Background
The MET oncogene encodes for a transmembrane tyrosine kinase, the receptor for Hepatocyte Growth Factor (HGF), endowed with pleiotropic functions, including regulation of cell proliferation, motility, invasion, and apoptosis [
1‐
4]. When genetic alterations leading to deregulated MET activation occur (mostly amplification and/or point mutations), MET initiates and maintains cell transformation (MET addiction) [
5‐
8]. MET genetic lesions have been described in different types of solid tumors, with an overall frequency of about 4% (
www.cbioportal.org).
MET gene amplification generates a MET receptor constitutively active due to overexpression. Copy number gain can result from focal amplification or polysomy, i.e. in the absence or in the presence of multiple copies of chromosome 7, including the MET gene. The level of MET gene amplification defines tumor cell dependence from the oncogene. Although experimental and clinical data indicate that MET-addiction is reached only in the presence of a high MET gene copy number, a clear cut-off point has still to be defined. MET gene amplification has been found in patients carrying gastric cancers, lung tumors, and in type 1 papillary renal cell carcinomas at a frequency around 10% [
9‐
11], and at a lower frequency (1–5%) in hepatocellular carcinomas, ovarian tumors, melanomas and type 2 papillary renal cell carcinomas [
11‐
14]. In addition to de novo condition, MET gene amplification represents a molecular mechanism responsible for resistance to Epidermal Growth Factor Receptor inhibitors. This has been described in cases of colorectal and lung carcinomas [
15‐
17]. MET gene copy number gain has been also found in cases of resistance to BRAF targeted therapies [
18,
19].
Non-synonymous activating point mutations in the MET gene were first described in hereditary and sporadic kidney cancers, involving residues located in the kinase domain [
20]. During the last 20 years, an increased number of MET gene mutations have been found in patients carrying different types of cancers (e.g. lung, breast, gastric, hepatocellular, head and neck carcinomas, cancer of unknown primary, see
www.cbioportal.org). These amino acid conversions are clustered, in addition to the kinase domain, also in the extracellular portion of the receptor (SEMA domain) or in the juxtamembrane region [
21]. While some of these amino acid changes have been functionally validated, the real activating function of others is still debated.
Finally, several genetic alterations have been identified in non-coding regions [
22]. These genomic modifications generate alternatively spliced MET mRNAs lacking the exon 14, a region involved in the negative regulation of the receptor enzymatic activity [
23]. They have been found in the 3% of Non-Small Cell Lung Cancers (NSCLC) [
24] and - at a smaller frequency - in other tumor types.
In light of the role of MET in cancer, several different MET inhibitors have been developed [
5,
25,
26]. In the past, we generated and characterized DN30, a murine monoclonal antibody (mAb) displaying both agonist and antagonist properties [
27]. In a bivalent antibody format, allowing simultaneous interaction with two distinct antigen molecules, the antibody stabilizes receptor complexes in a way similar to what is achieved by the natural ligand. This leads to partial activation of the MET kinase, which ultimately triggers some - but not all - MET-driven biological responses [
27]. On the other hand, DN30 mAb enhances MET proteolytic cleavage, at one single site located within the extracellular domain of the receptor, in a region close to the cell membrane [
28]. As a consequence, soluble MET receptors are released in the extracellular space (receptor ‘shedding’) where they can behave as a ‘decoy’, neutralizing HGF and forming inactive heterodimers with bona fide MET receptors which, eventually surviving the cleavage, are still exposed on the cell surface. Inside the cell, the transmembrane MET fragment derived from shedding is further cleaved to detach the kinase-containing cytoplasmatic portion from the membrane. The latter is then rapidly addressed to proteasomal degradation [
29]. As a consequence, MET is physically removed from the plasma membrane. This complex cascade of molecular events occurring upon DN30 mAb binding translates into inhibition of MET-driven biological activities. Thus, when DN30 mAb interacts with MET on the cell surface, the net biological response within the cell depends on the balance between two opposing functions: kinase activation and receptor shedding. Disassociation of the two activities has been achieved by conversion of the ancestral bivalent DN30 mAb into a monovalent Fab fragment [
30]. This unleashes the therapeutic potential of DN30, leading to a full antagonist molecule. Nevertheless, the short half-life of the Fab fragment limits its clinical development. Thus, to maintain the full antagonist behavior without losing the good pharmacokinetic properties of the immunoglobulins, the conversion from a bivalent to a monovalent antibody was obtained by deleting one of the two antibody arms through a molecular engineering approach. Here we describe the design and validation of the DN30 mAb humanized one-arm form (hOA-DN30), and present data demonstrating its robust anti-proliferative and tumor growth-inhibitory effects in multiple MET-addicted cell lines and xenograft cancer models, as well as its favorable pharmacokinetics and safety profile in non-human primates.
Methods
More details are provided in Additional File
1.
hOA-DN30 generation, production, and purification
hOA-DN30 has been provided by Metis Precision Medicine B-Corp (Torino, Italy). Humanization of the DN30 mouse antibody, conversion into the ‘one-arm’ format, production, and purification of hOA-DN30 has been done by Fair Journey Biologics.
ELISA assays
For affinity determination, the MET extracellular domain fused in frame with a human Fc domain (R&D System) was in the solid phase and pure MvDN30 or hOA-DN30 were in the liquid phase. ELISA signal was quantified by the multi-label reader VICTOR X4 (Perkin Elmer Instrument INC.).
For hOA-DN30 quantification in mouse serum, a goat anti-human IgG antibody (Sigma Aldrich) was in the solid phase. For hOA-DN30 quantification in monkey serum samples, neutravidin-coated plates were used to immobilize biotin-labeled goat anti-human IgG monkey adsorbed antibody (Southern Biotech). In both cases, pure hOA-DN30 and serum samples were in the liquid phase. The concentration of hOA-DN30 in serum was achieved by interpolating the optical density (OD) readings of the samples with the calibration curve fitted by a weighted (1/Y) 4 parameters regression function.
MET shedding and MET inhibition
For dose-dependent experiments, cells were incubated in serum-free medium for 48 h with the indicated concentration of hOA-DN30. For time-dependent experiments, cells were incubated in serum-free medium in the presence of hOA-DN30 (1 μM) and samples were collected at the indicated time points. Proteins in cell extracts or in cell culture supernatants were resolved by SDS-PAGE and analyzed by Western blotting.
For analysis of MET recovery after shedding, cells were incubated in serum-free medium in the presence of hOA-DN30 (1 μM); after 48 h, cell monolayers were extensively washed and fresh serum-free medium was added. Cell culture supernatants and cell lysates were collected at different time points and analyzed as described above.
For inhibition of MET and downstream signalling pathway activation, cells were treated with the indicated concentrations of hOA-DN30 for 24 h in serum-free medium. Cell monolayers were lysed and analyzed as described above.
Biological assays
For scatter assay, cells were treated for 24 h with HGF (8 ng/ml, R&D System), hOA-DN30, or DN30 mAb (both 200 nM). Cell scattering was determined by optical microscopy.
For cell growth assay, cells were treated with increasing concentration of hOA-DN30, and viability was evaluated after 72 h by CellTiter-Glo luminescent cell viability assay (Promega Corp.), according to the manufacturer’s instructions. Chemiluminescence was detected with VICTOR X4.
For proliferation assay, cells were treated with hOA-DN30 (1 μM). After 48 h cellular DNA synthesis was determined measuring EdU incorporation using the Click-iT® EdU Alexa Fluor® 647 Flow Cytometry Assay Kit (Thermo Fisher Scientific) according to manufacturer’s instruction.
For cell death assay, cells were treated as described for cell growth assay. Cytotoxicity was evaluated after 72 h by CellTox™ Green Cytotoxicity Assay (Promega Corp., Madison, WI), according to the manufacturer’s instructions. Fluorescence was detected with VICTOR X4 and data normalized over CellTiter-Glo luminescence.
For determination of M192 tumor organoid viability, cells were incubated for 5 h with Alamar blue (Life Technologies). The signal measured in the cell culture supernatants by VICTOR X4 was set as time 0, then freshly prepared medium with 1 or 5 μM hOA-DN30 was added to the culture. Every 3 days, the medium was replaced with a fresh one containing the antibody. After 9 days, a new assessment of the Alamar blue signal was performed. Cell growth was measured calculating the ratio between day 9 and day 0 for each sample.
For determination of Antibody-Dependent Cellular Cytotoxicity (ADCC), cells were treated with increasing concentrations of hOA-DN30, and effector cells were added. After 6 h, ADCC activity has been measured by ADCC Reporter Bioassay (Promega Corp.) according to the manufacturer’s instructions.
Evaluation of tumor growth inhibition in vivo
All procedures in mice were performed according to protocols approved by the Ethical Committee for animal experimentation of the Candiolo Cancer Institute and by the Italian Ministry of Health.
To generate Cell Derived Xenografts (CDX), cancer cells were injected subcutaneously into the right posterior flank of adult NOD-SCID mice. To generate Patient Derived Xenografts (PDX), tumor materials derived from human gastric tumor specimens expanded for at least 2 generations in mice were implanted in a subcutaneous pocket generated in the flank of NOD-SCID mice as described in [
31]. When tumors were established, animals were divided into experimental arms homogeneous for tumor size and randomly assigned to the different treatments. hOA-DN30 was administered by intravenous injection. Tumor size was evaluated periodically with a caliper. Tumor volume was calculated as described [
30].
Pharmacokinetics, pharmacodynamics, and toxicological studies
For pharmacokinetic (PK) evaluation in mice, adult male NOD/SCID mice bearing or not EBC-1 tumors were used. Animals were treated with 30 mg/kg of hOA-DN30 in a single intravenous administration. Animals were bled at different time points. hOA-DN30 levels in mouse serum were determined by ELISA assay (see above) by Accelera Srl. (Nerviano, Italy). PK analysis was performed according to standard non-compartmental and compartmental approach using Phoenix-WinNonlin package (v. 6.3, Pharsight Inc., Certara Company) by Accelera Srl.
For PK evaluation in Cynomolgus monkeys, the experiments have been performed by Accelera Srl. Three adult male monkeys were administered with hOA-DN30 as a single intravenous bolus at the dose of 11 mg/kg and animals were bled at different time points. hOA-DN30 concentrations in serum were determined by ELISA assay (see above). PK analysis was performed according to the standard non-compartmental approach as described above.
For pharmacodynamics (PD), the analysis was performed by Accelera Srl, applying an E
max (maximum kill rate) PK/PD model [
32] to tumor volumes measured in mice treated with the indicated doses of hOA-DN30. The threshold systemic (plasma) concentration for tumour stabilization (C
τ) was calculated from the equilibrium status of the first differential equation of the model.
Determination of hOA-DN30 tolerability was performed by Accelera Srl. The antibody was administered intravenously according to ascending doses (30, 90, and 180 mg/kg) one week apart, or at the dose of 180 mg/mg at weekly intervals for two times to two adult Cynomolgus monkeys (one male, one female). Animal’s healthy status, clinical signs observation, body weights, and food consumption were evaluated periodically. Standard hematological and serum chemical parameters were determined. At necropsy, body weight measurement and macroscopic examinations were done.
All the studies performed by Accelera Srl. have been sponsored by Metis Precision Medicine B-Corp.
Statistical analysis
Averages, standard deviations, and P values obtained by Student’s t-Test were calculated using Microsoft Office Excel 2010 software (Microsoft Corporation). To calculate Kd and Bmax, data from ELISA assay were analyzed and fitted according to nonlinear regression, one-site binding hyperbola curve, using GraphPad Prism software (GraphPad Software). To calculate IC50, data from growth assays were analyzed and fitted according to a nonlinear regression, sigmoidal dose-response curve, using GraphPad Prism software. P values obtained by One-way or Two-way Anova were calculated using GraphPad Prism software.
Discussion
From the outset of precision medicine, MET has been considered an attractive target for cancer therapy [
41]. As a consequence, a variety of molecules against either the MET receptor or its specific ligand, HGF, have been explored during the last 15 years [
5,
26]. While most of them stopped at the preclinical phase, a reasonable number of small molecules and antibodies entered clinical trials, with different grades of success.
Crizotinib, a multi-targeting MET Tyrosine Kinase Inhibitor (TKI) [
42], gave positive clinical outcomes in case reports of patients sharing the presence of MET-gene amplification in the tumor [
18,
43‐
45]. Crizotinib activity against MET high-amplified tumors has been also confirmed in a retrospective study in NSCLC patients re-classified according to MET/CEP7 ratio [
46]. Recently, positive results have been obtained in Phase II clinical trials conducted in NSCLC patients with Capmatinib and Tepotinib, two selective MET inhibitors [
47,
48]. These data show that a fraction of patients was responsive to the treatment, and this portion further increases in the case of naïve patients [
49,
50]. Nevertheless, the use of TKIs intrinsically brings some drawbacks. Tumors treated with small molecules inevitably and rapidly become resistant to the therapy [
51‐
53]. Moreover, small molecule discontinuation can induce hyper-activation of the kinase, leading to a disease flare [
54,
55]. Lastly, TKI toxicity, especially when combination regimens are required, is common [
56]. Thus, MET inhibitory antibodies have still the potential to provide a beneficial efficacy and safety outcome in patients.
The first MET antibody tested in the clinic was Onartuzumab, a potent HGF competitor antibody efficiently blocking HGF-dependent MET activation [
57]. So far, these studies, as well as those conducted with anti-HGF antibodies, resulted in poor or null benefits [
58‐
61]. These failures rely mainly on the mechanism of action of the above molecules. Indeed, MET-addicted tumors, which are those potentially eligible for a MET-targeted therapy, are characterized by the presence of activating genetic lesions (gene amplification or point mutations), making receptor activation independent from its ligand. In this condition, MET antibodies exclusively competing for ligand binding, as well as HGF-targeting antibodies, are intrinsically ineffective.
hOA-DN30 displays a novel mechanism of action completely different from the simple inhibition of ligand/receptor interaction; its binding epitope - the IPT4 region of MET [
35] - does not overlap with HGF binding sites [
27]. hOA-DN30 activity relies on the peculiar property, not shown for any other anti-MET drug, of enhancing MET shedding [
28]. Shedding, a physiologic cellular mechanism of protein degradation, is exploited by the cells to tightly regulate receptor signalling [
62]. MET shedding is predominantly operated by a surface metalloprotease - ADAM-10 - that cleaves the extracellular domain of MET, recognizing a specific sequence immediately upstream of the trans-membrane moiety [
63]. Upon shedding, a cascade of events takes place; the membrane-linked MET C-terminal fragment becomes the substrate of a transmembrane protease, γ-secretase, which detaches the intracellular kinase-domain from the cell membrane. This MET portion is rapidly shuttled towards the proteasome [
29]. Moreover, an alternative lysosomal-dependent second route of MET degradation is also activated [
64]. As a consequence of the above proteolytic events, the net number of MET receptors exposed at the cell surface is strongly reduced and, concomitantly, MET-ECDs are released in the surrounding space. As they are fully competent for ligand binding, MET-ECDs can sequester HGF in the extracellular environment. In addition, MET-ECDs can dimerize with ‘cleavage-survived’ MET receptors still exposed in the plasma membrane, impairing kinase trans-activation [
65]. In summary, the binding of hOA-DN30 to MET triggers a complex response acting on the HGF/MET axis at different levels. Thus, unlike other modes of action, the effects of hOA-DN30 on the HGF/MET axis may be deeper and longer sustained.
Currently, a new generation of MET antibodies featuring mechanisms of action different from Onartuzumab is under clinical evaluation.
Teliso-V/ABBV-399 is an antibody-drug conjugate that displays inhibitory effects against MET-overexpressing cells [
66] mainly - if not only - thanks to the activity of the anti-mitotic drug. Thus, the antibody represents a tool to deliver a cytotoxin to tumor cells characterized by high MET expression, regardless of their reliance on MET signalling. In our study, we demonstrate that the activity of hOA-DN30 is highly restricted to MET-addicted cancers. Although this condition is expected in a reduced proportion of tumors (see
www.cbioportal.org), this should be considered a
plus and not a limitation, as it assures a personalized response when a correct genetically-based patient selection is applied. On the contrary, patient selection based uniquely on the level of MET expression does not seem reliable [
67].
Other antibodies tested so far in humans - SAIT-301, Emibetuzumab, ARGX-111, and a mixture of two antibodies, Sym-015 - are able to block HGF binding and concomitantly induce MET internalization [
68‐
71]. These bivalent antibodies share a common feature: they interact with MET in a region overlapping the binding site for the HGF beta chain. This ligand/receptor interaction, although occurring with low affinity, is crucial as it drives the biological responses [
72]. Antibodies binding at this site productively interfere with HGF activity and, thanks to the ability to downregulate MET, also impair ligand-independent kinase activation. Nevertheless, they still retain some agonistic properties - measured as receptor phosphorylation and/or signalling pathway activation – due to induction, even transient, of receptor dimerization [
69‐
71,
73]. Such residual ligand mimetic activity can be considered a potential issue during the clinical application, as it might produce unwanted side effects. Therefore, only monovalent antibodies, like hOA-DN30, behaving as pure antagonists, guarantee a univocal outcome.
In addition to ligand competition and receptor downregulation properties, ARGX-111 and Sym-015 also unleash the immune response of the host against the tumor by triggering ADCC [
68,
74]. We proved that ADCC does not contribute to the anti-tumor effect of hOA-DN30. This inability is probably related to the mechanism of action of hOA-DN30. By inducing MET shedding, the antibody does not accumulate at the cell membrane in complex with its target, and thus it cannot mediate the interaction between killer cells and tumor cells. Importantly, the absence of ADCC activity represents a safety benefit, minimizing the risk of adverse immune effects against normal tissues.
Besides antibodies with a canonical structure, a biparatopic antibody, REGN-5093, is currently under clinical testing. This molecule not only blocks HGF/MET binding and induces receptor internalization, but also modulates receptor trafficking [
75]. Due to the biparatopic structure, the interaction between REGN-5093 and MET gives rise to large complexes that, not entering in the recycling tubules, are mainly included in the multivesicular endosomes and then degraded in the lysosomes. Nevertheless, receptor recycling is only partially affected; a quote of receptors - estimated as 25% of the total internalized molecules - still goes back to the membrane. This leakiness of the mechanism of action is reflected in the antibody efficacy; tumor regression is very slow and requires prolonged treatment [
75]. On the contrary, the therapeutic response observed upon hOA-DN30, relying on an enzymatic response, is quite fast and long-lasting.
In summary, hOA-DN30 possesses unique features when compared to anti-MET antibodies in clinical development. The molecule being monovalent is a pure antagonist. hOA-DN30 mechanism of action (i.e. MET shedding) is based on an initial proteolytic event, the MET cleavage, which prompts intracellular receptor destruction not dependent on receptor trafficking. This activity should be considered more robust and definitive compared with the receptor internalization induced by other MET inhibitory antibodies. In the case of internalization, the receptor can be recycled from the endosomes, while, in the case of shedding, the recovery of MET on the cell surface relies exclusively on new receptor synthesis. Thus, MET turn-over is rapid in the case of internalizing antibodies and, conversely, delayed in the case of hOA-DN30, assuring a longer-lasting effect. MET shedding not only removes MET from the cell surface but also intrinsically triggers other mechanisms able to block the activation of MET receptors that survived the cleavage. This occurs by releasing MET-ECDs in the microenvironment. The ‘decoy’ effect of MET-ECDs is multi-faceted, being effective on the side of ligand-independent MET-activation, by forming inactive receptor heterodimers, as well as on the side of ligand-dependent MET activation, sponging HGF. Notably, MET-ECDs represent the best option to neutralize the ligand, as they incorporate all the binding features included in the natural receptor.
From a pharmacological point of view, hOA-DN30 displays a considerably improved profile compared to its parental molecules (DN30mAb and MvDN30) [
76], not dissimilar from those reported for other therapeutic antibodies featuring the typical bivalent structure [
77]. Moreover, the one-arm structure of the molecule should not limit the duration of the therapeutic response, as no antibodies against the knob and hole regions have been detected, neither in monkeys nor in humans [
57]. The half-life and clearance values of hOA-DN30 indicate a low rate of antibody elimination from the systemic circulation. The volume of distribution measured in the monkey denotes a limited distribution in normal tissues, while the higher clearance rate in tumor-bearing mice compared to the healthy ones alludes to a promising accumulation of the antibody at the tumor site. Dose response and dose regimen experiments in mice suggest that hOA-DN30 can potentially operate with a wide therapeutic window.
From a toxicological point of view, the absence of adverse effects in the monkey at the highest dose tested (180 mg/kg) is extremely encouraging. The good tolerability of hOA-DN30 is in line with what was reported for other MET therapeutic antibodies in non-human primate preclinical models [
69,
71] and in clinical trials [
78‐
80] (Sym 015: Camidge R JCO 38 issue 15 suppl. ASCO 2020 abstract 9510). The only observed side effects were, in general, similar to those reported in patients treated with antibodies directed against other receptor tyrosine kinases, thus suggesting that MET targeting does not elicit major complications.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.