Three antibodies in atopic dermatitis trials don't bind OX40 or OX40L the same way. New AlphaFold3 models show rocatinlimab and amlitelimab block the OX40–OX40L handshake through physical steric occlusion, while telazorlimab breaks a single critical contact bond at the interface.

The work, published in the Journal of Investigative Dermatology, maps the exact epitopes of rocatinlimab (KHK4083/AMG-451), telazorlimab (GBR-830/ISB-830), and amlitelimab (KY1005) at atomic resolution for the first time. Atopic dermatitis affects up to 1 in 5 children and 1 in 20 adults worldwide — more than 200 million people, by recent counts. The chronic disease drives Th2 inflammation, skin barrier breakdown, and relentless pruritus.

The OX40–OX40L axis sits upstream of the type 2 inflammation that defines eczema. When OX40L on antigen-presenting cells meets OX40 on activated T cells, the interaction promotes Th2 expansion, prolongs T-cell survival, and blocks regulatory T-cell suppression — fueling the chronic, relapsing disease cycle that biologics aim to interrupt.

What does the modeling show?

Researchers generated cocomplex structures using AlphaFold3 and ran three independent 400-ns molecular dynamics replicates in NAMD for each antibody-protein pair. Binding-free energies of residues within 5 Å of each binding interface were then calculated using MM/PBSA molecular mechanics with per-residue decomposition, giving atomic-level insight into every contact.

Rocatinlimab's model converged on one principal epitope on OX40. Its antibody footprint physically covers the cognate OX40–OX40L interface through steric occlusion, blocking the receptor from engaging its ligand. Amlitelimab lands at a similar principal location on OX40L and blocks binding by the same mechanism.

Telazorlimab's predicted epitope mapped differently. It disrupts a critical OX40–OX40L contact bond rather than covering the entire interface surface, the authors report. The two OX40 antibodies and the one OX40L antibody thus hit overlapping but not identical sites, with telazorlimab reaching a distinct, secondary location.

The takeaway here is structural: each antibody knocks on the OX40–OX40L axis at its own molecular door.

What does this mean for the OX40/OX40L pipeline?

The structural differences may help explain varied efficacy and safety signals as the OX40/OX40L drug class matures and head-to-head comparisons begin to emerge in dermatology. Rocatinlimab sits in a phase 3 program run jointly by Kyowa Kirin and Amgen. Telazorlimab's phase 2b readout published in 2024. Amlitelimab has cleared phase 2 trials, with phase 3 results anticipated.

Distinct binding modes could matter when patients don't respond to one agent. They could also matter for safety, since the OX40–OX40L pathway figures in asthma, ulcerative colitis, and possibly psoriasis through shared Th2 biology. Two additional OX40 antibodies — IMG-007 and BAT6026 — sit in earlier trials, with rocatinlimab now tested in adolescents and in combination with topical calcineurin inhibitors.

The takeaway: epitope-level distinctions could shape sequencing choices once long-term efficacy and safety data land across the drug class.

What's next?

The full data package — antibody sequences, equilibration trajectories, MM/PBSA outputs, and visualization scripts — sits on GitHub at kanolden/OX40_OX40L_manuscript. That kind of public deposit is rare in antibody epitope work, where proprietary data often stays sealed.

"No human subjects participated in this study," the authors state. "No artificial intelligence or large language model tools were used to prepare this manuscript." Senior author CGB discloses past advisory or research ties to AbbVie, Apogee, Connect BioPharma, Eli Lilly, Incyte, LEO Pharma, Pfizer, Sanofi, and Regeneron. NIH/NIAMS R01 AR079428 funded the work, with first author KN supported by NIH T32-GM080202 through the Medical College of Wisconsin MSTP.

Researchers now plan to extend the framework to the next wave of OX40 antibodies in late-stage trials and to map OX40 agonists in oncology settings, where costimulation rather than blockade is the therapeutic goal.

The bottom line: three biologics, one pathway, three different molecular doors — and a new level of clarity for anyone tracking which antibody is which.