Introduction
When a bone breaks, the body usually repairs the gap between the broken ends. But for around 5-10% of patients, this never happens (Wildemann et al., 2021). This failure, known as non-union, leads to chronic pain, repeated surgery, and lasting disability. However, not all non-unions are equal: some form excessive or disorganised bone (hypertrophic), while others form too little (atrophic). Why fractures fail, and why they fail in different ways, remains poorly understood.
To answer these questions and reliably study this clinical problem, researchers rely on animal models to reproduce the healing process under controlled conditions. We hypothesised that controlling movement at the fracture site could determine whether healing results in successful repair or different forms of non-union. By defining these mechanical conditions, we developed a mouse model in which healing outcomes are predictable and can be investigated at the molecular level. Combined with spatial gene mapping, our work provides a platform to investigate not only whether a fracture fails to heal, but the biological mechanisms explaining why.
Methods
Female C57BL/6JRj mice (n=16, 17–25 weeks) received femoral defects stabilized by custom external fixators (ExFixs); miniature versions of the metal frames used to stabilize broken bones in patients. To test how defect size and mechanical stability affect healing, we created small (<0.9 mm), middle (0.9–1.2 mm), and large (>1.2 mm) defects and stabilized with either flexible (~5 N/mm) or stiffer (~15 N/mm) fixators. Healing was monitored in the living animal over ten weeks by repeated micro-computed tomography (micro-CT), a high-resolution imaging technique comparable to clinical CT scanning. A fracture was considered healed when newly formed mineralised tissue bridged the gap between the two bone ends. In parallel, we established RNA in situ hybridisation, a technique that shows where specific genes are active within tissues. Using bones from a recent non-union study (Schröder et al., 2025), we mapped two genes involved in bone healing: Dmp1, associated with bone-forming cells, and Acp5, associated with bone-resorbing cells.
Results
Fixation stiffness significantly influenced fracture healing: the flexible fixator produced a larger mineralised callus, the temporary tissue that forms around a healing fracture, than the stiffer one. Healing outcomes also depended on defect size and fixator stiffness. Small defects consistently healed and large defects consistently failed to heal. In contrast, defects within the middle range (0.9–1.2 mm) showed both healing and non-union outcomes, making them particularly sensitive to mechanical conditions. The sensitivity of this middle defect range therefore represents a valuable model for studying why some fractures heal while others fail. The dual gene-mapping method was successfully established and revealed distinct patterns of bone remodelling activity in healed and non-healed bones.
Discussion
By adjusting fixator stiffness within this sensitive intermediate defect range, the model allows different forms of non-union to be generated under controlled conditions. Combined with spatial gene mapping, it provides a platform for linking early molecular changes to the biological processes that ultimately determine whether a fracture bridges or fail, and why some never do.
Significance
Every year, millions of patients live with fractures that simply won’t heal. By improving our understanding of why fractures fail, this model provides a foundation for developing treatments targeted at the right biological cause overall reducing the need for repeated surgeries and improving patient quality of life.
References
Wildemann B, et al. (2021). Non-union bone fractures. Nature Reviews Disease Primers, 7(1), 1–21.
Wehrle E, et al. (2021). Scientific Reports, 11, 23037.
Chabot C, et al. (2025). 30th Congress of the European Society of Biomechanics, Zürich.
Schröder M, et al. (2025). Orthopaedic Research Society Annual Meeting, Phoenix.
Acknowledgements
Funded by AO Trauma.