Depleting MET-Expressing Tumor Cells by ADCC Provides a Therapeutic Advantage over Inhibiting HGF/MET Signaling

MET, the product of the c-MET proto-oncogene, is a transmembrane tyrosine kinase that is expressed by the majority of epithelial tissues and by endothelial, muscle, neuronal, hematopoietic, and immune cells (1). The high-affinity ligand of MET, hepatocyte growth factor (HGF), is a pleiotropic cytokine that is secreted by cells of mesenchymal origin, typically fibroblasts and macrophages (2). HGF signaling through MET promotes a peculiar biologic program that results from the fine orchestration of complementary biologic activities, including promigratory, mitogenic, and antiapoptotic cues (3). This process, known collectively as invasive growth, is essential for embryogenesis, during which HGF/MET signaling is responsible for migration, proliferation, and survival of various progenitor cells (4). In normal adult tissues, HGF-driven invasive growth is usually quiescent, but is resumed following tissue injury to promote wound healing and tissue regeneration (5).

Although HGF/MET signaling is tightly controlled in normal tissues, it is frequently subverted in tumors, which exploit the invasive growth program to grow within and to colonize the host organism (6). HGF overexpression in the tumor microenvironment correlates with increased tumor aggressiveness and invasion (7). HGF is also a powerful proangiogenic factor promoting tumor angiogenesis (8). Aberrant activation of MET is frequently observed in human cancer and can be due to different molecular mechanisms, including c-MET transcriptional activation, amplification or mutation, as well as autocrine or paracrine HGF stimulation (9). In the clinic, MET protein overexpression and c-MET gene amplification have been identified as negative prognostic markers for a variety of oncologic diseases, including lung (10), colon (11), breast (12), head and neck (13), and gastric (14) carcinoma, as well as multiple myeloma (15) and glioblastoma multiforme (16).

Several inhibitors have been developed so far to tackle HGF/MET activity in cancer, including small-molecule MET tyrosine kinase inhibitors, anti-HGF antibodies, and anti-MET antibodies (17–19). Surprisingly, and in striking contrast with the importance of HGF/MET signaling for survival of different cell types during embryogenesis, HGF- or MET-targeted drugs have, however, demonstrated cytostatic rather than cytotoxic activity on tumor cells, independently of their mechanism of action (20, 21). In fact, even in the most MET-dependent tumors displaying c-MET gene amplification, inhibition of MET activity leads to G₁ arrest and not to cell death, both in cellular studies and in xenografts (22). Consistent with this idea, HGF/MET signaling in cancer cells is typically associated with promotion of epithelial-to-mesenchymal transition (EMT) and with tumor progression rather than with tumor onset (1, 4, 6, 9).

To overcome this pharmacologic limitation, and to investigate whether MET-targeted cytotoxicity would confer a therapeutic advantage over plain HGF/MET inhibition, we engineered an antagonistic anti-MET antibody that, in addition to blocking both HGF-dependent and -independent MET activation, displays enhanced antibody-dependent cellular cytotoxicity (ADCC). At the same time, we also engineered an ADCC-dead version of the same antibody that inhibits HGF/MET signaling but is completely devoid of any effector function, and compared the antitumor and antimetastasis activity of this antibody pair in multiple mouse models of cancer. The results presented here suggest that killing MET-expressing cancer cells by ADCC is significantly more effective from a therapeutic viewpoint than simply inhibiting HGF/MET signaling.

A549, BxPC3, MDA-MB-231, NCI-H1437, NCI-H2122, NCI-H441, NCI-H596, Raji, SNU-5, TOV-112D, U87-MG, and 786-O cells were purchased from the ATCC. A498, L540, MKN-45, and U266 cells were purchased from the Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ). EBC-1 cells were purchased from the Japanese Collection of Research Bioresources. All cells were cultured according to the protocols provided by the supplier and used within 6 months after receipt or validated by short tandem repeat profiling using Cell ID System (Promega). LOC cells (23), immortalized human mammary fibroblasts (24) and CRCM tumors (25) have been described previously.

ADCC of MKN-45, NCI-H441, and A549 cells was determined with a standard ⁵¹Cr release assay (26) using human peripheral blood mononuclear cells from three different donors (Etablissement Français Du Sang) and a 50:1 effector:target ratio. All experiments were performed in triplicate. The percentage of lysis was analyzed versus antibody concentration, and the EC₅₀ and E_MAX values were calculated using Prism software (GraphPad). ADCC of MDA-MB-231 cells and of primary triple negative mammary carcinoma cells was analyzed using an ADCC Reporter Bioassay Kit (Promega). Single cell suspensions were stained with AldeFluor (Stemcell Technologies) and then sorted by FACS using a MoFlo XDP cell sorter (Beckman Coulter). ALDH-1⁺ cells were incubated with increasing ARGX-111 concentrations (0–667 nmol/L) and reporter effector cells (effector:target ratio 2.5:1), and ADCC was measured as described by the manufacturer. Data were analyzed and fit using Prism software (GraphPad).

U87-MG or MKN-45 cells (3 × 10⁶ cells/mouse) were injected s.c. into the right or left hind flank of 6- to 8-week-old CD-1 nude mice (Charles River Laboratories). When tumor volume approached 100 mm³, mice were randomly assigned to three treatment arms (U87-MG, 10 mice/group; MKN-45, 6 mice/group): irrelevant IgG1; WT52-E; and WT52-D. Antibodies were administered two times a week by i.p. injection, and tumor volume was monitored over time as described previously (24). At autopsy, tumor weight was determined on freshly explanted tumors (U87-MG). For analysis of MET and phospho-MET levels, mice bearing MKN-45 tumors were treated with a single bolus of WT52-E or irrelevant IgG1 (3 mice/group) and tumors were extracted for analysis 24 hours later. Tumor sections were analyzed by immunohistochemistry using anti-MET or anti–phospho-MET antibodies (both from LSBio).

For the neoadjuvant setting, MDA-MB-231-luc cells were mixed 4:1 with immortalized human mammary fibroblasts and then injected bilaterally (2.5 × 10⁶ cells/injection site) into the mammary fat pad of 6-week-old female CB17 SCID mice (Charles River Laboratories) as described previously (24). After 3 weeks, mice were stratified on the basis of the tumor volume and randomly assigned to the following treatment arms (n = 8): IgG1; WT52-E; WT52-D. Antibodies were administered two times a week by i.p. injection at a dose of 5 mg/kg IgG1. Tumor growth was monitored over time as described previously (24). After 4 weeks of treatment, tumors were surgically removed, and neoadjuvant therapy was interrupted. Two weeks after surgery, mice were injected with luciferin, sacrificed, and subjected to autopsy. Metastatic dissemination was determined by bioluminescence analysis of isolated lungs and femurs as described previously (24). The number of circulating tumor cells (CTC) was determined at the indicated time points by measuring luciferase activity on mouse blood. Approximately 100 μL of whole blood was cleared from erythrocytes using red blood cell lysis buffer (BD Biosciences), and CTCs were collected by centrifugation. The cell pellet was washed in PBS and then lysed in 20 μL luciferase assay buffer (Promega). Luciferase activity was measured on 10 μL of cell lysate using a Luciferase Reporter Assay System kit (Promega). Samples were analyzed with a GloMax 96 Microplate Luminometer (Promega). Estimation of CTC number was calculated using a standard curve of luciferase-expressing MDA-MB-231 cells. For the adjuvant setting, MDA-MB-231-luc cells were injected orthotopically into the mammary fat pad of 6-week-old female CB17 SCID mice along with human mammary fibroblasts as described above, and CTCs were determined at regular time intervals. Primary tumors were removed by surgery approximately 5 weeks after cell injection. Two days after surgery, mice were randomized on the basis of the CTC number and assigned to three treatment arms (10 mice/group) as above (IgG1, WT52-E, WT52-D; 5 mg/kg). After 4 weeks of treatment, mice were sacrificed and metastatic dissemination was assessed by bioluminescence analysis of whole body and isolated organs as described previously (24).

All procedures involving the use of animals and/or human samples were authorized by the competent authorities. Please see online Supplementary Methods for more information.

The supplementary online material section comprises Supplementary Figs. S1–S7, Supplementary Tables S1–S7 and Supplementary Methods.

We recently reported the generation of a wide panel of chimeric llama-human antagonistic anti-MET antibodies (24). The most antagonistic molecule within this panel (WT52) blocks both HGF-dependent MET activity (by competing with HGF for binding to MET; Supplementary Fig. S1A and S1B) and HGF-independent MET activity (by promoting MET internalization; Supplementary Fig. S1C and S1D). WT52 was engineered to an ADCC-enhanced version (WT52-E) by introducing the S239D/I332E amino acid substitutions in the CH2 domain (27). As a control, an ADCC-inactive (“dead”) mutant version of WT52 (WT52-D) was also engineered that contained the E233P/L234V/L235A amino acid substitutions in the CH2 domain (26). ELISA analysis of Fc receptor binding revealed that WT52-E displays increased affinity for human FcγRIIIa and its mouse orthologue FcγRIV, as well as for human and mouse FcγRI compared with WT52 (Supplementary Table S1). WT52-D displayed very low affinity for human FcγRI and no affinity for human FcγRIIIa, mouse FcγRI or mouse FcγRIV. In in vitro ADCC assays using effector cells from three distinct donors, WT52-E displayed a tumor cell killing activity directly proportional to MET expression levels and invariably superior to that of WT52 (Supplementary Table S2). WT52-D did not display any ADCC activity in any cell line tested.

To determine whether ADCC enhancement provides an advantage over plain HGF/MET inhibition in mouse models of cancer, we tested the WT52-E–WT52-D antibody pair in both HGF-dependent and -independent xenograft systems. U87-MG cells, which display an autocrine HGF/MET loop (28), were injected s.c. into CD-1 nude mice [that have functional natural killer (NK) cells and are, therefore, ADCC competent] to induce the formation of experimental tumors. Tumor-bearing mice were randomly assigned to three treatment arms (irrelevant IgG1; WT52-E; WT52-D), and antibodies were administered at a dose of 5 mg/kg. Tumor burden analysis revealed that WT52-E and WT52-D inhibited tumor growth by 88% (P = 0.0003) and 51% (P = 0.0159), respectively (Fig. 1A). At the end of the experiment, the mean weight of control tumors was 1819 ± 331 mg, whereas that of antibody-treated tumors was 295 ± 103 mg (WT52-E) and 930 ± 215 mg (WT52-D; Fig. 1B). The activity of WT52-E and WT52-D was also tested in an HGF-independent mouse model of cancer. MKN-45 human gastric carcinoma cells, which display constitutive, ligand-independent MET activation (29), were injected s.c. into CD-1 nude mice. Tumor-bearing mice were randomized into three treatment arms as above (IgG1; WT52-E; WT52-D), and antibodies were administered at a dose of 1 mg/kg. In this model, WT52-E and WT52-D inhibited tumor growth by 85% (P = 0.0018) and 54% (P = 0.0355), respectively (Fig. 1C). In a separate experiment using CD-1 nude mice bearing MKN-45 tumors, a single bolus of WT52-E (5 mg/kg) resulted in a significant reduction of both phospho-MET and total MET levels (Fig. 1D). Taken together, these results suggest that ADCC enhancement provides a therapeutic advantage over plain HGF/MET signaling inhibition in both HGF-dependent and -independent xenograft models.

We recently reported that anti-MET antagonistic antibodies inhibit tumor growth and suppress metastatic dissemination in an orthotopic model of mammary carcinoma (24). To explore the therapeutic potential of MET-targeted ADCC in this model, we injected MDA-MB-231 human triple-negative mammary carcinoma cells expressing luciferase (MDA-MB-231-luc) into the mammary fat pad of SCID mice (which have functional NK cells and are, therefore, ADCC competent). Because mouse HGF does not activate human MET, tumor cells were injected along with HGF-secreting, immortalized human mammary fibroblasts (24). Tumor-bearing mice were randomly assigned to three treatment arms as above (IgG1; WT52-E; WT52-D), and antibodies were administered at a dose of 5 mg/kg. After 4 weeks of treatment, tumors were surgically removed, and neoadjuvant therapy was interrupted. Two weeks after surgery, mice were injected with luciferin, sacrificed, and subjected to autopsy. Metastatic dissemination was determined by bioluminescence analysis of isolated lungs and femurs (Fig. 2A). This analysis revealed that WT52-E and WT52-D reduced tumor growth by 57% (P < 0.0001) and 20% (P = 0.0271), respectively (Fig. 2B). Interestingly, however, the effect of anti-MET antibodies was greater on metastases. In fact, WT52-E and WT52-D reduced lung metastases (Fig. 2C) by 91% (P = 0.0090) and 74% (P = 0.0294), respectively, and femur metastases (Fig. 2D) by 95% (P = 0.0047) and 79% (P = 0.0158). We also monitored the number of CTCs during the whole experiment by measuring luciferase activity in the blood. Before surgery, we observed neither a correlation between CTC number and tumor volume nor a significant difference in CTC number among the various experimental groups (Fig. 2E). However, interestingly, once the primary tumors were removed, mice treated with WT52-E displayed significantly reduced CTC number and metastases compared with both the WT52-D and the control group (Fig. 2F).

Metastasis onset following tumor resection is typically due to the seeding of CTCs at secondary anatomical sites. Unfortunately, surgery itself is often a cause of tumor cell spreading. Prompted by this consideration, we explored the therapeutic potential of MET-targeted ADCC in an adjuvant model of metastatic mammary carcinoma in which treatment started after tumor resection. MDA-MB-231-luc cells and HGF-secreting human fibroblasts were injected orthotopically into the mammary fat pad of SCID mice as described above, and CTCs were determined at regular time intervals. Primary tumors were removed by surgery approximately 5 weeks after cell injection. Two days after surgery, mice were randomized on the basis of the CTC number and assigned to three treatment arms as above (IgG1; WT52-E; WT52-D; 5 mg/kg). After 4 weeks of treatment, mice were sacrificed and metastatic dissemination was assessed by bioluminescence analysis of whole body and isolated organs (Fig. 3A). Consistent with our work hypothesis, CTC number burst soon after surgery and then returned to lower levels in approximately a week (Fig. 3B). However, although CTC number in mice receiving WT52-E remained low over time, it tended to climb up again in mice treated with control IgG1 or with WT52-D. At sacrifice, the WT52-E group displayed 89% less CTCs compared with the IgG1 arm (P = 0.0274) and 81% less CTCs compared with the WT52-D arm (P = 0.0383; Fig. 3C). In vivo imaging (Supplementary Fig. S2) revealed that WT52-E dramatically reduced whole body bioluminescence compared with both IgG1 (86%, P = 0.0035) and WT52-D (81%, P = 0.0397; Fig. 3D). Bioluminescence analysis of explanted lungs unveiled that WT52-E potently suppressed pulmonary metastases compared with both IgG1 (93%, P = 0.0361) and WT52-D (91%, P = 0.0499; Fig. 3E). Similarly, WT52-E–treated mice displayed reduced bone metastases compared with both control mice (88%, P = 0.0277) and mice of the WT52-D arm (82%, P = 0.0446; Fig. 3F). Interestingly, CTC number in each mouse showed a direct correspondence with metastatic dissemination (Supplementary Fig. S3). We conclude that killing CTCs by MET-targeted ADCC may be an effective strategy to prevent tumor recurrence and to suppress metastatic dissemination.

The results obtained in mouse models of cancer suggest that MET-targeted ADCC achieves superior tumor inhibition and metastasis suppression compared with HGF/MET signaling blockade. Prompted by these data, we set to develop an ADCC-enhanced anti-MET antibody for clinical use. The llama WT52 VH and VL framework regions display 88% average human identity. Using phage display-based affinity enrichment, we selected a molecule with 95% human framework identity (G52; Supplementary Fig. S4). Germlining did not affect the major biochemical and biologic characteristics of the antibody (Supplementary Table S3). G52 was further optimized by introducing the H433K/N434F substitutions in the CH3 region, generating the G52-HN antibody. These substitutions are known to increase binding to human FcRn at acidic pH while not affecting its affinity at neutral pH, thus enhancing antibody recycling from the sorting endosome (30). Surface Plasmon Resonance confirmed that G52-HN has a K_D for human FcRn at pH 6.0 that is approximately six times lower compared with G52 (0.255 ± 0.01 nmol/L vs. 1.37 ± 0.11 nmol/L). G52-HN was produced in a GS-CHO cell line lacking alpha-1,6-fucosyltransferase, thus obtaining an afucosylated G52-HN antibody (hereafter referred to as ARGX-111). Antibody afucosylation is known to increase affinity for human FcγRIIIa, thus resulting in enhanced ADCC (31). ELISA analysis confirmed that afucosylation increased antibody binding to both V158 FcγRIIIa and F158 FcγRIIIa by at least 25 times, whereas it did not significantly affect hFcγRI binding (Supplementary Table S4). The major biochemical and pharmacologic characteristics of ARGX-111 are summarized in Supplementary Fig. S5.

The ability of ARGX-111 to inhibit HGF-dependent MET activity was determined in a variety of biochemical and biologic assays. Flow-cytometry analysis on a panel of human tumor cell lines expressing different MET levels unveiled that ARGX-111 binds to native MET with an EC₅₀ ranging from 0.1 nmol/L to 0.7 nmol/L (Supplementary Table S5). In ELISA assays, ARGX-111 competed with HGF displaying an IC₅₀ of 1.3 nmol/L and an I_MAX of 95% (Fig. 4A). In MET activation experiments using A549 cells, ARGX-111 inhibited HGF-induced MET autophosphorylation with an IC₅₀ of 5.3 nmol/L and an I_MAX of 74% (Fig. 4B). Similar results were obtained using NCI-H1437 NSCLC cells (IC₅₀ = 13.2 nmol/L; I_MAX = 60%; Fig. 4C). In an HGF-dependent cell migration assay using A549 cells seeded in a Boyden chamber, ARGX-111 completely neutralized the promigratory effect of HGF (Fig. 4D). In an anchorage-independent growth assay using NCI-H1437 cells seeded in soft agar, ARGX-111 inhibited HGF-dependent colony formation by 78% (Fig. 4E). Finally, in an HGF-induced branching morphogenesis assay using SV40 T-antigen–transformed LOC human kidney epithelial cell spheroids (23) and increasing antibody concentrations, ARGX-111 significantly inhibited HGF-induced branched tubule formation already at a 1:1 molar ratio with recombinant HGF (Fig. 4F).

We next determined whether ARGX-111 could affect HGF-independent MET activity in tumor cells displaying constitutive MET activation. In phospho-MET ELISA assays, ARGX-111 showed dose-dependent inhibition of MET autophosphorylation in both MKN-45 cells (IC₅₀ = 0.2 nmol/L, I_MAX = 63%; Fig. 5A) and EBC-1 human lung carcinoma cells (IC₅₀ = 12.7 nmol/L, I_MAX = 69%; Fig. 5B). Flow-cytometry analysis of MET expression revealed that ARGX-111 promoted a rapid decrease of MET expression levels in both MKN-45 cells (Fig. 5C) and EBC-1 cells (Fig. 5D). Interestingly, this effect was also observed in MDA-MB-231 human mammary carcinoma cells (Fig. 3E), which express normal MET levels and retain ligand sensitivity. ARGX-111–mediated MET downregulation was dose-dependent, reaching its maximal effect already at 10 nmol/L (Fig. 5F). Western blot analysis of conditioned medium (Supplementary Fig. S6A) and of cell lysates (Supplementary Fig. S6B) demonstrated that ARGX-111 does not promote receptor shedding, but reduces total MET protein levels, suggesting that antibody binding to MET results in receptor internalization followed by protein degradation. ARGX-111–induced MET downregulation also occurred in the presence of a highly specific MET tyrosine kinase inhibitor (JNJ-38877605; ref. 32) in MKN-45 (Fig. 5G), EBC-1 (Fig. 5H), and MDA-MB-231 cells (Fig. 5I), suggesting that this process does not require MET kinase activity. In contrast, MET kinase inhibition did block HGF-induced MET internalization in MDA-MB-231 cells (Fig. 5I). Western blot analysis confirmed that the concentration of JNJ-38877605 used in these experiments achieved complete MET kinase blockade in all cells (Supplementary Fig S6C).

We investigated whether ARGX-111 could kill cancer cells by ADCC using a representative panel of human tumor cell lines expressing different MET levels. Cells were incubated with increasing concentrations of ARGX-111 (0-1,000 nmol/L) in the presence of NK cells, and tumor cell lysis was determined by a standard ⁵¹Cr-release assay. An irrelevant IgG1 and the fucosylated G52 antibody were used as controls. Each assay was repeated using effector cells derived from three different healthy donors (Supplementary Table S6). This analysis revealed that ARGX-111 induces dose-dependent lysis of all tumor cell lines tested with an efficacy directly proportional to MET protein expression levels. The fucosylated G52 antibody displayed only a modest ADCC activity on MKN-45 cells, which express very high MET levels, but did not have any effect on the other cell lines tested (Supplementary Fig. S7A). We also analyzed ARGX-111-induced antibody-dependent cellular phagocytosis (ADCP) using MKN-45 cells or 786-O human renal cell carcinoma cells and human monocyte-derived macrophages. Consistent with our finding that afucosylation does not significantly change antibody affinity for FcγRI (Supplementary Table S4), ARGX-111 did not display increased ADCP compared with G52 in any condition tested (Supplementary Fig. S7B).

The results obtained in the orthotopic mammary carcinoma models indicate that MET-targeted ADCC impairs the ability of CTCs to colonize distant organs and to give rise to metastases. Because only tumor cells with stem-like characteristics are competent to originate secondary colonies, we hypothesized that the mammary carcinoma stem cell population may express MET. To test this hypothesis, we analyzed coexpression of MET with established breast cancer stem cell markers (33) in MDA-MB-231 tumors as well as in a vital library of patient-derived breast xenografts (CRCM tumors; ref. 25) generated by implantation of primary mammary carcinomas into immunodeficient mice (Fig. 6A). Interestingly, all triple-negative tumors from this library expressed MET (Fig. 6B). Flow-cytometry analysis of CD24, CD44 and MET unveiled that the majority of cells in the CD44⁺/CD24^lo population express MET in both the MDA-MB-231 cell line (99%) and the primary breast carcinoma samples (78%–98%; Supplementary Table S7). Similarly, dual staining with AldeFluor reaction and anti-MET antibodies revealed that ALDH-1–positive tumor cells express MET in all samples analyzed (Fig. 7A). Prompted by these data, we determined whether MET-targeted ADCC could kill mammary carcinoma stem cells in vitro. To this end, cells derived from MDA-MB-231, CRCM168, CRCM174, and CRCM389 tumors were stained with AldeFluor and then sorted by FACS. Sorted cells were incubated with increasing ARGX-111 concentrations (0–667 nmol/L) in the presence of bioluminescent ADCC reporter cells (34), and ADCC was determined by analyzing luciferase activity. Remarkably, ARGX-111 promoted dose-dependent ADCC of all stem cell populations analyzed (Fig. 7B). We conclude that MET-targeted ADCC may be an effective method to eliminate breast cancer stem cells.

HGF and its receptor MET have been attracting increasing interest in the last two decades as appealing targets for cancer therapy. This interest has led to the development of numerous targeted drugs, some of which have reached the clinic. The initial enthusiasm has been partially mitigated, however, by the realization that anti-HGF/MET drugs display cytostatic rather than cytotoxic activity in most systems, including cell-based assays, mouse models of cancer, and oncologic patients. In this study, we report the engineering of an antagonistic anti-MET antibody that, in addition to inhibiting HGF/MET signaling, kills MET-expressing tumor cells by enhanced ADCC. The data presented here suggest that depleting MET-positive cancer cells is therapeutically more effective than simply inhibiting HGF/MET signaling.

These results must be interpreted at the light of the role played by MET in cancer biology. It has been shown that tumors take advantage of HGF/MET signaling in two different situations, determined by the number of c-MET gene copies contained in their genome (22). Tumors bearing normal c-MET gene copy-number exploit paracrine HGF stimulation to proliferate in adverse environmental conditions and to invade adjacent tissues, but do not strictly depend on HGF/MET signaling for their own survival. In this setting, inhibition of HGF/MET signaling attenuates proliferation and invasion, but does not eliminate MET-expressing cancer cells per se. On the other hand, tumors displaying high-grade focal c-MET amplification totally rely on constitutive HGF-independent MET signaling, resulting from MET protein overexpression, to sustain their proliferation and EMT phenotype (35, 36). In these tumors, MET inhibition leads to G₁ arrest and to transient reversion of EMT; however, both proliferation and EMT can be resumed upon removal of MET blockade (20). Therefore, whether we are hindering HGF-mediated oncogenic “expedience” in tumors bearing normal c-MET gene copy-number or MET-dependent oncogenic “addiction” in c-MET–amplified lesions, we are temporarily preventing tumor cells to carry out their malignant program, but we are not eradicating them.

This notion acquires high therapeutic relevance in light of the emerging evidence that MET is expressed and functionally required in the stem cell compartment of several tumor types, including head and neck carcinoma (37), colorectal carcinoma (38), prostate cancer (39), breast cancer (40), and glioblastoma multiforme (41). In analogy with normal tissue stem cells, cancer stem cells represent a multipotent tumor cell subpopulation that possesses the ability to replicate indefinitely and to give rise to secondary tumor colonies with the full differentiation spectrum of the primary lesion. Circulating cancer cells with stem-like characteristics are the only cell type that can give rise to metastases and are often drug resistant (42). Following therapy interruption, residual cancer stem cells that survived treatment may resume their proliferative cycle, reconquer space in the host organism, and give rise to malignant lesions, resulting in new metastases or tumor recurrence (43). From a therapeutic viewpoint, killing rather than simply inhibiting the proliferation of cancer stem cells may be associated with a greater patient benefit.

The possibility to intercept MET-positive CTCs with stem-like properties generates a new paradigm for the treatment of MET-positive tumors. So far, MET-targeted therapy has been envisioned either for causing the regression of established HGF/MET-dependent lesions or for overcoming HGF/MET-driven resistance to a second anticancer drug (44). The data presented here suggest that these strategies may not exploit the full therapeutic potential of an ADCC-enhanced anti-MET antibody, and point at the neoadjuvant and adjuvant settings as valid alternative approaches. Indeed, CTCs are ultimately responsible for tumor recurrence or metastasis onset following surgery, independently of how well the original lesion has responded to a previous therapy. On the basis of the results obtained in our neoadjuvant and adjuvant orthotopic breast cancer models, it would be reasonable to expect that eradication of CTCs by MET-targeted ADCC is associated with prolonged disease-free survival. Importantly, this opportunity is unique to an ADCC-enhanced anti-MET antibody. In fact, inhibition of HGF/MET signaling using a conventional HGF/MET–targeted drug, whether an antibody or a small molecule, would probably be able to prevent the escape of cancer cells from a primary lesion, but would have no effect on the survival of metastasis-initiating cells once they are in circulation, especially if these cells have become resistant to anoikis during their journey from the home epithelium to the breakout vessel.

Although MET plays its most relevant physiologic role during embryo development and remains mostly silent during the adult stage, it is widely expressed in epithelial and endothelial tissues of virtually all organs. This broad expression pattern could raise the concern that MET-targeted ADCC may cause damage to normal tissues, resulting in systemic toxicity. However, toxicologic studies conducted in cynomolgus monkeys (Macaca fascicularis) indicate that this is not the case. In fact, although ARGX-111 does not cross-react with mouse MET, it binds with high affinity to simian MET and to simian FcγRIIIa (Supplementary Table S4). In a large study involving 40 adult animals, the no-observed adverse event level of ARGX-111 corresponded to a dose of 30 mg/kg administered weekly (data on file). These results suggest that MET-targeted ADCC may be well tolerated by normal tissues expressing MET and may, therefore, not interfere with housekeeping physiologic functions at doses that are active against MET-dependent tumors. More definitive data on ARGX-111 safety will emerge from the phase I study that has started in January 2014 (NCT02055066).

A. Hultberg is a senior scientist and has ownership interest (including patents) in arGEN-X. C. Blanchetot is an employee in arGEN-X. M. Saunders is a senior director in arGEN-X. A. Thibault has ownership interest (including patents) in arGEN-X. H. De Haard is a CSO in arGEN-X BVBA. P. Michieli is a consultant/advisory board member in arGEN-X BVBA. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Hultberg, V. Morello, C. Blanchetot, F. Maina, P. Brouckaert, M. Saunders, A. Thibault, T. Dreier, H. de Haard, P. Michieli

Development of methodology: A. Hultberg, V. Morello, L. Huyghe, N. De Jonge, C. Blanchetot, F. Lamballe, F. Maina, P. Brouckaert, H. de Haard, P. Michieli

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Hultberg, V. Morello, L. Huyghe, V. Hanssens, K. Silence, R. Heukers, B. Roux, F. Lamballe, C. Ginestier, E. Charafe-Jauffret, P. Brouckaert, H. de Haard, P. Michieli

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Hultberg, V. Morello, L. Huyghe, G. De Boeck, K. Silence, E. Festjens, F. Lamballe, F. Maina, P. Brouckaert, H. de Haard, P. Michieli

Writing, review, and/or revision of the manuscript: A. Hultberg, V. Morello, N. De Jonge, V. Hanssens, K. Silence, R. Heukers, F. Maina, P. Brouckaert, M. Saunders, A. Thibault, H. de Haard, P. Michieli

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): V. Hanssens, E. Festjens, H. de Haard

Study supervision: A. Hultberg, F. Maina, H. de Haard, P. Michieli

The authors thank Brigitte Bonnin and John Wijdenes for the ADCP and ADCC assays, Sylvie Richelme for HGF-dependent assays, Sjudry-llona Osepa for llama-human chimera construction, and U-protein Express for production and purification of antibodies.

This work was supported by IWT (Agentschap voor Innovatie door Wetenschap en Technologie) grants # IWT090297 and IWT100440, by Italian Association for Cancer Research (AIRC 2010 Special Program in Molecular Clinical Oncology 5‰ Project n. 9970 and AIRC 2012 IG grant n. 12798), and the University of Torino/Compagnia di San Paolo (Progetti di Ricerca di Ateneo 2012).

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