Modelling selective cell targeting through avidity to inform bispecific antibody discovery and development
Anis Krache, Joao NS Pereira, Kyra J. Cowan
Merck KGaA
Objectives:
Avidity is the phenomena referring to an increasing probability of a bivalent binder to bind a second target after binding the first one. Avidity is responsible for the very high apparent affinity of monoclonal antibodies. When applied to bispecific antibodies (BsAb), “selectivity through avidity” can be leveraged to achieve selective cell targeting and avoid potential toxicities induced by monospecific therapies [1]. However, finding the optimal combinations of monovalent affinities to achieve selective cell targeting is a multi-factorial problem. We have developed an ODE modelling framework to systematically screen system parameters (target densities, cell numbers) to inform development of novel combinations of bispecific antibodies. Here we developed an avidity model applied to a known dataset (CD4 x CD70, binding in vitro data from Mazor et al.) running on RStudio using RXODE package and model fitting using metaheuristic package.
Methods:
The avidity model was implemented in two parts: (A) Based on the Kareva et al. ODE system, a dynamic avidity factor was estimated depending on CD4 & CD70 effective target concentration, cell densities and binding affinities (B) An ODE system was developed to reproduce in vitro experimental setting where the BsAb interacts with CD4+/CD70+, CD4+/CD70- and CD4-/CD70+ cells. Model parameters included: Cell concentration, cell radius, drug radius and CD4 and CD70 KD. Model fitting was performed on CD4 & CD70 KD parameters using metaheuristic optimizer. To confirm the model performance and applicability, we used the same model to simulate different published affinity variants of the same bispecific molecule as a test dataset.
Results:
The current model describes the different entities such as target concentration, drug concentration, monovalent antibody (BsAb bound to one target) and bivalent antibody (BsAb bound to two targets on the same cell). A relationship between measured mean fluorescence intensity (MFI, experimental data) and target occupancy (model output) was derived from the model to allow fitting. Affinity constants were fitted for each binding arm (0.251 nM and 1.42 nM for CD4 and CD70) and the model was able to reproduce the in vitro cellular binding to CD4+/CD70+, CD4+/CD70- and CD4-/CD70+ cells. The avidity model was then used to model the “test dataset” consisting of different BsAb variants with decreased CD4 KD arm (11, 51, 110 nM) and the model reproduced the general trend of the data. We were able to show the applicability of this avidity model to explain selective cell targeting with a known dataset and the model can be applied to novel target combinations within the same experimental setup.
Conclusions:
Our aim was to explore the applicability of an ODE based avidity model on a published dataset. Overall, the avidity model was able to fit the experimental data and correctly described the data for different variants that were used as test dataset. Beyond the parameters that were expected to impact the simulations such as affinity, target density, cell numbers, we have also found the model to be sensitive on parameters such as drug radius or to the in silico experimental incubation time. The current model can be a suitable platform to guide BsAb drug discovery and development starting with in vitro experimental conditions and work as platform for in vivo extrapolations in combination with full-body QSP models.
References:
[1] Jiabing et al., Bispecific Antibodies: From Research to Clinical Application, 2021.
[2] Kareva et al., practical guide to calculating impact of avidity in models of bispecific antibodies with two membrane-bound targets, 2022.
[3] Mazor et al., Insights into the molecular basis of a bispecific antibody's target selectivity, 2015.