Hinge-region Cys-to-Ser mutation enables homogeneous DAR6 anti-c-Met biparatopic ADCs with tolerable O-glycosylation

Abstract

Background

Antibody–drug conjugates (ADCs) with homogeneous drug-to-antibody ratios (DARs) are increasingly favored for their improved physicochemical and pharmacokinetic properties; however, strategies to reliably generate homogeneous DAR6 ADCs remain limited.

Methods

In this study, we investigate hinge-region cysteine engineering as an approach to constrain conjugation stoichiometry, using a potent biparatopic anti–c-Met antibody as a model. Various biological and physicochemical assays were used for its characterization.

Results

We show that substitution of a hinge-region cysteine within a CPPC (Cysteine - Proline - Proline - Cysteine) motif reduces the number of accessible conjugation sites, allowing reproducible generation of homogeneous DAR6 ADCs. During characterization of Ser- or Thr-substituted hinge variants, we observed additional mass species consistent with mucin-type O-linked glycosylation. Site-specific mass spectrometry analyses localized these modifications to a serine residue within hinge-region and identified the core structures of the O-glycans. Analysis of additional antibodies carrying analogous mutations demonstrated that hinge-associated O-glycosylation is not restricted to a single antibody scaffold.

Conclusions

These results demonstrate that hinge-region cysteine engineering can serve as a rational and adaptable approach for DAR6 ADC development while highlighting the need for case-by-case evaluation of associated post-translational modifications during antibody engineering.

Statement of Significance These findings demonstrate that hinge-region cysteine engineering can enable homogeneous DAR6 ADCs and highlight O-glycosylation as an associated molecular feature of this strategy that appears functionally tolerated in ADC-relevant assays.

Introduction

Antibody–drug conjugates (ADCs) have emerged as a promising therapeutic modality in oncology, combining the specificity of monoclonal antibodies with the cytotoxic potency of small-molecule drugs. One prominent example is DS-8201a (trastuzumab deruxtecan), a HER2-targeting ADC conjugated with a cleavable GGFG-deruxtecan linker–payload (GGFG-DXd), which has demonstrated remarkable clinical efficacy [1–6]. Clinical experience has highlighted that ADC efficacy, safety, and pharmacokinetics (PKs) are strongly influenced by conjugation chemistry and drug-to-antibody ratio (DAR).

c-Met is a key oncogenic driver frequently upregulated in lung and gastroesophageal cancers, while exhibiting restricted expression in normal tissues, making it an attractive therapeutic target [7–13]. Although c-Met–directed tyrosine kinase inhibitors demonstrate clinical activity, only approximately 4%–5% of non-small-cell lung cancer tumors harbor relevant MET alterations, limiting the eligible patient population [14, 15]. Their therapeutic benefit is further constrained by both primary and acquired resistance mechanisms, including mutations within MET itself or compensatory signaling pathways [16–18]. ADCs targeting c-Met offer the potential to overcome these limitations by delivering cytotoxic payloads independently of MET kinase signaling. Several c-Met–targeting ADCs, including telisotuzumab vedotin (ABBV-399) and telisotuzumab adintocatecan (ABBV-400), have demonstrated anti-tumor activity in clinical studies, although safety concerns have also been reported [19–24].

ADCs with homogeneous DARs generally exhibit more favorable structural and functional properties compared with heterogeneous conjugates [25]. Different antibody–payload combinations may require optimized DAR values, typically ranging from DAR2 to DAR8. For highly potent payloads, excessive DARs can increase hydrophobicity and clearance, whereas lower DARs may compromise efficacy, necessitating careful optimization. For certain payload classes, including exatecan-based toxins, intermediate DAR values such as DAR6 may provide a favorable balance between potency and tolerability. Trastuzumab rezetecan (Avida) is an approved ADC that employs a DAR of 5.7 appropriate for its linker-payload; however, it comprises heterogeneous species rather than being a homogeneous ADC [26]. While established strategies exist for generating homogeneous DAR2, DAR4, or DAR8 ADCs, methods enabling homogeneous DAR6 ADCs remain limited, and most DAR6 conjugates rely on random cysteine conjugation, resulting in heterogeneous mixtures of species bearing 0–8 payloads [27–33].

Biparatopic ADCs that target two non-overlapping epitopes of c-Met represent a promising strategy compared with monospecific ADCs, as they can enhance receptor engagement and internalization [34, 35]. IgG1 antibodies contain four interchain disulfide bonds within or close to the hinge region, providing eight solvent-accessible cysteine residues for conjugation following reduction. We hypothesized that selective disruption of one interchain disulfide bond could reduce the number of accessible conjugation sites from eight to six, thereby constraining conjugation stoichiometry and enabling homogeneous DAR6 ADCs. Hinge-region cysteine engineering following defined CPPX or XPPC motifs represents a rational approach to test this hypothesis.

Protein glycosylation is a widespread post-translational modification and occurs primarily as N-linked or O-linked glycosylation [36]. All IgG subclasses contain a conserved N-linked glycosylation site in the Fc region [36–39], and the structure of N-glycans has been extensively characterized [40, 41]. O-Glycans, in contrast, are attached to serine or threonine residues and lack a strict consensus sequence, complicating their prediction and analysis [36, 42]. The most common O-linked glycans include O-GalNAc and O-GlcNAc modifications [37, 43], which are catalyzed by distinct glycosyltransferases [36, 44–46]. O-Glycosylation is relatively rare in IgG1 antibodies but has been observed in the hinge regions of certain immunoglobulin isotypes, including IgA1, IgD, and IgG3 [47–50]. Advances in liquid chromatography–mass spectrometry (LC–MS) have enabled increasingly detailed characterization of glycoproteins and glycopeptides [51–53]. In this study, we investigate hinge-region cysteine engineering as a strategy to enable homogeneous DAR6 ADCs and to characterize molecular features associated with this approach. Using a potent biparatopic anti-c-Met antibody as a model system, we introduce defined hinge-region CPPX mutations to constrain conjugation stoichiometry. During analytical characterization, we observe O-linked glycosylation associated with hinge Ser/Thr substitution and localize this modification to a specific hinge residue using mass spectrometry. Finally, we assess whether variation in hinge O-glycosylation is tolerated in ADC-relevant assays, including stability, in vitro activity, PKs, and in vivo efficacy.

Materials and methods

Reagents for O-glycosylation analysis

Chemicals used for sample preparation and analysis included Urea, dipotassium hydrogen phosphate, potassium chloride, ammonium acetate, trifluoroacetic acid (TFA), dithiothreitol (DTT), iodoacetic acid (IAA), guanidine hydrochloride (GuHCl), acetonitrile (ACN), iodoacetamide (IAM), and Tris (Sigma: U5378, V900041, V900068, 102,612,648, 302,031, D9779, I4386, G3272, 34,851, T6066, I1149). Proteolytic enzymes used in this study included trypsin, PNGase F, IdeZ (FabRICATOR), and O-glycoprotease (IMPa), obtained from commercial suppliers (Promega: V5111; Roche: 11365193001; New England Biolabs: P0770S and P0761S). Desalting columns (Bio-Rad: 732-6222), FA (ThermoFisher: A117) and isopropanol (Merck: 1.01040.4008) were used as indicated. The trypsin used for electron-transfer/higher-energy collision dissociation (EThcD) experiments was obtained from T&G Biotech (YJ-O-068-100). Tris (2-carboxyethyl) phosphine (TCEP) is supplied by Adamas-beta (61820E). The desalting column used in the EThcD experiments was sourced from CDS Analytical (2215-C18). All reagents were used according to manufacturers’ instructions unless otherwise specified.

Cell lines and culture conditions

All cell lines used in this study were routinely tested for mycoplasma contamination and confirmed to be negative. Human cancer cell lines MKN45, EBC-1, NCI-H226, NCI-H1975, NCI-H2170, and HT-29 were obtained from commercial repositories (Cobier Biosciences: CBP60488; JCRB: JCRB0820; Chinese National Cell Bank: SCSP-5073, SCSP-597, and ATCC: CRL-5928, HTB-38). Cells were maintained in RPMI-1640 medium supplemented with 10%–20% fetal bovine serum, depending on cell line requirements, and cultured under standard conditions (37°C, 5% CO₂).

Animals

All animal experiments were conducted in compliance with relevant ethical regulations and the ARRIVE guidelines. Immunization, PK studies, and in vivo efficacy studies were reviewed and approved by the Animal Ethics Committee of Shandong Boan Biotechnology Co., Ltd. (Approval numbers: 2024-TS0001-10-4, 2024-TS0001-11-1, and 2025-TS0001-2-1).

All mice were housed in a specific-pathogen-free barrier facility under a 12-h light/12-h dark cycle in individually ventilated cages with 5 animals per cage. Cages contained autoclaved corncob bedding, which was changed twice weekly. The ambient temperature was maintained at 22–24°C with a relative humidity of 40%–60%. Animals had ad libitum access to autoclaved standard rodent chow and sterile water. Ambient noise was controlled below 60 dB, and ammonia concentration was monitored to be less than 14 ppm.

Antibody production

Heavy chain variable region and light chain variable region were amplified (2× Phanta Max Master Mix, Vazyme, P515-01) using the positive clones screened from the library as templates. Overlap PCR was conducted to assemble variable region and signal peptide. Purified gene fragments were separately fused (ClonExpress II One Step Cloning Kit, Vazyme, C112-01) into the linearized pcDNA3.4 vectors with human IgG1 constant regions. The recombinant plasmids were prepared for production. Antibodies were expressed with Expi-CHO Expression system (A29133, Gibco) for 12 days and the supernatants were harvested and purified by a AT Protein A Diamond (AA0272, Bestchrom).

ADC preparation

A solution of naked mAb (5–10 mg/ml in PBS (7.2) + 2 mM EDTA) was treated with 6 molar equivalents of TCEP for 3 h at 25°C. Then 8 molar equivalents of linker-payload (from a 2 mM DMSO stock solution) were added to the fully reduced antibody, while keeping residual DMSO concentration below 15% (v/v). The mixture was incubated at room temperature for 2 h, and excess N-acetylcysteine then added to quench any unreacted linker-drug.

ELISA-based binding assay

Human C1q protein (Abcam: AB282858) was coated on high binding ELISA plates with 3 μg/ml at 4°C overnight, and then the plates were blocked with 3% skim milk powder in PBST (PBS containing 0.05% Tween-20) at 37°C for 1 h. After washing two times with PBST, serially diluted antibodies were added to each well, incubated at 37°C for 1 h. Plates were washed two times and then HRP-goat anti-human IgG mAb (474-1006, KPL) was used to detect antibodies binding to C1q. Experiments were performed in duplicate, Data was presented as Mean ± SD.

Cell-based binding assay

MKN45 and EBC-1 cells were harvested and washed by FACS buffer (0.2% BAS in PBS) two times. Then the cells were incubated with gradient diluted antibody at 4°C for 1 h. After washing twice with FACS buffer, cells were stained with Goat Anti-Human IgG Fc in 1 μg/ml at 4°C for 30 min. Cells were washed twice and then resuspended in 100 μl FACS buffer for analysis by NovoCyte 2060R flow cytometry. Experiments were performed in duplicate. Data were presented as Mean ± SD.

FACS-based internalization assay

EBC-1 cells were harvested and washed by cold PBS. Then the cells were resuspended in PBS. The antibody was diluted with PBS and labeled with SwiftStainTM Human pH-red Antibody Labeling Dye (SinoBiological: SSL001). Then the labeled antibody was incubated with cells. The control group was incubated at 4°C, while the test group was incubated at 37°C for different hours. Cells were washed twice and then resuspended in 100 μl cold PBS to analyze the fluorescence of live cells by NovoCyte 2060R flow cytometry. Experiments were performed in duplicate. Data were presented as Mean ± SD.

Cell viability assay

Cell viability was assessed using a luminescent cell viability assay on a microplate reader (Biotek Synergy™ Neo2). NCI-H226 or NCI-H1975 cells were seeded into 96-well plates and allowed to adhere prior to treatment. ADCs were added in serial dilutions ranging from 0.0016 to 5 μg/ml. After 4 days of incubation, cell viability was measured using CellCounting-Lite 2.0 (Vazyme Biotech, DD1101). Experiments were performed in duplicate, and data are presented as mean ± standard deviation (SD).

Bio-layer interferometry competitive binding assay

Competitive binding experiments were conducted using a ForteBio Octet Red96 system based on bio-layer interferometry (BLI). Human c-Met-His protein (Sino Biological, 10692-H08H) was immobilized on HIS1K biosensors (ForteBio, 18–5121). Sensors were sequentially incubated with the first antibody (15 μg/ml) or PBST control for 100 s, followed by incubation with the second antibody (15 μg/ml) for an additional 100 s. Binding responses were analyzed using ForteBio Data Analysis software (version 9.0).

Aggregation analysis by size-exclusion chromatography

Aggregation levels were determined by size-exclusion chromatography (SEC) using a TSKgel G3000SWXL column (Tosoh Bioscience, P/N 0008541) on a Waters E2695 HPLC system with UV detection at 280 nm. Samples (50 μg per injection) were eluted using a mobile phase consisting of 200 mM dipotassium hydrogen phosphate and 200 mM potassium chloride (pH 6.8) containing 15% (v/v) isopropanol, at a flow rate of 0.75 ml/min.

DAR analysis by hydrophobic interaction chromatography

DAR was determined by hydrophobic interaction chromatography (HIC) using a TSKgel Butyl-NPR column (4.6 × 35 mm, 2.5 μm; Tosoh Bioscience) on a Waters E2695 HPLC system with UV detection at 280 nm. Samples (50 μg per injection) were separated at a flow rate of 0.5 ml/min using mobile phase A (20 mM phosphate buffer containing 25% isopropanol, pH 7.0) and mobile phase B (20 mM phosphate buffer containing 1.5 M ammonium sulfate, pH 7.0). The sample was eluted in a gradient of 26%–100% mobile phase A over 18 min. The column temperature was set at 25°C. DAR values were calculated based on peak areas.

Two-week toxicity study in B-hMET/hHGF/hEGFR mice

Toxicity study was evaluated using MET humanized mice (B-hMET/hHGF/hEGFR, male, Catalog number 113292), which were purchased from Beijing Biocytogen Co., Ltd. Six mice were randomly assigned into two groups: the PR304(CPPS)-68B group (n = 3) and PR304-68B group (n = 3). A single dose of 150 mg/kg was administered to all groups via tail vein injection. After administration, body weights were measured to evaluate the toxicity of the ADCs. Data were expressed as mean ± standard error of the mean (SEM) and analyzed using GraphPad Prism 8.0 software. Student's t-test was used to compare body weights between the two groups at each time point. P < .05 indicated statistically significant differences. Abbreviations: ^*P < .05, ^**P < .01, ^***P < .001.

Intact mass analysis by LC–MS

Samples were deglycosylated using PNGase F at an enzyme-to-protein ratio of 1:5 (U:μg) at 37°C for 2 h. Intact mass analysis was performed using an Agilent 1290 HPLC system coupled to an Agilent 6545XT Q-TOF mass spectrometer. Separation was achieved on a Waters Polyphenyl column (2.1 × 150 mm, 2.7 μm; P/N 186008946) maintained at 75°C. The mobile phase consisted of 0.1% TFA, 99.9% water (MPA, Mobile Phase A) and 0.1% TFA, 20.0% isopropanol, 70.0% ACN, 9.9% water (MPB, Mobile Phase B). A linear elution gradient from 39% B to 48% B over 22.5 min was applied. Mass spectra were acquired in positive-ion mode (m/z 100–1000) and deconvoluted to identify mass shifts consistent with post-translational modifications.

Subunit mass analysis

For subunit analysis, 50 μg of protein was digested with IdeZ according to the manufacturer’s instructions. Subunit mass analysis was conducted using a Waters Polyphenyl column (2.1 × 150 mm, 2.7 μm; P/N 186008946) at 75°C coupled to an Agilent 1290 HPLC/6545XT Q-TOF system operating at 0.4 ml/min with mobile phases consisting of 0.1% TFA in water (MPA) and 0.1% TFA in ACN (MPB). A linear elution gradient from 30% B to 50% B applied over 45 min was applied. Resulting fragments were analyzed using the same LC–MS platform described above. Subunit-level analysis was used to localize mass modifications to specific antibody domains.

Peptide mapping

Samples were denatured with GuHCl/Tris, reduced with DTT, and alkylated with IAA. Then, it was buffer-exchanged into Tris buffer using a desalting column, followed by trypsin digestion at 37°C for 4 and 20 h, respectively. The mobile phase consisted of water containing 0.05% TFA (MPA) and ACN containing 0.05% TFA (MPB) with a flow rate of 0.3 ml/min. An Ultimate 3000 UHPLC system coupled to a Q-Exactive mass spectrometer was used. Separation was achieved on a Waters CSH130 C18 column (2.1 × 150 mm, 1.7 μm; P/N 186005294) maintained at 60°C. A linear elution gradient from 2% B to 33% B applied over 66 min was applied. Data-dependent MS/MS acquisition was used to identify peptide sequences and post-translational modifications.

O-Glycopeptide mapping

For O-glycopeptide analysis, samples were digested using trypsin in combination with 10 μl O-glycoprotease (IMPa). The model and parameters of the experimental instrumentation and chromatography column used in this study were consistent with those employed in the peptide mapping analysis. IMPa cleavage specificity was used to assist in site localization of mucin-type O-glycosylation. LC–MS analysis was performed under peptide mapping conditions. Glycan assignments were made at the core-structure level based on accurate mass and diagnostic oxonium ions.

EThcD fragmentation

Samples were denatured with Urea, reduced with TCEP, alkylated with IAM, and buffer-exchanged prior to digestion with trypsin. The samples were separated using a ThermoFisher U3000 RSLC nano-nanoliter HPLC system with a flow rate of 300 nl/min, which employed a C18 column (Dr. Maisch GmbH, 200 mm × 75 μm i.d., 1.9 Å particle size) at 40°C. The mobile phase consisted of water containing 0.1% FA (MPA) and ACN containing 0.1% FA (MPB). A linear elution gradient from 8% B to 25% B applied over 30 min was applied. An ThermoFisher Orbitrap Lumos mass spectrometer with a m/z scan range of 750–2000 in positive ion mode was used. Data were acquired in DDA (Data-Dependent Acquisition) mode with EThcD as the fragmentation method.

Native SEC-Q-TOF-MS

Samples were digested using PNGase F at a ratio of 1:5 (U:μg) at 37°C for 2 h. The native SEC-Q-TOF-MS analysis was conducted using a Waters ACQUITY UPLC Protein BEH SEC column (4.6 × 150 mm, 1.7 μm; P/N 186005225). The mobile phase consisted of 100 mM ammonium acetate, delivered at a flow rate of 0.15 ml/min with the column maintained at 30°C. The mass spectrometer operated in positive-ion mode (scan range: m/z 100–10 000) for molecular weight analysis.

iCIEF analysis

Analysis by iCIEF was carried out using an iCE3 instrument from ProteinSimple (San Jose, CA). The samples were diluted to a final concentration of 0.4 mg/ml in a solution containing 3% carrier ampholytes (50% Pharmalytes [GE Healthcare] pH 8–10.5, 50% Pharmalytes pH 3–10), 0.35% methylcellulose, and pI markers. The sample-ampholyte mixture was introduced into the capillary under pressure and then focused at 1500 V for 1 min followed by 8 min at 3000 V, at which time a charge-coupled device camera captured an image of the entire capillary illuminated by a 280 nm light source. The software iCE CFR from ProteinSimple provided by the instrument manufacturer converted the image into an electropherogram showing absorbance as a function of the position in the capillary or of pI based on standards.

Differential scanning calorimetry

Analysis by differential scanning calorimetry (DSC) was carried out using a Microcal VP-Capillary instrument from GE. Pipette 500 μl of the sample solution at a concentration of 0.5 mg/ml into the DSC sample crucible. Set the sample holder temperature to 10°C for equilibration prior to scanning. Define the scanning range with a start temperature of 25°C and an end temperature of 100°C. Apply a heating rate of 90°C/h and a cooling rate of 60°C/h as required by the thermal cycle. Configure the data acquisition frequency to 8 s per point. Post-run, export the thermogram for analysis of thermal transitions.

PK studies in mice

Female BALB/c mice (8–10 weeks old) were used for the PK study and purchased from Shandong Pengyue Laboratory Animal Technology Co., Ltd. Mice received a single intravenous dose of antibodies and ADCs at 10 mg/kg (n = 5 and n = 3 per group, respectively). Blood samples were collected by retro-orbital bleeding at predefined time points until 21 days post-dose. After blood collection, the mice were euthanized by carbon dioxide (CO₂) asphyxiation. Serum concentrations of total antibody and ADC were quantified by ELISA. PK parameters were calculated using Phoenix™ WinNonlin software and are reported as exploratory measurements.

In vivo efficacy studies

In vivo efficacy was assessed using an HT-29 xenograft model in BALB/c-nude mice and an MKN45 xenograft model in NPG immunodeficient mice. Mice were stratified into vehicle and treatment groups according to the average tumor volume, with 5 mice in each group. All agents were administered via a single tail-vein injection, with dosages shown in the figure. Tumor volumes were measured twice weekly, and tumor growth inhibition was calculated as described. At the experimental endpoint, all mice were euthanized by carbon dioxide inhalation. These studies were designed to assess gross efficacy differences rather than establish definitive therapeutic equivalence.

Statistics and reproducibility

Statistical analyses were performed using GraphPad Prism 8. Data are presented as mean ± SD or mean ± SEM, as indicated in figure legends. No data points were excluded from analysis, and no formal sample size calculations were performed.

Results

Identification of a biparatopic anti-c-Met antibody suitable for ADC development

To establish a suitable model system for evaluating hinge-region engineering strategies, anti–c-Met antibodies were generated by immunizing BA-huMab® mice with human c-Met protein, followed by phage display screening. Two antibodies (Ab1 and Ab2) sharing a common light chain were selected and assembled into a 1 + 1 format biparatopic antibody, PR304 (Fig. 1a).

Figure 1.

Identification of a highly potent anti-c-Met biparatopic antibody PR304. (a) Schematic diagram of the structure of the anti-c-Met biparatopic antibody PR304. The common light chain and the two heavy chains are displayed in the schematic diagram, respectively. (b) Competitive binding of anti-c-Met antibodies PR304-Ab1, PR304-Ab2, and ABT-700 to human c-Met. <20%: High competition, overlapping or same epitope; >60%: No competition, different epitope. (c) Binding activity of anti-c-Met antibodies PR304, ABT-700 and REGN5093 to MKN45 cells by flow cytometry. Experiments were performed in duplicate. Data were presented as Mean ± SD. (d) Internalization activity of anti-c-Met antibodies PR304, ABT-700, and REGN5093 on EBC-1 cells at different time using flow cytometry. Experiments were performed in duplicate. Data were presented as mean ± SD. (e and f) Tumor cell proliferation inhibitory activity of anti-c-Met antibodies PR304, ABT-700, and REGN5093 conjugated with VcMMAE against NCI-H226 cells (e) and NCI-H1975 cells (f). Experiments were performed in duplicate. Data were presented as mean ± SD. IC₅₀ values were got by calculation using GraphPad Prism 8.

Epitope competition analysis using BLI demonstrated that Ab1, Ab2, and an in-house analog of ABT-700 bound non-overlapping epitopes on human c-Met (Fig. 1b). PR304 was selected as a model antibody based on its robust cell binding and internalization properties, which were superior to those of ABT-700 and REGN-5093 in MKN45 gastric cancer cells and EBC-1 lung squamous carcinoma cells, respectively (Fig. 1c and d).

PR304 and comparator antibodies were conjugated to mc-vc-PAB-MMAE to assess baseline ADC performance. PR304-VcMMAE showed modestly enhanced inhibition of tumor cell proliferation in NCI-H226 and NCI-H1975 cells relative to comparator ADCs (Fig. 1e and f). These results supported the suitability of PR304 as a representative scaffold for subsequent hinge engineering studies, rather than constituting a primary discovery objective.

Hinge-region cysteine engineering constrains conjugation stoichiometry to DAR6

For highly potent payloads, such as exatecan derivatives, excessively high DARs can increase hydrophobicity and compromise exposure, motivating the development of intermediate DAR ADCs. However, homogeneous DAR6 ADCs are difficult to achieve using conventional conjugation approaches. To rapidly screen hinge-region mutations capable of constraining conjugation stoichiometry, the in-house antibody CA149 was first used as a practical screening scaffold. The two solvent-accessible cysteines within the hinge CPPC motif were selectively mutated (CPPS, CPPA, SPPC, APPC), thereby disrupting one interchain disulfide bond and reducing the number of accessible cysteines from eight to six (Supplementary Fig. S1a–d).

HIC analysis demonstrated that the CA149 (CPPS) variant conjugated to GGFG-DXd exhibited a dominant DAR6 species with reduced heterogeneity compared with other variants and with randomly conjugated CA149 ADCs (Supplementary Fig. S1e and f). These results informed selection of the CPPS mutation for evaluation in a therapeutically relevant antibody scaffold. In addition to the CPPS mutation, two additional CPPX mutations (CPPT and CPPG) were also introduced into PR304 for comparative evaluation (Fig. 2a). Size-exclusion chromatography revealed that PR304 (CPPS) displayed lower aggregation levels than PR304 (CPPT) and PR304 (CPPG) (Fig. 2b–e and Supplementary Table S1). PR304 (CPPS) was therefore selected for further characterization. Conjugation of PR304 (CPPS) with the PEG2-COOH-(p-Glucuronide)-PAB-Exatecan (68B) linker–payload produced ADCs with a single dominant species by both HIC and SEC (Fig. 2f and g), consistent with a homogeneous DAR6 configuration. The linker–payload 68B comprises a cleavable β-glucuronic acid-based linker and the topoisomerase 1 inhibitor exatecan, and is conjugated to the antibody via a maleimide moiety. Exatecan inhibits topoisomerase 1, leading to the accumulation of DNA strand breaks and ultimately causing cell death. Upon internalization into the target cells, the β-glucuronic acid-based linker is cleaved to release exatecan. Compared with PR304-68B (DAR8), PR304 (CPPS)-68B (DAR6) exhibited much lower body weight loss and improved safety in mouse toxicity studies (Fig. 2h), demonstrated superior PK profiles including longer half-life, higher drug exposure and slow ADC clearance in mouse PK studies (Fig. 2i and Supplementary Table S2).

Figure 2.

Cys-to-Ser mutation in the hinge region for the development of homogeneous DAR6 ADCs. (a) Schematic diagram of the structure of the anti-c-Met biparatopic antibody PR304 with CPPX mutation. (b–e) Size exclusion chromatography (SEC) of PR304 (b) and PR304 with CPPS mutation (c), CPPT mutation (d), and CPPG mutation (e). (f) Hydrophobic interaction chromatography (HIC) of PR304 (CPPS) (black) and its counterpart ADC (red). (g) Size exclusion chromatography (SEC) of PR304 (CPPS) conjugated with the linker-payload 68B. (h) Body weight changes over time in c-Met/EGFR humanized mice (B-hMET/hHGF/hEGFR) following administration of PR304 (CPPS)-68B (DAR6) and PR304-68B (DAR8). (i) Mean serum concentration–time profiles of total antibody (Tab) and ADC for PR304 (CPPS)-68B (DAR6) and PR304-68B (DAR8) in mice.

Hinge Ser/Thr substitution is associated with O-linked glycosylation

During intact mass analysis of PNGase F treated PR304 (CPPS), additional mass species were observed that were absent in the wild-type PR304 (CPPC). These species were detected in antibodies produced by both stable and transient expression systems (Fig. 3a–c). The observed mass shifts (+203, +365, +657, +860, +1022, and +1314 Da) were consistent with the presence of GalNAc, Core 1, and sialylated Core 1 O-glycan structures (Fig. 3d). Subunit-level analysis following IdeZ digestion localized these mass additions to the F(ab′)₂ region (Supplementary Fig. S2 and Supplementary Table S3). These data indicate that hinge-region Ser/Thr substitution is associated with the appearance of mucin-type O-glycosylation.

Figure 3.

O-Glycosylation analysis of the de-N-glycosylated intact PR304 using RPLC–Q-TOF-MS at intact level. (a) Deconvoluted spectrum of PR304 (CPPS)-S produced from a stable cell line. (b) Deconvoluted spectrum of PR304 (CPPS)-T produced by transient expression. (c) Deconvoluted spectrum of PR304 (CPPC). (d) O-glycan modifications introduced by mutation. GalNAc, N-acetyl-d-galactosamine; Gal, Galactose; Neu5Ac, N-acetylneuraminic acid; C1, GalNAc-Galactose; C1S1, GalNAc-Gal-Neu5Ac.

O-Glycosylation occupancy was estimated by deconvolution of intact mass spectra. PR304 (CPPS) produced by transient expression, designated "PR304 (CPPS)-T," exhibited a higher proportion of O-glycosylation than that produced from a stable cell line, designated "PR304 (CPPS)-S (Table 1). These differences likely reflect expression-system-dependent variation rather than intrinsic differences in antibody sequence.

Table 1.

O-Glycosylation ratios of PR304 (CPPS) produced in different expression system as measured by RPLC−Q-TOF-MS.

	O-Glycan	Theoretical mass (Da)	Observed mass (Da)	Error (Da)	Peak area	O-Glycosylation (%)
PR304 (CPPS)-S	Aglycosylation	144 089	144 087	−2	551 870 848	27.4
	GalNAc	144 292	144 289	−3	157 864 339
	C1	144 455	144 451	−4	67 333 772
	C1S1	144 749	144 741	−8	83 680 665
	GalNAc/C1S1	144 952	144 947	−5	42 128 630
	C1/C1S1	145 111	145 106	−5	30 545 353
	C1S1/C1S1	145 402	145 399	−3	39 682 515
PR304 (CPPS)-T	Aglycosylation	144 089	144 088	−1	243 243 923	48.3
	C1	144 455	144 452	−3	50 727 243
	C1S1	144 749	144 744	−5	136 325 071
	C1/C1S1	145 111	145 110	−1	72 681 162
	C1S1/C1S1	145 402	145 399	−3	148 615 123

The proportion of O-glycosylation (Table 1) was calculated from LC–MS intact mass deconvolution peak area percentage using the following equation:

where Area₀ is the sum of the peak areas of PR304 (CPPS) without O-glycan, Area₁ is the sum of the peak areas of PR304 (CPPS) with one O-glycosylation, Area₂ is the sum of the peak areas of PR304 (CPPS) with two O-glycosylations and Area_total is the sum of Area₀, Area₁, and Area₂.

Site localization and core-type characterization of hinge O-glycosylation

Tryptic peptide mapping of PR304 (CPPS)-S identified O-glycosylation on the heavy-chain hinge peptide ₁THTCPPSPAPELLGGPSVFLFPPKPK₂₆ spanning residues 1–26. This region contains three potential O-glycosylation sites (Thr1, Thr3, and Ser7). NetOGlyc-4.0 (https://services.healthtech.dtu.dk/services/NetOGlyc-4.0/) was used as in silico assay to predict the O-glycosylation sites of the sequence "THTCPPSPAPELLGGPSVFLFPPKPK," and only Ser7 was predicted as a positive site. To localize the modification site by experiment, O-glycoprotease (IMPa) digestion was employed, exploiting its specificity for O-glycosylated serine or threonine residues. IMPa treatment generated a diagnostic peptide consistent with cleavage N-terminal to Ser7 (Fig. 4a–c). Extended trypsin digestion confirmed that this peptide was not generated by nonspecific proteolysis (Supplementary Fig. S3).

Figure 4.

O-Glycosite identification using IMPa and LC–MS/MS after trypsin digest for 4 h. (a) The restriction enzyme cleavage site of IMPa. (b and c) Sequence coverage map of tryptic O-glycopeptide without (b) or with IMPa (c). Sequence coverage maps are got by PepFinder 2.0. (d) EThcD fragmentation mass spectrum of ₁THTCPPSPAPELLGGPSVFLFPPKPK₂₆ with C1 from PR304 (CPPS)-S: The position of Ser7 is occupied by O-linked glycosylation. (e) HCD MS/MS spectrum of ₁THTCPPSPAPELLGGPSVFLFPPKPK₂₆ with a mono-sialylated trisaccharide from PR304 (CPPS)-S digested by trypsin for 4 h. Oxonium ions GalNAc at m/z 204.0865; Neu5Ac at m/z 274.0917 and 292.1024; and C1 at m/z 366.1397.

While IMPa enzymatic digestion suggested Ser7 as an O-glycosylation site, direct experimental evidence remains lacking. For this reason, the data-dependent acquisition with EThcD fragmentation modes was employed. The spectrum of PR304 (CPPS)-S (Fig. 4d) contained two peaks at m/z 357.19 and 711.33, corresponding to the c3-ion and c6-ion without O-glycosylation (Supplementary Table S4), indicating that Thr1 and Thr3 are not O-glycosylation sites. Meanwhile, the c8-ion + C1 was detected. This provided evidence for O-glycosylation of Ser7. The EThcD fragmentation spectrum of PR304 (CPPS)-T shows similar fragmentation patterns (Supplementary Fig. S4a and Supplementary Table S4).

Figure 4e shows the HCD fragmentation spectrum of peptide ₁THTCPPSPAPELLGGPSVFLFPPKPK₂₆ carrying a mono-sialylated trisaccharide from PR304 (CPPS)-S. The spectrum showed characteristic glycosylation-related oxonium ions at lower m/z values: 204.0865, 274.0917, 292.1024, and 366.1397 Da, which result from glycosidic bond dissociation. In the HCD fragmentation mode, although precise identification of modification sites still presents technical challenges, the detection of characteristic oxonium ions (e.g. 366.1397 Da) provides definitive evidence for confirming the mucin-type core 1 structure of O-linked glycans. The HCD fragmentation spectrum of PR304 (CPPS)-T shows similar fragmentation patterns (Supplementary Fig. S4b). Glycan structures were assigned at the core-type level (GalNAc, Core 1, sialylated Core 1); fine structural isomers were not resolved using the current fragmentation strategy.

Quantitative glycopeptide analysis showed that PR304 (CPPS)-T exhibited higher overall O-glycosylation than PR304 (CPPS)-S, consistent with intact mass analysis (Table 2 and Supplementary Fig. S5).

Table 2.

Glycoform profiles at Ser7 for PR304 (CPPS) produced in different expression systems.

	Peptide sequence	Glycoform	Theoretical [M + 3H]3+ (Da)	Observed m/z (Da)	Error (ppm)	RT (min)	Peak area (%)	O-Glycosylation (%)
PR304 (CPPS)-S	THTCPPSPAPELLGGPSVFLFPPKPK	GalNAc	992.5123	992.5130	1	52.10	3.05	25.87
		C1	1046.5299	1046.5319	2	51.33	3.69
		C1S1	1143.5617	1143.5643	2	52.53	19.13
		Aglycosylation	924.8192	924.8208	2	53.62	74.14
PR304 (CPPS)-T		GalNAc	992.5123	992.5147	2	52.63	1.22	42.07
		C1	1046.5299	1046.5333	3	51.85	8.12
		C1S1	1143.5617	1143.5652	3	53.00	32.73
		Aglycosylation	924.8192	924.8212	2	54.17	57.93

Functional tolerance of hinge O-glycosylation under ADC-relevant assays

To investigate whether the hinge O-glycosylation affects antibody or ADC stability and function, PR304 (CPPS)-S and PR304 (CPPS)-T, which differ in their levels of O-glycosylation, were first compared across multiple physicochemical and biological endpoints. Both antibody variants showed satisfactory stability under stress conditions as assessed by SEC, iCIEF, and reduced CE-SDS (Fig. 5a and b; Supplementary Figs S6 and S7; and Supplementary Tables S5–S7). DSC was employed to measure the heat capacity required to induce structural changes. The DSC curves for PR304 (CPPS)-S and PR304 (CPPS)-T differ from that of PR304, possibly due to mutation. The similar thermal unfolding profiles and thermal transition midpoint temperatures indicate that the thermal stability and conformation of the PR304 (CPPS)-S and PR304 (CPPS)-T are similar (Supplementary Fig. S8 and Supplementary Table S8). HIC analysis of the corresponding ADCs demonstrated a single dominant DAR6 species for both variants (Fig. 5c–e). SEC-MS analysis was used to monitor molecular weight changes and DAR values of the antibodies and ADCs under the stress condition. The results showed that no significant changes in molecular weight were observed for the antibodies after incubation at 40°C for 14 days, nor were there any obvious changes in DAR values for the ADCs after incubation at 37°C for 14 days (Supplementary Tables S9 and S10). For the ADC samples stored at 37°C for 14 days, the measured molecular weight was greater than that at the initial time point (0 day), which is speculated to be due to the ring-opening reaction of the succinimide linker [54]. The average DAR values for PR304 (CPPS)-S-68B and PR304 (CPPS)-T-68B at day 0 were comparable, at 5.8 and 5.9, respectively (Supplementary Fig. S9 and Supplementary Table S10).

Figure 5.

Characterization and comparison of physicochemical and biological properties of PR304 samples with low and high level O-glycosylation. (a and b) SEC of PR304 (CPPS)-S (a) and PR304 (CPPS)-T (b) at day 0 and after 37°C incubation for 14 days. (c) Schematic diagram of the structure of the PR304 (CPPS) ADC. (d and e) HIC of PR304 (CPPS)-S (d) and PR304 (CPPS)-T (e) and their counterpart ADCs. (f) Mean serum concentration–time profiles of PR304, PR304 (CPPS)-S and PR304 (CPPS)-T in mice. (g) In vitro anti-tumor activity of PR304 (CPPS)-S-68B and PR304 (CPPS)-T-68B against NCI-H2170 cells. Data were presented as Mean ± SD. (h) In vivo anti-tumor activity of ABT700 (CPPS)-68B, PR304 (CPPS)-S-68B, and PR304 (CPPS)-T-68B in HT29 xenograft model. Data were presented as Mean ± SEM. (i) In vivo anti-tumor activity of PR304 (CPPS)-S-68B and ABBV-399 analog in MKN45 xenograft model. Data were presented as Mean ± SEM.

PK analysis for PR304, PR304 (CPPS)-S, and PR304 (CPPS)-T in BALB/c mice showed dose-normalized exposure was maintained within a physiologically relevant range (Fig. 5f and Supplementary Table S11). These PK data are exploratory and were not powered to detect subtle differences. In vitro cytotoxicity assays against NCI-H2170 cells, PR304 (CPPS)-S/T-68B exhibited comparable anti-tumor activity (Fig. 5g). In an HT-29 xenograft model, the two conjugates demonstrated anti-tumor activity comparable to or slightly higher than ABT700 (CPPS)-68B, which uses the c-Met antibody from ABBV-399(Fig. 5h). In the MKN45 xenograft model, PR304 (CPPS)-S-68B outperformed the ABBV-399 analog at both dose levels, achieving statistical significance at the lower dose.

To further investigate the functional tolerance of the hinge O-glycosylation, PR304 (CPPS)-S, PR304 (CPPS)-T, and the parental PR304 were compared across multiple aspects. PR304 (CPPS)-S/T and PR304 demonstrated comparable cell binding abilities and internalization potency on EBC-1 cells, c-Met affinity, and Fc receptor affinity (Supplementary Fig. S10a–c and Supplementary Tables S12 and S13). Possibly due to differences in the glycoform profiles of N-glycans and O-glycans, the C1q-binding activity of PR304 (CPPS)-T was reduced. Following conjugation, PR304 (CPPS)-S-68B and PR304 (CPPS)-T-68B also demonstrated comparable cell binding activities and internalization potency on EBC-1 cells (Supplementary Fig. S10d and e).

Occurrence of O-glycosylation in additional CPPX-engineered antibodies

To evaluate whether hinge-associated O-glycosylation was restricted to CPPS mutation, additional CPPX mutations were examined. PR304 (CPPT) similarly exhibited O-glycosylation, whereas PR304 (CPPG) did not (Supplementary Fig. S11a and b and Supplementary Table S14), further strengthening the link between CPPT/S mutations and O-glycosylation.

Three additional antibodies targeting distinct antigens (CA149, CA608.1, and hCA450-21.1) were engineered with the CPPS mutation and expressed transiently. Intact mass analysis of these antibodies revealed O-glycosylation species consistent with those observed for PR304 (Supplementary Fig. S11c–e and Supplementary Table S14). These findings indicate that hinge-associated O-glycosylation can occur across multiple antibody scaffolds carrying CPPS or CPPT mutations, although the extent may vary depending on the sequence and expression system. To determine whether different antibodies with the CPPS mutation can consistently and reliably yield DAR6 products, the linker–payload 68B was conjugated to the three antibodies—CA149 (CPPS), CA608.1 (CPPS), and hCA450-21.1 (CPPS)—with distinct frameworks. HIC analysis revealed that all three antibody conjugates produced homogeneous species consistent with the expected DAR6 products (Supplementary Fig. S12).

Discussion

ADCs with homogeneous DAR are increasingly recognized for their favorable physicochemical properties and more predictable PK behavior. While established strategies exist to generate homogeneous ADCs with DAR2, DAR4, or DAR8, methods enabling homogeneous DAR6 ADCs remain comparatively underdeveloped, despite growing interest in intermediate DAR configurations for highly potent payloads. In this study, we focused on hinge-region cysteine engineering as a practical strategy to constrain conjugation stoichiometry and enable reproducible generation of homogeneous DAR6 ADCs.

By selectively disrupting one interchain disulfide bond within the IgG1 hinge region, we reduced the number of accessible cysteine conjugation sites from eight to six, thereby constraining drug loading to DAR6. Screening of CPPX and XPPC hinge variants using a practical antibody scaffold identified the CPPS mutation as particularly favorable with respect to conjugation homogeneity and aggregation propensity. Application of this mutation to a therapeutically relevant biparatopic anti-c-Met antibody enabled generation of ADCs with a dominant DAR6 species, as confirmed by HIC and native SEC-MS. These findings demonstrate that hinge-region cysteine engineering can serve as a rational and adaptable approach for DAR6 ADC development.

During analytical characterization of CPPS- and CPPT-engineered antibodies, we observed the appearance of additional mass species consistent with mucin-type O-linked glycosylation. Importantly, we emphasize that the data support an association between hinge Ser/Thr substitution and O-glycosylation, rather than direct enzymatic induction. Introduction of serine or threonine residues into a flexible, proline-rich hinge region likely creates a permissive structural context for O-GalNAc glycosylation.

The use of orthogonal analytical approaches—including intact and subunit-level LC–MS, O-glycoprotease cleavage specificity, and EThcD mass spectrometry detection—provides robust support for site localization. Comprehensive mass spectrometry analyses localized the O-glycosylation to a single serine residue within the hinge region and identified the attached glycans at the core-structure level, including GalNAc, Core 1, and sialylated Core 1 species. We explicitly note that fine structural isomers and linkage-specific details were not resolved using the current HCD-based fragmentation strategy, and therefore refrain from higher-order structural claims.

To assess the potential functional relevance of hinge-associated O-glycosylation, we compared antibody and ADC variants differing in O-glycosylation abundance across a range of ADC-relevant assays. Across these assays, we observed no detectable differences in stability under stress, in vitro activity except for C1q binding activity, PKs, or in vivo efficacy under the conditions tested. Importantly, these evaluations were exploratory in nature, particularly for PK and in vivo studies, which were not powered to detect subtle differences. Moreover, long-term stability was not assessed and may warrant further investigation. High-resolution structural analyses have established that the C1q-binding site on IgG is located adjacent to the hinge region (PDB: 6FCZ), in close spatial proximity to the CPPS mutation. Consequently, O-glycans introduced in this region may directly interfere with C1q binding to the Fc segment through steric hindrance, or indirectly modulate accessibility of the C1q-binding epitope by altering the local conformational dynamics of the hinge. Additionally, it has been reported that galactosylation on antibody N-glycans can enhance C1q binding and promote CDC activity [55]. Glycoform analysis showed that the galactosylation level of PR304 (CPPS)-T was lower than that of PR304 (CPPS)-S (data not shown). Therefore, the higher O-glycosylation rate (Table 1) and lower galactosylation level of N-glycans observed in PR304 (CPPS)-T compared to PR304 (CPPS)-S may account for the reduced C1q binding activity. For the anti-c-Met ADC, the primary cell-killing mechanism is direct cytotoxicity mediated by the payload exatecan, which is released following ADC internalization and lysosomal degradation to disrupt DNA replication and trigger apoptotic cell death. Both ADCC and CDC represent Fc-dependent effector functions that may provide supplementary anti-tumor activity; however, for ADCs bearing highly potent topoisomerase I inhibitors, these are considerably less potent and generally regarded as auxiliary rather than dominant mechanisms. Consequently, although diminished CDC activity may theoretically impair overall therapeutic efficacy and offer a clinical safety benefit given that c-Met is expressed on normal tissues, but the impact on clinical efficacy and safety is expected to be minimal or negligible. Furthermore, a head-to-head comparison of the homogeneous DAR6 ADC versus a heterogeneous DAR6 ADC in terms of efficacy and safety would provide stronger evidence for the advantages of site-specific conjugation. These studies will be pursued in future work to further substantiate this benefit.

The extent of hinge-associated O-glycosylation varied between antibodies produced by transient versus stable expression systems, and three factors may account for this discrepancy. First, host cell differences: CHO-S was used for transient transfection, whereas CHO-K1 was used for the stable cell line. Despite their common CHO origin, distinct adaptation and cultivation histories, coupled with the inherent genetic instability of CHO cells, may yield divergent O-glycosylation profiles. Second, cell-specific productivity: the non-replicating antibody gene in transient transfection results in lower productivity, alleviating endoplasmic reticulum congestion and enabling more complete glycosylation. Stable cell lines, with higher productivity, may saturate O-glycosylation capacity, reducing modification efficiency. Third, medium composition: Mn²⁺ concentration generally modulates enzyme activity, and the differing media used in the two systems may have altered glycosyltransferase activity, affecting O-glycosylation levels. In summary, the observed O-glycosylation differences likely stem from combined effects of host cell, expression system, and medium. Future work will further validate the CPPS strategy across additional platforms (e.g. CHO, HEK293). Additionally, observation of O-glycosylation in multiple antibody scaffolds carrying CPPS or CPPT mutations indicates that this phenomenon is not restricted to a single antibody sequence, although the degree of modification may vary depending on sequence context and production system. These findings underscore the importance of analytical characterization when applying hinge-engineering strategies across different antibody platforms.

Taken together, this study demonstrates that hinge-region cysteine engineering can enable homogeneous DAR6 ADCs and identifies hinge-associated O-glycosylation as a molecular feature that may accompany this strategy. O-Glycosylation typically occur within flexible, proline-rich regions, and introduction of serine or threonine residues into the IgG1 hinge region may create a permissive context for O-glycosylation. Within the scope of the assays performed, this modification appears to be functionally tolerated. These results provide a framework for rational DAR6 ADC design while highlighting the need for case-by-case evaluation of associated post-translational modifications during antibody engineering.

Author contributions

Deyong Song (Conceptualization [equal], Data curation [equal], Investigation [equal], Project administration [lead], Writing—original draft [lead]), Zhenduo Shen (Investigation [equal], Supervision [equal]), Xiaolin Zhu (Investigation [equal], Methodology [equal]), Haifeng Zhao (Methodology [equal], Validation [equal]), Yingdi Wang (Investigation [equal], Methodology [equal]), Junmei Zhang (Investigation [equal], Methodology [equal]), Rui Li (Investigation [equal], Methodology [equal]), Hongguang Xu (Investigation [equal]), Zengjin Cheng (Investigation [equal]), Xiaoju Ma (Investigation [equal], Methodology [supporting]), Haixia Lin (Investigation [equal], Methodology [supporting]), Changlin Dou (Funding acquisition [equal], Supervision [equal]), Min Xiao (Conceptualization [equal], Project administration [equal], Supervision [equal], Writing—review & editing [lead]).

Conflict of interest

Z.S., X.Z., H.Z., Y.W., J.Z., R.L., H.X., Z.C., X.M., H.L., and C.D. are employees of Shangdong Boan Biotechnology Co., Ltd. The authors declare no other competing financial interests.

Funding

This work was supported by the Yantai Science and Technology Innovation Development Plan Projects (2023ZDCX030), Shandong Boan biotechnology Co., Ltd., and Intramural Joint Program Fund of State Key Laboratory of Microbial Technology (SKLMTIJP-2025-03).

Data availability

All relevant data in the article, supplementary information are available from the corresponding author M.X., upon reasonable request. The sequences of biparatopic antibody PR304 (CPPS) have been deposited in GenBank with the accession numbers PX095975 and PX095974 for heavy chains of Ab1 and Ab2, and PX095976 for the common light chain, respectively. Source data are provided in this paper. The other source data generated in this study are provided in the supplementary information and the Source Data file.

Ethics and consent statement

Not applicable.

Animal research statement

All animal experiments were conducted in compliance with relevant ethical regulations and the ARRIVE guidelines. All procedures were approved by the Animal Ethics Committee of Shandong Boan Biotechnology Co., Ltd. (Yantai, China).