Preclinical Profile of BYON3521 Predicts an Effective and Safe MET Antibody–Drug Conjugate

Abstract

MET, the cell-surface receptor for the hepatocyte growth factor/scatter factor, which is widely overexpressed in various solid cancer types, is an attractive target for the development of antibody-based therapeutics. BYON3521 is a novel site-specifically conjugated duocarmycin-based antibody–drug conjugate (ADC), comprising a humanized cysteine-engineered IgG1 monoclonal antibody with low pmol/L binding affinity towards both human and cynomolgus MET. In vitro studies showed that BYON3521 internalizes efficiently upon MET binding and induces both target- and bystander-mediated cell killing. BYON3521 showed good potency and full efficacy in MET-amplified and high MET–expressing cancer cell lines; in moderate and low MET–expressing cancer cell lines good potencies and partial efficacy were observed. In mouse xenograft models, BYON3521 showed significant antitumor activity upon single-dose administration in multiple non-MET–amplified tumor types with low, moderate, and high MET expression, including complete tumor remissions in models with moderate MET expression. In the repeat-dose Good Laboratory Practice (GLP) safety assessment in cynomolgus monkeys, BYON3521 was well tolerated and based on the observed toxicities and their reversibility, the highest non-severely toxic dose was set at 15 mg/kg. A human pharmacokinetics (PK) model was derived from the PK data from the cynomolgus safety assessments, and the minimal efficacious dose in humans is estimated to be in the range of 3 to 4 mg/kg. In all, our nonclinical data suggests that BYON3521 is a safe ADC with potential for clinical benefit in patients. A first-in-human dose-escalation study is currently ongoing to determine the maximum tolerated dose and recommended dose for expansion (NCT05323045).

Introduction

The hepatocyte growth factor receptor (HGFR, MET) is a tyrosine kinase encoded by the MET proto-oncogene, which under normal circumstances plays a key role in maintaining tissue homeostasis by controlling a variety of important cellular responses, including proliferation, angiogenesis, apoptosis, motility, and morphogenesis (1). Aberrations in the hepatocyte growth factor (HGF)/MET pathway leading to the overexpression and/or constitutive activation of MET promote the development and progression of various cancer types and are associated with a significantly worse clinical outcome (1, 2).

Therapeutic compound classes that have been and are being explored extensively either as monotherapy or in combination with chemotherapeutics or other targeting agents, are selective and nonselective small-molecule inhibitors of MET signaling and monoclonal antibodies that bind HGF or compete with HGF for binding to MET (2–4). Several small-molecule inhibitors have been approved, but their clinical benefit has been underwhelming suggesting that antagonizing the MET/HGF signal transduction pathway in patients with MET overexpression is not a sufficiently effective strategy. Recent insights learned that the success of these therapies appeared to be more dependent on the constitutive activation of this pathway rather than MET overexpression alone, as is seen for example in tumors with exon 14 skipping mutations in MET which prevent MET degradation (5–7). For this reason, the new generation selective MET TKIs, capmatinib and tepotinib have only been approved for the treatment of patients with metastatic non–small cell lung cancer (NSCLC) with exon 14 skipping mutations (8, 9).

Designing anti–MET or -HGF antibodies for use in targeted therapy in patients with cancer also proved to be challenging. To date no antibody specific for MET or HGF has been approved for use in cancer therapy. Most programs have been terminated because of lack of efficacy or safety concerns (3, 4). The only approved biological that targets MET is the bispecific antibody RYBREVANT (Amivantamab-vmjw) known to bind to both EGFR and MET. Accelerated approval status was granted by the FDA after positive results in a phase I study. Amivantamab has been approved for the treatment of adult patients with locally advanced or metastatic NSCLC with EGFR exon 20 insertion mutations. These tumors are driven by EGFR and/or MET signaling (10, 11). Recent data also showed encouraging results in patients with NSCLC with MET exon 14 skipping mutations in a separate cohort of the CHRYSTALIS study (NCT02609776; ref. 12).

Antibody–drug conjugates (ADC) are a promising alternative therapeutic strategy because they employ a different modus operandi: their efficacy depends on MET expression on the cell surface to mediate binding and cellular uptake of the conjugated toxins, rather than on downstream signaling. Because elevated MET expression is a much more widespread phenomenon in cancer compared with the prevalence of MET aberrations leading to MET signaling addiction, it is therefore still a very promising cell surface protein for the targeted delivery of toxic payloads. Several MET–targeting ADCs composed of different antibodies, linker systems and payload classes have shown promising preclinical activity and some have entered clinical development. Telisotuzumab vedotin (ABBV-399) composed of an anti–MET antibody ABT-700 conjugated to the microtubule inhibitor monomethylauristatin E (13) is the most advanced program, and the FDA recently granted this program the Breakthrough Therapy Designation for patients with previously treated MET overexpressing NSCLC. Nonetheless, the therapeutic window for most ADCs approved to date is still quite narrow as the MTDs are seemingly dictated by the payloads (14). To broaden the therapeutic window, it is therefore important to optimize ADC features that can improve efficacy and/or tolerability. In this regard, the preclinical profile of BYON3521, composed of a novel anti–MET antibody site-specifically conjugated (15) with the vc-seco-DUBA linker payload, holds promise. BYON3521 showed good efficacy in cell lines and in vivo tumor models with low, moderate, and high MET expression, while at the same time exhibiting an encouraging safety profile in a 4-cycle Good Laboratory Practice (GLP) toxicity study in cynomolgus monkey, predicting a therapeutic window that will allow efficacious and safe treatment of patients with MET–positive cancers. A clinical phase I dose-escalation study is currently underway (NCT05323045).

Materials and Methods

Generation of BYON3521

Murine anti–MET antibodies were generated by Nanotools (Teningen, Germany). Humanization was performed at Byondis, during which also the heavy chain 41C (HC-41C; Kabat numbering) mutation was introduced. The sequence is published in WO 2022/214517 A1 (16). Antibody mAb3b (SYD2884) was selected to produce BYON3521 and is composed of HCVR SEQ ID NO: 16, LCVR SEQ ID NO: 20, HCCR SEQ ID NO: 22, and LCCR SEQ ID NO: 23; ref. 16). Antibodies were site-specifically conjugated using a proprietary conjugation method as described in the paper of Coumans and colleagues (15). BYON3521 has a drug-to-antibody ratio (DAR) of about 1.8.

Cell lines

Human tumor cell lines Hs746T, NCI-H441, HCC1954 were obtained from ATCC (Rockville, MD). The human tumor cell lines BxPC-3, HepG2, PC-3, and MKN-45 were obtained from German Collection of Microorganisms and Cell Cultures (DSMZ-Leibniz). The human tumor cell line Jurkat Nuclight Red was obtained from Essen bioscience (Sartorius). The human tumor cell line A2780 was obtained from The European Collection of Authenticated Cell Cultures (Salisbury, United Kingdom) and EBC-1 and KP4 from RIKEN BRC (Tsukuba, Japan). Cell line authentication was performed using the short tandem repeat profiling service of Eurofins.

The cells were maintained in the culture medium recommended by the supplier, supplemented with 10% (v/w) FBS [heat-inactivated (HI); Gibco-Life Technologies; Carlsbad, CA], only for HCC1954 10% v/w FBS Qualified was used and 1% (v/w) penicillin–streptomycin (Lonza); except for PC-3, which was cultured in RPMI, and the A2780 and BxPC-3 which were cultured w/o glutamine. The culture medium of the Jurkat NucLight Red cells was supplemented with 0.5 μg/mL Puromycin.

Master and working cell banks were generated in up to two passages. When freezing the master and working cell banks Mycoplasma testing was performed at Minerva Analytix GmbH using qualitative real-time PCR. Every three months all cell lines in culture were again tested for Mycoplasma contamination. Cell lines were kept in culture for a maximum of 3 months after thawing of the working cell bank vial.

Transient expression of MET in mammalian cells

MET antigens (full length and MET ECD) were cloned in the CMV:BGHpA expression cassette of the commercially available (Thermo Fisher Scientific) mammalian expression vector pcDNA3.4. Large-scale preparation of the final expression vectors for transfection was performed using Maxi- or Megaprep kits (Qiagen).

Commercially available ExpiCHO-S cells (Thermo Fisher Scientific) were transfected with the vectors expressing full length MET using the ExpiFectamine CHO transfection agent according to the manufacturer's instructions. One day after transfection, the cell cultures were used for dose-dependent cell binding analysis or cryopreserved for later use in binding assays.

Quantification of cellular MET antigen expression

MET antigen expression on the surface of the human tumor cell lines was determined with a mouse anti-human HGFR/MET antibody (R&D systems, Bio-Techne) and a FITC-conjugated goat anti-Mouse F(ab')2 as detection antibody as described in the manufacturer's protocol of the Qifi kit (DAKO, Agilent).

Cellular binding assays

Cellular binding was evaluated on a set of 12 cell lines with varying MET expression levels as well as to ExpiCHO-S cells expressing recombinant human and cynomolgus MET. Cells were harvested at 80% of confluency. Staining was performed in 96-well round-bottomed microtiter plates (100,000 cells/well using chilled reagents/solutions at 4°C to prevent the modulation and internalization of surface antigens). Cells were incubated with a serial dilution of antibody or ADC followed by a 30-minute incubation (protected from light) with R-PE-conjugated F(ab')2 fragment goat anti-Human IgG, (1:400; Jackson ImmunoResearch) or Allophycocyanin (APC) AffiniPure F(ab’)2 fragment goat anti-human IgG (1:200; Jackson ImmunoResearch). Antibody or ADC binding was evaluated in FACS buffer in 96-well black clear bottom plates (Pelkin Elmer) on the EnVision plate reader.

Cellular binding of BYON3521 to human and cynomolgus MET expressed in ExpiCHO-S cells was evaluated using the same procedure with the exception that the APC-conjugated secondary F(ab’)2 goat anti-Human IgG (Fc fragment specific) antibody (Jackson ImmunoResearch) was further diluted to 1:6,000.

To evaluate the potential interference of HGF with cellular binding, cells were preincubated with 320 ng/mL recombinant human (rh)HGF (30 minutes at 4°C; Sino Biological; mixture of pro-HGF and activated HGF) prior to incubation with antibody or ADC.

Cell viability assays

Cell viability was assessed in 384-well culture plates after 6 and/or 9 days of incubation by measuring the DNA content of living cells assessed with the fluorescence based CyQuant Cell Proliferation Assay (Invitrogen, Thermo Fisher scientific) according to the manufacturer's instructions.

Cells were plated at different optimized densities (EBC-1 600 cells/well, Hs746T 1,000, MKN-45 650, NCI-H441 2,000, BxPC-3 450, HCC1954 1,500, Hep-G 2,500, KP4 400, A2780 4,000, and Jurkat Nuclight Red 600 cells/well). Cell viability was expressed relative to the average mean survival of untreated cells (only growth medium). For DUBA (free toxin), cells cultured in the presence of 1% DMSO were used as control. When the incubation times were extended to 9 days, cell densities were adjusted for some cell lines (NCI-H441 1,000 cells/well, HEP-G2 1,250, and MKN-45 3,500 cells/well). Due to the low cell densities for the PC-3 cells, the assay was performed in 96-well plates instead, in which 600 and 1,000 cells/well were seeded and incubation periods were respectively 9 and 6 days.

To evaluate the potential interference of HGF with cellular binding, cells were incubated with antibody, ADC or free toxin in the presence of 32 ng/mL rhHGF (Sino Biological).

In addition, interference by soluble MET (sMET) was mimicked by repeating above-mentioned cell viability assays with the MET high EBC-1 cells the presence of recombinant MET ECD (Acro Biosystems). Cells were incubated with of ADCs or DUBA in the presence of 1 μg/mL solution of recombinant MET ECD. In a separate experiment, the recombinant MET ECD and the ADCs were preincubated together for 30 minutes at 37°C, prior to administration to the cells. Cell viability was assessed with the CellTiterGlo Assay (Promega) according to the manufacturer's instructions.

Biolayer interferometry

The impact of the presence of fully activated recombinant human rhHGFrhH (R&D Systems) versus the inactive recombinant human pro-HGF (Sigma) in the binding of SYD2884 to MET was also studied using Biolayer interferometry (BLI). To confirm this, nonreducing SDS-PAGE was performed using the Criterion TGX Stain Free Precast Gel system (4%–20%; Bio-Rad). Samples were 15 incubated for 15 minutes at 60°C loading buffer containing N-ethyl maleimide (Sigma) and Laemmli sample buffer (Bio-Rad) and 1 μg of protein was loaded on a 4% to 20% Criterion TGX Stain free precast gel and run at 300 V for 24 minutes in Tris-Glycine-SDS buffer (TGS buffer, Bio-Rad) to separate the proteins based on their molecular weight. For visualization of the proteins the gels were placed in a GelDoc EZ stainfree imager (Bio-Rad).

SAX sensortips (Sartorius AG, Göttingen, Germany) were loaded with AVI-tagged biotinylated human MET (approximate MW 140 kDa; ACRO Biosystems, Newark, Delaware) with a contact time of 300 seconds at 45 nmol/L (6.25 μg/mL) to a level of 1.5 nm. All dilutions were made in 0.01 mol/L HEPES pH 7.4, 0.15 mol/L NaCl, 3 mmol/L EDTA, 0.005% v/v Surfactant P20 (HBS-EP+, Cytiva, Malborough, Massachusetts). Remaining streptavidin binding sites were blocked with 27 μmol/L (10 μg/mL) Biocytin (ThermoFisher, Waltham, Massachusetts) for 300 seconds.

Prior to performing the binding assay, a fixed amount of SYD2884 (2.5 nmol/L) was preincubated with increasing concentrations of rhHGF (range 0–250 nmol/L) from R&D Systems consisting of fully activated HGF, and rhHGF from Sigma consisting of only pro-HGF; Supplementary Fig. S1). Each analysis included a 120-second baseline assessment, a binding association phase and was followed by a 60-second dissociation time. The binding of the SYD2884 + HGF mixtures was recorded for 300 seconds. SYD2884 binding was monitored using an anti-hIgG-HRP (The Binding Site, Birmingham, United Kingdom; AP003.M, 5 minutes, 1:500), HGF binding was detected using a polyclonal anti-HGF antibody (Abnova, Taipei City, Taiwan; H0003082-DOIP, 10 minutes, 20 μg/mL).

Lysosomal routing and internalization kinetics

Internalization kinetics studies were performed with MKN-45 cells. SYD2884 was labeled with the Alexa Fluor 488 antibody labeling kit (Molecular Probes). The labeled antibody (SYD2884-AF488) was concentrated using 10K Amicon filters (Millipore). The concentration was estimated by A280 measurements using the NanoDrop Lite (Thermo Scientific) and the average degree of labeling was estimated to be 5.4 by A494 measurements using a UV-VIS spectrophotometer (Shimadzu). Translocation kinetics of cell-bound Alexa-Fluor488-labeled SYD2884 (SYD2884-AF488) from cell surface to cytoplasm was monitored by fluorescence imaging. A series of cell images taken over time were analyzed using the quantitative image analysis algorithm in Python v3.7, which analyses pixel locations in different cellular compartments. The internalization percentage at each time-point was calculated by dividing the fluorescence signal from the cell cytoplasm through the total fluorescence signal in the cell (cell cytoplasm + cell membrane). Background corrections were performed by subtracting the average background signal per pixel from each pixel in cytoplasm and membrane regions. Subsequently, the internalization rate (k_int) and internalization half-life were calculated.

To follow the intracellular trafficking of SYD2884 and BYON3521 towards the lysosomes, a pH-sensitive dye was used. Outside the cells at neutral pH the dye is nonfluorescent, and it becomes increasingly brighter when the pH drops in case the labeled antibodies travel from early endosomes to the lysosomal compartments. MKN-45 cells (15,000 cells/well) were incubated overnight with SYD2884, BYON3521 and a nonbinding isotype control ADC that were pre-labeled for 15 minutes with IncuCyte Human FabFluor-pH Red (molar ratio 1:3), while being protected from light. Plates were imaged in the IncuCyte S3 at 30-minute intervals for at least 24 hours. Percentage of lysosomal routing was expressed as the Red Area (Red positive cells)/Phase Area (Total cell confluency).

Bystander activity

The ability of BYON3521 to induce bystander killing was mimicked by coculturing MET–negative Jurkat NucLight Red and MET–positive tumor cells (1:1 ratio; total of 10,000 cells/well; 96-well plates coated with 0.1 mg/mL Poly-L-Ornithine) and treating the mixed cell population with BYON3521. The impact on the MET–negative cell population was visualized by following the proliferation of Jurkat NucLight Red cells in real time using the IncuCyte S3. A non-bystander control, an anti–MET (SYD2884)-mcMMAF ADC and a nonbinding ADC control were included in the experiments.

To determine the minimum proportion of MET–positive target cells required for BYON3521-mediated bystander killing, different ratios of MET–positive MKN-45 and MET–negative Jurkat NucLight Red cells were cocultured in 96-well plates pre-coated with Poly-L-Ornithine (total of 10,000 cells for coculture and 5,000 cells single culture in 90 μL/well), resulting in 100%, 80%, 60%, 40%, 20%, and 0% MET–positive cells.

Antibody-dependent cellular cytotoxicity and complement-dependent cytotoxicity

The immunological effector functions of SYD2884 and BYON3521 were evaluated using the antibody-dependent cellular cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC) assays.

The CDC assay was performed using EBC-1, Hs746T, MKN-45 (20,000 cells/well) and Daudi cells (15,000 cells/well). Cells were preincubated for 15 minutes at room temperature with 10 μg/mL mAb or ADC after which (HI) pooled normal human serum (NHS, Sanquin, Amsterdam) from healthy donors was added as source of complement (10%) and cells were incubated for 1 hour at 37°C. Cell viability was assessed using the CellTiter-GloTM (CTG) luminescent assay kit (Promega Corporation, Madison, WI). The controls included were a nonbinding isotype control ADC and its unconjugated antibody. CDC induction by rituximab using CD20-expressing Daudi cells as target cell population was taken as a positive assay control to validate the complement activity in the pooled human serum.

BYON3521 and SYD2884 were also evaluated in the ADCC assay using peripheral blood mononuclear cells (PBMC) as effector cells and MKN-45 as MET–expressing target cells. MKN-45 cells (20,000 cells/well; 50 μL/well) were preincubated for 15 minutes at 37°C, 5% CO₂ in the presence of BYON3521, SYD2884, nonbinding isotype control ADC and its isotype control antibody. PBMCs were isolated from human whole blood obtained from healthy donors using a Lymphocyte Separation Medium (Lonza). Cryopreserved PBMCs were recovered overnight at 37°C, 5% CO₂ in RPMI1640 medium containing 10% v/w FBS HI (Gibco-Life Technologies). PBMCs were added to the antibody- or ADC-treated MKN-45 cells at an effector-to-target ratio of 30:1. After 4 and 24 hours of incubation cell supernatant was transferred to a 96-well plate to determine the amount of released lactate dehydrogenase (LDH) using a colorimetric assay (Roche, Basel, Switzerland). Maximum release is determined by incubating with lysis buffer and determine all LDH released from effector and target cells separately. Spontaneous release by these cells is determined in the presence of only culture medium. The experimental release is the LDH released when all antibody solution, effector and target cells are mixed. Percent lysis of target cells is then calculated as follows:

Binding kinetics studies using the LigandTracer green

Binding kinetics of SYD2884 to MET was studied using a fluorescent real-time cellular binding assay (LigandTracer Green; Ridgeview Instruments AB, Sweden). By measuring the association and dissociation of different concentrations of SYD2884-AF488, the k_a, k_d, and K_D can be determined. This was done on three cell lines with different MET expression levels, EBC-1, NCI-H441 and PC-3. Cells (1.5 × 10⁶) were seeded onto two quadrants of a 2 × 2 Multidish (Ridgeview Instruments AB, Sweden) and grown until 90% to 00% confluency. The dish is placed onto a rotating, tilted platform in the LigandTracer Green instrument (Ridgeview Instruments AB, Sweden). The antibody SYD2884 was labeled with the Alexa Fluor 488 as described earlier. Cellular binding was subsequently monitored in real-time for 0.5, 0.97, 1, 1.7, 2, 2.47, 4, 4.94, 5, 8, and 12 nmol/L SYD2884-AF488 added to the medium with a blue-green detector (Ex: 488 nm, Em: 535 nm). Prior to the measurements cells were preincubated for 15 minutes at room temperature in culture medium supplemented with 10 nmol/L TMI-1 (Sigma) to inhibit metalloprotease activity to prevent MET shedding, and 10 mmol/L SRI31215 (MedChemExpress) to inhibit activation of pro-HGF to active HGF (17, 18). During the measurements first a baseline was measured for 15 minutes before one concentration of the concentrations was added to the medium and association to MET on adherent cells was measured for 30 minutes or 1 hour.

All real-time cell-binding assays were executed at room temperature and as all measurements were executed in 2×2 Multidishes (Ridgeview Instruments) two individual experimental conditions could be measured in the same dish. Specific binding is calculated by subtracting the signal measured in the control quadrants from the target cells-containing quadrants.

In addition, the impact of exogenous recombinant human MET-ECD-Fc (approximate MW 130 kDa; HGFR, Sino Biological) on the binding of SYD2884 to the cells was studied. Cells in the Multidishes were preincubated for 30 minutes in the presence of 7.7 nmol/L (1 μg/mL) recombinant MET ECD.

Data were analyzed using TraceDrawer (v. 1.9.1, Ridgeview Instruments AB) after baseline correction. A 1:1 Langmuir model was used to estimate the kinetic parameters and K_D.

Mouse xenograft studies

The performance of BYON3521 in vivo was evaluated in the gastric adenocarcinoma MKN-45 cell line–derived xenograft model (CDX; MET amplified), a mouse clinical trial (MCT) in patient-derived xenograft (PDX) models, as well as in two single rising dose studies. All in vivo studies were performed in carboxylesterase 1c (Ces1c) knockout SCID mice (CES1c KO SCID mice). The Ces1c is a rodent-specific enzyme which turned out to remove the linker-drug from the ADCs, a process which does not occur in monkey or man (19). After primary tumors have reached a volume of approximately 100 to 200 mm³, tumor-bearing animals were randomized over the treatment groups according to tumor volumes and were dosed on the day of randomization or on the following day with a single intravenous dose injection of 10 mg/kg ADC.

A MCT was performed in non-MET–amplified PDX models with low, moderate, and high MET expression. Included were PDX models of NSCLC, breast cancer, gastric cancer, head & neck cancer, and pleural mesothelioma. In addition, four melanoma PDX models were included; these tumors do not express MET. Mice were dosed with a single intravenous injection in the tail vein of 10 mg/kg BYON3521, and in the case of MAXF574, 3 mg/kg because this is a model known to be highly sensitive to DUBA. We used 3 mice per treatment group and the change in tumor volume was calculated relative to the average tumor volume at the moment of randomization and dosing.

Dose–response studies were performed in two PDX models, the LXFL1176, a large cell lung carcinoma derived from a lymph node metastasis, and HNXF1905, a primary squamous cell head and neck carcinoma from the supraorbital area. The MET gene is not amplified in these tumors and IHC staining showed MET 3+ expression on the cell surface in the LXFL1176 model, and MET 2+ expression in the HNXF1905 model. The mice were dosed on the day of randomization or on the following day with a single injection of 0.3, 1, 3, or 10 mg/kg BYON3521 or a nonbinding isotype control ADC.

Antitumor activity of ADCs was assessed by measuring tumor sizes twice weekly after single-dose treatment. The length and width of the tumor were measured with calipers and the volume of the tumor estimated by the formula:

Changes in tumor volume were calculated relative to the tumor volumes at the time of randomization. In cases where an antitumor response was apparent the maximum observed response was considered, in cases where tumors showed progressive disease the change in tumor volume was assessed at 21 days posttreatment.

A complete response (CR) is achieved when the change in tumor volume is −100%. When tumors do not respond to BYON3521 and continue growing, tumor response can be > 100% (in the waterfall plot the bars are maximized at 100%). In the cases where a reduction in tumor volume is observed of > 30% but < 99% this was considered a partial response (PR). When the tumor volume does not increase > 20% and is not reduced > 30% this was considered stable disease. To assess whether the effects are mediated through MET binding, nonbinding isotype control ADCs with similar DAR were used. Activity of the isotype control ADCs indicates that tumor cells have been exposed to the toxin, which can be achieved through for instance the extracellular cleavage of the vc-seco-DUBA or the nonspecific uptake of the ADC through micropinocytosis by the cells.

In the dose–response studies blood samples were taken from the mice by retro-orbital sinus puncture under anesthesia at different time points after dosing, cooled on ice water and processed to K2-EDTA plasma as soon as possible. Subsequently, plasma samples were snap frozen in liquid nitrogen and stored at –70°C until bioanalysis.

All animal studies were conducted at Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC)-certified CROs (Reaction Biology, Freiburg, Germany, for the MKN-45 model, Charles River, Freiburg, Germany for the PDX models), using well-established protocols developed and maintained at these CROs and have been approved by the Institutional Animal Care and Use Committee.

IHC

Cynomolgus tissues were collected at autopsy in the BYON3521 DRF study. A multi-normal human tissue microarray was purchased from Abcam (reference ab178224). Human eye tissues were obtained via Tissue For Research Ltd. (dba Accio TX Biobank Online Ltd.). Tissue was procured from three donors through rapid autopsy at Restore Life USA (Elizabethton, TN). This nonprofit organization (501c3) provides postmortem non-transplantable human tissues for medical research and education purposes through a volunteer whole-body donation program. Tissues were collected and processed by a specialist certified through the American Association of Tissue Banks board. Donor IDs were anonymized and written informed consent was given. Postmortem serum testing (Hepatitis B, C, HIV-1/-2, and Syphilis) was performed at VRL Eurofins (Centennial, CO).

All IHC procedures were performed at Cerba Research (Montpellier, France). IHC was performed on the Ventana Discovery XT automated platform.

Tissue sections of formalin-fixed, paraffin-embedded tissues were subjected to a 40-minute antigen retrieval in 0.05% citraconic acid. Sections were then incubated for 12 hours at room temperature with anti–MET antibody (Clone SP44, rabbit monoclonal, 2.2 μg/mL; Eurobio, reference M3440) or anti-phosphorylated MET (pMET) tyr1234/1235 antibody (Clone D26, rabbit monoclonal, 1:50; Cell Signaling Technology, reference 3077S) and antibody binding was visualized with the OmniMap RB HRP detection kit (16-minute incubation; Ventana-Roche, reference 760–4311) and the ChromoMap DAB chromogenic kit (Ventana-Roche, reference 760–159). The isotype control used was rabbit IgG (Abcam, reference ab172730).

Frozen sections were stained on the Benchmark Ultra platform (Roche-Ventana). Sections were fixed for 10 minutes in 10% neutral buffered formalin and incubated for 32 minutes with 0.5 or 1 μg/mL of the FITC-conjugated anti–MET antibody, SYD2884. Antibody binding was visualized with the mouse anti-FITC antibody (1:250 dilution, 32 minutes; Jackson ImmunoResearch, reference 200062037) and the UltraviewDAB detection kit (Ventana Medical Systems, reference 760–500).

In vitro plasma stability

The ADC BYON3521 was spiked in pooled mouse CES1c knockout mouse, cynomolgus monkey (Macaca fascicularis) and human K2-EDTA plasma at a concentration of 100 μg/mL and incubated at 37°C. After 0, 1, 6, 24, 48, and 96 hours of incubation, plasma samples were snap frozen and stored at –70°C until bioanalysis. On the basis of the time-concentration data, the in vitro plasma half-life (t_½) was calculated using the equation t_½ = 0.693/b, where b is the slope found in the linear fit of the natural logarithm of the fraction remaining of total antibody or conjugated antibody versus incubation time.

Also, BYON3521 was studied in vitro for its propensity of hemolysis and RBC clumping in human whole blood as well as precipitate formation in plasma.

BYON3521 specificity screen

To confirm the binding selectivity of BYON3521, the human cell array technology of Retrogenix (Charles River Laboratories) was employed. This screening assay tests ligands for binding against 5484 full-length human cell surface membrane proteins and cell surface-tethered secreted proteins that are individually overexpressed in the human cell line HEK293.

Safety evaluation

A four cycle GLP toxicity study was performed in which four groups of naïve cynomolgus monkeys (Macaca fascicularis, 4–5 animals/sex/group) received four intravenous 30-minute infusions of 0, 5, 15, or 25 mg/kg, each 3 weeks apart. Two animals/sex in the control group and 3 animals per sex per dose group were sacrificed 3 weeks after the last dose, whereas 2 animals per sex per group were allowed an additional recovery period of 6 weeks and were then sacrificed. Mortality/moribundity, clinical signs, body weight, food consumption, safety pharmacology (respiratory and cardiovascular safety endpoints), clinical pathology, urinalysis (including urine chemistry) and antidrug antibodies were monitored before dosing and throughout the study. Upon terminal sacrifice, macroscopic and microscopic evaluation of selected tissues was performed.

The toxicity studies in cynomolgus monkeys were approved by the animal welfare commission of the North Rhine-Westphalia authority (LANUV, Germany), and were performed in compliance with the German Animal Welfare Act and Directive 2010/63/EU of the European Parliament, the councel of 22 September 2010 on the protection of animals used for scientific purposes, and the Commission Recommendation 2007/526/EC on guidelines for the accommodation and care of animals used for experimental and other scientific purposes (Appendix A of Convention ETS 123).

Bioanalytical assays

All measurements were performed in K2-EDTA plasma. A validated LC-MS/MS based method was used for quantification of active DUBA payload. In addition, an LC-MS method was developed to measure total sMET. This method detects both free sMET and sMET bound to SYD2884 or BYON3521 (LLOQ 200 ng/mL).

A specific conjugated antibody assay was developed, in which only BYON3521 molecules with at least one complementarity-determining region (CDR) not bound by sMET are detected. An electrochemiluminescence immunoassay (ECLIA) method was used for the quantification of BYON3521 in mouse plasma. BYON3521 was captured with a biotinylated mouse monoclonal antibody directed against the toxin (Byondis, clone A15–5-4). Immune complexes were then captured on streptavidin QuickPlex plates (Meso Scale Discovery, Rockville) and detection was performed with a SULFOtag labeled anti-idiotype fragment directed towards the CDR of BYON3521. Plates were read with a MESO QuickPlex SQ 120 instrument (Meso Scale Discovery, Rockville).

Enzyme-linked immunoassay (ELISA) based procedures were used for human and cynomolgus monkey plasma. Due to presence of sMET in both human and cynomolgus monkey plasma, chicken plasma lacking interfering sMET (BioChemed Services, Winchester) was used as a surrogate matrix for preparation of BYON3521 calibration curves. BYON3521 was captured with the biotinylated anti-toxin antibody and immune complexes were captured on EvenCoat streptavidin plates (R&D systems, Minneapolis). Detection was performed with an HRP labeled anti-idiotype fragment, TMB (TeBu-Bio, Le Perray-en-Yvelines, France) was used as a substrate and the color reaction was stopped with H₂SO₄. The plate was read at 450 and 630 nm with a Varioskan 3001 plate reader (ThermoFisher Scientific, Waltham).

A specific total antibody assay (hybrid IA/LC-MS/MS method) was developed that quantifies total (conjugated and unconjugated) BYON3521. BYON3521 with at least one CDR not bound by sMET was captured in plasma by using the biotinylated anti-idiotype antibody coupled to magnetic streptavidin beads (Dynabeads MyOne Streptavidin C1, ThermoFisher Scientific, Waltham). A signature peptide (VLIFGATNLADGVPSR) from the light chain of BYON3521 generated through tryptic digestion was selected for detection. The internal standard that was used for quantification was a stable isotope labelled analogue of the signature peptide containing two cleavable groups (at N- and C-terminus, Pepscan, Lelystad, the Netherlands) offering partial correction for variations in trypsin digestion efficiency. After the digestion resulting in two signature peptides from BYON3521 and the IS, a sample clean-up using solid phase extraction (Oasis HLB 96-well μElution Plate, Waters, Milford) was performed prior to LC-MS/MS analysis (ACQUITY UPLC Peptide BEH C18 Column, Waters, in combination with API 6500+, SCIEX, Framingham). Chicken plasma lacking interfering sMET (BioChemed Services, Winchester) was used as a surrogate matrix for the preparation of BYON3521 calibration curves for the analysis of human and cynomolgus plasma samples.

The specific conjugated antibody assay is DAR sensitive, which implies that a similar total antibody concentration of DAR1 and DAR2 leads to half of the response of DAR1 as compared with DAR2 in the free conjugated BYON3521 assay.

Pharmacokinetics/Pharmacodynamics analysis

To relate the in vivo exposure of BYON3521 [Pharmacokinetics (PK)] to efficacy [Pharmacodynamics (PD)] and to study the PK properties of BYON3521 in tumor bearing mice, BYON3521 concentrations were determined at different time points in the PDX models. BYON3521 was dosed via an intravenous bolus injection into the caudal vein, and blood samples were collected by retro-orbital sinus puncture under anesthesia.

Single and repeat dose PK of BYON3521 were determined in cynomolgus monkey a GLP toxicity study (5, 15 and 25 mg/kg, Q3WKx4; male and female animals). BYON3521 was dosed via a 30-minute intravenous infusion and blood samples were collected from a suitable vein other than what was used for administering BYON3521. Total and conjugated antibodies were quantified in mouse and cyno PK samples as described below. PK parameters were calculated in WinNonlin version 8.2 using the non-compartmental analysis for single intravenous bolus or infusion injection.

Modeling human PK and prediction of human minimal efficacious dose

The software package Phoenix version 8.2, with NLME version 7 (Certara) was used in the analysis. The applied estimation method in NLME was by First-Order Conditional Estimation (Extended Least Squares) or FOCE ELS, with initial fitting using naïve pooling by MAP-NP, with a proportional error model for residual variability on non-transformed data. For linear population PK estimation including interindividual variability, FOCE ELS method with proportional error was applied. The Phoenix WinNonlin version 8.2 program was used for the exploratory analysis including non-compartmental analysis, post-processing of NLME output, and making of graphs. Excel was used for scaling parameter estimation.

Data availability

The data generated in this study are available within the article and its Supplementary Data file.

Results

Binding, in vitro cytotoxicity and bystander activity

The humanized anti–MET antibody SYD2884 was site-specifically conjugated via cysteine residues (HC-41C) in the antibody with vc-seco-DUBA to deliver the well-defined ADC BYON3521 with a DAR of about 1.8 (Supplementary Fig. S2). Antibody-related features such as binding affinity, lysosomal routing, and induction of ADCC activity in vitro were not affected by conjugation of vc-seco-DUBA (Figs. 1A–C). SYD2884 and BYON3521 showed no CDC activity. Moreover, BYON3521 binding to cynomolgus MET, transiently expressed on ExpiCHO-S cells, was comparable with the binding to human MET (Fig. 1D), which supports the relevance of the safety studies in cynomolgus monkeys.

Figure 1.

Antibody-related features of unconjugated SYD2884 and BYON3521. A, Cellular binding of SYD2884 and BYON3521 to MET–expressing MKN-45 cells. (Data show the average RFU ± SD of two experiments tested in duplicate). B, Lysosomal trafficking kinetics of BYON3521, SYD2884 and nonbinding isotype control ADC into MKN-45 cells acquired in the IncuCyte S3 live-cell imaging instrument using the Fabfluor-pH Red antibody (Graph presents the internalization of two independent experiments). C, ADCC activity of SYD2884 and BYON3521 using the MKN-45 cell line as target cell (Data show the percentage of lysis ± SD for three donors tested in duplicate). D, Binding to ExpiCHO-S cells transiently expressing human or cynomolgus MET (Data show the mean ± SEM of two experiments performed in duplicate).

Binding of SYD2884 and BYON3521 to cell surface expressed MET was also evaluated in a panel of cell lines with different MET expression levels. Binding EC₅₀’s of antibody and ADC were highly similar irrespective of the MET expression level of the cell lines (Supplementary Table S1).

Upon binding of the antibody SYD2884, the receptor-antibody complex traffics to the lysosomal compartment for cleavage and release of the drug. Using MKN-45 cells, the k_int of SYD2884 was estimated to be 0.021 min⁻¹ yielding an internalization half-life of approximately 30 minutes.

Surface plasmon resonance appeared not to be suitable to quantify the K_D of SYD2884, due to the extremely slow dissociation of the antibody. Therefore, the LigandTracer was employed to study binding kinetics using live cells with different MET expression levels to estimate both association and dissociation and establish a K_D. These experiments confirmed high affinity binding of SYD2884-AF488 to EBC-1 (K_D 28 pmol/L), PC-3 (K_D 11 pmol/L), and NCI-H441 (K_D 0.11 nmol/L) and that this can indeed be attributed to slow dissociation of the antibody. Examples of the sensorgrams are presented in Supplementary Fig. S3.

All cell lines used were sensitive to the active toxin DUBA (Supplementary Table S1). BYON3521 showed good potencies in the MET–expressing cell lines with IC₅₀ values ranging from 1.5 to 18.2 ng/mL and even in cell lines with lower MET expression (HepG2 and KP4) BYON3521 retained its activity (Fig. 2; Supplementary Table S1). In most cell lines, the maximum efficacy achieved was between 50% and 70%. Extending the incubation period from 6 to 9 days further enhanced the efficacy (Supplementary Fig. S4). Only high concentrations of 10 μg/mL BYON3521 and its nonbinding isotype control ADC were able to kill MET–negative A2780 or Jurkat NucLight cells (that are sensitive to the payload), indicating that the induction of cytotoxicity by low concentrations of BYON3521 in MET–expressing cell lines is mediated through MET.

Figure 2.

In vitro cytotoxicity of BYON3521 in a subset of human tumor cell lines with varying levels of MET expression. Cells were treated for 6 days with BYON3521 and a nonbinding isotype control ADC. Cytotoxicity was determined by CyQUANT direct cell proliferation assay according to the manufacturer's instruction. The cell survival is presented as mean ± SEM of two experiments performed in triplicate.

The cleavability of the vc-seco-DUBA linker-drug, combined with cell permeability of the toxin, gives BYON3521 the intrinsic capacity to kill neighboring MET–negative tumor cells as well, also referred to as bystander activity. Cocultured MET–negative Jurkat NucLight Red and MET–positive tumor cells (1:1 ratio) were treated with BYON3521 and proliferation of Jurkat NucLight Red cells was followed in time. In those cocultures, BYON3521 was able to induce not only cell death in MET–positive cells, but also induced cell killing of the MET–negative Jurkat NucLight Red cells (Fig. 3A). SYD2884-mcMMAF was included as an ADC that is not able to induce bystander activity. It lacks bystander activity as the ADC contains non-cleavable linkers and its payload cys-mc-MMAF is not able to freely diffuse across cell membranes. As expected, no target-negative cells were killed in the presence of SYD2884-mcMMAF after 6 days of incubation (Fig. 3A).

Figure 3.

Bystander activity of MET–negative Jurkat NucLight Red cells cocultured with MET–positive MKN-45 cells. A, Cells were cultured in a 1:1 ratio in the presence of only culture medium (black), BYON3521 (orange), the nonbinding isotype control ADC (grey), or the non-cleavable control ADC SYD2884-mcMMAF (blue). The MET–negative Jurkat NucLight Red cell population was monitored in real time with the incuCyte S3 instrument. Insert: Images of the cocultures taken after 6 days of incubation. B, Cocultures of different ratios of MET–positive MKN-45 and MET–negative Jurkat NucLight red cells. The fraction of MKN-45 cells cocultured with the MET–negative Jurkat NucLight Red cells was reduced from 100 to 0% in steps of 20%. The MET–negative Jurkat NucLight Red cell population was monitored in real time with the incuCyte S3 instrument.

To evaluate to what extent bystander killing of BYON3521 contributes to overall cytotoxicity, the ratio of cocultured MET–positive MKN-45 and MET–negative Jurkat cells was stepwise decreased. In the presence of as little as 20% of MET–positive MKN-45 cells, killing of MET–negative bystander cells was observed in the presence of BYON3521 (Fig. 3B).

HGF/sMET interference in binding and efficacy

Tumor cells can produce HGF (20, 21) and shed sMET (22), which both can interfere with the binding of BYON3521/SYD2884 to the cell surface of tumor cells.

In the BLI experiments increasing the molar excess of active HGF and the inactive pro-HGF decreased the detectable amount of SYD2884 binding to the recombinant human MET ECD immobilized on the surface of the sensor tip (Fig. 4A). Simultaneously, an increase in the detectable amount of rhHGF on the surface of the sensor tip was observed. A 50-fold excess of active rhHGF was needed to fully compete with SYD2884. Similar experiments were performed with pro-HGF (Fig. 4A), however, approximately a 100-fold molar excess of pro-HGF was needed to fully displace SYD2884.

Figure 4.

Competition experiments with rhHGF and MET ECD. A, BLI experiment to assess competition of rhHGF for MET ECD binding of SYD2884. Presented are the binding curve of SYD2884 visualized with an HRP-conjugated anti-hIgG (red line, solid squares) and rhHGF visualized using a polyclonal anti-HGF antibody (blue line, solid triangles). The detectable amount of SYD2884 binding to the recombinant human MET ECD, immobilized on the surface of the sensor tip, decreases whereas at the same time the detectable amount of rhHGF increases. The data are presented as the correlation between the measured signals versus the molar excess of fully activated rhHGF (left graph) and pro-HGF (right graph). B, Cell viability of EBC-1 cells treated with increasing concentrations of BYON3521 or nonbinding isotype control ADC in the presence or absence of 1 μg/mL MET ECD, either administered together directly, or after 30-minute preincubation of BYON3521 or isotype control ADC and MET ECD (Premix). (Approximate MW of BYON3521 is 149 kD, of SY2884 144 kD, and MET ECD 140 kD).

To study the effect of the presence of rhHGF on the cellular binding and/or efficacy of BYON3521, binding and viability assays were performed in the presence of rhHGF using MKN-45 with high MET expression, and NCI-H441 and PC-3 cells both with moderate MET expression. Preincubation of BYON3521 with rhHGF reduced the binding potency of BYON3521 to MET on MKN-45 and PC-3 cells (Supplementary Table S2) as well as the potency of BYON3521 in the cell viability assay (Supplementary Table S3). There was no impact of rhHGF on the binding of BYON3521 to NCI-H441 cells, an autocrine cell line with a constitutively activated MET receptor and which actively produces HGF. However, the potency in the cell viability assay was reduced in the presence of rhHGF.

sMET can serve as a sink for BYON3521 and negatively affect its antitumor effect. A cytotoxicity assay was performed with BYON3521 in MET–positive EBC-1 cells in the presence or absence of 1 μg/mL recombinant MET ECD, a clinically relevant concentration (22). The presence of recombinant human MET ECD interfered with the cytotoxicity of BYON3521 whenever the ratio of MET ECD/BYON3521 (w/w) was 1 or higher (Fig. 4B). The findings in the LigandTracer experiments are in line with these observations. At SYD2884-AF488 concentrations of 8 nmol/L (approximately 1.2 μg/mL) and higher, in the presence of 1 μg/mL MET-ECD, binding to the EBC-1 cells became evident (Supplementary Fig. S5).

In vivo efficacy of BYON3521

MET expression was determined in all xenograft models employed. Examples of the IHC staining and scoring criteria are presented in Supplementary Fig. S6.

The MKN-S45 xenograft model is an aggressive tumor model associated with severe ulceration. In this tumor model, BYON3521 shows good antitumor activity after a single dose of 10 mg/kg (Fig. 5A). The antitumor activity of BYON3521 was also illustrated by the striking reduction in the number of mice that had to be euthanized because of excessive tumor ulceration (Fig. 5B).

Figure 5.

Antitumor activity of BYON3521 in mouse xenograft models. A, Antitumor activity of BYON3521 in the MKN-45 CDX model (MET-amplified with high MET expression levels; see insert). B, Survival plot presenting the number of mice remaining on study due to reduced tumor ulceration in the vehicle control group and BYON3521-treated group. In the BYON3521 group, no mice were euthanized prematurely due to tumor ulceration (N = 12 mice/group, presented is the mean ± SEM). C and D, Efficacy of BYON3521 and corresponding levels of conjugated antibody (with at least 1 linker-drug), as measured with the specific conjugated antibody assay (ECLIA), in the non-MET–amplified head & neck cancer PDX model HNXF1905 (C), and lung cancer PDX model LXFL1176 (D). The arrow indicates the moment of randomization and dosing of the mice. Concentrations are presented in μg/mL, N = 10 mice/group, mean ± SEM.

A MCT was performed to explore the antitumor activity of BYON3521 in multiple tumor indications. The MCT results are presented as a waterfall plot in Supplementary Fig. S7. BYON3521 showed an antitumor response in 52% of MET–expressing tumor models (7% stable disease, 31% PR, 14% CR, independent of the level of MET expression or the tumor type. No antitumor response was observed in the tumor models that did not express MET.

In the PK/PD studies in the HXNF1905 (MET 2+) and LXFL1176 (MET 3+) PDX models, BYON3521 showed dose-dependent antitumor activity (Fig. 5C and D). Tumor stasis was achieved at a dose of 3 mg/kg in both models, whereas tumor remission was observed in the LXFL1176 model treated with 10 mg/kg. The nonbinding control ADC showed mild antitumor activity, most likely due to bystander activity as a result of extracellular cleavage of the linker-drug from the ADC in the tumor environment.

Mouse PKs

In vitro plasma stability studies showed that BYON3521 is stable in CES1c KO mouse, cynomolgus, and human plasma (Supplementary Fig. S8).

The concentration-time data in mouse PDX studies showed that conjugated antibody is cleared slowly and that there is low variation between individual mice (Fig. 6; Supplementary Fig. S9). The few deviating concentrations in individual animals that occurred were observed at the lower dose levels and are most likely caused by differences in target-mediated drug disposition (TMDD) due to individual differences in tumor volume. The plasma curves and derived PK parameters show a slight nonlinearity for lung tumor model LXFL1176. For this model, clearance decreased with increasing doses, which could be explained by TMDD via binding of BYON3521 to human MET expressed in the tumor. Corresponding trends in TMDD are visible for area under the curve (AUC_last) and terminal half-life (t_½) for LXFL1176. Maximum concentration (C_max) levels increased dose proportionally in general. Values for volume of distribution (V_ss) suggest that BYON3521 is mainly confined to the plasma compartment. PK parameters for the mouse are presented in Supplementary Data Table S4.

Figure 6.

Mean PK curves in cynomolgus monkey. A, Conjugated antibody concentrations in male and female monkeys after repeated dosing with 5, 15, or 25 mg/kg BYON3521 (N = 5 monkeys/group). B, Concentration of total antibody, conjugated antibody and free toxin (DUBA) in male and female monkeys after SD administration of 25 mg/kg BYON3521 intravenously.

Safety evaluation and toxicokinetics in cynomolgus monkey (Macaca fascicularis)

The specificity screen performed using the human cell array technology Retrogenix revealed only binding of BYON3521 to Fc gamma receptors and MET, indicating that BYON3521 binds to MET as intended target with high specificity.

The binding of BYON3521 to human and monkey MET is comparable (Fig. 1D) and therefore the toxicity profile observed in this species is considered relevant for human risk assessment. In the 4-cycle toxicity study in monkeys, 5, 15, or 25 mg/kg BYON3521 given once every 3 weeks was generally well tolerated. There was a slight dose-related decreased body weight gain in male monkeys. Single-dose and multiple-dose toxicokinetic (TK) parameters are presented in Tables S5A and S5B, respectively. Plasma levels of total and conjugated antibody were comparable and levels of the free toxin DUBA were very low (Fig. 6), indicating high stability of the ADC in monkey plasma. Exposure to BYON3521 was high throughout the entire dosing period of 12 weeks, with exposure multiples of 10- to 13-fold based on the AUC_last (0–3 weeks) in cynomolgus monkeys (males and females) treated with 25 mg/kg and the AUC_last (0–3 weeks) of mice treated with 3 mg/kg BYON3521 in the pharmacologic models.

The target organs identified were kidney, lung, uterus, skin, eye, eyelid, lacrimal gland, thymus, and bone marrow. Most observed side effects were dose dependent, and partial or complete recovery was noted in the 6-week recovery period in low and mid-dose groups. In the high dose group, no recovery was apparent for the effects in lung and kidney. Despite the mild histopathologic effects observed in the lung, no functional effects of BYON3521 were noted on the respiration rate. In addition, there were no effects on the cardiovascular system (electrocardiogram, blood pressure) and central nervous system.

On the basis of the absence of recovery of the effects in the lung and kidney at 25 mg/kg, the highest non-severely toxic dose (HNSTD) was estimated to be 15 mg/kg. At this dose, the exposure (AUC 0–3 weeks, male and female animals) to conjugated antibody is 41150 μg.h/mL, which is a factor of 5–7 higher than the exposure of 5920 to 7980 μg.h/mL (AUC 0–3 weeks) observed in the mouse PK/PD studies at the effective dose of 3 mg/kg.

Prediction of human PK and minimal efficacious dose

The nonlinear PK in monkey was well described using a TMDD model from Luu and colleagues (23). The used model parameters are listed in Supplementary Table S6A.

The monkey PK model was then translated to a human PK model by weight-based allometric scaling of the clearance and disposition parameters (k_el, k₁₂ and k₂₁) and the volume of distribution (V) by weight-based allometric scaling using an allometric exponent of 1 for volume of distribution (V), and an exponent of -0.15 for disposition rate constants k_el, k₁₂, and k₂₁ which was derived from an allometric exponent of 0.85 for clearance (Supplementary Table S6B; ref. 24), and by incorporating the target binding affinities to human and monkey MET. With the derived human PK model (Supplementary Table S6A), simulations were performed to predict human dose levels that would give similar exposure as observed at efficacious dose levels in the mouse tumor growth inhibition (PDX) models which was obtained at an AUC range of 5920 to 7980 h. μg/mL and an average concentration (C_avg) of 12 to 16 μg/mL within 21 days after a single intravenous dose (Supplementary Fig. S10).

Sensitivity analyses showed that the predicted human PK is impacted significantly if antigen levels vary (MET and sMET combined; Supplementary Fig. S9). Assuming antigen concentrations of 0.3 and 0.75 μg/mL, the minimal efficacious dose for human is respectively 2 to 3 mg/kg or 3 to 4 mg/kg based on a single-dose schedule.

The effect that sMET could have on target binding and safety of BYON3521 in human is not yet known. However, the PK data that were used to develop the human TMDD model were based on assays that detect the conjugated antibody fraction capable of binding MET (CDRs not occupied by sMET), and thereby taking potential sMET interference already into account in the human predictions.

Discussion

The preclinical profile of BYON3521 shows that site-specific conjugation of vc-seco-DUBA to a MET–targeting antibody has the potential to deliver an ADC with a favorable therapeutic window in patients. BYON3521 is very stable in human and cyno plasma and capable of inducing both target- and bystander-mediated cell killing in MET-amplified as well as non-MET–amplified tumors with low, moderate, and high MET expression. Even though it is remarkable that some tumors with low MET expression respond well to BYON3521, it is not yet clear why these are sensitive. Studies are ongoing to elucidate what drives the sensitivity of cancer cells, or lack thereof, to DUBA.

The prospect of a successful MET targeting ADC depends on a delicate balance between its stability, favorable binding characteristics, its ability to internalize, and sensitivity of the tumor cells to the toxin. Several MET targeting ADCs, exhibiting different payloads and cleavable or non-cleavable linkers, are in preclinical or clinical development. Even though ADCs show promising results in preclinical efficacy models, the most important hurdle to overcome when bringing an ADC to patients is that of being able to dose high enough to achieve antitumor activity at a well tolerable dose level.

In the 4-cycle tox and PK study in cynomolgus monkey, BYON3521 was well tolerated. The number of target organs in the cynomolgus monkey was limited and the observed adverse events were relatively mild and, in many cases, reversible, resulting in an HNSTD of 15 mg/kg/Q3W. At this dose, the BYON3521 ADC exposure was approximately 5 to 7 times higher compared with the exposure needed for antitumor activity in the mouse xenografts, which is indicative of a substantial therapeutic window. This was interesting given the high affinity of BYON3521 for cynomolgus MET and the fact that several tissues quite evidently express MET on the cell surface including the GI tract, liver, skin, and eyes (Supplementary Fig. S11).

The modest eye toxicity observed for BYON3521 is quite remarkable considering the fact that HER2 expression in eye tissues is also quite prominent and that eye-related adverse effects are frequently dose-limiting in patients treated with SYD985, the DUBA-based HER2-targeting ADC (25). The findings for BYON3521 appear to be more in line with the observations in patients treated with the B7-H3-targeting DUBA-based ADC MGC018 from Macrogenics; despite evident B7-H3 expression in the eye (26), no dose-limiting eye toxicities were reported in the nonclinical safety evaluation or clinical studies (27, 28).

Dose-limiting ocular toxicities have also been reported for ADCs making use of other payloads such as for instance DM4 and MMAF (29), and even ADCs for which no target expression in the eyes could be confirmed; e.g., the folate receptor alpha-targeting DM4-based ADC mirvetuximab soravtansine, are known to induce severe eye toxicity (30). These observations, together with the fact that the adverse events are quite diverse in nature, exemplify that it is very challenging to discern whether ocular side effects are target-mediated, antibody-related or caused by systemically circulating payload.

Why healthy cells expressing MET appear to be less sensitive to BYON3521 than tumor cells with low to moderate MET expression remains elusive. As indicated earlier, healthy cells may be better equipped to deal with inflicted DNA damage than cancer cells in which the DNA damage repair systems are usually not functioning properly (31), but an alternative explanation may be related to the fact that MET on normal cells is mostly silent/not activated as evidenced by the absence of MET phosphorylation in healthy MET–expressing tissues (Supplementary Fig. S12; refs. 32, 33). Under these conditions MET appears to be present at the cell surface as a complex with pro-HGF (34), an inactive form of HGF continuously produced by all cellular sources of HGF. When pro-HGF is synthesized, it is transported to the cell surface where it is sequestered by heparan sulphate proteoglycan structures and binds MET with high affinity keeping it in its ‘silenced’ state (34). The BLI binding competition studies showed that the antibody part of the ADC BYON3521 (SYD2884) could be displaced from MET ECD with both the active rhHGF and the pro-HGF, suggesting that the presence of pro-HGF may hinder BYON3521 binding to MET on healthy cells. Only upon trauma, causing the release of proteases from lysosomes or infiltrating leukocytes or in an invasive tumor microenvironment where extracellular proteases are present, HGF is generated from its pro-form, leading to the activation of MET (35). More studies are needed to understand why BYON3521 induces such limited toxicity in healthy tissue.

Another interesting finding were the modest effects seen on RBC, leukocyte and platelet counts. These bone marrow-related toxicities are typically considered as class effect toxicities for different payloads, most notably for microtubule inhibitors (29). Also DNA alkylating drugs are known to be cytotoxic to bone marrow stem cells (36), which is line with the fact that leukopenia, neutropenia and thrombocytopenia have been reported as dose-limiting toxicities (37, 38). Evidently, the systemically circulating DUBA levels detected after administration of BYON3521, or any DUBA which in theory could be released locally from ADCs in the bone marrow environment (39), have limited impact on the bone marrow.

It has been argued that ADC MTDs are mostly related to the normalized amount of (conjugated) payload dosed (14), independent of the antigen targeted or the antibody employed. PK/PD modeling estimates the minimal efficacious dose for BYON3521 to be 3 to 4 mg/kg and it is anticipated that this can readily be achieved within the predicted therapeutic window. Given the properties of BYON3521 it is predicted, and therefore will be interesting to learn, whether a higher dose of conjugated payload will indeed be tolerated. A first-in-human dose-escalation study is currently ongoing to determine the MTD and recommended dose for phase II (NCT05323045).

Supplementary Material

Supplementary Data

Supplementary tables and figures

Authors' Disclosures

P.G. Groothuis reports a patent for WO 2022/214517 pending to Byondis bv. R.G.E. Coumans reports a patent for WO2015/177360 pending and issued to Byondis B.V. and a patent for WO2017/137628 pending and issued to Byondis B.V. M. Blomenröhr reports a patent for WO 2022/214517 pending. M.M.C. Van der Lee reports a patent for WO 2022/214517 pending to Byondis B.V. No disclosures were reported by the other authors.

Authors' Contributions

P.G. Groothuis: Conceptualization, supervision, investigation, methodology, writing–original draft, project administration, writing–review and editing. D.C.H. Jacobs: Formal analysis, supervision, validation, investigation, methodology, writing–original draft, writing–review and editing. I.A.T. Hermens: Validation, investigation, writing–original draft, writing–review and editing. D. Damming: Formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. K. Berentsen: Formal analysis, validation, investigation, visualization, methodology, writing–review and editing. E. Mattaar-Hepp: Investigation, methodology, writing–original draft, writing–review and editing. M.E.M. Stokman: Investigation, methodology, writing–original draft, writing–review and editing. T. Van Boekel: Formal analysis, supervision, validation, investigation, methodology, writing–review and editing. M. Rouwette: Supervision, validation, investigation, methodology, writing–original draft. M.A.J. Van der Vleuten: Software, investigation, writing–review and editing. A. Sesink: Formal analysis, supervision, validation, investigation, methodology, writing–original draft, writing–review and editing. F.A. Dijcks: Conceptualization, writing–review and editing. R.G.E. Coumans: Conceptualization, validation, investigation, methodology, writing–original draft, writing–review and editing, compound design, analysis and supply. J. Schouten: Validation, investigation, methodology, writing–original draft, writing–review and editing. D.H. Glaudemans: Validation, investigation, writing–original draft, writing–review and editing. D. van Wijk: Validation, investigation, writing–original draft, writing–review and editing. M. Blomenröhr: Conceptualization, supervision, project administration, writing–review and editing. W.A. Kappers: Data curation, software, investigation, methodology, writing–original draft, writing–review and editing. R. Ubink: Resources, supervision, investigation, methodology, writing–review and editing. M.M.C. Van der Lee: Resources, supervision, investigation, methodology, writing–review and editing. W.H.A. Dokter: Resources, supervision, writing–review and editing.