Trabecular bone remodeling in the ageing mouse: a micro-multiphysics agent-basedin silicomodel using single-cell mechanomics

Abstract

Bone remodeling is regulated by the interaction between different cells and tissues across many spatial and temporal scales. In silico models have been of help to further understand the signaling pathways that regulate the spatial cellular interplay. We have established a 3D multiscale micro-multiphysics agent-based (micro-MPA)in silicomodel of trabecular bone remodeling using longitudinalin vivodata from the sixth caudal vertebra (CV6) of PolgA(D257A/D257A)mice, a mouse model of premature aging. Our model includes a variety of cells as single agents and receptor-ligand kinetics, mechanotransduction, diffusion and decay of cytokines which regulate the cells' behavior. The micro-MPA model was applied for simulating trabecular bone remodeling in the CV6 of 5 mice over 4 weeks and we evaluated the static and dynamic morphometry of the trabecular bone microarchitecture. We identified a configuration of the model parameters to simulate a homeostatic trabecular bone remodeling. Additionally, our simulations showed different anabolic, anti-anabolic, catabolic and anticatabolic responses with an increase or decrease by one standard deviation in the levels of osteoprotegerin (OPG), receptor activator of nuclear factor kB ligand (RANKL), and sclerostin (Scl) produced by the osteocytes. From these results, we concluded that OPG inhibits osteoclastic bone resorption by reducing the osteoclast recruitment, RANKL promotes bone resorption by enhancing the osteoclast recruitment and Scl blocks bone formation by inhibiting osteoblast differentiation. The variations in trabecular bone volume fraction and thickness (BV/TV and Tb.Th) were relatively higher with variations of one standard deviation in the levels of RANKL compared to OPG and Scl. This micro-MPA model will help us to better understand how cells respond to the mechanical signal by changing their activity in response to their local mechanical environment.

Mechanostat parameters estimated from time-lapsedin vivomicro-computed tomography data of mechanically driven bone adaptation are logarithmically dependent on loading frequency

“…Eventually, these factors were vital to producing relatively smooth RmV curves and enabling consistent and plausible piecewise linear, and hyperbola fits. In any case, as previous work has focused on group average results both in vivo (Schulte et al 2013, Razi et al 2015, San Cheong et al 2020) and in silico (Levchuk et al 2014, San Cheong et al 2020, Boaretti et al 2023), our analysis still aligns with such standard practices. Second, contrasting with conditional probability-based approaches, (re)modeling events are no longer characterized separately since our approach yields a single curve representing the average RmV describing the net effect of (re)modeling events for a given mechanical signal.…”

“…Furthermore, the decreased anabolic response observed in the RmV curves for high strains may also be linked to a decrease in mechanosensitivity resulting from increased cell stiffness, as previously reported for such high strain values (Nawaz et al 2012). Therefore, the trends estimated with the mechanostat remodeling velocity curves could be leveraged by in silico simulations that also rely on time-lapsed in vivo micro-CT data as input, such as novel agent-based models that simulate individual cell populations in 3D (Tourolle 2019, Boaretti et al 2022) and with that, improve the accuracy of their predictions with respect to in vivo data. In this regard, our results demonstrating that several parameters estimated from the mechanostat also follow a logarithmic relationship with loading frequency can help to calibrate such models and investigate loading frequency-dependent responses in silico .…”

Mechanostat parameters estimated from time-lapsed in vivo micro-computed tomography data of mechanically driven bone adaptation are logarithmically dependent on loading frequency

“…Eventually, these factors were vital to producing relatively smooth RmV curves and enabling consistent and plausible piecewise linear, and hyperbola fits. In any case, as previous work has focused on group average results both in vivo ( Schulte et al, 2013 ; Razi et al, 2015 ; San Cheong et al, 2020a ) and in silico ( Levchuk et al, 2014 ; San Cheong et al, 2020a ; Boaretti et al, 2023 ), our analysis still aligns with such standard practices. Second, contrasting with conditional probability-based approaches, (re)modeling events are no longer characterized separately since our approach yields a single curve representing the average RmV describing the net effect of (re)modeling events for a given mechanical signal.…”

“…In agreement with these studies and with potential relevance for future in silico models, we chose to also focus our analysis on velocity rather than change in bone mass, spatially resolving the displacement of surface voxels between time-points. In fact, recent advances in the context of in silico single-cell mechanomics ( Boaretti et al, 2023 ) have associated positive and negative velocities with the activity of osteoblasts and osteoclasts, respectively. Still, converting the output to other units of interest, including a change in bone volume as originally proposed, is equally possible.…”

“…Furthermore, the decreased anabolic response observed in the RmV curves for high strains may also be linked to a decrease in mechanosensitivity resulting from increased cell stiffness, as previously reported for such high strain values ( Nawaz et al, 2012 ). Therefore, the trends estimated with the mechanostat (re)modeling velocity curves could be leveraged by in silico simulations that also rely on time-lapsed in vivo micro-CT data as input, such as novel agent-based models that simulate individual cell populations in 3D ( Tourolle, 2019 ; Boaretti et al, 2023 ) and, with that, improve the accuracy of their predictions with respect to in vivo data. For this reason, we chose to express RmV curves using effective strain, as this signal has already been implemented successfully in such in silico models.…”