A quantitative modular modeling approach reveals the consequences of different A20 feedback implementations for the NF-kB signaling dynamics
Signaling pathways involve complex molecular interactions and are controlled by non-linear regulatory mechanisms. If details of regulatory mechanisms are not fully elucidated, they can be implemented by different, equally reasonable mathematical representations in computational models. The study presented here focusses on NF-{kappa}B signaling, which is regulated by negative feedbacks via I{kappa}B and A20. A20 inhibits NF-{kappa}B activation indirectly through interference with proteins that transduce the signal from the TNF receptor complex to activate the I{kappa}B kinase (IKK) complex. We focus on the question how different implementations of the A20 feedback impact the dynamics of NF-{kappa}B. To this end, we develop a modular modeling approach that allows combining previously published A20 modules with a common pathway core module. The resulting models are based on a comprehensive experimental data set and therefore show quantitatively comparable NF-{kappa}B dynamics. Based on defined measures for the initial and long-term behavior we analyze the effects of a wide range of changes in the A20 feedback strength, the I{kappa}B feedback strength and the TNF stimulation strength on NF-{kappa}B dynamics. This shows similarities between the models but also model-specific differences. In particular, the A20 feedback strength and the TNF stimulation strength affect initial and long-term NF-{kappa}B concentrations differently in the analyzed models. We validated our model predictions experimentally by varying TNF concentrations applied to HeLa cells. These time course data indicate that only one of the A20 feedback models appropriately describes the impact of A20 on the NF-{kappa}B dynamics.\n\nAuthor summaryModels are abstractions of reality and simplify a complex biological process to its essential components and regulations while preserving its particular spatial-temporal characteristics. Modelling of biological processes is based on assumptions, in part to implement the necessary simplifications but also to cope with missing knowledge and experimental information. In consequence, biological processes have been implemented by different, equally reasonable mathematical representations in computational models. We here focus on the NF-{kappa}B signaling pathway and develop a modular modeling approach to investigate how different implementations of a negative feedback regulation impact the dynamical behavior of a computational model. Our analysis shows similarities of the models with different implementations but also reveals implementation-specific differences. The identified differences are used to design and perform informative experiments that elucidate unknown details of the regulatory feedback mechanism.