A Highly Potent and Specific MET Therapeutic Protein Antagonist with Both Ligand-Dependent and Ligand-Independent Activity

This article is featured in Highlights of This Issue, p. 2283

Targeting of the MET oncogenic pathway has emerged as a promising therapeutic strategy for the treatment of multiple cancers. MET is a receptor tyrosine kinase, which upon binding of its only known ligand, hepatocyte growth factor (HGF), leads to the stimulation of proliferative, migratory, and survival pathways implicated in tumor development (1–7). A link between MET pathway dysregulation and poor patient prognosis has been reported for several cancers (8–14). Apart from ligand-dependent signaling of MET, the pathway can be induced in cancers through MET gene amplification/mutation; receptor overexpression or through cooperation with other membrane bound proteins or receptors to drive HGF-independent tumorigenesis (15, 16). Thus, optimal targeting of the pathway benefiting the broadest patient population requires a drug with both ligand-dependent and ligand-independent efficacy.

Several experimental drugs targeting various aspects of the MET pathway are currently undergoing clinical evaluation. These include small molecule kinase inhibitors; antibodies that bind HGF and antibodies targeting MET. However, each of the current drug classes being tested has confounding characteristics. The small molecule receptor tyrosine kinase inhibitors (TKI) are hampered by their inability to specifically target and/or antagonize this pathway, which makes patient selection difficult. There have also been preclinical reports of acquired resistance to MET targeting TKIs via receptor mutation (17). Although molecules such as XL184, ARQ197, and crizotinib have all showed varying degrees of efficacy in MET positive patients, it is difficult to apportion this solely to their inhibition of a MET phenotype due to their broader inhibitory profile against other kinases (18–22). Anti-HGF antibodies, such as AMG102, bind circulating ligand and although they may benefit a subset of patients they cannot impact HGF-independent receptor activation (23, 24). The generation of antibodies against the receptor has proved difficult as the bivalent structure of immunoglobulins leads to receptor dimerization and activation in the absence of any ligand present (2). There are several bivalent antibodies targeting MET receptor, undergoing preclinical or early clinical evaluation. Although these antibodies may have ligand-independent activity, they have the potential to act as partial agonists through receptor clustering thereby creating a potential safety risk. A monovalent form of DN-30 mAb was engineered to eliminate agonist activity and although it does not directly inhibit HGF binding it causes receptor cleavage and shedding in preclinical models (25). The one-armed Ab, MetMab, was also developed in an effort to counteract this agonism (26). In a phase II clinical trial, MetMab led to a significant improvement in overall survival in NSCL cancer patients with Met-positive tumors (27, 28). Its therapeutic mechanism of action is via HGF antagonism, thus questioning its impact in patients with a ligand-independent disease (26). Therefore, we developed a MET antagonist based on a novel biotherapeutic drug platform called Anticalins to block both ligand-dependent and ligand-independent signaling, which was also devoid of any agonistic activity.

Anticalins are engineered human lipocalins and have been developed to address limitations of existing drug platforms (29). The lipocalins form a family of structurally conserved human proteins involved in binding and transporting diverse molecules (small molecules and large proteins; 30, 31). The β strands that make up the lipocalins and Anticalins are connected by 4 loop regions that can be engineered (randomized) to generate Anticalins with exquisite specificity for molecular targets (31). Furthermore, Anticalins display robust pharmaceutical properties required in a therapeutic product candidate. Targeting specific amino acid residues, we have rationally engineered PRS-110, a MET-specific monovalent full antagonist. Herein, we describe the discovery and functional properties of PRS-110 displaying efficacy in both ligand-dependent and ligand-independent cancer models. We further show a dual mechanism of action for PRS-110 in terms of both blockade of receptor activation and downregulation of receptor highlighting its potential as a novel therapeutic modality for the treatment of MET driven tumors.

Standard methodology was used to characterize the biophysical properties of PRS-110. Briefly, for SE-HPLC analysis, Anticalin samples (0.1 mg/mL, 25 μL injection) were run on an HP1100 (Agilent Technology, A280 UV detection) using a Superdex 75 PC3.2/30 (GE Healthcare) column equilibrated in phosphate-buffered saline (PBS; Gibco) at a flow rate of 0.1 mg/mL for 40 minutes. The area under the curves was calculated and expressed as percentage. In addition, a SEC-RALLS (right angle laser light scattering) detector was used to show absence of aggregates. To determine the melting temperature of PRS-110, capillary nanoDSC (Q2000; TA Instruments) was used. Stability studies were carried out with PRS-110 in PBS as standard formulation buffer. Recovery of the protein after 12 weeks at 25°C was assessed by quantitative ELISA.

A549 human lung carcinoma cells, Caki-1 human renal carcinoma cells, HT29 colorectal cancer cells, and U87-MG human glioblastoma cells were obtained from American Type Culture Collection. The cell lines were authenticated by STR analysis at the DKFZ in 2012. A2780 human ovarian carcinoma cells were purchased from European Collection of Cell Cultures CAMR Centre for Applied Microbiology & Research and tested by STR-PCR at the IRCC (Torino, Italy) in 2013. Human umbilical vein endothelial cells (HUVEC) were purchased from PromoCell. A quality control including immunohistochemical analyses of cell-type-specific markers was conducted by the provider. All cells were cultured in recommended media with 10% FBS in a humidified incubator at 37°C and 5% CO₂ and were not passaged for longer than 6 months.

Selectivity of PRS-110 for MET was tested using a direct ELISA-based receptor binding assay. High-binding fluorescence plates were coated with 5 μg/mL of MET-Fc (358-MT/CF; R&D), Tie2-Fc (313-TI-100; R&D), TrkA (11073-H08H; Sino Biological), and Ron (1947-MS-050; R&D) tyrosine kinase receptors in PBS pH 7.4 overnight at 4°C. After washing, the wells were blocked with PBS/2% bovine serum albumin (BSA)/0.1% Tween 20 at pH 7.4 for 1 hour at room temperature. PRS-110 was added in a dilution series ranging from 0.0002 to 100 nmol/L and allowed to bind for 1 hour at room temperature. Bound PRS-110 was detected by a mixture of rabbit anti-TLc pAb (antitear lipocalin) obtained by immunization of rabbits with a mixture of TLc muteins (BioGenes) and anti-rabbit IgG-HRP (Abcam; 1:5,000) for 1 hour at room temperature. All incubation steps were conducted in PBS/0.1% Tween 20 at pH 7.4. Between the steps, the ELISA plate was washed for 5 times with PBS, 0.05% Tween 20. Quanta Blu (Thermo) was added and resulting relative fluorescence units (RFU) at λEx = 320 nm and λEm = 430 nm were measured after 15 minutes of incubation using the microplate reader Genios Plus (Tecan). RFU values were plotted against PRS-110 concentrations and EC₅₀ values were fitted with a single-sites binding model [RFU = (RFU_max•c_PRS-110)/(EC₅₀ + c_PRS-110)] using Prism 5 software (GraphPad).

An Meso Scale Discovery (MSD) plate was coated with 1 μg/mL human HGF mAb (R&D) in PBS at pH 7.4 overnight at 4°C. The wells were blocked with PBS/2% BSA/0.1% Tween 20 at pH 7.4 (assay buffer) for 1 hour at room temperature. For capture, 0.5 μg/mL human HGF (R&D) was added and incubated for 1 hour at room temperature. Before addition to the assay plate, 2 nmol/L biotinylated human MET (Sino Biological, biotinylated at Pieris) was preincubated in solution with varying concentrations of PRS-110 or Anticalin control ranging from 0.008 to 500 nmol/L. After 1 hour incubation at room temperature, the reaction mixture was transferred to the MSD plate to detect unbound/nonneutralized MET for 20 minutes at room temperature. To allow for transformation of ECL signals into free MET concentrations, standards of biotinylated MET with varying concentrations were incubated for 20 minutes on the same MSD plate as well. Bound biotinylated MET was detected by 1 μg/mL Streptavidin Sulfo-Tag (Meso Scale Discovery) for 1 hour at room temperature.

All incubation steps were conducted in assay buffer. Between the steps, the MSD plate was washed with PBS 0.05% Tween 20 for 5 times. MSD Read Buffer T with Surfactant (Meso Scale Discovery) was added and the resulting ECL signals were measured within 15 minutes using the SECTOR Imager 2400 (Meso Scale Discovery). ECL signals were converted into free MET concentrations using the MET-bio standard curve and IC₅₀ values were fitted with a single-sites binding model [c_{MET free} = c_METtotal/(1 + c_PRS-110/IC₅₀)] with Prism 5 software (GraphPad).

Binding affinity was determined using a kinetic exclusion assay (KinExA) at Sapidyne Instruments. Two fixed concentrations of PRS-110 (i.e., 1 and 10 nmol/L) were incubated with variable concentrations of soluble MET (Sino Biological) at room temperature until the equilibrium was reached (≥6.5 hours). To detect the fraction of free (unbound) PRS-110 at equilibrium, the solutions were briefly exposed to biotinylated MET captured on Streptavidin PMMA beads. Anticalin bound to the beads was detected with rabbit anti-TLc pAb (Biogenes) followed by a Cy5-labeled anti-rabbit Ab (Jackson Immunoresearch). Detected fluorescence signals were directly proportional to the concentration of free Anticalin in the equilibrated sample. Normalized signals of 2 Anticalin concentrations were plotted against molar concentrations of MET and Kd and active binding concentration were simultaneously fitted by a 1:1 reversible bimolecular interaction model.

CHO cells were stably transfected with expression vector pcDNA3.1 containing either the human or the cynomolgus monkey cDNA for MET. Mock-transfected cells served as negative control. CHO cells were trypsinized at a subconfluent stage, washed in ice cold PBS 2% fetal calf serum (FCS), and incubated with Anticalins on ice for 2 hours. Cells were washed twice and a rabbit lipocalin scaffold detection antibody was applied for 30 minutes on ice. After a wash step, the secondary PE-labeled anti-rabbit antibody was added. Measurement of surface staining was carried out by means of a FACSCalibur flow cytometer (BD Biosciences).

A library of chemically linked peptides on scaffolds were synthesized by Pepscan (Pepscan) using the sequence of the MET extracellular domain (MET_HUMAN; Uniprot accession number: P0858; residues 25-932) as the template. Peptide synthesis and epitope mapping experiments were conducted as previously described with the following modifications for the PRS-110 molecule (32, 33). The peptide arrays were incubated with the PRS-110 molecule at concentrations from 1 to 10 μg/mL at 4°C overnight in a solution containing the competing proteins horse serum and ovalbumin. Following a wash step, the array was incubated for 1 hour at 25°C with either rabbit antitear lipocalin 1 antibody (1/10,000 dilution; BioGenes) or a biotinylated goat-antitear lipocalin 1 antibody (0.2 μg/mL; R&D systems). Following incubation with the secondary antibody, the array was washed and incubated for 1 hour at 25°C with a 1/1,000 dilution of either a swine-anti-rabbit antibody horseradish peroxidase (HRP) conjugate (Dako) or a Streptavidin–HRP conjugate (Southern Biotech). After washing, the peroxidase substrate 2, 2′-azino-di-3-ethylbenzthiazoline sulfonate and 2 μL/mL of 3% H₂O₂ were added and after 1 hour color development was measured. Color development was measured using a charge coupled device camera and image-processing system.

Cell lines expressing different MET receptor densities were grown in normal growth media with 10% FCS and trypsinized at a subconfluent stage. A washing step with PBS and blocking in ice cold PBS 2% FCS for 30 minutes followed. Blocked cells were incubated with a PE-labeled anti-MET antibody (R&D systems) for 1 hour on ice. Cells were washed twice in PBS 2% FCS and measurement of surface staining was carried out by means of a FACSCalibur flow cytometer (BD Biosciences).

Cells were seeded at a density of 2,000 cells/well in a 96-well plate. After growing the cells for 24 hours in normal growth medium with 10% FCS, medium was aspirated and replaced by serum reduced medium. After 3 days supernatants were centrifuged and analyzed in an HGF ELISA (Meso Scale Discovery) according to the instructions of the manufacturer.

Total RNA was extracted from subconfluent cell monolayers with miRNAeasy mini Kit (Qiagen) according to manufacturer's instructions. One microgram of total RNA was used as a template for cDNA synthesis with random hexamer primers and the High Capacity Reverse Transcription kit (Applied Biosystem) according to manufacturer's instructions. Real-time PCR was conducted on ABI PRISM 7900HT sequence detection system (Applied Biosystem) with standard conditions. To evaluation MET expression, real-time PCR was conducted on template cDNA using GoTaq qPCR Master Mix (Promega) and the QuantiTect primer 3-5′mix QT00023408 (Qiagen). To evaluate HGF expression, real-time PCR was conducted using TaqMan PCR Master Mix (Applied Biosystem) and the predesigned primers and probes set Hs00300159_m1 (Applied Biosystem). All samples were analyzed in triplicate. Average values were normalized using endogenous gene expression (cyclophilin for MET reaction and GAPDH for HGF reaction). Values are calculated as relative to A549 for MET and to U87-MG for HGF.

HUVEC were seeded in 96-well plates at a density of 4,000 cells per well with 100 μL endothelial cell growth (EBC) medium containing 2% FCS, 1 ng/mL basic fibroblast growth factor, 500 pg/mL epidermal growth factor, 90 μg/mL heparin, and 1 μg/mL hydrocortisone (PromoCell). Following incubation at 37°C and 5% CO₂ for 24 hours, cells were washed once and starved overnight in EBC medium without supplements. After blocking the cells in starving medium with 0.5% BSA for 3 hours, Anticalins were preincubated for 30 minutes at 37°C and 5% CO₂. Stimulation with 25 ng/mL human recombinant HGF (R&D systems) followed for 20 minutes. Cell lysis and phospho/total protein assays were conducted according to the instructions of the manufacturer (Meso Scale Discovery). Briefly, cell lysates were incubated on multispot MSD plates coated with antibodies against phospho (Y1349) and total Met, phospho (Ser473) and total AKT, and phospho (Thr202/Tyr204; Thr185/Tyr187) and total ERK1/2 Incubation with sulfo-tag labeled detection antibodies followed and chemiluminescence was measured using an MSD-ELISA reader (Meso Scale Discovery).

For the assessment of total MET levels in xenograft models, resected tumors were snap frozen and ground under liquid nitrogen using a Freezer/Mill 6770 (SPEX Sample Prep). The resulting powder was resuspended in Tris lysis buffer containing proteinase and phosphatase inhibitors (Meso Scale Discovery). Protein concentrations were determined using a BCA kit (Pierce) and total MET levels were measured using the ELISA MSD protocol as described earlier.

A total of 2.5–8 × 10⁴ cells (A549 or HT-29) were seeded on 8 chamber glass slides (BD) or coverslips coated with 1% gelatin (Sigma Life Science) and allowed to adhere for 24 hours. Cells were then washed and incubated for 1 hour on ice with serum free medium 0.3% BSA containing 500 nmol/L PRS-110, Anticalin control or 50 ng/mL HGF (R&D Systems). As a control, a sample was incubated with medium only. After binding, cells were incubated for 15 minutes or 1 hour at 37°C to allow internalization. Then, the cells were washed twice in cold PBS and washed with 0.2 M acetic acid, 0.5 M NaCl, pH 2.8 for 6 minutes on ice. Alternatively, 250 nmol/L PRS-110 was added to the cells for 30 minutes in growth conditions. Following incubation, the cells were fixed in 4% paraformaldehyde in PBS for 15 minutes at room temperature, permeabilized in PBS 0.5% Triton X-100, and blocked for up to 1 hour in PBS 1% BSA (optionally with 10% donkey serum). Primary antibodies were incubated for either 1 hour at room temperature or overnight at 4°C. Lysotracker blue (Life Technologies) was added 2 hours before fixation to a final concentration of 50–100–500 nmol/L. A murine anti-MET antibody (3D4) recognizing the c-terminal tail of the MET receptor was purchased from Invitrogen, the anti-EEA-1 antibody was purchased from Santa Cruz Biotechnology, and anti-LAMP-1 rabbit polyclonal antibody was purchased from Sigma Life Science. A rabbit antitear lipocalin 1 antibody (BioGenes) was used to detect PRS-110.

Appropriate Alexa-Fluor-tagged secondary antibodies (Life Technologies) were incubated from 30 minutes to 1 hour at room temperature and nuclei were counterstained with DAPI 0.1 μg/mL (Roche Applied Science). Immunofluorescence was analyzed using a Leica TCS SP2 AOBS confocal laser-scanning microscope (Leica Microsystems) or a Leica SP5 DNI6000 CS Confocal Microscope. Internalization of PRS-110 was also assessed in a flow cytometric assay. Briefly, 1 × 10⁶ cells (HT-29 or A2780) were seeded in 6-well plates 16 hours before incubation with 100 nmol/L PRS-110 for 1 hour at either 4°C or 37°C. After an acid strip (pH 3) for 5 minutes at 4°C, cells were washed in PBS, fixed and permeabilized (Invitrogen), and an anti-tear lipocalin 1 antibody (BioGenes) was used to detect PRS-110 on a FACScanto II flow cytometer (BD).

Subconfluent U87-MG glioblastoma cells were resuspended in PBS/Matrigel (1:1) at 5 × 10⁷ cells/mL. Female CD-1 nude mice (Charles River) were injected subcutaneously with 100 μL of the suspension. When mean tumor volumes had reached 130 mm³ mice were randomized into treatment groups with 10 mice per group (study day 0). Daily intraperitoneal treatment was initiated on day 0 as indicated. Tumor growth inhibition was calculated as TGI = 100 − [(mean tumor volume treated/mean tumor volume control) × 100]. At study, termination tumors were excised and snap-frozen in liquid nitrogen for evaluation of total MET.

For the Caki-1 xenograft model, female Balb/c nude mice (Sino-British SIPPR/BK Lab. Animal Co. Ltd.) were implanted subcutaneously in the right flank with 100 μL of 5 × 10⁶ Caki-1 renal clear cell carcinoma cells. When tumors reached an average volume of 100 to 150 mm³, mice were randomized into treatment groups with 10 mice per group (study day 0). Daily intraperitoneal treatment was initiated on day 0 as indicated. Tumor growth inhibition was calculated as TGI = 100 − [(mean tumor volume treated/mean tumor volume control) × 100]. At study, termination tumors were excised and snap-frozen in liquid nitrogen for evaluation of total MET. All animal studies were approved by a local Ethics Committee for Animal Experiments. These studies were conducted in compliance with the Animal Welfare Act.

The elimination rate constant and volume of distribution of PRS-050 (a VEGF-A-specific Anticalin conjugated to same 40 kDa PEG as PRS-110) determined in a clinical phase I trial were used to estimate the biweekly human dose for PRS-110 required to maintain plasma levels above 789 nmol/L using the equation D = (C_ss,min × V × (1 − r))/r, where r = e^{−kel × tau} and D (dose), C_ss,min (trough level at steady state), V (volume of distribution), kel (elimination rate constant), and τ (dosing interval; ref. 34).

All values are expressed as the mean ± SEM. Dose–response curves were generated using nonlinear curve regression (GraphPad Prism 5). Differences between PRS-110 and control-treated groups for in vitro and in vivo experiments were analyzed using the unpaired Student t test or ANOVA with P < 0.05 considered as statistically significant.

The MET-specific Anticalin, PRS-110, was selected from a combinatorial library of tear lipocalin (TLc and also referred to as Lcn1) variants using the extracellular domain of the MET receptor as a panning reagent (35). The cysteine free mutein was expressed in soluble form in the periplasm of Escherichia coli, essentially as described in Breustedt and colleagues (36). After intermediate purification, a free cysteine, introduced through rationale engineering to facilitate half-life extension, was modified with a maleimide activated branched 40 kDa PEG (2× 20 kDa PEG,NOF) yielding a highly pure monoPEGylated product referred to as PRS-110 (Fig. 1A). Thermal stability assessment by nanoDSC revealed a single cooperative melting transition with an onset of melting of 55°C and a thermodynamic melting point (T_m) of 69°C (Fig. 1B). An accelerated stress study (25°C for 12 weeks) showed that the molecule tolerated storage with no loss of binding activity over the time period as shown by quantitative ELISA (Fig. 1C). Full product recovery was consistently observed without formation of aggregates as visualized by superimposing of chromatograms from analytical SEC-RALLS (Fig. 1D and E).

The impact of PRS-110 on MET receptor activation was assessed using an ECL-based phosphorylation assay. Serum starved HUVECs were exposed to PRS-110 or control reagents prior to detection of pMet (Y1349). HGF, the natural ligand of MET, and the bivalent anti-MET antibody (5D5) both led to a strong activation of the receptor as previously reported (37), whereas PRS-110 had no detectable agonistic activity (Fig. 2A). Furthermore, PRS-110 was shown to be a full antagonist of HGF-induced activation of MET in HUVECs with an IC₅₀ of 35 ± 8 nmol/L (n = 3; Fig. 2B). To monitor the effect of PRS-110 on signal transduction pathways downstream of the multisubstrate docking site of MET receptor, phosphorylation of ERK1/2 (Thr202/Tyr204; Thr185/Tyr187) and AKT (Ser473) was also measured. The HGF-induced activation of both kinases, which are central in MET signaling, was blocked by PRS-110 (Fig. 2C and D).

Selectivity of PRS-110 for MET was investigated by a direct ELISA. Although binding to MET was confirmed, the absence of binding to the human tyrosine kinase receptors Tie2-Fc, TrkA, and Ron was also showed (Fig. 2E). Furthermore, flow cytometric analysis showed that PRS-110 exclusively binds to target positive cells. PRS-110 binds to CHO cells transfected with MET but not to mock-transfected cells (data not shown). Full inhibition of receptor–ligand interaction by PRS-110 was observed with an IC₅₀ of 3.2 ± 0.4 nmol/L when assessed by ELISA (Fig. 2F). The binding affinity of PRS-110 for MET was determined to be 0.6 ± 0.1 nmol/L (n = 3) by kinetic exclusion assay. An immunogenicity screen (EpiScreen) was conducted on ex vivo human donor PBMCs to predict clinical immunogenicity of PRS-110. Analysis of multiple human donor samples showed PRS-110 did not result in cell proliferation or cytokine stimulation (data not shown). It has a profile similar to nonimmunogenic controls, which reduces the risk of observing clinical immunogenicity (Supplementary Table S1). Flow cytometry analysis showed cynomolgus monkey cross-reactivity of PRS-110 (data not shown).

Epitope mapping was used to determine the binding site of the PRS-110 on the MET extracellular domain. Using a library of partially overlapping peptides, PRS-110 was shown to bind to the sequences LTEKRKKRST (aa 300-309) and WDFGFRRNNKFD (aa 586-597). In the MET ectodomain structure, the PRS-110 epitope (LTEKRKKRST) is located in a loop connecting the last β strand of the fourth β-propeller blade and the first β strand of the fifth β-propeller blade. This is also the location of the furin cleavage site between the MET α and β chains. This loop is disordered in the crystal structure and is presumably highly flexible (38–42). In addition, PRS-110 binds to an extended β-hairpin of the IPT1 domain (WDFGFRRNNKFD). Both the SEMA and IPT regions of MET are known to be important for HGF binding and receptor activation (40). The SEMA domain is also involved in ligand-independent activation of MET via receptor dimerization (Fig. 3; refs. 38, 39).

To facilitate the selection of representative/disease relevant cell lines to assess ligand-dependent and ligand-independent activity of PRS-110, we characterized expression of MET and HGF (Supplementary Fig. S1A–S1D). A549, an adenocarcinomic human alveolar basal epithelial cell line, showed significant expression of MET at both mRNA and protein levels. However, A549 cells did not express HGF at a detectable level as shown by RT-PCR and ELISA. Caki-1, a human kidney clear cell carcinoma cell line, featured the highest MET expression but no detectable levels of the ligand HGF. In contrast, the glioblastoma cell line U87-MG was shown to express moderate amounts of MET but relatively high levels of HGF. Therefore, A549 and Caki-1 cell lines were used to assess ligand-independent activity of PRS-110 and U87-MG cells were used to study ligand-dependent activity.

Flow cytometry was used to assess internalization of PRS-110 upon binding to MET receptor. HT-29 cells were exposed to PRS-110 (at 37°C or 4°C) and analyzed by flow cytometry. It was shown that PRS-110 binds to MET receptor and is internalized on HT29 cells incubated at 37°C whereas no internalization occurred at 4°C (Fig. 4). No internalization was observed in the MET negative control cell line A2780 at either 37°C or 4°C (Fig. 4).

Based on the data revealing that PRS-110 is internalized, the ability of PRS-110 to trigger MET internalization was investigated. To this end, A549 cells were incubated with PRS-110, HGF, or appropriate controls for 15 minutes before confocal microscopy analysis. The staining pattern observed with PRS-110 was different to that seen with the natural ligand. A predominantly granular intracellular staining pattern for MET was evident for cells treated with HGF compared to diffuse staining observed with PRS-110 or control-treated cells (Fig. 5A). The MET receptor was also found to colocalize with the early endosomal marker EEA-1 in HGF-treated cells but not PRS-110 or control-treated cells. To further examine the impact of PRS-110 on MET fate, A549 cells were costained for the late lysosomal marker LAMP-1 after 1 hour incubation. Confocal microscopy showed that PRS-110 leads to an accumulation of the MET receptor in intracellular compartments. PRS-110 resulted in clustering of LAMP-1 positive vesicles with apparent colocalization of MET and LAMP-1. Figure 5B, iv,) shows high signal intensity of MET receptor staining (in green) localizing to intracellular vesicles following incubation with PRS-110 (when compared to control panels in Fig. 5B, i). Counterstaining with LAMP-1 (red) shows a colocalization (yellow) of MET receptor with terminal endo-lysosomal vesicles when exposed to PRS-110 (Fig. 5B, vi). Immunoflourescent microscopy also showed the trafficking of MET to lysosomal vesicles in HT29 colorectal cancer cells in response to PRS-110. Confocal images showed an enhanced colocalization of MET receptor with lysotracker positive vesicles in PRS-110–treated samples relative to controls (Supplementary Fig. S2).

The in vivo efficacy of PRS-110 was assessed by its ability to inhibit growth of the U87-MG xenograft model. The model was selected based on expression levels of both receptor and ligand allied to its known dependence on HGF/MET autocrine signaling for growth (26). Tumor bearing athymic mice were treated with increasing doses of PRS-110 or control (daily intraperitoneal). As shown in Fig. 6A, a dose-dependent decrease in mean tumor volume was observed throughout the study when compared to controls. Mice receiving 0.8 or 2.5 mg/kg PRS-110 had significant tumor growth inhibition (57% and 79%, respectively, on day 21), whereas treatment with 7.5 mg/kg resulted in tumor regression. Based on our receptor trafficking studies and epitope analysis, we asked if significant reduction of MET occurred in vivo. Analysis of tumor biopsy material from mice treated with PRS-110 showed a significant reduction in total MET receptor compared to control mice consistent with our in vitro observations (Fig. 6C; P < 0.001).

To assess the ligand-independent activity of PRS-110, a Caki-1 renal cell carcinoma xenograft model was established (Caki-1 cells do not express HGF but have constitutive activation of the MET pathway). Athymic mice were subcutaneously implanted with Caki-1 cells and were randomly assigned to treatment groups once tumors reached an average volume of 100 to 150 mm³. As seen in Fig. 6B, a statistically significant reduction of tumor volume was observed in PRS-110 compared to the control group at doses starting at 0.8 mg/kg and going up to 30 mg/kg. To evaluate the possible mechanism of action for the observed tumor growth inhibition, tumor biopsies from satellite mice treated with PRS-110 were examined for total MET. Consistent with the observations from the U87-MG xenograft model, analysis of Caki-1 tumor biopsy material showed a significant reduction in total MET receptor in mice treated with PRS-110 compared to control mice (Fig. 6D; P = 0.01).

Pharmacokinetics studies of PRS-110 were conducted in three preclinical species (mice, rats, and cynomolgus monkey) to estimate the dosing regimen in humans. The PRS-110 drug profiles followed a characteristic biphasic pattern, entering the terminal linear phase between 8 and 24 hours, which could be well described with a two-compartmental pharmacokinetics (PK) model (Fig. 7A). Basic PK parameters scale with animal weight as expected for this molecule (Fig. 7B) and suggest a half-life of approximately 5 days in humans based on allometric scaling of clearance and volume of distribution in the 3 preclinical species (Table 1). This is in line with the observed terminal half-life of approximately 6 days during the phase I clinical trial of the VEGF-A Anticalin antagonist PRS-050 bearing the identical PEG moiety (34).

To determine biologically relevant human dose levels, predose plasma drug levels were assessed at steady state in xenograft mice receiving therapeutically effective regimens. Analysis across multiple independent in vivo studies determined that a mean trough level (C_min,ss) of 789 nmol/L would provide effective dosing. Interestingly, these drug levels correspond to the IC₉₀ level in cell-based assays. To approximate a human dosing regimen required to achieve similar drug trough levels, we used the human PK parameters (elimination rate constant and volume of distribution) determined for PRS-050 (VEGF-A targeting Anticalin) during its phase I clinical trial to extrapolate trough levels for PRS-110 (34). PRS-050 belongs to the same drug class as PRS-110. It is based on the human tear lipocalin scaffold and is conjugated to the same branched 40 kDa PEG (2× 20 kDa PEG, NOF) as PRS-110, which is likely to be the key factor determining the PK parameters of these drugs.

Under the assumption of a 2-week dosing interval, the approach described by Rowland and Tozer predicts a dosing requirement of 10 mg/kg to maintain drug levels above 789 nmol/L (43). Thus, this calculation supports the feasibility of administering acceptable doses of PRS-110 as part of a routine chemotherapy regimen to achieve efficacious human dosing.

Deregulated activation of HGF/MET signaling has been implicated in the development and progression of aggressive cancers (9–14). For this reason, the pathway has been identified for potential therapeutic targeting with multiple test agents undergoing clinical evaluation. Although HGF is a key initiator of MET signaling, recent preclinical data have shown the importance of HGF-independent activation of MET. Mitra and colleagues showed that activation of MET by fibronectin and α5β1-integrin regulates ovarian cancer invasion and metastasis (16). They postulate that ligand-independent signaling potentially forms an important first step in ovarian cancer metastasis (16). Dulak and colleagues also describe ligand-independent activation of MET via cooperation with EGFR in non–small cell lung carcinoma, suggesting that lateral communication with MET is required to maximize the tumorigenic impact of EGFR signaling (15). The complexity of ligand-independent MET activation is highlighted by its ability to behave as a promiscuous receptor that can potentially cooperate with many cell surface proteins including Erb3, EGFR, α6β4 integrin, the semaphorin receptors of the plexin B family, and CD44v6 to induce cell signaling or tumorogenesis (15, 16, 44, 45). Optimal therapeutic inhibition of the MET pathway should therefore target both ligand-dependent and ligand-independent mechanisms of MET activation. Although there are multiple experimental therapeutics undergoing clinical evaluation, each current drug class has its limitations. Development of small molecule MET inhibitors has proved challenging due to specificity/selectivity profiles. Moreover, they can lead to stabilization of their target receptor on the cell surface, which implies the risk of prolonged signal transduction (8). Biologics (bivalent or one armed Abs) are limited by their inability to impact on ligand-independent activation of the receptor and/or residual agonistic activity (2, 24–28).

Leveraging a novel biotherapeutic platform with robust pharmaceutical properties, we hypothesized that Anticalins would be “fit for purpose” in light of their monovalency and smaller molecular weight for targeting the MET pathway. Here we describe the development and characterization of a MET inhibitor, PRS-110, which acts as both a HGF antagonist and an inhibitor of ligand-independent tumorigenesis, with no agonistic activity. Furthermore, PRS-110 displays highly desirable biophysical properties (e.g., target specificity, ease of expression, storage stability), which together with its efficacy profile support its further development. PRS-110 is a full antagonist of HGF-induced phosphorylation of MET, whereas it also inhibits key downstream effector kinases, ERK1/2 and AKT (Fig. 2B–D). The safety profile of Anticalin-based therapeutics is supported by the data generated in a recent phase I study with Angiocal (a VEGF targeting Anticalin). In a 26 patient study, the Anticalin was well tolerated and there were no detectable anti–drug antibodies (ADA) identified, following repeat dosing for up to 10 months (34). Furthermore, PRS-110 was shown in an ex vivo assay to have a projected immunogenicity profile similar to licensed monoclonal antibodies with the lowest immunogenicity signatures on the market. These findings suggest that PRS-110 has a suitable safety profile for clinical evaluation.

The binding characteristics of PRS-110 were evaluated in protein-based binding assays, which showed highly specific binding to MET with no recognition of related receptor tyrosine kinases even at micromolar concentrations (Fig. 2E). Flow cytometry was also used to show PRS-110 binds to MET positive cells and is internalized (Fig. 4). To study the impact of PRS-110 binding and internalization on the fate of MET receptor multicolor, confocal microscopy was conducted. It was shown that upon exposure to PRS-110, the MET receptor trafficked from the plasma membrane to intracellular compartments. We also wished to assess whether PRS-110-induced internalization differed from the classical ligand-induced internalization. It is known that upon HGF stimulation MET is endocytosed in clathrin-coated vesicles and directed to early endosome compartments so this was examined (46). As anticipated, our data shows that HGF leads to colocalization of the receptor with early endosomal vesicles (EEA-1 positive). We do not however observe colocalization of MET in EEA-1 positive vesicles in cells treated by PRS-110, suggesting that MET internalization induced by HGF and by PRS-110 are two distinct phenomena and follow different pathways. Dual staining with LAMP-1 showed that after one-hour incubation, PRS-110 led to an accumulation of the MET receptor in terminal endo-lysosomal vesicles, indicating a ligand-independent influence of PRS-110 on the MET receptor.

To further elucidate the mechanism of action of PRS-110, epitope mapping was conducted. Two putative binding epitopes were identified, one in the Sema domain and the second in the IPT domain of MET. Both the Sema and IPT regions of MET are known to be involved in HGF binding and receptor activation. The resolved crystal structure of the ligand and receptor complex identified the interaction of MET Sema domain with the β chain of HGF (38). Multiple indirect lines of evidence support the interaction of IPT domain with the α chain of HGF (39). It is believed that these distinct binding sites cooperate to induce maximal receptor signaling. Based on previous domain characterization it is likely that PRS-110 prevents HGF binding to MET through interaction with the Sema domain epitope. This epitope may also have a ligand-independent role in blocking receptor dimerization (38, 39). It is also possible that through its combined binding to both the Sema domain and the IPT domain epitopes PRS-110 locks MET receptor in a specific conformation thus preventing rotation of the SEMA domain on the PSI axis. This unique binding pattern of PRS-110 may also help explain its impact on receptor internalization.

The in vivo efficacy of PRS-110 was shown in both ligand-dependent and ligand-independent xenograft models. Treatment of established HGF-dependent U87-MG tumor xenografts with PRS-110 led to a dose-dependent reduction in tumor volume and tumor regression at doses of 2.5 and 7.5 mg/kg. Treatment of established HGF-independent Caki-1 tumor xenografts with PRS-110 led to significant tumor growth inhibition even with doses as low as 0.8 mg/kg. To gain further insight into the mechanism of action of PRS-110, tumors from animals treated with PRS-110 were evaluated for pharmacodynamic markers. A significant reduction in total MET was observed for PRS-110-treated mice in both U87-MG and Caki-1 models. This ligand-independent mechanism of action is supported by the in vitro confocal microscopy findings where PRS-110 leads to MET colocalization with lysosomal markers. It suggests that the Anticalin may have broader antitumor activity (e.g., in settings were MET activation is ligand independent) than several clinical candidates including those targeting HGF or MET inhibitors whose activity is limited to the HGF-dependent cancers. These findings also suggest that monitoring of total MET may provide a useful early pharmacodynamics readout in the clinic. To further assess the clinical utility of PRS-110, human dosing estimates were extrapolated from preclinical studies and pharmacokinetic observations of PRS-050 (another tear lipocalin derived Anticalin with identical half-life extension format). Based on steady-state serum levels of PRS-110 required to achieve efficacy in xenograft models allied to IC₉₀in vitro drug concentrations it was estimated that biological effective dosing would be achieved at approximately 10 mg/kg every 2 weeks. This is in line with therapeutic regimens currently in use for other targeted biologics.

In summary, we describe a novel highly specific and potent MET inhibitor with both ligand-dependent and ligand-independent activity that induces receptor degradation in vivo. Furthermore, PRS-110 shows potent antitumor efficacy in a ligand-independent setting. These data suggest that PRS-110 displays highly desirable drug properties that support its evaluation in patients.

H. Gille has ownership interest (including patents) in Pieris. A.M. Hohlbaum has ownership interest (including patents) in Pieris. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S.A. Olwill, C. Joffroy, H. Gille, M. Hülsmeyer, K. Jensen, A.M. Hohlbaum, L. Audoly

Development of methodology: S.A. Olwill, H. Gille, E. Vigna, M. Hülsmeyer, L. Audoly

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.A. Olwill, G. Matschiner, A. Allersdorfer, J. Jaworski, J.F. Burrows, C. Chiriaco, M. Hülsmeyer, L. Audoly

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.A. Olwill, C. Joffroy, H. Gille, E. Vigna, G. Matschiner, A. Allersdorfer, B.M. Lunde, C. Chiriaco, M. Hülsmeyer, K. Jensen, A.M. Hohlbaum, L. Audoly

Writing, review, and/or revision of the manuscript: S.A. Olwill, C. Joffroy, H. Gille, A. Allersdorfer, H. Christian, M. Hülsmeyer, S. Trentmann, A.M. Hohlbaum, L. Audoly

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.A. Olwill, C. Joffroy, A. Allersdorfer, H. Christian, L. Audoly

Study supervision: S.A. Olwill, L. Audoly

The authors thank all technical staff at Pieris for their contributions to this body of work. In addition, thanks to L. Lanzetti for her help with analysis and interpretation of the confocal microscopy work. Thanks also to J. Benchop for his assistance with epitope mapping.

This project was in part-funded by the German Federal Ministry of Education and Research within the Leading-Edge Cluster “m4 – Personalized Medicine” in Munich (Grant number 01EX1122 awarded to A.M. Hohlbaum). All additional work was funded by Pieris.

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.