Evidence-based network modelling to simulate nucleus pulposus multicellular activity in different nutritional and pro-inflammatory environments

Baumgartner, Laura; Sadowska, Aleksandra; González Ballester, Miguel Ángel, 1973-; Wuertz-Kozak, Karin; Noailly, Jérôme

doi:https://doi.org/10.3389/fbioe.2021.734258

Evidence-based network modelling to simulate nucleus pulposus multicellular activity in different nutritional and pro-inflammatory environments

License

Citation

Baumgartner L, Sadowska A, Tío L, González Ballester MA, Wuertz-Kozak K, Noailly J. Evidence-based network modelling to simulate nucleus pulposus multicellular activity in different nutritional and pro-inflammatory environments. Front Bioeng Biotechnol. 2021;9:734258. DOI: 10.3389/fbioe.2021.734258

Abstract

Initiation of intervertebral disc degeneration is thought to be biologically driven. This reflects a process, where biochemical and mechanical stimuli affect cell activity (CA) that compromise the tissue strength over time. Experimental research enhanced our understanding about the effect of such stimuli on different CA, such as protein synthesis or mRNA expression. However, it is still unclear how cells respond to their native environment that consists of a “cocktail” of different stimuli that might locally vary. This work presents an interdisciplinary approach of experimental and in silico research to approximate Nucleus Pulposus CA within multifactorial biochemical environments. Thereby, the biochemical key stimuli glucose, pH, and the proinflammatory cytokines TNF-α and IL1β were considered that were experimentally shown to critically affect CA. To this end, a Nucleus Pulposus multicellular system was modelled. It integrated experimental findings from in vitro studies of human or bovine Nucleus Pulposus cells, to relate the individual effects of targeted stimuli to alterations in CA. Unknown stimulus-CA relationships were obtained through own experimental 3D cultures of bovine Nucleus Pulposus cells in alginate beads. Translation of experimental findings into suitable parameters for network modelling approaches was achieved thanks to a new numerical approach to estimate the individual sensitivity of a CA to each stimulus type. Hence, the effect of each stimulus type on a specific CA was assessed and integrated to approximate a multifactorial stimulus environment. Tackled CA were the mRNA expressions of Aggrecan, Collagen types I & II, MMP3, and ADAMTS4. CA was assessed for four different proinflammatory cell states; non-inflamed and inflamed for IL1β, TNF-α or both IL1β&TNF-α. Inflamed cell clusters were eventually predicted in a multicellular 3D agent-based model. Experimental results showed that glucose had no significant impact on proinflammatory cytokine or ADAMTS4 mRNA expression, whereas TNF-α caused a significant catabolic shift in most explored CA. In silico results showed that the presented methodology to estimate the sensitivity of a CA to a stimulus type importantly improved qualitative model predictions. However, more stimuli and/or further experimental knowledge need to be integrated, especially regarding predictions about the possible progression of inflammatory environments under adverse nutritional conditions. Tackling the multicellular level is a new and promising approach to estimate manifold responses of intervertebral disc cells. Such a top-down high-level network modelling approach allows to obtain information about relevant stimulus environments for a specific CA and could be shown to be suitable to tackle complex biological systems, including different proinflammatory cell states. The development of this methodology required a close interaction with experimental research. Thereby, specific experimental needs were derived from systematic in silico approaches and obtained results were directly used to enhance model predictions, which reflects a novelty in this research field. Eventually, the presented methodology provides modelling solutions suitable for multiscale approaches to contribute to a better understanding about dynamics over multiple spatial scales. Future work should focus on an amplification of the stimulus environment by integrating more key relevant stimuli, such as mechanical loading parameters, in order to better approximate native physiological environments.