Background DuoBody-PD-L1×4-1BB (GEN1046) is a class-defining bispecific antibody, designed to elicit an anti-tumor immune response by simultaneous and complementary blockade of PD-L1 on tumor cells and conditional stimulation of 4-1BB on T-cells and NK cells. Optimizing target engagement for a bispecific antibody is challenging, as it involves binding with two targets, and predicting trimer levels in tumors based on affinity of individual arms and target expression. Here we describe a semimechanistic, physiologically based pharmacokinetic/pharmacodynamic (PK/PD) model that predicts a dosing regimen for DuoBody-PD-L1×4-1BB, which results in the formation of maximum levels of a therapeutically active 4-1BB-bispecific antibody-PD-L1 trimolecular complex (trimer), and optimal PD-L1 receptor occupancy (RO).

Methods An integrated semimechanistic PK/PD model that describes the distribution of DuoBody-PD-L1×4-1BB into central and peripheral compartments and partitioning into tumor/lymph nodes was developed. The model used PK/PD data and physiological parameters from the literature for parameterizations of PD-L1 and 4-1BB expression levels and T-cell trafficking. The model incorporates dynamic binding of DuoBody-PD-L1×4-1BB to its targets to predict trimer formation and RO for PD-L1 in tumors. Model parameters were calibrated to match in vitro PD studies, such as analyses of T-cell proliferation and cytokine release, as well as clinical PK data. Sensitivity to model assumptions were assessed by varying PK/PD parameters, and assessing their impact on trimer formation and PD-L1 RO. The model was subsequently used to explore in vivo trimer levels and PD-L1 RO in tumors at various dosing regimens.

Results The model was able to adequately describe the PK of DuoBody-PD-L1×4-1BB in the central compartment. Simulations showed a bell-shaped response for average trimer levels in tumors that peaked at 100 mg every 3 weeks (Q3W), with doses >100 mg resulting in reduced trimer formation. Average PD-L1 receptor occupancy at the 100 mg dose was predicted to be approximately 70% over 21 days and increased at higher doses. Based on these model predictions, and available safety, anti-tumor activity, and PD data from the ongoing GCT1046-01 trial (NCT03917381), 100 mg Q3W was chosen as the expansion dose for further evaluation in Part 2 of the study.

Conclusions This semimechanistic PK/PD model provides a novel approach for dose selection of bispecific antibodies such as DuoBody-PD-L1×4-1BB, by using preclinical and clinical PK/PD data to predict formation of optimal trimer levels and PD-L1 receptor occupancy.

Acknowledgements The authors thank Friederike Gieseke and Zuzana Jirakova at BioNTech SE; Kalyanasundaram Subramanian at Applied Biomath LLC for their valuable contributions.

Trial Registration Written informed consent, in accordance with principles that originated in the Declaration of Helsinki 2013, current ICH guidelines including ICH-GCP E6(R2), applicable regulatory requirements, and sponsor policy, was provided by the patients.