Targeting Transforming Growth Factor-beta to Enhance the Fracture Resistance of Bone

BODY: Introduction: Finite element analysis (FEA) can assess the strength of bone because quantitative computed tomography (qCT) scans can be converted to finite element (FE) models. The usefulness of qCT-FEA is evident in the growing number of clinical studies reporting that strength predictions derived from CT scans of hip, spine, or radius can differentiate fracture patients from non-fracture patients, though some overlap exists across the cohorts [1-2]. Much of the validation behind the failure criteria in these FE model predictions came from correlations with strength measurements as determined by whole bone testing of cadaveric tissue for which empirical relationships exist to convert mineral density to material properties. Given the relatively small size of mouse bones, especially vertebral bodies (VBs), no such relationships exist even though μCT-FEA could also be useful in pre-clinical and genetic studies involving mice. Since there is a dearth of evidence establishing the ability of μCT-FEA to assess mouse VB strength, we performed a parametric analysis to determine whether an elastic μCT-FEA with simple failure criteria could reasonably predict VB strength as Abstract Central Abstract Proof PopUp http://ors2013.abstractcentral.com/abstract?TAG_ACTION=V... 1 of 4 1/9/13 9:46 AM determined by compression tests. Methods: Study design: We assessed the strength of VBs from 2 studies in which differences in BV/TV were expected from prior work: a TGFβ-neutralizing Ab (1D11) for 4 weeks increased trabecular bone volume fraction (BV/TV) [3] and mice lacking activating transcription factor 4 (ATF4) have a low BV/TV compared to wild-type mice [4]. Therefore, we treated 13-wk, male mice (FVB strain) with a control antibody (13C4, n=7) or 1D11 (n=8) for 4 weeks at the same dose (10 mg/kg 3x per wk). We aged Atf4+/+ (n=10) and Atf4-/(n=10) mice (male and female on a FVB background) to 17 weeks. After euthanasia, L6 VBs were harvested and stored frozen in PBS. μCT-FEA: All VBs were scanned at 12 μm voxel size using the same settings (μCT40, Scanco Medical) and hydroxyapatite (HA) phantom calibration. To calculate BV/TV, contours were manually drawn for the region between the end plates (Fig. 1), whereas to generate the FE model, a circle with a constant radius was copied into each image and positioned to transect the transverse processes, which do not bear load in the compression test (Fig. 2). For the VBs of the 1D11 study, a threshold of 421.4 mgHA/cm3 segmented bone from soft tissue and air. Then, using Scanco FEA elastic solver software (brick elements), the equivalent strains were determined for high-friction, axial compression loading of each VB at 1% apparent strain, elastic modulus (E) of 10 GPa, and υ of 0.03 for all elements. The reaction force was determined for initial failure criteria in which 2% of the model volume exceeds 0.7% equivalent strain (Base Model), which were the criteria in a validation study for human bone [5]. Parametric study: Keeping the failure criteria of the Base Model constant, the elastic modulus was increased to 18 GPa and 22 GPa. Next, E was held constant while varying the percentage of volume that must exceed 0.7% strain before failure. Instead of having one value for E across all VBs, modulus was also based on the tissue mineral density (TMD) of each VB using a conversion derived by Wagner et al. [7]. As one final change, we investigated the effect of increasing the threshold to 538.9 mgHA/cm3 (Fig. 2). To verify applicability across studies, the strength of VBs from the ATF4 study were determined for a threshold of 450.8 mgHA/cm3, 18 GPa, 0.7% failure strain, and a critical volume of 2%. Compression tests: Each hydrated VB was subjected to axial compression at 3 mm/min in which the supporting platen had a rough surface and a moment relief to minimize slippage and off-axis loading. Two VBs were removed due to translation. Statistical analysis: The ability of μCT-FEA to predict mouse VB strength was ascertained by linear regression and the root mean squared error (RMSE). Results: The Base Model FEA was not effective in predicting mouse VB strength (Fig. 3a) as the RMSE was rather high and the 95% confidence interval (CI) of the slope was above 1. However, increasing E of bone tissue reduced the error and led to a CI of 0.67 to 2.09, while not affecting the R2 (Fig. 3b,c). Using an E based on mean TMD for each VB increased the ability of FEA to predict experimental strength (Fig. 3d). Varying the failure volume from 1% to 4% had modest effects on R2 and RMSE: 1% (66.9% and 5.8), 2% (62.3% and 5.0), 3% (62.0% and 5.5), and 4% (61.9% and 6.3). Interestingly, whether using a constant E or VB-specific E, increasing the threshold to reduce the number of bony elements in the model improved the R2 to 70.7% and 73.1%, respectively (Fig. 3b,e). Building confidence in this approach to μCT-FEA, the predicted strength also strongly correlated with experimental strength of the VBs from the Genetic study with low error (Fig. 3f). Even though the FE model includes the thin cortex, the ability of FEA to explain the variance in experimental strength was similar to that of trabecular BV/TV (1D11: R2 = 62.4% & Genetic: R2 = 83.6%). Discussion: A μCT-derived elastic FEA can reasonably assess the strength of mouse VBs when the model has uniform isotropic material properties typical of bone (18 GPa) and uses simple failure criteria (2% of volume exceeds 0.7% equivalent strain). The choice of the threshold does influence the ability of μCT-FEA to predict VB strength, and as such, segmented images should be viewed for multiple VB scans across experimental groups. Using an E based on TMD can improve the prediction. Whether further improvements are gained with an inhomogeneous distribution of E and non-linear material behavior are yet to be determined. Compression tests of mouse VBs are prone to error owing to their small size and irregular shape. Abstract Central Abstract Proof PopUp http://ors2013.abstractcentral.com/abstract?TAG_ACTION=V... 2 of 4 1/9/13 9:46 AM As such, μCT-FEA is an attractive way to assess experimental effects on VB strength, especially if histology is desired. However, the predicted strengths should be interpreted in the context of the FEA assumptions, namely material behavior. If differences are expected at the material level, then compression tests should be carefully conducted. Significance: This study provides evidence that μCT-FEA can predict mouse vertebral strength in pre-clinical studies assessing anti-fracture efficacy. Acknowledgements: CDMRP Concept Award provided support. References: [1] Orwoll et al. J Bone Miner Res 2009 [2] Vilayphiou et al. J Bone Miner Res 2011 [3] Edwards et al. J Bone Miner Res 2010 [4] Yang et al. Cell 2004 [5] Pistoia et al. Bone 2002 [6] Wagner et al. Bone 2011.