Assessing the Binding Properties of the Anti-PD-1 Antibody Landscape Using Label-Free Biosensors

Michael E. Brown, Daniel Bedinger, Asparouh Lilov, Palaniswami Rathanaswami, Maximiliano Vásquez, Stéphanie Durand, Ian Wallace-Moyer, Lihui Zhong, Juergen H. Nett, Irina Burnina, Isabelle Caffry, Heather Lynaugh, Melanie Sinclair, Tingwan Sun, John Bukowski, Yingda Xu, Yasmina Noubia Abdiche

PLoS ONE 15, e0229206. DOI: 10.1371/journal.pone.0229206 

Overview 

Programmed cell death protein 1 (PD-1) is one of the most important immune checkpoint proteins in cancer immunotherapy and a validated therapeutic target, with over thirty unique monoclonal antibodies (mAbs) in clinical development or approved across the US, EU, and China. Label-free biosensors employing surface plasmon resonance (SPR) detection are widely used to characterize antibody binding interactions in terms of kinetic rate constants, affinities, and epitope diversity, yet the impact of sensor chip surface chemistry on the accuracy of these measurements is not well established. This industry-wide collaboration set out to comprehensively characterize the binding properties of a large panel of anti-PD-1 mAbs using multiple SPR platforms and chip types to determine which surface chemistry most closely approximates solution-phase affinity measurements. 

Key hypotheses and objectives 

The choice of sensor chip surface chemistry has a significant impact on the accuracy of SPR-derived kinetic constants relative to solution-phase measurements, and high-throughput and lower-throughput SPR platforms will yield comparable results when the same chip type is used. 

The objectives of the study were to: 

• Compare binding kinetics and affinity measurements obtained on high-throughput and lower-throughput SPR platforms using the same panel of antibodies. 

• Determine which sensor chip type, flat or 3D-hydrogel, produces affinities that more closely match those determined by solution-based methods. 

• Characterize the epitope diversity and PD-L1 ligand competition profiles of the anti-PD-1 antibody panel using high-throughput SPR epitope binning. 

Approach and techniques 

Thirty-five anti-PD-1 mAbs were selected from the patent and WHO-INN literature to represent a diverse range of binding affinities and epitope specificities. They were all expressed as IgG4 isotypes analogs as this was the most common format. Binding kinetics using anti-human Fc capture assays were measured by SPR on two platforms: the high-throughput Carterra LSA, which enables parallel measurement of all 35 mAbs in a single 384-ligand array format, and the lower-throughput Biacore 8K. Both platforms were tested with multiple flat and 3D hydrogel chip types. Solution-phase affinities were determined independently using Meso Scale Discovery (MSD) and the Kinetic Exclusion Assay (KinExA) as reference methods. Epitope binning and PD-L1 ligand competition studies were performed using the Carterra LSA instrument in a high-throughput sandwich assay format. An N-terminal domain-deleted PD-1 fragment was also used to probe epitope differences related to the nivolumab binding site.  

Major findings  

• Broad affinity range across the anti-PD-1 landscape: Mean apparent affinities spanned a range >4,000-fold, from single-digit picomolar to 424 nM, challenging the dynamic range of the SPR methods and reflecting the substantial diversity of binding properties across the clinical anti-PD-1 antibody landscape. 

• Platform concordance on flat chips: When using flat chip formats, the Carterra LSA and Biacore 8K platforms yielded near-identical kinetic rate and affinity constants, with ka and kd values generally within twofold across platforms for a panel of 33 mAbs spanning a broad affinity range. 

• Flat chips more closely approximate solution-phase affinities: Across both platforms, 3D-hydrogel chip surfaces systematically produced slower apparent on-rates and therefore weaker apparent affinities compared with flat chips, with minimal effect on off-rates. Affinities from flat chips were on average within fivefold of solution-phase MSD values, whereas 3D-hydrogel affinities were on average within eightfold. This finding challenges the prevailing consensus that 3D-hydrogel chips better approximate solution-phase conditions for antibody-antigen interactions. 

• More than ten unique epitope bins identified: High-throughput epitope binning revealed more than ten distinct competitive binding profiles within the anti-PD-1 mAb panel. The majority of mAbs clustered into a single major bin (Bin A). A sub-bin (Bin A*) emerged for nivolumab and one additional mAb that uniquely required the N-terminus of PD-1 for binding, consistent with published structural data. Pembrolizumab and mAbs from several other bins were able to displace one another, indicating closely adjacent or minimally overlapping epitopes. Three mAbs were unable to block PD-L1 binding, while the majority of the remainder mAbs blocked PD-L1, with one mAb identified as binding a bridging epitope between PD-L1-blocking and non-blocking antibodies. 

• Widely cited affinity values for nivolumab and pembrolizumab are inaccurate: Using the flat chip SPR, the affinities of the nivolumab and pembrolizumab analogs were determined to be 7.2 nM and 3.9 nM, respectively. These values are inconsistent with sub-nanomolar or picomolar values reported in much of the published literature. A unit error in an EMA document is likely responsible for the widely cited picomolar affinity of nivolumab. The provenance of the picomolar affinity frequently cited for pembrolizumab remains unanswered. 

Conclusions and impact 

This study provides the most comprehensive characterization of the binding kinetics, affinities, and epitope profiles of the anti-PD-1 antibody landscape to date. The finding that flat chip SPR surfaces more faithfully recapitulate solution-phase affinities than 3D-hydrogels establishes a new consideration for the choice of chip type in antibody binding studies. The kinetic, affinity, and epitope diversity revealed across this panel of mAbs demonstrates that there is no single binding profile that defines a clinically successful anti-PD-1 antibody. The unique binding properties of many of the antibodies studied may translate to similarly unique biological activity.  

Implications for therapeutic antibody development 

This study offers practical guidance on platform and chip selection for SPR-based characterization of antibody-antigen interactions. The demonstration that flat chip surfaces produce affinities more consistent with solution-phase measurements has direct implications for how SPR data should be collected and interpreted in antibody discovery and lead optimization programs. The correction of widely cited but inaccurate affinity values for nivolumab and pembrolizumab is also of practical relevance, as these molecules are frequently used as benchmarks and positive controls in both academic and industry settings.