Development of a mathematical model to elucidate the cross-linking of GSK-057 and ADAs and binding to TNFR1 receptor – filling the gap between hypothesis and reality.
Tamara van Steeg [1], Anne Keunecke [1] and David Fairman [2]
[1] LAP&P, Leiden, the Netherlands, [2] GSK Medicines Research Centre, Stevenage, Hertfordshire, SG1 2NY, UK1LAP&P, Leiden, the Netherlands. 2GSK Medicines Research Centre, Stevenage, Hertfordshire, SG1 2NY, UK
Objectives: The variable heavy (VH) chain domain antibody (dAb) GSK1995057 (GSK-057) was developed to selectively block TNF-α receptor 1 (TNFR1), thereby antagonising TNF-α signalling. Recently, pre-existing auto-antibodies (ADAs) have been discovered in approximately 50% of the GSK-057-naïve healthy human subjects which together with the dAb were found to cross-link and activate the TNFR1 membrane bound receptor (mTNFR1), leading to symptoms of cytokine release [1]. The aim of the current investigation was to explain the apparent reversible agonistic effects of dAbs in subjects expressing pre-existing ADAs by development of a mathematical framework for the interaction between dAb, ADA and both soluble (sTNFR1) and membrane-bound TNFR1 (mTNFR1).
Methods: A phase I trial investigating the safety and tolerability of GSK-057 in humans provided data on PK, free and total sTNFR1 levels after administration of a wide dose range of GSK-057 (single dose administration, 0.0004 up to 2.0 mg/kg). Moreover, these GSK-057 treated subjects were monitored for the existence of ADAs. Analyses for the Phase I data were performed in NONMEM (Version 7.3, method FOCE with interaction). Model extension based on theoretical considerations, simulations and all graphical explorations were performed using R (version 3.3.2) and Rstudio (version 1.0.44). Relevant model parameters were estimated or obtained from theoretical concepts, literature or in-vitro experiments.
Results: A full TMDD model was used to capture the observed profiles in time for unbound dAb, free and total sTNFR1 following administration of GSK-057. The clearance and inter-compartmental clearance for GSK-057 were estimated 0.0502 and 0.00306 L/h/kg and the estimated volumes of distribution were in the normal range (e.g. Vc = 0.0692 L/kg). The interaction with the target was captured using the in-vitro affinity (KD = 0.0143 nM) and an estimated kon of 0.778 nM-1.h-1. Finally, the baseline sTNFR concentration, the degradation rate and the internalization rate were estimated 0.0541 nM, 0.709 and 0.108 h-1, respectively. Although, the biomarker profiles were captured adequately, this analysis revealed that the interaction with the membrane-bound receptor could not be derived from the data. For that reason, the model was extended to capture mTNFR1 binding, GSK-057-ADA complex formation and subsequent mTNFr1 crosslinking, based on theoretical concepts. The final model included (i) the synthesis and degradation of mTNFR1 (ii) the production of sTNFR as a result of shedding of mTNFR1, (iii) the internalization and shedding of the dAb-mTNFR1 complex, (iv) dAb-ADA binding and (v) binding of the dAb-ADA complex to mTNFR1 and, thus, reflected all relevant binding processes involved in the activation of the mTNFR1 by dAbs in individuals with pre-existing ADAs. An iterative simulation process provided insights into the existing knowledge gaps. Simulation results were discussed with the clinical/preclinical project teams regularly, thereby challenging the model against the available knowledge on the system. Finally, a sensitivity analysis further strengthened the parameter space.
Overall, the simulations revealed a bell-shaped binding curve for the dAb-ADA-mTRFR1 complex, thereby demonstrating the reversible nature of the agonistic effects when increasing the dose of the dAb. Furthermore, the simulations provided clear insights into the parameters, which governed the reversible nature of the agonist effects. For example, the pre-existing ADA were shown to be approximately 10-fold lower in affinity to GSK-057 than it is to TNFr1 and the maximum ADA baseline estimated to be approximately 0.1 nM.
Conclusions: Simulations with the developed mathematical model showed that a reversible agonist effect is indeed to be expected based the underlying biological binding principles and the parameters, thereby filling the gap between hypothesis and reality.
References:
[1] Holland MC, Wurthner JU, Morley PJ, Birchler MA, Lambert J, Albayaty M, et al. Autoantibodies to variable heavy (VH) chain Ig sequences in humans impact the safety and clinical pharmacology of a VH domain antibody antagonist of TNF-α receptor 1. Journal of Clinical Immunology. 2013; 33(7):1192–1203.
[2] Wilson MR, Wakabayashi K, Bertok S, Oakley CM, Patel BV, O’Dea KP, et al. Inhibition of TNF receptor p55 by a domain antibody attenuates the initial phase of acid-induced lung injury in mice. Frontiers in Immunology. 2017; 8: 1–12.This was a GSK sponsored study. Study number 110951 (NCT01476046)
