A framework for evaluating unexpected reactions in antibody-drug conjugates during clinical development: a case study of a promiscuous payload Authors: Eef H. C. Dirksen, Marijn Buurman, Jeroen Kooistra, Benno Ingelse, Ruud Ubink Published: 2026 ABSTRACT ABSTRACTAntibody-drug conjugates (ADCs) are complex biotherapeutic molecules, comprising an antibody moiety and a small molecule linker-drug. The product quality of an ADC is controlled through monitoring of critical quality attributes to ensure safety and efficacy. During development, unexpected changes may occur in the product quality of an ADC, for example under stress or physiological conditions. This should prompt a systematic evaluation to determine the cause and impact on product quality. A structured approach for such an evaluation is not well established. Here, a strategy for the detection, identification, and assessment of unexpected reactions in an ADC during clinical development is presented encompassing physicochemical and functional analytical characterization, mechanistic confirmation, and evaluation of the relevance in vivo. The strategy is illustrated by a real-world example of an ADC in clinical development, SYD1875, exhibiting an unexpected decrease in biological activity following 3 months of storage at the stressed condition. Comprehensive analytical characterization revealed an intramolecular cross-link between the linker-drug and a glutamic acid in the light chain of the antibody. Mechanistic studies confirmed that this cross-link abolishes the cytotoxicity of the ADC. Subsequent bioanalytical and pharmacokinetic investigations demonstrated that, although the cross-link is also formed in vivo, its overall impact on systemic exposure was limited. The generalizable framework described in this report aims at facilitating the evaluation of anomalies in the product quality of an ADC and subsequent (clinical) risk assessment, thereby supporting decision making. Introduction Antibody-drug conjugates (ADCs) are a class of therapeutic molecules that combine the specificity of monoclonal antibodies with the potency of small molecule drugs, allowing for specific delivery of payloads to cells to selectively kill them. In addition to the ADCs that are currently available for the treatment of various oncological indications, such as Kadcyla® (ado-trastuzumab emtansine), Enhertu® (fam-trastuzumab deruxtecan), and Trodelvy® (sacituzumab govitecan), hundreds of ADCs are being evaluated in clinical investigations. By using a monoclonal antibody to deliver a potent small molecule to cells expressing a particular surface antigen, an intricate vehicle is created that requires thorough analysis by a set of orthogonal analytical methods to ensure the desired product quality, assuring efficacy and safety for use in patients. For this purpose, guidelines have been established by, among others, the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH) and regulatory agencies (e.g., the US Food and Drug Administration (FDA) and the European Medicines Agency). Given the complexity of the manufacturing and purification processes of ADCs and their intrinsic molecular heterogeneity, a series of analytical tests is employed to monitor the quality of the ADC during chemistry, manufacturing and controls (CMC) development, as represented by an ensemble of critical quality attributes (CQAs). CQAs are physical, chemical, biological, or microbiological properties or characteristics that should be within an appropriate limit, range, or distribution to ensure the desired product quality.1 For a typical ADC, these include identity, purity, drug-to-antibody ratio (DAR) and drug distribution, charge distribution, product- and process-related impurities, microbial quality, and biological activity. The latter is usually monitored using two assays: 1) an enzyme-linked immunosorbent assay (ELISA)-based method focused on target binding of the antibody moiety, and 2) a cell-based cytotoxicity assay that provides a representation of the mechanism of action of the ADC, capable of providing insight into the potency of the ADC. As data for the CQAs of the ADC at the time of release of the product are acquired, formulated ADCs are also stored under various conditions, for example, according to the ICH guideline Q1A,2 to assess the (long-term) stability of the product. Typically, such stability studies comprise three storage conditions: 1) the anticipated long-term temperature storage condition, 2) an accelerated temperature storage condition, and 3) a stressed temperature storage condition. The latter two are included to accelerate potential degradation pathways, resulting in the formation of degradation products. This approach helps building product knowledge and allows for extrapolation of stability data obtained at the long-term condition to support justification of the shelf life (expiry date) of pharmaceutical products in the early stages of pharmaceutical development. Despite the advancements in analytical control strategies, unexpected reactions can occur within an ADC product during development. In contrast to the guidelines that exist for the (bio)analytical evaluation of metabolites in either small molecule or protein drugs, limited guidance from ICH3–5 or FDA6 exists when it comes to the (bio)analysis of ADCs that combine both. Upon encountering a change in the product quality of an ADC, it is important to initiate a tiered approach that allows a systematic evaluation of the observation, leading to a mechanistic understanding of the root cause and an informed assessment of its clinical relevance. Such an approach includes six key elements: 1) detection of a change in a CQA using a stability-indicating analytical method, 2) evaluation of complementary analytical data obtained for other CQAs using physicochemical methods, 3) in-depth characterization using advanced analytical techniques, such as mass spectrometry (MS), 4) empirical confirmation of the underlying mechanism, 5) evaluation of the occurrence and impact of the identified root cause in vivo, and 6) integration of all data to assess clinical relevance. Here, the approach described above is illustrated by an example involving the ADC drug product SYD1875. SYD1875 comprises an IgG1-type monoclonal antibody directed against the trophoblast glycoprotein 5T4, to which two duocarmycin-based linker-drugs (Figure 1) are conjugated site-specifically via two engineered cysteines7 in the heavy chain of the antibody. Upon binding to the surface antigen, the SYD1875-receptor complex is internalized and, after lysosomal degradation of the ADC, the linker-drug is activated and translocated to the nucleus where the duocarmycin moiety intercalates into the DNA and subsequently alkylates adenine residues at the N3 position.8 This eventually leads to cell death through DNA-damage-induced apoptosis.9 Figure 1.Chemical structure of the duocarmycin-containing linker-drug used in SYD1875. The top left part contains the maleimide group that reacts with the engineered cysteine in the antibody, coupled to an enzymatically cleavable valine-citrulline linker. The right-hand side of the molecule contains the DNA-alkylating and DNA-binding moieties that are part of the payload.Two-dimensional line-bond structural formula with atom labels O, N, Cl, H and OH and no charges, arrows, or wedge or dash stereobonds. At the far left, a five-membered ring contains one N and two C double bond O groups and one C double bond C bond. The ring N is single bonded to a three-carbon chain that contains two O atoms as ether links and ends at a carbonyl C double bond O connected to an NH. That NH is connected through a chain of amide linkages and carbon atoms. One carbon bears a single-bond substituent drawn as a short line and later a side chain ends at an NH connected to a carbonyl C double bond O that is single bonded to NH subscript 2. A six-membered benzene ring appears mid-structure, attached through an NH to a carbonyl C double bond O. The benzene ring is substituted by a single-bond methylene chain leading to an O, then a carbonyl C double bond O, then an N that also bears a single-bond methyl substituent. This N is connected by a two-carbon chain to another N. From this second N, one branch goes downward through a three-carbon chain to an O, then a two-carbon chain ending in OH. The other branch goes right through a carbonyl C double bond O and an O to a fused polycyclic carbon framework that includes a five-membered ring containing one N. A side chain from this region ends in Cl. Continuing right, the five-membered N-containing ring is connected through a carbonyl C double bond O to a fused bicyclic ring system containing two N labels. This is linked by an NH to a carbonyl C double bond O attached to a terminal six-membered benzene ring bearing an OH substituent.Two-dimensional structural formula of the duocarmycin-containing linker-drug used in SYD1875. Chemical structure of the duocarmycin-containing linker-drug used in SYD1875. The top left part contains the maleimide group that reacts with the engineered cysteine in the antibody, coupled to an enzymatically cleavable valine-citrulline linker. The right-hand side of the molecule contains the DNA-alkylating and DNA-binding moieties that are part of the payload. Following 3 months of storage at the stressed temperature storage condition (25 °C), SYD1875 was found to exhibit a decrease in potency. Combining the stability data obtained with mechanistic studies and a bioanalytical assessment identified a linker-drug mediated intramolecular cross-link as the root cause and provided insight into the qualitative and quantitative occurrence of cross-linking in vivo and the clinical relevance thereof. By presenting this work in a structured manner, we aim to share a generalizable framework for the evaluation of unexpected reactions in ADCs during clinical development. Materials and methods SYD1875 was manufactured at Byondis B.V., Nijmegen, The Netherlands, and formulated to a concentration of 10 mg/mL. Aliquoted ADC material was stored in glass vials at −20°C (long-term storage condition) for up to 36 months, at 5°C (accelerated storage condition) for up to 6 months and at 25°C (stressed storage condition) for up to 6 months, respectively, and analyzed at the start of the stability study (t = 0) and at predefined time points (e.g., t = 1, 3, and 6 months). All animal-derived materials were obtained from BioIVT (Westbury, NY, USA) and originated from studies conducted in accordance with applicable institutional and national guidelines for animal welfare, with protocols approved by the relevant Institutional Animal Care and Use Committee (IACUC) or equivalent ethical review body. Human plasma samples were obtained from BioIVT (Westbury, NY, USA), which collects and provides human biological materials under Institutional Review Board (IRB)-approved protocols, with informed consent obtained from all donors, in accordance with the Declaration of Helsinki. Biological activity by cell-based assay The cell-based assay was performed by measuring SYD1875-dependent cytotoxicity toward the human breast carcinoma cell line MDA-MB-468 by quantitating adenosine triphosphate levels using PrestoBlue and a fluorescence measurement using a Varioskan LUX. The assay is aimed at establishing similarity between the curves obtained for the sample of interest and those of the SYD1875 reference material (relative potency). Multi-attribute method by RPLC-MS/MS Multi-attribute method (MAM) was performed using 50 µg of lyophilized sample that was reconstituted in 8 M urea and 10 mM dithiothreitol (DTT) in 20 mM ammonium bicarbonate. Following an incubation at 70°C for 3 min while shaking, 1 μg of endoproteinase Lys-C was added, after which the mixture was left at 37°C for 15 min in a thermomixer shaking at 1200 rpm. Then, the sample was diluted to less than 1 M urea using 20 mM ammonium bicarbonate, after which 2.5 μg of trypsin was added. The mixture was incubated for another 15 min at 37°C followed by acidification with trifluoroacetic acid to quench the digestion. The resulting proteolytic peptide mixture was separated using reversed-phase liquid chromatography (RPLC) with a gradient of 0.1% formic acid (FA) in acetonitrile in 0.1% FA in ultrapure water (Milli-Q). The injection amount was 3 μg, and eluting peptides were detected using UV absorbance at 214, 280, and 330 nm, as well as MS. The latter was performed using a Thermo Fisher Scientific Orbitrap Fusion mass spectrometer operating in positive ion mode using electrospray ionization with MS detection in the Orbitrap at 120 K resolution over the scan range m/z 400 to 1,600 and further analyzed using data-dependent tandem MS (MS/MS) using higher-energy collisional dissociation (HCD) activation in the quadrupole and subsequent mass analysis in the ion trap on the most abundant precursor ions. Incubation of SYD1875 in plasma for cross-linking evaluations SYD1875 was diluted to 4 µg/mL in mouse, cynomolgus, and human EDTA plasma, aliquoted, and stored at ≤−65°C. For stability testing, samples were incubated at 37°C, retrieved at pre-set time points (0, 1, 3, 4, 7, 11, 14 d) and subsequently stored at −70°C. More details are provided in Supplementary Information Section S1. Generation of a fully cross-linked SYD1875 standard Non-crosslinked SYD1875 was extracted from temperature-stressed SYD1875 by ligand-based extraction using streptavidin-coated beads and a biotinylated anti-hapten monoclonal mouse IgG2A antibody (custom-made by BioGenes GmbH). A detailed description of the method is provided in Supplementary Information Sections S2 and S3. ELISAs for total antibody, conjugated antibody (specific and non-specific) The bioanalytical assay capable of measuring concentrations of conjugated SYD1875 (CAb) in human plasma uses a capture antibody (VHH; single-domain antibody) directed against the DNA-binding part of the linker-drug (custom-made by QVQ, Utrecht, The Netherlands) (for details, please refer to Supplementary Information Section S4). As such, this assay is unable to distinguish cross-linked from non-cross-linked SYD1875. Therefore, an ELISA using a VHH directed toward the DNA-alkylating part of the linker-drug (custom-made by QVQ, Utrecht, The Netherlands) was developed, establishing specificity toward non-cross-linked SYD1875 (NC-CAb). This method is described in Supplementary Information Section S5. The total antibody (TAb) ELISA assay employs a SYD1875‑specific anti‑idiotype VHH (custom-made by QVQ, Utrecht, The Netherlands) as both capture and detection reagent, and is described in Supplementary Information Section S6. Relative quantitation of SYD1875 cross-linking using targeted MS (bioanalytical assay) Following ligand-based extraction of SYD1875, the ADC was reduced using DTT and subsequently digested by means of sequential Lys-C/tryptic proteolysis at 37°C overnight. This procedure also results in hydrolysis of the maleimide group and of the intramolecular cross-link, yielding identifiers that distinguish cross-linked from non-cross-linked SYD1875. Quality control samples (diluted in biological matrix) and calibration samples (diluted in buffer) were spiked with known amounts of cross-linked SYD1875. Samples were analyzed by LC-MS on a reversed-phase column using data-independent acquisition and full MS scanning. The relative amount of SYD1875 cross-linking was derived from the corresponding identifier signal responses. A detailed description of the method is provided in Supplementary Information Section S7. Pharmacokinetics in monkey Pharmacokinetic data were obtained from the females of a repeat-dose toxicity study in naïve purpose-bred cynomolgus monkeys (Macaca fascicularis) of Vietnamese origin (Labcorp, Münster, Germany). The animals were 39–50 months old and weighed 2.4–4.4 kg. They were assigned to treatment groups based on existing social groups and stratified body weights, housed in pairs or groups, acclimated to study procedures for at least 2 weeks prior to dosing. They were considered suitable for study participation following veterinary examination. Animals (n = 4 or n = 5 for the highest dose) received intravenous 30-min infusions of 3, 9 or 18 mg/kg SYD1875 once every 3 weeks; treatment administration was not blinded. Blood samples were collected from all individual animals at predefined time-points, processed to plasma, and analyzed for Tab, CAb, and NC-CAb. Pharmacokinetic parameters for each individual animal were determined using all time-points after the first dose by non-compartmental analysis in Phoenix WinNonlin version 8.2 (Certara). The toxicity study in cynomolgus monkeys was approved by the Landesamt für Verbraucherschutz und Ernährung Nordrhein-Westfalen, Germany (license number 81–02.042018.A041), and was performed in compliance with the German Animal Welfare Act and Directive 2010/63/EU of the European Parliament, the Council 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). Results The investigation of a potency decrease observed for an ADC during stability testing is presented here as a stepwise workflow to illustrate a comprehensive strategy for evaluating unexpected findings during ADC development. Each step in the workflow, from initial detection to assessment of clinical relevance, comprises approaches that can be applied to address similar challenges occurring during the development of other ADCs. Detection of an unexpected change in a critical quality attribute of an ADC A batch of SYD1875 that was envisioned to be investigated in a Phase 1 clinical study to assess its safety was monitored for stability according to ICH guidelines2 using a panel of stability-indicating analytical methods at pre-defined time points. During this stability study, a decrease in the potency of SYD1875 was observed after 3 months of storage at 25°C, as evidenced by a shift in the EC50, the concentration of the ADC that induces a biological response halfway between the baseline and maximum, to higher SYD1875 concentration (Figure 2). Figure 2.Overlay of the cell viability as a function of SYD1875 concentration curves obtained in the cell-based cytotoxicity assay for the SYD1875 reference material used (purple curve) and the SYD1875 stability sample (25°C, 3 months, orange curve).Graph: Cell viability vs. SYD1875 concentration; reference and stability samples both show a decrease. Overlay of the cell viability as a function of SYD1875 concentration curves obtained in the cell-based cytotoxicity assay for the SYD1875 reference material used (purple curve) and the SYD1875 stability sample (25°C, 3 months, orange curve). The shift observed was such that the curve obtained for the SYD1875 stability sample differed substantially from that of the product-specific reference material, hindering the generation of a relative quantitative number for its biological activity. Both SYD1875 stored at the anticipated long-term condition (−20 °C) and the accelerated condition (5 °C) for 3 months did not exhibit a decrease in potency: those curves were found to overlap with that obtained for the reference standard, indicating similar relative potency. Although the stressed condition (25 °C) of the stability study was included to speed up any degradation, the decrease in biological activity observed was not expected. Therefore, the nature of this decrease was investigated using a generalizable strategy, in accordance with the limited ADC-specific regulatory guidance available. Assessment of data obtained for other physicochemical CQAs To identify the root cause of the observation, other data obtained for the SYD1875 stability sample were interrogated, for example those obtained for average DAR and drug distribution by hydrophobic interaction chromatography (HIC), for purity by RP-LC, and for purity by reduced capillary electrophoresis-sodium dodecyl sulfate (CE-SDS). The MAM was also included in the panel of stability assays. MAM is based on peptide‑mapping LC-MS/MS analysis and allows for in-depth characterization of protein-based molecules, such as ADCs, at the individual amino acid level.10 The MAM stability data processing was initially used in a targeted way,11 i.e., allowing for the evaluation of a defined set of post-translational modifications, such as methionine oxidation and asparagine deamidation. Even though some of the results of the aforementioned assays revealed signs of increased heterogeneity in SYD1875 upon storage at 25°C, as shown in Table S1, the results provided insufficient information to pinpoint the cause of the decrease in relative potency observed. This triggered a more detailed look into the integrity of the ADC. In-depth analysis of mass spectrometric data In addition to the data obtained for the relative levels of methionine oxidation and asparagine deamidation, the MAM LC-MS/MS data obtained for the SYD1875 stability sample could contain more information regarding the increase in product heterogeneity suggested by the physicochemical analytical methods used to monitor the stability of SYD1875. Unbiased analysis of the MAM data did not reveal any modifications in peptides harboring the complementarity-determining regions of the antibody moiety that recognize the antigen, so that could be ruled out as a reason for the decrease in potency observed. Further investigation of the MAM results focused on modifications of the payload. This was aided by the UV absorbance data that was also acquired as part of the analysis, as the duocarmycin-based linker-drug specifically absorbs UV light at a wavelength of 330 nm. Next to the peak representing the expected linker-drug-conjugated tryptic peptide, eluting at 61.41’ (Figure 3), an additional peak was observed at 51.97’ in the UV absorbance (330 nm) chromatogram of the SYD1875 stability sample, as compared to that of the reference material. This suggested that this ‘new’ compound was related to, or conjugated with, a linker-drug. Figure 3.Mass spectrometric (base peak chromatograms, top panels) and UV absorbance (330 nm, bottom panels) data obtained for the SYD1875 reference material (left panel) and the SYD1875 stability sample exhibiting decreased potency following 3 months of storage at 25°C (right panel).Images A and B display base peak chromatograms with Time (min) from 10 to 65 and Relative Abundance from 0 to 100. Image A peaks at 12.89, 14.53, 16.39, 20.86, 21.66, 27.80, 30.76, 35.43, 36.05, 40.53, 41.11, 41.98, 47.42, 49.37, 54.71, 55.17, 56.07, 57.03, 61.41, with highest at 30.76 and 56.07. Image B peaks at 12.82, 14.47, 20.83, 21.67, 27.81, 30.76, 35.44, 36.06, 40.52, 41.16, 42.00, 47.42, 49.40, 54.72, 55.17, 55.61, 56.03, 57.05, 61.12, 61.40, 61.41, with highest at 55.61 and 56.03. Images C and D show UV absorbance graphs with Time (min) from 10 to 65 and UV 330nm, ranging 0 to 3500 for C and 0 to 3000 for D. Image C peaks sharply at 61.41, while D peaks at 51.97, 60.68, 61.02, 61.21, 61.41, with dominant at 61.41. Both chromatograms share peaks at 30.76, 47.42, 55.17 and absorbance graphs peak at 61.41. All graphs use continuous lines.Four line graphs showing base peak chromatograms and ultraviolet 330 nanometer absorbance for the sample and the reference. Mass spectrometric (base peak chromatograms, top panels) and UV absorbance (330 nm, bottom panels) data obtained for the SYD1875 reference material (left panel) and the SYD1875 stability sample exhibiting decreased potency following 3 months of storage at 25°C (right panel). The peak at retention time (tR) 51.97’ was found to represent a peptide with an m/z of 606.9306 (3+), which is approximately 18 Dalton lighter than the expected conjugated peptide originating from SYD1875 (tR 61.4’ (Figure 3), m/z of 612.9192 (3+)). The newly formed peptide was also found to be less hydrophobic than the expected conjugated peptide, as evidenced by the shift in retention time of approximately 9 min in the RPLC run. Strikingly, MS/MS fragmentation of the alleged conjugated peptide that was only detected in the SYD1875 stability sample stored at 25°C yielded an intense fragment ion at m/z 509.26 (1+), which is distinctly different from the diagnostic fragment ions at m/z 491.26 (1+) and 527.20 (1+) formed upon MS/MS of the expected conjugated peptide (Figure 4). The latter two fragment ions represent variants of the fragment containing the DNA-alkylating moiety of the linker-drug: one containing a chlorine atom, the other a cyclopropyl group (Figure S1). Figure 4.Fragmentation (MS/MS) spectra of the conjugated peptide generated from a proteolytic digest of the SYD1875 reference material (m/z of 612.92 (3+), top spectrum) and of the peptide detected in the SYD1875 stability sample stored at the stressed condition (m/z of 606.93 (3+), bottom spectrum).Two stacked mass spectra. Top spectrum: The y-axis is labeled, Relative Abundance, from 0 to 100. The x-axis is labeled, m slash z, from 0 to 1900. Labeled peaks occur at 210.1770, 298.1700, 385.7650, 456.7600, 491.2580, 527.2010, 617.3190, 715.3500, 816.4100, 913.5180, 1044.5841 and 1211.4760. The tallest labeled peaks are at 491.2580 and 527.2010 near the top of the scale. Bottom spectrum: The y-axis is labeled, Relative Abundance, from 0 to 100. The x-axis is labeled, m slash z, from 0 to 1900. Labeled peaks occur at 212.1570, 298.1400, 394.7340, 456.7640, 509.2640, 617.3470, 697.3201, 833.4370, 913.4360, 1044.5701, 1155.7101, 1310.3330, 1426.8660 and 1628.6951. The tallest labeled peak is at 509.2640 near the top of the scale.Two tandem mass spectra comparing conjugated peptide fragmentation, with different dominant fragment peaks. Fragmentation (MS/MS) spectra of the conjugated peptide generated from a proteolytic digest of the SYD1875 reference material (m/z of 612.92 (3+), top spectrum) and of the peptide detected in the SYD1875 stability sample stored at the stressed condition (m/z of 606.93 (3+), bottom spectrum). The MS and MS/MS data suggested that, under the stressed condition, the DNA-alkylating moiety in SYD1875 was perturbed due to the elevated temperature it was exposed to during storage. The fragment (m/z 509.26) is 18 Da heavier than the fragment containing a cyclopropyl group (m/z 491.26) and 18 Da lighter than the variant containing the chlorine atom (m/z 527.20), which suggested a hydroxylation of the DNA-alkylating moiety of the linker-drug. Since such a modification had not been observed before in other ADCs containing this linker-drug,12 it was considered unlikely that the DNA-alkylating moiety of the duocarmycin linker-drug is directly hydroxylated in SYD1875 in formulation buffer. Hence, the hydroxylated variant was suspected to be a product of another reaction. The MAM data were further interrogated to find evidence of this hypothesis. Moreover, SYD1875 in formulation was incubated at 40°C to induce the unwanted reaction(s) underlying the loss of potency and subjected to MAM analysis to allow for a more in-depth assessment of the phenomenon. An unbiased search of the MAM dataset obtained for SYD1875 stressed at 40°C versus that obtained for the corresponding, non-stressed, SYD1875 reference material, revealed several other ions that were uniquely observed in the stressed sample, including those at m/z 788.15 (5+) and at m/z 791.75 (5+). The MS/MS spectrum obtained for the precursor ion at m/z 788.15 (5+) was found to contain a fragment ion characteristic of the linker-drug used: m/z 491.18 (1+) (Figure 5). Moreover, a short sequence tag could be deduced from the MS/MS fragmentation data: glutamic acid–serine–valine–threonine (Figure 5), which originates from the constant region of the light chain of SYD1875 and is in tryptic peptide V153DNALQSGNSQESVTEQDSK172. Figure 5.Fragmentation (MS/MS) spectrum of the tryptic peptide with an m/z of 788.15 (5+) observed in the data obtained for SYD1875 stored at 40°C for 35 d. A series of y ions are observed which corresponds to a sequence tag of the protein’s light chain. Based on this, the peptide was found to represent the linker-drug-cross-linked dipeptide: heavy chain (39–43)/linker-drug/light chain (153–172), where the linker-drug has formed a cross-link toward glutamic acid at position 168 in the light chain.A mass spectrum plot with many vertical peaks and annotated fragment ions. The x-axis label is m slash z, ranging from 100 to 1500 with major ticks every 100. The y-axis label is Relative Abundance, ranging from 0 to 100 with major ticks every 5. Labeled peaks include 115.03, 158.09, 215.05, 234.11, 298.09, 301.10, 329.15, 349.17, 372.17, 400.18, 468.23, 477.17, 491.18, 492.24, 513.29, 589.24, 617.28, 674.25, 731.29, 771.34, 816.35, 834.38, 893.40, 912.43, 913.39, 939.46, 956.42, 973.40, 1049.49, 1096.54, 1197.53, 1198.53, 1237.50, 1296.55, 1297.60, 1383.61, 1384.68, 1385.59, 1495.66 and 1512.67. The tallest labeled peak is at 491.18 near 100 relative abundance; another high peak is at 329.15 near 90. Blue annotations mark y2 at 234.11, y3 at 349.17, y4 at 477.17, y5 at 1096.54, y6 at 1198.53, y7 at 1296.55, y8 at 1384.68 and y9 at 1512.67. A peptide sequence is printed at the top as V D N A L Q S G N S Q E S V T E Q D S K, with y9 through y2 labels above the right side. A horizontal annotation reads E plus 490.2 Da aligned with the sequence region.A tandem mass spectrum of relative abundance versus mass to charge ratio with labeled y ion peaks and a sequence tag containing the modified glutamic acid residue. Fragmentation (MS/MS) spectrum of the tryptic peptide with an m/z of 788.15 (5+) observed in the data obtained for SYD1875 stored at 40°C for 35 d. A series of y ions are observed which corresponds to a sequence tag of the protein’s light chain. Based on this, the peptide was found to represent the linker-drug-cross-linked dipeptide: heavy chain (39–43)/linker-drug/light chain (153–172), where the linker-drug has formed a cross-link toward glutamic acid at position 168 in the light chain. Adding the theoretical mass of the tryptic light chain peptide 153–172 to the mass of the conjugated heavy chain peptide QAC(+linker-drug) GK and subtracting 36 Da to account for the loss of HCl resulted in a cross-linked species of 3934.72 Da that corresponds to the ion with an m/z of 788.15 at a charge state of 5 +. Upon further investigation of the MS/MS spectrum of this cross-linked peptide, more of the SYD1875 light chain peptide sequence could be deduced, including a fragment containing the glutamic acid residue at position 168 with an added mass of 490.2 Da (Figure 5). This mass increase reflects a part of the duocarmycin payload (Figure S2) and pinpoints the exact location and nature of the cross-link: an ester with a carboxylic acid. The ion detected at m/z 791.75 (5+), exhibiting a mass increase of 18 Da, further substantiated the identification, as this represented the maleimide-hydrolyzed version of the cross-linked peptide, a reaction frequently observed on maleimide-conjugated ADCs in circulation.13 The identification and MS/MS fragmentation of these SYD1875-related peptides confirmed that a covalent bond was formed between the conjugated engineered cysteine residue of the heavy chain and the side chain of the glutamic acid residue in position 168 of light chain of the antibody moiety of SYD1875 via the linker-drug; see Figure S3 for a modeled representation of the antigen-binding domain of SYD1875 showing the proximity of the reactive site in the payload and the glutamic acid residue in position 168 of the light chain. This intramolecular cross-link was expected to be the underlying cause of the reduction in potency as observed by the cell-based assay. No evidence of intramolecular cross-linking to other amino acids in SYD1875 was found in either the UV (330 nm) absorption chromatogram, or in the extracted ion chromatogram of the diagnostic fragment ion that was found to be representative of linker-drug-mediated cross-link formation (m/z 509.26; 1+). Moreover, data obtained by both non-reduced CE-SDS and SEC for the SYD1875 stability sample following 3 months of storage at the stressed condition (25 °C) provided no evidence of intermolecular cross-link formation. Neither an increase in post-peaks nor in high‑molecular‑weight species was observed (see Supplementary Table S1). This indicated that intramolecular cross-linking at the identified site (Glu168) in the light chain of SYD1875 is the sole reaction occurring. Taking the lack of reduction in potency for the SYD1875 stability samples stored at the long-term and accelerated conditions into account, it was concluded that cross-link formation is a chemical reaction driven by temperature. Due to the conditions used during sample preparation of the MAM, the cross-link is hydrolyzed, resulting in the peptide containing the hydroxylated version of the linker-drug as the predominant form that is representative of the intramolecular cross-link. The observations made in the stability data indicate that a subset of SYD1875 is converted to a species containing one intact linker-drug and one that has been inactivated as a result of intramolecular cross-linking upon storage at the stressed condition. Even though this leads to a reduction in the overall activity of the ADC, the variant carrying one cross-linked linker-drug is in theory still capable of contributing to cell killing through target binding and intracellular payload release following internalization of the unaffected linker-drug located on the other heavy chain of the ADC that has an average DAR of approximately 2. A relative quantitative estimation of the amount of intramolecular cross-linking in SYD1875 in formulation by MAM yielded a level of approximately 15% following 3 months of storage at 25°C, which was found to further increase to approximately 39% after 6 months of storage at 25°C (data not shown). Consequently, it was postulated that this modification leads to loss of its cell-killing ability of the linker-drug as substantiated by the unexpected results obtained in the cell-based cytotoxicity assay. Mechanistic confirmation of the premise To confirm the hypothesis of the intramolecular linker-drug-mediated cross-linking being the root cause of the decrease in potency observed in the stability sample that was stored at the stressed condition, hydroxylated versions of the payload and linker-drug were synthesized chemically. As synthesizing a cross-linked variant of SYD1875 is nearly impossible, the hydroxylated versions of the payload and linker-drug were considered the variants containing a modified DNA-alkylating domain, similar to the way the DNA-alkylating domain is altered in the intramolecular cross-link. Subsequently, ADCs were prepared using this modified linker-drug. The cellular cytotoxicity of these molecules was investigated in an in vitro assay using MDA-MB-468 cells. As can be seen from the results shown in Figure 6, modifying the part of the payload that was found to be affected by intramolecular cross-linking had a detrimental effect on its cell-killing ability compared to that of the intact payload, as evidenced by the shift to a higher concentration (4–5 orders of magnitude) of the curve obtained for the hydroxylated variant of the payload, as compared to the intact payload in Figure 6(A). A highly similar observation was made for the potency of the corresponding ADCs: where molecules containing the active linker-drug showed good dose–response curves, no cell killing was observed for ADCs carrying the modified versions of the payload, as evidenced by the horizontal curves in Figure 6(B). The results of this in vitro experiment confirmed the hypothesis that intramolecular cross-linking abolishes the potency of SYD1875. This, in turn, raised concerns about potential intramolecular cross-linking occurring at physiological conditions. Figure 6.In vitro potency (cell viability) evaluation of different variants of the payload and the corresponding ADCs. (A): in vitro potency of intact payload (orange trace) and the hydroxylated variant (purple trace) in MDA-MB-468 cells. (B): in vitro potency of ADCs produced with intact (orange trace) and hydroxylated (purple trace) payloads in MDA-MB-468 cells.Image A shows a dose-response graph with error bars. The x-axis is labeled Concentration (M) with tick marks from -12 to -3. The y-axis is labeled percent Survival, ranging from 0 to 120. Two curves are plotted: Curve 1 starts near 105 at -11, drops to about 65 at -10.6, 20 at -10.2 and stabilizes between 5 to 10 from -10 to -8. Curve 2 starts near 110 from -7.5 to -6.5, then drops to 95 at -6, 55 at -5.7, 30 at -5.3 and approaches 0 by -4.2. Image B also shows a dose-response graph with error bars. The x-axis is labeled Concentration (ng/mL) with tick marks at 0.1, 1, 10, 100, 1000. The y-axis is labeled percent Survival, ranging from 0 to 120. Two curves are plotted: One remains between 105 to 110 across all concentrations. The other starts near 100 to 105 around 1 to 10, decreases to 85 near 10, 60 near 30, 35 near 60, 25 near 100 to 200 and 20 near 500.Two line graphs showing percent survival versus concentration for payload variants and corresponding ADCs. In vitro potency (cell viability) evaluation of different variants of the payload and the corresponding ADCs. (A): in vitro potency of intact payload (orange trace) and the hydroxylated variant (purple trace) in MDA-MB-468 cells. (B): in vitro potency of ADCs produced with intact (orange trace) and hydroxylated (purple trace) payloads in MDA-MB-468 cells. Translation of the findings to the in vivo context: in vivo occurrence The exposure to this newly identified metabolite in vivo was investigated in plasma as per current guidelines.3,6 For this purpose, an affinity extraction of SYD1875 from the plasma of a cynomolgus monkey that had been exposed to the ADC for 21 d (504 h) as part of a preclinical toxicity study was performed, using an immobilized anti-human Fc antibody. The resulting SYD1875 material was analyzed using an LC-MS/MS method aimed at identifying conjugated species that yield the diagnostic fragment ion of the hydrolyzed linker-drug at m/z 509.21 upon MS/MS fragmentation. Such a species was observed in the affinity-extracted material at m/z 612.93 (3+) (see MS/MS spectra in Figure S4), which corresponds to the peptide conjugated with a hydroxylated linker-drug of which the maleimide moiety was hydrolyzed. Moreover, the dipeptide cross-linked through the linker-drug containing a hydrolyzed maleimide moiety was also observed (m/z 791.55 (5+), see MS/MS spectra in Figure S4). Since SYD1875 that was used in the preclinical toxicity study did not exhibit any intramolecular cross-linking at the time of administration, the data confirmed that linker-drug-mediated cross-linking also occurs in vivo, which can be expected as this involves exposure of SYD1875 to elevated temperature. Translation of the findings to the in vivo context: quantitation by LC-MS/MS To allow for quantitative assessments of the amount of cross-linking occurring in SYD1875 in vivo, a bioanalytical assay was developed using affinity extraction from plasma and targeted MS. Initially, the ‘bottom-up’ approach involving peptide‑mapping LC-MS/MS of the affinity-extracted material yielded several peptides, representing cross-linked and hydrolyzed variants. Optimizing the sample preparation conditions eventually led to the predominant formation of the variant containing the hydroxylated linker-drug and the hydrolyzed maleimide species (Figure S5). The LC-MS/MS method was aimed at relative quantitation of those species in plasma obtained from animals that had been administered SYD1875. Method performance was assessed using SYD1875 samples containing well-defined levels of intramolecular cross-linking (Figure S6 and Table S2) and was found to be fit for its intended purpose, not only based on the small differences observed between nominal and actual levels of cross-linked species but also on the analytical variation observed between replicate analyses of the same plasma samples spiked with cross-linked SYD1875. Furthermore, the assay demonstrated parallelism across matrices. Subsequently, plasma samples taken from cynomolgus monkeys during the preclinical toxicology study were analyzed using the bioanalytical LC-MS/MS method, such that the samples, calibration standards, and quality control samples were analyzed under comparable capture conditions, to ensure equivalent capture efficiency for cross-linked and non-cross-linked species. The amount of cross-linking in vivo on average was found to increase over time, to a maximum of approximately 10% at 504 h after the first dose and approximately 15% after 84 d (2016 h) of exposure to SYD1875 in cynomolgus monkeys (Figure S7). The amounts of cross-linking reported here are based on peptides originating from reduced species, while intact SYD1875 comprises a heterotetramer of protein chains of which two are conjugated with linker-drug. Statistically, the subset of SYD1875 carrying a cross-linked linker-drug is expected to still contain one active linker-drug that is capable to contribute to inducing cell death. Translation of the findings to the in vivo context: in vivo occurrence: retrospective assessment on impact on bioanalytical methods As the data obtained from the aforementioned preclinical toxicology study were also used to model the pharmacokinetics (PK) of SYD1875 in preparation of the first-in-human clinical study, these data were re-evaluated taking the newly formed metabolite into account. Since the initial PK modeling was founded on data obtained by an ELISA-based assay that was unable to distinguish the fully active and cross-linked variants of SYD1875, the amount of active, conjugated antibody could have been overestimated in the original assessment. Therefore, an additional, drug‑load-sensitive ELISA-based assay was developed that only recognized the non-cross-linked version of SYD1875, allowing for a more accurate estimation of conjugated antibody. As can be seen from an overlay of the data obtained for total antibody (TAb), conjugated antibody (CAb) using the non-specific assay and conjugated data of the newly developed specific ELISA presented in Figure 7, all three traces overlap for all dosing groups that were included in the preclinical toxicology study. Figure 7.Overlays of pharmacokinetic profiles of total (TAb), conjugated (CAb) and non-cross-linked-conjugated (NC-CAb) antibody in plasma from female cynomolgus monkeys after dosing SYD1875 at 3, 9 or 18 mg/kg (n = 4 or 5, +/− standard error of the mean).A graph shows concentration versus time for three series: TAb, CAb and NC-CAb, across doses of 3 mg/kg, 9 mg/kg and 18 mg/kg. TAb uses circular markers, CAb uses circular markers with a central mark and NC-CAb uses inverted triangle markers. The horizontal axis is labeled ′Time (h)′ ranging from 0 to 504, with ticks at 0, 168, 336 and 504. The vertical axis is labeled ′Concentration (ng/mL)′ on a log scale from 10 to 1,000,000, with ticks at 10, 100, 1,000, 10,000, 100,000 and 1,000,000. At 504 h, concentrations are: 18 mg/kg group at 30,000 for all series; 9 mg/kg group at 600 for TAb, 1,500 for CAb and 1,200 for NC-CAb; 3 mg/kg group at 15 for CAb and NC-CAb. All series decrease from initial values between 100,000 to 1,000,000 to lower concentrations by 504 h. At 504 h, the 18 mg/kg group maintains around 30,000, the 9 mg/kg group ranges from 600 to 1,500 and the 3 mg/kg group is around 15 for readable series.Graph of plasma antibody levels for TAb, CAb, NC-CAb at 3, 9, 18 mg/kg over time. Overlays of pharmacokinetic profiles of total (TAb), conjugated (CAb) and non-cross-linked-conjugated (NC-CAb) antibody in plasma from female cynomolgus monkeys after dosing SYD1875 at 3, 9 or 18 mg/kg (n = 4 or 5, +/− standard error of the mean). The PK data suggested that the impact of the intramolecular cross-linking occurring in SYD1875 in vivo is limited. This was confirmed by calculating the total exposure (in area under the curve, AUC) of cross-linked SYD1875 relative to the CAb exposure and NC-CAb in AUC. For the CAb assay, this was found to be maximally 2.7% relative after the first cycle, increasing toward 6.6% after the fourth (last) cycle (see Table S3). For the NC-CAb assay this was maximally 3.1% (cycle 1 data only). The levels of the cross-linked SYD1875 metabolite remained well below the 10% relative drug-related exposure that is set as a threshold in a guidance for industry on how to characterize and quantitatively evaluate metabolites.6 Taking the quantitative LC-MS/MS data and the data obtained by ELISA into consideration, the overall impact of intramolecular cross-linking on exposure of SYD1875 and on the PK evaluations based on the GLP Tox study was considered to be negligible. Translation of the findings to the in vivo context: species dependence Finally, the species dependence of SYD1875 intramolecular cross-linking was evaluated. Following incubation of SYD1875 in mouse, cynomolgus monkey, and human plasma at 40°C, the extent of intramolecular cross-linking was determined at various time points using affinity extraction and LC-MS/MS analysis. Neither a difference in the rate at which, nor in the extent to which, cross-linking occurred was seen between SYD1875 incubated in the different types of plasma (Figure S8), which confirmed intramolecular cross-linking to be a chemical reaction that is merely driven by temperature. Overall assessment of clinical relevance ELISA-based quantitation of CAb and TAb and subsequent PK modeling, as well as data obtained by a specifically developed LC-MS/MS-based quantitative bioassay, revealed that the impact of intramolecular cross-linking on PK and SYD1875 exposure was limited. Combined with the highly similar kinetics of cross-link formation in various species, including humans, intramolecular cross-linking was concluded not to meaningfully affect systemic exposure or the interpretation of PK data for this ADC. Consequently, the clinical trial for SYD1875 continued as planned. Discussion ADCs are subject to scrutiny due to their complex chemical and higher‑order structure. Therefore, product quality attributes of an ADC are monitored using a comprehensive set of analytical methods. Eventually, selected quality attributes that are known or expected to affect product quality, safety, immunogenicity, or efficacy should meet pre-defined acceptance criteria before the product is administered to patients. For ADCs, a substantial number of CQAs relate to the linker-drug. Despite prior knowledge on the attributes of an ADC, new, and critical, quality attributes can become apparent during the early stages of product development. Here, linker-drug-mediated intramolecular cross-linking was observed, causing a decrease in potency of SYD1875, as detected by the cell-based cytotoxicity assay. Using advanced analytical assays, the cross-link was found to consist of an ester formed between the DNA-alkylating moiety of the linker-drug and the carboxylic acid in the side chain of a specific glutamic acid residue in the light chain of SYD1875. Synthesis of the hydroxylated variant of the payload, linker-drug, and the associated ADC revealed that this modification, which is considered to be representative of intramolecular cross-linking, negatively impacted the potency of these therapeutic entities, confirming the causality of these observations. As per FDA guidance,6 the in vivo exposure of this newly identified metabolite in SYD1875 was investigated. Following confirmation of the presence of linker-drug-mediated cross-linking in SYD1875 affinity‑extracted after 3 weeks following administration to cynomolgus monkeys, a quantitative analysis of this modification in the plasma of monkeys that were dosed with the ADC in a preclinical toxicology study yielded maximum relative exposure levels (based on AUC) well below the threshold that would require more strict control, monitoring, and reporting. ELISA-based quantitation and subsequent PK modeling, as well as data obtained by a specifically developed LC-MS/MS-based quantitative bioassay, revealed that the impact of intramolecular cross-linking on PK and SYD1875 exposure is limited. Combined with the highly similar kinetics of cross-link formation in various species, including humans, intramolecular cross-linking was concluded not to meaningfully affect systemic exposure or the interpretation of PK data for this ADC. Importantly, the investigation described here reveals that the occurrence of a functionally impactful modification does not inherently translate into clinical relevance. While targeted methods are indispensable for routine control of known CQAs, they may inherently overlook novel degradation pathways. These, in turn, can unveil new CQAs and instigate the development of an analytical method to enable adequate monitoring. The relative quantitation of intramolecular cross-linking in SYD1875 by MAM was added to the panel of analytical assays used to monitor the stability of the ADC. Data obtained revealed no presence of linker-drug-mediated cross-linking exceeding the reporting threshold of 1% (data not shown) in SYD1875 when stored at −20°C for a prolonged period of time. Conclusions The case reported here describes a systematic and step-wise approach to investigating the root cause and impact of an unexpected observation in a CQA of an ADC: the potency analyzed by a cell-based cytotoxicity assay. The combination of assessing complementary analytical data, performing an unbiased in-depth characterization, mechanistic confirmation of causality, and a quantitative assessment helped to uncover and understand a previously unknown degradation route. Although intramolecular cross-linking negatively impacted the function of the ADC, its occurrence was found not to be clinically relevant. This underscores the importance of thorough, multi-disciplinary characterization of ADCs and their degradation pathways in order to establish a comprehensive analytical control strategy. As ADCs continue to evolve and diversify, such integrated strategies will become increasingly important to ensure sufficient product understanding and regulatory confidence. The risk of off-target reactivity within the ADC does not only hold for the example of duocarmycin in SYD1875 reported here: the underlying principle is generally applicable to linker-drugs bearing a reactive moiety. Under certain conditions, including favorable three-dimensional orientation, exposure to light, or elevated temperature, reactive groups may interact with amino acid side chains in the primary structure of the antibody. This could result in interactions or modifications that impact the safety and efficacy of the ADC. Strategies aimed at delaying payload activity to reduce systemic off-target toxicity by means of, for example prodrug designs or self-eliminating systems9 that require activation within the tumor microenvironment, are unable to completely prevent unintended linker-drug reactions from occurring before the ADC reaches its intended target. Data and information obtained in a variety of preclinical and analytical studies executed throughout early clinical development should result in a robust set of analytical methods that allow for the monitoring of the product’s CQAs and the determination of adequate acceptance criteria, to ensure the quality of the product and above all, patient safety. Supplementary Material DirksenEHC_Supplementary Information.docx