The investigation used a micro-Multiphysics Agent-based model, or micro-MPA, developed to connect the mechanics of bone tissue with the cellular and molecular events that shape regeneration. The framework combines three domains: micro-finite element analysis for local mechanical strain energy density, micro-finite difference calculations for cytokine reaction, diffusion and decay, and discrete agents representing cells that migrate, divide, differentiate and die. Osteocytes act as the principal mechanical sensors, changing production of sclerostin, receptor activator of nuclear factor kappa-B ligand and osteoprotegerin according to their mechanical environment. Low mechanical strain promotes signals associated with osteoclast recruitment, whereas sufficiently strong stimulation can suppress sclerostin and permit osteoblast-mediated bone formation.

The model was based on longitudinal micro-computed tomography data from female PolgA mice with a mitochondrial mutation that produces premature ageing. A 0.5-millimetre osteotomy was stabilised with an external fixator. After the protocol-defined bridging point at approximately three weeks after surgery, animals received either habitual loading through the fixator or an individualised cyclic loading regimen at 10 hertz for 3,000 cycles three times per week. Weekly scans at a 10.5-micrometre isotropic resolution tracked mineralised tissue across the defect. The young cohort underwent surgery at 12 weeks and was evaluated at 19 weeks, while the aged cohort underwent surgery at 35 weeks and was evaluated at 42 weeks.

Mechanical boundary conditions were first calibrated in young animals through a hierarchical grid search designed to reproduce bone-volume fraction in both habitual and cyclic loading conditions. Habitual loading matched held-out young control animals with a 3.8% root mean square error. Calibrating cyclic loading against the young loading group produced a 4.5% error and reproduced the prominent increase in lateral bone formation, while remaining close to the experimental medial response. In the calibrated mechanical environment, the fixator did more than stabilise the fracture: it shared part of the applied force with the healing tissue and introduced bending that shielded the lateral cortex during habitual loading. Additional cyclic loading partly offset this asymmetry by generating high strain in the lateral callus.

Once these conditions were fixed, the researchers tested two biological scenarios in the aged animals. In the first, native cortical osteocytes and osteocytes newly embedded in the callus both received a degenerated phenotype. In the second, the native cortex retained its aged parameters, but the callus osteocytes received the youthful baseline. The degenerated state doubled both the assumed sclerostin production rate and the strain-energy-density threshold required for mechanosensitivity. These were parameter assumptions adapted from descriptions of age-related impairment, not values directly measured in the mice. Osteocyte density and embedding rate were also held constant across conditions, so the comparison isolated selected aspects of cellular function rather than differences in cell number.

The distinction between the two scenarios was pronounced under cyclic loading. With degenerated callus osteocytes, the simulation produced widespread resorption and little formation despite elevated local strain energy. At week 42, it predicted a bone-volume fraction of 49.0% plus or minus 6.2% in the defect centre, compared with 58.4% plus or minus 8.1% in vivo. The aged loading dataset therefore showed significant deviation between simulation and observation. The model also generated a transient bone-volume peak that was absent from the experimental data. By contrast, the rejuvenated-callus scenario predicted 56.2% plus or minus 8.3% in the defect centre, close to the measured value, and its longitudinal bone-volume curves did not differ detectably from the experimental curves.

Cell-level outputs offered a mechanistic explanation for this difference. In the degenerated scenario, osteocytes exposed to substantial mechanical stimulation continued to produce elevated levels of sclerostin, suppressing osteoblast recruitment even where strain exceeded the threshold for osteogenic differentiation. The model’s qualitative outputs suggest that sclerostin dysregulation was the more proximal driver, although the study did not separately test the effects of sclerostin production and the elevated mechanosensitivity threshold. In the rejuvenated scenario, cyclic loading lowered sclerostin within the lateral callus and created a permissive environment for osteoblasts. Bone formed only where strain was sufficient, reproducing the location and scale of the experimental anabolic response while leaving the adjacent aged cortex comparatively unresponsive.

The findings support an “osteochronology” interpretation of skeletal ageing: the functional age of newly generated tissue may differ from that of the organism that produced it. The aged callus in this model behaved like a locally young unit embedded within an aged skeleton, able to detect and translate mechanical signals even though the surrounding cortex retained a quiescent state. The authors describe this as in silico evidence that the regenerative potential of a fracture callus is modulated by the functional capacity of its embedded osteocytes rather than being determined solely by the systemic environment. However, the model cannot establish that the same tissue-age separation occurs naturally in all older animals or in human fracture healing.

Several limitations constrain the conclusion. PolgA mice exhibit accelerated ageing driven by mitochondrial dysfunction and may not reproduce the full range of skeletal changes associated with natural ageing. The model treats osteocyte function as a binary choice between rejuvenated and degenerated states, although real cell populations are likely to occupy a continuum. Revascularisation was omitted because all included defects had bridged before the simulated period, leaving reduced perfusion as an untested alternative explanation for reduced native-cortex responsiveness. The mechanical conditions calibrated in young animals were assumed to apply to aged animals, although differences in body weight, activity and load partitioning could alter the true stimulus. Statistical comparisons were also underpowered, with four simulated animals per aged group, and a non-significant difference should not be interpreted as proof of equivalence.

The computational setup adds further caveats. Tissue stiffness was derived from mineral density through a linear relationship, so regions with equal density received identical elastic properties despite possible differences in collagen organisation. Osteoclast activity and the remodelling surface were simplified, producing visual discrepancies from the experimental micro-CT structures even when total bone-volume fraction was predicted accurately. The model also represented a single calibrated mechanical boundary condition for each cohort, rather than fitting the effective axial-to-bending ratio of each animal. Within those constraints, the study provides a testable explanation for a striking observation: a smaller aged callus and higher local strains were not sufficient to produce the load-driven formation observed in vivo when osteocytes were modelled as degenerated, whereas assigning the callus a youthful functional phenotype substantially improved agreement with the aged biological data.

Subject of Research: Modelling how osteocyte functional age regulates mechanical remodelling in aged mouse femur defects

Article Title: Agent-based modelling of late-stage mechanical loading of femur defects in prematurely ageing PolgA mice

Article References: Kendall, J. J., Ledoux, C., Marques, F. C., Kuhn, G. A., Schulte, F. A., & Müller, R. (2026). Agent-based modelling of late-stage mechanical loading of femur defects in prematurely ageing PolgA mice. Frontiers in Bioengineering and Biotechnology, 14, Article 1879777. https://doi.org/10.3389/fbioe.2026.1879777

Image Credits: AI Generated

DOI: 10.3389/fbioe.2026.1879777

Keywords: bone regeneration, osteocytes, agent-based modelling, mechanical loading, fracture healing, callus remodelling, cellular ageing, micro-CT, sclerostin, PolgA mice, Agent-based, modelling