Drug treatments against osteoporosis are commonly divided into anti-catabolic and anabolic. The former act to reduce bone turnover and achieve the increase in bone mass mainly from mineralisation of the existing bone matrix. The latter increase bone mass by enhancing osteoblastic activity resulting in new bone formation. The duration of treatments is often limited to a few years due to reported side effects, but discontinuation of treatment might pose significant risk for fracture in some drugs such as denosumab. Switching to a different drug is the most commonly adopted strategy; however, it is not clear what is the best combination of a dual-drug therapy, the lapse between treatments and other parameters defining the combination that needs to be studied.

Clinical trials are long and costly and may have ethical implications that can be avoided with in-silico trials. In this work we have simulated two drug treatments: denosumab (anti-catabolic) and romosozumab (with a dual effect anabolic and anti-catabolic). We evaluated BMD gain and fracture risk by incorporating a damage model into the bone remodelling algorithm.

Our results showed that greater BMD gain is achieved by starting with romosozumab. The reason for this is that anti-catabolic treatment decreases bone turnover rate and the population of osteoblast precursors. The main action of romosozumab is to increase the proliferation of these precursors, so that their population should be as high as possible for a better efficacy of the drug. Therefore, prior administration of an anti-catabolic drug may be counterproductive to the effectiveness of romosozumab.

An optimum treatment should minimise fracture risk at the lowest possible dose to prevent adverse effects of drugs. For this reason, we propose here an optimisation procedure to obtain an optimum sequential treatment (romosozumab + denosumab) using a patient-specific model of bone remodelling including a pharmacokinetics-pharmacodynamics model of both drugs.

While mechanobiology has been mainly studied in terms of vascular porosity changes and trabecular orientations, we here present a combined experimental-computational approach to explore the interplay of mechanical forces and biological structures at the scale below, i.e. at the textural orientations of the extracellular bone matrix (ECM).

As a model system, we consider developing murine femoral bone adjacent to a resorbing cylindrical magnesium implant of 1.6 mm diameter, one month after surgery. Hereby the implant is approximately located in the middle of the femur and orthogonal to the axis of the femoral shaft.

The system is documented in terms of micro Computed Tomography (μ-CT) scans, Small-Angle X-ray Scattering Tensor Tomography (SAXS-TT) and a back-scattered electron micrograph (BSE). These datasets are mutually registered, yielding a real-scale digital-twin data space which spans over multiple length scales. Hereby a general review on the creation of such a digital twin is given with methods for the identification of landmark features like the implant axis, the femoral bone axis, and cross-sectional parameters of an interpenetrating cylindrical model. In the latter, the implant is represented by the geometrical object “solid circular cylinder” and the femoral cortical bone by a “cylindrical annulus” fitted to associated voxels in the µ-CT scans.

The musculoskeletal loading acting onto the femur during physiological locomotion, is reconstructed from a pertinent inverse-dynamic model [Wehner 2010], and mapped to this geometrical model with the means of an extension to classical Bernoulli-Euler beam theory. This extension takes into account that the cross-sections vary along the femur shaft axis and thereby activate shear flows usually not considered in beam theory.

Principal and maximum shear stress directions obtained by the aforementioned modeling approach are compared to the SAXS-TT textural orientations of the ECM. This enlarges our understanding of mechanical stimulation in bone mechanobiology.

[Wehner 2010] DOI: 10.1016/j.jbiomech.2010.05.028

Study of preclinical animal models of osteoporosis (OP) are typically conducted using micro computed tomography (microCT) at selected bones sites and have provided invaluable insights into bone physiology and pathology. Among these, the mouse tibia loading model has contributed to the current understanding of the link between bone structure and function.

One challenge of using these models is the generation and processing of large amounts of data including image- segmentation, registration and analysis together with performing in-silico mechanical assessment using finite element analysis (FEA) to identify highly loaded bone regions. The final step in this analysis pipeline is performing statistical analyses to determine which OP intervention strategies are most effective. Manual execution of these tasks is time consuming and costly, but also makes reproducibility of results difficult due to user-specific parameters.

In this contribution, we develop a deep-learning based approach for automating the above tasks [1]. We demonstrate the performance of this approach based on the data of Sugiyama et al. 2008 [2], i.e. mice treated with different doses of parathyroid hormone (PTH) and mechanical loading (ML). We analyse tibia response both in trabecular and cortical regions both for whole bone quantities (bone area, marrow area, BV/TV) and local quantities (cortical thickness) together with automated in-silico mechanical analysis indicating bone regions experiencing highest strains.

References

[1] A. Lagzouli et al., DBAHNET: Dual-branch attention-based hybrid network for high-resolution 3D micro-CT bone scan segmentation, IEEE 21^st International Symposium of Biomedical Imaging (ISBI), 2024, pp1-5.

[2] T. Sugiyama et al., Mechanical loading enhances the anabolic effects of intermittent parathyroid hormone (1–34) on trabecular and cortical bone in mice, Bone,43, 2008, pp238-248.

Osteoarthritis (OA) is an age-related degenerative joint disease with known mechanical aetiology^1,2. The distinction between the mechanical loads essential in preserving healthy joint ageing, versus those that drive OA onset and progression remains, however, ill-defined. Here, we perform hierarchical characterisation of full-field biomechanical strains across whole tibial epiphyses in a murine OA model and normal intact joints, under physiological load.

Whole hindlimbs from OA-prone STR/Ort and healthy parental CBA mice (N=3) at ages before OA onset (10-weeks) and advanced OA (40-weeks) were subjected to in-situ physiological loading. A series of high-resolution synchrotron computed tomography images (1.45 µm/voxel) of each joint were acquired after stepwise displacement-controlled loading (European Synchrotron Radiation Facility, Grenoble, BM05-beamline). Following tomographic reconstruction, 3D strain magnitudes and patterns were quantified by digital volume correlation (DVC) across all anatomical zones (Avizo 3D, XDVC).

3D displacements and strain fields were quantified across entire tibial epiphyses with nanoscale displacement accuracy of <90nm (0.06voxels) and strain precision of <350µstrain (0.035%). We found load-induced compressive and tensile strains were effectively transferred between articular-to-metaphyseal regions in all joints. In STR/Ort joints, however, additional regionalised foci of high magnitude compression and tension were observed across articular zones, which were completely absent in all CBA joints. Global average strain quantification revealed higher tension (P=0.0001) and compression (P=0.0089) in STR/Ort versus CBA at 10-weeks. At 40-weeks, tensile but not compressive strains were higher in STR/Ort (P=0.0107).

Our data indicate that altered microarchitecture in OA-prone joints is linked to detrimental strain concentration at articular regions. In healthy joints however, load-induced strains are efficiently transferred to remote epiphyseal regions. We speculate that strain patterns arising at pre-OA stages, predetermine the pathological architectural changes that arise with advancing OA. Quantification of localised, region-specific strains and correlation with shifting joint microarchitecture will aid in deciphering these pathological mechanisms.

Bone remodeling, controlled by the interaction of osteoblasts, osteoclasts, and osteocytes within basic multicellular units (BMUs), is a dynamic process crucial for mineral homeostasis and bone maintenance. Cell communication within BMUs is regulated by the RANK-RANKL-OPG signalling pathway controlling activation and survival of osteoclasts and, due to coupling, bone formation and resorption (1). Additionally, factors such as TGF-β and parathyroid hormone (PTH), have been identified to affect bone remodelling (2).

Many mathematical models have been developed either at the whole systems scale (3) or at the bone tissue scale (4) to better understand the bone remodeling process. A major assumption in these models is that hormonal concentrations (or ligand concentrations in general) are constant and bone cell responses can be described by a simple 1-state receptor ligand model which delivers Hill type control functions. Here, we develop a multiscale bone remodeling model, where dynamic PTH glandular secretion is accounted for together with formulation of a 2-state receptor model (5, 6). The formulation of a cellular responsiveness function from the 2-state receptor model is linked to the bone cell activities in bone remodeling. Using this formulation, we investigate cases of bone diseases related to the parathyroid gland including hypoparathyroidism and hyperparathyroidism.

References.

1. B. F. Boyce, L. Xing, Archives of Biochemistry and Biophysics 473, 139–146 (2008).
2. S. Khosla, Endocrinology 142, 5050–5055 (2001).
3. V. Lemaire, F. L. Tobin, L. D. Greller, C. R. Cho, L. J. Suva, Journal of Theoretical Biology 229, 293–309 (2004).
4. P. Pivonka et al., Bone 43, 249–263 (2008).
5. Y. Li, A. Goldbeter, Biophysical Journal 55, 125–145 (1989).
6. D. Martonova et al., PLOS ONE 18, 1–21 (2023).