of activation of the protein signaling architecture using RPPA requires very small amounts of lysates, which makes it ideal to combine with LCM where pure tumor cell populations are isolated [36,37]. However, biological replicates of LCM tumor samples could not be obtained (Fig. 3D), because each resistant tumor evolved as a single independent event. Instead, cell lines were generated from each resistant tumor to serve as a more statistically powerful in vitro tool to study mechanisms of resistance, provided that the derived cells retain the characteristics of their parental tumors. Subsequently, results from the cell line RPPA analyses was compared with the tumor RPPA results. As shown in Figure 3 and Supplementary Tables S2 and S3, many of the phosphoproteomic changes observed in the cell lines were also detected in the tumors, confirming that the generated cell lines resembled the resistant tumors and suggesting that the cell lines have retained the rewired signaling profile of the resistant tumors.

An interesting finding from the RPPA analysis is that both resistant cell lines demonstrate elevated AKT and S6RP activation (Fig. 3), indicating increased signaling through the PI3Kinase pathway. This correlates with a previous study where increased PI3K p110α gene expression was observed in a SNU-5 model resistant to MET TKI treatment [34]. Together, these findings suggest that PI3Kinase is likely to be involved in resistance to MET-targeting therapeutics, indicating the potential of combined MET/PI3Kinase inhibition as an effective therapeutic strategy. However, combining PI3K- and MET-targeting did not add to the efficacy of PI-103 alone. In the current

study only 4626 was sensitive to PI3Kinase pathway inhibition while both resistant cell lines were sensitive to mTOR-targeting (Fig. 4A and Suppl. Table S4). This difference might be explained by loss of the tumor suppressor PTEN in 7333, which results in increased AKT phosphorylation and may thereby render the cells resistant to PI3Kinase/AKT-targeting, in line with previous clinical findings [43]. The PTEN deletion could also contribute to escape from treatment in other ways, for example through delayed intracellular receptor sorting and degradation as observed for EGFR [44]. Furthermore, AXL and VEGFR2 levels were upregulated in 7333 cells (Fig. 3), but this did not translate into sensitivity to TKIs targeting either protein (Suppl. Table S4). This lack of sensitivity to RTK-targeting may again be due to the loss of PTEN downstream of AXL and VEGFR2, which makes the cells less dependent of growth factor stimulation and activation to grow. RPPA and Simple Western analyses demonstrated that pPTEN and PTEN levels were downregulated in the 7333 tumor compared with control-treated tumors (Fig. 3D and 4D). This suggests that the PTEN deletion was already present in the resistant tumor prior to cell line generation.

The slower growth of 4626 compared with SNU-5 and 7333 (Fig. 2B and C), together with the increased activation of MET in 4626 cells (Fig. 2D), suggests that 4626 still depends partly on MET. In fact, both 7333 and 4626 were still sensitive to treatment with crizotinib and cabozantinib (Suppl. Table 4), which could indicate sustained MET-dependency, though to a lesser extend than in SNU-5. This finding is in line with observations reported for MET-amplified lung cancer cell lines resistant to MET-blockade by DN30 which remained MET-addicted [35]. However, crizotinib and cabozantinib also target other RTKs, so it cannot be ruled out that the sensitivity of 7333 and 4626 to these TKIs is due to non-MET related inhibition. Nevertheless, escape from treatment through a mutation in the emibetuzumab binding site on MET was suspected, as single targeting antibodies like emibetuzumab might be more likely to lose efficacy due to single mutations in binding sites than a mixture of antibodies like Sym015. However, in this study, the sustained MET activation in 4626 could not be explained by MET amplification or mutation in MET nor by increased HGF production, contrasting with previous findings showing increased MET gene copy number and increased MET activation as mechanisms of resistance to MET-inhibition [30,35]. Instead it was speculated that the increased MET activation and increased signaling throught the PI3K pathway was due to increased receptor crosstalk between MET and other RTKs [3–6], indicated by increased levels of phosphorylated RTKs such as VEGFR2, RET, AXL, and the HER- family (Suppl. Table S3). This is supported by the ERBB3 copy number gain observed in 4626 cells (Fig. 4C), together with increased HER3 levels in SNU-5 cells upon emibetuzumab treatment

(Suppl. Fig. S5), indicating increased cross-talk between MET and HER3, possibly through heterodimerization [6]. However, in 4626 VEGFR2, RET, AXL, or HER3-targeting in combination with PI3Kinase inhibition was equally effective to single agents. Nevertheless, it is likely that the increased signaling through the PI3Kinase pathway involves more RTKs contributing to the resistance. Finally, amplification of the transcription factor MYC was also observed in 4626, and may be a critical factor leading to increased signaling through the PI3Kinase signaling pathway (Fig. 4E), via upregulated expression of multiple RTKs and treatment escape. Indeed, MYC has been shown to mediate MET-addicted cell growth and to contribute to resistance to a MET- targeting TKI in lung cancer cells, while also being associated with escape from PI3Kinase/mTOR inhibition [33,45].

Antibody mixtures targeting more than one epitope on a single RTK have previously been demonstrated to more effectively inhibit tumor growth in vivo compared to single antibodies [46– 48]. This was also observed here, where acquired resistance to emibetuzumab, but not Sym015, emerged in vivo (Fig. 1C-D). These findings correlate with previous studies showing efficacy of Sym015 in emibetuzumab-resistant models and a model with both MET and ERBB2 amplification [20, 51]. The superior efficacy of antibody mixtures may be due to increased receptor crosslinking, facilitating increased receptor degradation [48,49], and it can be speculated if this mechanism of Sym015 contributes to delayed development of resistance in SNU-5. Another mechanism that distinguishes emibetuzumab and Sym015 is the ability to induce ADCC (Fig. 5). Emibetuzumab is an ADCC-silent IgG4 antibody, whereas the IgG1 backbone of the Sym015 antibodies is capable of inducing ADCC. Sym015 and the Fc muted Sym015 LALA variant were equally effective at eradicating tumors in parental SNU-5 xenografts (Suppl. Fig. S7), indicating that the ADCC effect does not play a role in the parental SNU-5 setting. However, in the resistant settings of 7333 and 4626, ADCC plays an important role, as indicated by the difference in efficacy between Sym015 and Sym015 LALA in these xenograft models (Fig. 5B). In 7333, ADCC seems to be the primary mode of action in vivo, whereas receptor downmodulation contributes less to efficacy. This is shown by the fact that Sym015 was more effective at inhibiting tumor growth than Sym015 LALA (Fig. 5B). Interestingly, the lack of significant receptor downmodulation or signaling inhibition in 7333 upon treatment with Sym015 and Sym015 LALA (Fig. 5C), could indicate that this model can grow relatively independently of MET signaling or that it has found a way to resist Sym015 and retain MET expression and activation. Furthermore, the lack of tumor regression in 7333 may be explained by the tumor inhibition being solely driven by ADCC. Since 7333 tumors grow faster

than 4626 (Fig. 2C), ADCC alone may not induce cell death to a degree sufficient to cause tumor shrinkage. In contrast, Sym015 LALA had some tumor growth inhibitory effect in 4626, most likely owing to receptor downregulation and signaling inhibition, as supported by the ability of both Sym015 and Sym015 LALA to effectively downregulate tumor MET, pMET, and pERK1/2 protein levels (Fig. 5C). Both of these emibetuzumab-resistant cases are in contrast to the parental SNU-5 setting, where Sym015 and Sym015 LALA were equally effective in eradicating the tumors (Suppl. Fig. S7), thus underlining the changes in the resistant compared to the parental setting and the broad mode of action of Sym015.

Antibodies used in the clinic are often combined with chemotherapy, which may postpone development of resistance, but combination of MET-targeting antibodies and chemotherapy has only received little attention so far [15]. However, the recently failed phase III study evaluating chemotherapy and trastuzumab+/-pertuzumab in HER2+ gastroesophageal adenocarcinoma [50], indicates that not all combinations are good combinations, and thus underlines the need for rational identification of combination partners to targeted agents.

In conclusion, our long-term in vivo modeling of acquired resistance to the MET-targeting antibody emibetuzumab identified two potential mechanisms of resistance both involving PI3Kinase pathway activation. One owing to PTEN loss and the other to increased RTK activation, potentially through MYC and ERBB3 copy number gains. Furthermore, the antibody mixture Sym015 effectively prevented and overcame these resistances, due to its broader mechanism of action. Although only one MET-amplified model was used for generation of resistance in this study, these findings add important information to the limited knowledge of mechanisms of resistance to MET-targeting antibodies. Furthermore, they imply that more effective targeting of MET with antibody mixtures may provide more effective MET inhibition than single antibodies such as emibetuzumab.

Acknowledgments

We acknowledge statistical support from Linlin Luo and Walid Chatila, who were supervised by Dr. Boca and proof-read of the manuscript by Stephen Gilliver (TFSCRO, Sweden).

References

[1] Trusolino L, Bertotti A, Comoglio PM. MET signalling: principles and functions in development, organ regeneration and cancer. Nat Rev Mol Cell Biol 2010;11:834–48.

[2] Bradley CA, Salto-tellez M, Laurent-puig P, Bardelli A. Targeting c‑MET in gastrointestinal tumours: rationale, opportunities and challenges. Nat Rev Clin Oncol 2017;14:562–76.

[3] Jo M, Stolz DB, Esplen JE, Dorko K, Michalopoulos GK, Strom SC. Cross-talk between epidermal growth factor receptor and c-Met signal pathways in transformed cells. JBiolChem 2000;275:8806– 11.

[4] Sheng-Hui Lan, Shan-Ying Wu, Giri Raghavaraju N-HC, Liu and H-S. The Crosstalk of c-MET with Related Receptor Tyrosine Kinases in Urothelial Bladder Cancer. Adv Sci Eval Bl Cancer Mol Basis Diagnosis Treat 2013:21–39.

[5] Qian F, Engst S, Yamaguchi K, Yu P, Won K-A, Mock L, et al. Inhibition of tumor cell growth, invasion, and metastasis by EXEL-2880 (XL880, GSK1363089), a novel inhibitor of HGF and VEGF receptor tyrosine kinases. Cancer Res 2009;69:8009–16.

[6] Tanizaki J, Okamoto I, Sakai K, Nakagawa K. Differential roles of trans-phosphorylated EGFR , HER2 , HER3 , and RET as heterodimerisation partners of MET in lung cancer with MET amplification. Br J Cancer 2011;105:807–13.

[7] Inokuchi M, Otsuki S, Fujimori Y, Sato Y, Nakagawa M, Kojima K. Clinical significance of MET in gastric cancer. World J Gastrointest Oncol 2015;7:317–27.

[8] Wadhwa R, Taketa T, Sudo K, Blum-Murphy M, Ajani J a. Ramucirumab: a novel antiangiogenic agent. Future Oncol 2013;9:789–95.

[9] Boku N. HER2-positive gastric cancer. Gastric Cancer 2014;17:1–12.

[10] Yano S, Wang W, Li Q, Matsumoto K, Sakurama H, Nakamura T, et al. Hepatocyte growth factor induces gefitinib resistance of lung adenocarcinoma with epidermal growth factor receptor-activating mutations. Cancer Res 2008;68:9479–87.

[11] Bean J, Brennan C, Shih J, Riely G, Viale A, Wang L, et al. MET amplification occurs with or without T790M mutations in EGFR mutant lung tumors with acquired resistance to gefitinib or erlotinib. PNAS 2007;104:20932–7.

[12] Okamoto W, Okamoto I, Arao T, Kuwata K, Hatashita E, Yamaguchi H, et al. Antitumor action of the MET tyrosine kinase inhibitor crizotinib (PF-02341066) in gastric cancer positive for MET amplification. Mol Cancer Ther 2012;11:1557–64.

[13] Smolen GA, Sordella R, Muir B, Mohapatra G, Barmettler A, Archibald H, et al. Amplification of MET may identify a subset of cancers with extreme sensitivity to the selective tyrosine kinase inhibitor PHA-665752. Proc Natl Acad Sci United States 2006;103:2316–21.

[14] Kwak EL, LoRusso P, Hamid O, Janku F, Kittaneh M, Virgil D, et al. Clinical activity of AMG 337, an oral MET kinase inhibitor, in adult patients (pts) with MET-amplified gastroesophageal junction (GEJ), gastric (G), or esophageal (E) cancer. J Clin Oncol 2015;33.

[15] Liu L, Zeng W, Wortinger MA, Yan SB, Cornwell P, Peek VL, et al. LY2875358, a neutralizing and internalizing anti-MET bivalent antibody, inhibits HGF-dependent and HGF-independent MET

activation and tumor growth. Clin Cancer Res 2014;20:6059–70.

[16] Merchant M, Ma X, Maun HR, Zheng Z, Peng J, Romero M, et al. Monovalent antibody design and mechanism of action of onartuzumab, a MET antagonist with anti-tumor activity as a therapeutic agent. Proc Natl Acad Sci U S A 2013;110:E2987-96.

[17] Nishio M, Horiike A, Nokihara H, Horinouchi H, Nakamichi S, Wakui H, et al. Phase I study of the anti-MET antibody onartuzumab in patients with solid tumors and MET-positive lung cancer. Invest New Drugs 2015;33:632–40.

[18] Rosen LS, Goldman JW, Algazi AP, Turner PK, Moser B, Hu T, et al. A First-in-Human Phase I Study of a Bivalent MET Antibody, Emibetuzumab (LY2875358), as Monotherapy and in Combination with Erlotinib in Advanced Cancer. Clin Cancer Res 2016;23:1910–9.

[19] Grandal MM, Poulsen TT, Koefoed K, Havrylov S, Dahlman A, Galler GR, et al. Simultaneous targeting of two distinct epitopes on MET effectively inhibits MET- and HGF-driven tumor growth by multiple mechanisms. Mol Cancer Ther 2017;16:2780–91.

[20] Poulsen TT, Grandal MM, Skarved NJØ, Hald R, Alifrangis L, Koefoed K, et al. Sym015, a highly effective antibody mixture for targeting MET-amplified tumors. Clincal Cancer Res 2017;23:5923– 35.

[21] Petrelli A, Circosta P, Granziero L, Mazzone M, Pisacane A, Fenoglio S, et al. Ab-induced ectodomain shedding mediates hepatocyte growth factor receptor down-regulation and hampers biological activity. PNAS 2006;103:5090–5.

[22] Wang J, Goetsch L, Tucker L, Zhang Q, Gonzalez A, Vaidya KS, et al. Anti-c-Met monoclonal antibody ABT-700 breaks oncogene addiction in tumors with MET amplification. BMC Cancer 2016;16:105.

[23] Shah MA, Wainberg ZA, Catenacci DVT, Hochster HS, Ford J, Kunz P, et al. Phase II Study Evaluating 2 Dosing Schedules of Oral Foretinib (GSK1363089), cMET/VEGFR2 Inhibitor, in Patients with Metastatic Gastric Cancer. PLoS One 2013;8:1–9.

[24] Camidge DR, Moran T, Demedts I, Grosch H, Di Mercurio J-P, Mileham KF, et al. A randomized, open-label, phase 2 study of emibetuzumab plus erlotinib (LY+E) and emibetuzumab monotherapy (LY) in patients with acquired resistance to erlotinib and MET diagnostic positive (MET Dx+) metastatic NSCLC. ASCO Annu. Meet., 2016.

[25] Chung HC, Satoh T, Oh D-Y, Park SH, Kadowaki S, Wacheck V, et al. A non-randomized, open- label, single-arm, phase 2 study of LY2875358 in Asian patients with MET diagnostic positive, advanced gastric cancer. Mol. Targets Cancer Ther., 2015.

[26] Shah MA, Bang Y-J, Lordick F, Alsina M, Chen M, Hack SP, et al. Effect of Fluorouracil, Leucovorin, and Oxaliplatin With or Without Onartuzumab in HER2-Negative, MET-Positive Gastroesophageal Adenocarcinoma The METGastric Randomized Clinical Trial. JAMA Oncol 2017;3:620–7.

[27] Catenacci D, Tebbutt N, Davidenko I, Murad A, Al-Batran S, Ilson A, et al. Rilotumumab plus epirubicin, cisplatin, and capecitabine as first-line therapy in advanced MET-positive gastric or gastro-oesophageal junction cancer (RILOMET-1): a randomised, double-blind, placebo-controlled, phase 3 trial. Lancet Oncol 2017.

[28] Qi J, Mctigue MA, Rogers A, Lifshits E, Christensen JG. Multiple Mutations and Bypass Mechanisms Can Contribute to Development of Acquired Resistance to MET Inhibitors. Cancer Res

2011:1081–91.

[29] Corso S, Ghiso E, Cepero V, Sierra JR, Migliore C, Bertotti A, et al. Activation of HER family members in gastric carcinoma cells mediates resistance to MET inhibition. Mol Cancer 2010;9:1–13.

[30] Cepero V, Sierra JR, Corso S, Ghiso E, Casorzo L, Perera T, et al. MET and KRAS gene amplification mediates acquired resistance to MET tyrosine kinase inhibitors. Cancer Res 2010;70:7580–90.

[31] Kentsis A, Reed C, Rice KL, Sanda T, Rodig SJ, Tholouli E, et al. Autocrine activation of the MET receptor tyrosine kinase in acute myeloid leukemia. Nat Med 2012;18:1118–22. doi:10.1038/nm.2819.Autocrine.

[32] Henry RE, Barry ER, Castriotta L, Ladd B, Beran G, Ren Y, et al. Acquired savolitinib resistance in non-small cell lung cancer arises via multiple mechanisms that converge on MET- independent mTOR and MYC activation. Oncotarget 2016;7:1–16.

[33] Shen A, Wang L, Huang M, Sun J, Chen Y, Shen YY, et al. C-Myc alterations confer therapeutic response and acquired resistance to c-Met inhibitors in MET-addicted cancers. Cancer Res 2015;75:4548–59.

[34] Ji F, Liu X, Wu Y, Fang X, Huang G. Overexpression of PI3K p110a contributes to acquired resistance to MET inhibitor, in MET-amplified SNU-5 gastric xenografts. Drug Des Devel Ther 2015;9:5697–704.

[35] Martin V, Corso S, Comoglio PM, Giordano S. Increase of MET gene copy number confers resistance to a monovalent MET antibody and establishes drug dependence. Mol Oncol 2014;8:1561– 74.

[36] Paweletz CP, Charboneau L, Bichsel VE, Simone NL, Chen T, Gillespie JW, et al. Reverse phase protein microarrays which capture disease progression show activation of pro-survival pathways at the cancer invasion front. Oncogene 2001;20:1981–9.

[37] Pierobon M, Wulfkuhle J, Liotta L, Petricoin E. Application of molecular technologies for phosphoproteomic analysis of clinical samples. Oncogene 2015;34:805–14.

[38] Smyth GK. Limma: Linear Models for Microarray Data. Bioinforma. Comput. Biol. Solut. Using R Bioconductor, 2005, p. 397–420.

[39] Smyth GK. Linear Models and Empirical Bayes Methods for Assessing Differential Expression in Microarray Experiments Linear Models and Empirical Bayes Methods for Assessing Differential Expression in Microarray Experiments. Stat Appl Genet Mol Biol 2004;3:1–25.

[40] Bahcall M, Sim T, Paweletz CP, Patel JD, Alden RS, Kuang Y, et al. Acquired MET D1228V Mutation and Resistance to MET Inhibition in Lung Cancer. Cancer Discov 2016;6:1334 LP-1341.

[41] Tseng Y-Y, Bagchi A. The PVT1-MYC duet in cancer. Mol Cell Oncol 2015;2:e974467-1–3.

[42] Nishio M, Horiike A, Murakami H, Yamamoto N, Kaneda H, Nakagawa K, et al. Phase I study of the HER3-targeted antibody patritumab (U3-1287) combined with erlotinib in Japanese patients with non-small cell lung. Lung Cancer 2015;88:275–81.

[43] Juric D, Castel P, Griffith M, Griffith OL, Won HH, Ellis H, et al. Convergent loss of PTEN leads to clinical resistance to a PI(3)Kalpha inhibitor. Nature 2015;518:240–4.

[44] Shinde SR, Maddika S. PTEN modulates EGFR late endocytic trafficking and degradation by

dephosphorylating Rab7. Nat Commun 2016;7:1–11.

[45] Ilic N, Utermark T, Widlund HR, Roberts TM. PI3K-targeted therapy can be evaded by gene amplification along the MYC-eukaryotic translation initiation factor 4E (eIF4E) axis. Proc Natl Acad Sci 2011;108:E699–708.

[46] Yamashita-Kashima Y, Iijima S, Yorozu K, Furugaki K, Kurasawa M, Ohta M, et al. Pertuzumab in Combination with Trastuzumab Shows Significantly Enhanced Antitumor Activity in HER2-Positive Human Gastric Cancer Xenograft Models. Clin Cancer Res 2011;17:5060–70.

[47] Pedersen MW, Jacobsen HJ, Koefoed K, Hey A, Pyke C, Haurum JS, et al. Sym004: A Novel Synergistic Anti-Epidermal Growth Factor Receptor Antibody Mixture with Superior Anticancer Efficacy. Cancer Res 2010;70:588–97.

[48] Pedersen MW, Jacobsen HJ, Koefoed K, Dahlman A, Kjaer I, Poulsen TT, et al. Targeting Three Distinct HER2 Domains with a Recombinant Antibody Mixture Overcomes Trastuzumab Resistance. Mol Cancer Ther 2015;14:669–80.

[49] Friedman LM, Rinon A, Schechter B, Lyass L, Lavi S, Bacus SS, et al. Synergistic down-regulation of receptor tyrosine kinases by combinations of mAbs: implications for cancer immunotherapy. Proc Natl Acad Sci U S A 2005;102:1915–20.

[50] Tabernero J, Hoff PM, Shen L, Ohtsu A, Shah MA, Cheng K, et al. 616O - Pertuzumab (P) + trastuzumab (H) + chemotherapy (CT) for HER2-positive metastatic gastric or gastro-oesophageal junction cancer (mGC/GEJC). ESMO Congr., 2017.

[51] Bouquin T, Pedersen MW, Jacobsen HJ, Poulsen TT, Grandal MM, Koefoed K, Kragh M, Eriksen KW, and Conrotto P. Anti-MET Antibodies and Compositions, patent issued 2016, WO 2016/042412 A1.

Figure 1. MET-dependent SNU-5 cells develop resistance to emibetuzumab, but not Sym015, upon long-term treatment in vivo. A. Dose-dependent sensitivity of SNU-5 cells to treatment with emibetuzumab or Sym015 compared with cells treated with a negative control antibody for 96 h, as measured using WST-1. B. MET and pMET(Y1234/1235) levels in SNU-5 cells upon treatment with 50 µg/mL emibetuzumab, 50 µg/mL Sym015 or a negative control antibody for 24 h, as measured by Simple Western. Data represent means ± SEM (n=3). *p < 0.05, **p < 0.01. Protein levels are normalized to pan-actin and are presented as percentage signal intensity relative to SNU-5 negative control. C-D. Tumor growth curves for SNU-5-xenografted BALB/c nu/nu mice dosed with 50 mg/kg emibetuzumab (green, n=16), 50 mg/kg Sym015 (red, n=16), or vehicle (black, n=8) by injection i.p. 3 times/week starting from day 13. Tumor volumes were measured using a caliper 3 times/week for >200 days. Three tumors with resistance to emibetuzumab (7333, 7360, and 4626) were isolated for characterization, and cell lines were successfully generated from two of the isolated tumors (7333 and 4626). E. MET, pMET(Y1234/1235), and pERK1/2(T202/Y204) levels in control-treated SNU-5 tumors (n=2) and resistant tumors, as measured by Simple Western. MET and pMET levels are normalized to pan-actin and are presented as percentage signal intensity relative to SNU-5 negative control. pERK levels are normalized to ERK1/2 and pan-actin levels and are presented as percentage signal intensity relative to SNU-5 negative control.

Figure 2. In vivo-derived emibetuzumab-resistant cell lines retain resistance in vitro. A. Dose- dependent response of 7333 and 4626 to treatment with emibetuzumab compared with cells treated with a negative control antibody for 96 h, as measured using WST-1 (compare with Fig. 1A for effect in non-resistant SNU-5 cells). Data represent means ±SEM (n=3) B. Cell proliferation based on live cell imaging of SNU-5, 7333, and 4626 cells with or without addition of 50 µg/mL emibetuzumab (Emi) for 160 h. Data represent means ±SEM (n=6) and are normalized to the individual cell line’s starting confluency. Representative pictures of the cell lines show confluency at 160 h. One-way ANOVA was used for statistical comparizon, *p < 0.05. C. BALB/c nu/nu mice inoculated with SNU-5, 7333, or 4626. 7333 and 4626 were dosed i.p. with 50 mg/kg emibetuzumab 3 times/week from the day of cell inoculation. SNU-5 was not dosed. Data represent means ±SEM (n=3-7). D. MET, pMET(Y1234/1235), and pERK1/2(T202/Y204) levels in SNU-5 cells with and without emibetuzumab (Emi) treatment (50 µg/mL, 24 h) and in the two resistant cell lines under continuous emibetuzumab treatment pressure (50 µg/mL), as measured by Simple Western. MET and pMET levels are normalized to pan-actin and are presented as percentage signal intensity relative to SNU-5 negative control. pERK levels are normalized to ERK1/2 and pan-actin

levels and are presented as percentage of signal intensity relative to SNU-5 negative control. Data represent means ± SEM (n=3). Significance illustrates effect compared with negative control treated SNU-5: **p < 0.01.

Figure 3. Protein and signaling pathway activation mapping reveals two different resistance profiles. A. Volcano plots of phosphoproteomic profiles of emibetuzumab-resistant cell lines (7333 and 4626), showing the relationship between fold change (x-axis) and significance (y-axis) for each of the 162 protein or phosphoprotein levels detected in the RPPA cell line analysis (n=3). The baseline is SNU-5 cells treated with emibetuzumab. B. Network analysis showing the direct interactions of the outliers marked with red (downregulated) or green (upregulated) in the volcano plots (A). C. VEGFR2, AXL, pS6 ribosomal protein(S235/236), and Ras protein levels in SNU-5 with or without emibetuzumab treatment (50 µg/mL, 24 h) and in the two emibetuzumab-resistant SNU-5 cell lines under continuous emibetuzumab treatment pressure (50 µg/mL), as measured by Simple Western. Data represent means ± SEM (n=3). Significance illustrates effect compared with emibetuzumab-treated SNU-5 cells: *p < 0.05, ****p < 0.0001. Protein levels are normalized to pan-actin and are presented as percentage signal intensity relative to SNU-5 negative control. pS6RP levels are normalized to S6RP and pan-actin levels and are presented as percentage of signal intensity relative to SNU-5 negative control. D. Heatmap with hierarchical clustering showing relative up- and downregulation of selected RPPA analysis results from LCM samples of control- treated SNU-5 (n=2), 7333, and 4626 tumors.

Figure 4. PI3Kinase pathway inhibition partly overcomes emibetuzumab resistance. A. Metabolic activity of SNU-5, 7333, and 4626 upon treatment for 96 h with 312.5 nM PI-103 or 2.5 nM sirolimus compared with untreated and DMSO-treated control cells, as measured using WST-1. Results are based on means of triplicates in two or three individual experiments. Significance illustrates effect compared with DMSO negative control: *p < 0.05, **p < 0.01, and p < 0.001. B. pPI3Kinase p85(Y458), pmTOR(S2448), and pS6 ribosomal protein(S235/236) levels in 7333 and 4626 cells upon treatment for 24 h with 312.5nM PI-103 or 2nM sirolimus compared with DMSO negative control, as measured by Simple Western. Data represent means ± SEM (n=3). *p < 0.05, **p < 0.01, ***p < 0.001. Protein levels are normalized to total protein levels and pan-actin and are presented as percentage signal intensity relative to the corresponding DMSO control. C. Log2 ratios for MYC, PVT1, and ERBB3 copy number gains in 4626 cells, and PTEN copy number deletion in 7333 cells. Each point represents a sequence that is either gained or deleted compared to SNU-5. D.

MYC, HER3, PTEN, and pPTEN(S380) levels in control-treated SNU-5 tumors (n=2) and resistant tumors, as measured by Simple Western. Protein levels are normalized to pan-actin and are presented as percentage signal intensity relative to SNU-5 negative control. E. Proposed mechanism of resistance via PI3Kinase pathway activation, mediated by PTEN loss in 7333 and by MYC and ErbB3 copy number gains and increased RTK phosphorylation in 4626.

Figure 5. Sym015 overcomes emibetuzumab resistance in vivo mainly due to ADCC. A. In vitro ADCC in SNU-5, 7333, and 4626 upon coincubation of Sym015, Sym015 LALA, emibetuzumab, or negative control antibody for 4 h with PBMCs. ADCC lysis is presented as a percentage of maximum lysis, with spontaneous lysis subtracted. Data represent means ±SEM (n=3). B. BALB/c nu/nu mice inoculated with 7333 or 4626 were dosed i.p. with 50 mg/kg emibetuzumab 2 times/week from the day of cell inoculation. Randomization of emibetuzumab, Sym015, and Sym015 LALA treatments (50 mg/kg 2 times/week) started at an average tumor size of 114 mm3 in 7333 and 145 mm3 in 4626. Data represent means ±SEM (n=7-8). Two tumors from each group were isolated where indicated. C. MET, pMET(Y1234/1235), and pERK1/2(T202/Y204) protein levels in 7333 and 4626 tumors from the mice described in B. isolated the day after the second dose of emibetuzumab, Sym015, or Sym015 LALA, as measured by Simple Western. Data represent means ± SEM (n=2). *p < 0.05, ***p < 0.001. MET and pMET levels are normalized to pan-actin and are presented as percentage signal intensity relative to emibetuzumab-treated tumors; pERK1/2 levels are normalized to ERK1/2 and pan-actin and are presented as percentage of signal intensity relative to emibetuzumab-treated tumors.

Acquired resistance to a MET antibody in vivo can be overcome by the MET antibody mixture Sym015

Sofie Ellebæk Pollmann, Valerie S. Calvert, Shruti Rao, et al.

Mol Cancer Ther Published OnlineFirst March 15, 2018.

Updated version Access the most recent version of this article at: doi:10.1158/1535-7163.MCT-17-0787

Supplementary Access the most recent supplemental material at: Material http://mct.aacrjournals.org/content/suppl/2018/03/15/1535-7163.MCT-17-0787.DC1

Author Author manuscripts have been peer reviewed and accepted for publication but have not yet Manuscript been edited.

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://mct.aacrjournals.org/content/early/2018/03/15/1535-7163.MCT-17-0787. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.