Multiscale Characterization of Ovariectomized Rat Femur

Abstract

Estrogen deficiency activates bone resorbing cells (osteoclasts) and to a lesser extent bone forming cells (osteoblasts), resulting in a gap between resorption and formation that leads to a net loss of bone. These cell activities alter bone architecture and tissue composition. Thus, the objective of this study is to examine whether multiscale (10⁻² to 10⁻⁷ m) characterization can provide more integrated information to understand the effects of estrogen deficiency on the fracture risk of bone. This is the first study to examine the effects of estrogen deficiency on multiscale characteristics of the same bone specimen. Sprague-Dawley female rats (6 months old) were obtained for a bilateral ovariectomy (OVX) or a sham operation (sham). Micro-computed tomography of rat femurs provided bone volumetric, mineral density, and morphological parameters. Dynamic mechanical analysis, static elastic and fracture mechanical testing, and nanoindentation were also performed using the same femur. As expected, the current findings indicate that OVX reduces bone quantity (mass and bone mineral density) and quality (morphology, and fracture displacement). Additionally, they demonstrated reductions in amount and heterogeneity of tissue mineral density (TMD) and viscoelastic properties. The current results validate that multiscale characterization for the same bone specimen can provide more comprehensive insights to understand how the bone components contributed to mechanical behavior at different scales.

1. INTRODUCTION

About 50% of the postmenopausal female population older than 50 years of age shows symptoms of osteoporosis and has a particularly higher risk for fracture than the male population and other age groups (Eastell et al. 2016; Seeman 2003a; 2003b; van Staa et al. 2001). Bone mineral density (BMD) has been widely used to diagnose osteoporosis and estimate the fracture risk of bone (Kanis et al. 1994). However, BMD, which measures mineral in a given bone volume that includes empty spaces, cannot fully explain bone fragility, as both bone architecture and composition play key roles in determining bone strength (Heaney 2003; McCreadie and Goldstein 2000). In contrast, tissue mineral density (TMD) is the mineral content in a unit volume of bone matrix only (Tassani et al. 2011; Yao et al. 2007). Bone cells involved in the high bone turnover due to estrogen deficiency produce heterogeneous bone matrix at different points in time during postmenopause and inevitably change the TMD distribution (Ames et al. 2010; Marcus 1996; Yao et al. 2007). As the mineral contents are associated with mechanical responses of bone matrix (Mulder et al. 2008; Mulder et al. 2007), it is likely that the abnormal osteoporotic bone matrix has different elastic, plastic, and time-dependent viscoelastic properties from normal healthy bone matrix. These characteristics of bone matrix are critical in maintaining its mechanical stability to resist static and dynamic loading at the organ level of bone (Bart et al. 2014; Donnelly et al. 2010; Kim et al. 2017; Kim et al. 2015; Kim et al. 2018; Shah et al. 2018).

Traditional methodologies directly characterize the irregular shape and heterogeneous composition of bone by measuring elastic stiffness, fracture force, elastic modulus and plastic hardness at the different levels of bone (Donnelly 2011). Recently, we have developed additional novel technologies to analyze TMD distribution and characteristics including static and dynamic elastic, viscoelastic, and plastic mechanical properties at the multiscale of bone in the same individual (Kim et al. 2015; Kim et al. 2018). Thus, the objective of this study is to examine whether the multiscale (10⁻² to 10⁻⁷ m) characterization can provide more integrated information to understand the effects of estrogen deficiency on the fracture risk of bone. We used a rat model because ovariectomized (OVX) rats have been widely accepted to investigate the etiology of osteoporosis resulting from estrogen deficiency (Chen et al. 2015; Comelekoglu et al. 2007; Fox et al. 2006; Liu et al. 2015).

2. MATERIALS AND METHODS

2.1. Specimen preparation

Following the protocol approved by Institutional Animal Care and the Use Committee of The Ohio State University, 40 Sprague-Dawley female rats were obtained after a bilateral OVX or a sham operation (sham) at 6 months of age (n=20 for each group) by Harlan Laboratories (Harlan Laboratories Inc., Indianapolis, IN, USA). After 2 months post-OVX, the rats were euthanized and a femur was randomly obtained from each rat.

2.2. Micro-computed tomography (micro-CT)

These femurs were scanned using micro-computed tomography (SkyScan1172-D, Kontich, Belgium) with 27×27×27 μm³ voxel size under an identical scanning condition of 70 kV, 141 μA, 0.4° rotation per projection, 6 frames averaged per projection, and 210 ms exposure time (Fig. 1). Gray value of each voxel was calibrated to tissue mineral density (TMD) using known density hydroxyapatite (HA) phantoms (1220 and 1540 mg HA/cm³ with the same dimension of Ø4 × 5.5 mm). Bone voxels were segmented from non-bone voxels using a heuristic algorithm (Kim et al. 2012). Bone volume (BV) was obtained by counting the whole bone (WB) voxels. The BMD was assessed by dividing a total sum of TMD in each bone voxel by the total volume (TV) including bone, pores and marrow cavity. Cortical bone (CB) was digitally separated from trabecular bone (TB) by subtracting the masked regions from the whole femur (Buie et al. 2007; Kim et al. 2018; Kim et al. 2012) (Fig. 1a). Mean, standard deviation (SD), coefficient of variation (COV=SD/Mean), Low and High (Low₅ and High₅ for the lower and upper 5^th percentile values, respectively) were determined using TMD histograms for each region (Fig. 1b,c,d). The femur length (Length) from femoral head to distal end was measured by counting axial slices in the micro-CT image. A cross-sectioned cortical region at 55% (CB₅₅) of the femoral length from the head was digitally cut with 50 voxels (1 mm) thickness (Fig. 1a). The parameters of CB₅₅ were assessed for TMD distribution, cortex bone volume (BV₅₅), total volume (TV₅₅), fraction (BV₅₅/TV₅₅), thickness (Ct.Th), periosteal perimeter (Periosteal Perimeter55), endosteal perimeter (Endosteal Perimeter55), outer and inner diameters of anterior-posterior axis and medial-lateral axis (D_{AP_outer}, D_{AP_inner}, D_{ML_outer}, and D_{ML_inner}), outer diameter ratio (AP/ML), maximum inertia (I_max), and minimum inertia (I_min). In particular, the axis of minimum inertia is the direction of least bending resistance, which is the AP direction of bending used in the current study. Further, layers of CB₅₅ were segmented by one voxel thickness to obtain TMD parameters from periosteum to endosteum (Fig. 2). A trabecular bone (TB) region (0.97×0.97×2.03 mm³) above the growth plate at the distal femoral condyle was digitally isolated to compute TB morphology. Trabecular morphologies including trabecular bone fraction (BV/TV_TB), surface-to-volume ratio (BS/BV), number (Tb.N), thickness (Tb.Th), and separation (Tb.Sp) were computed.

Fig. 1.

(a) Steps of compartmentation for the 3D micro-CT image of a femoral bone. Masking was performed to identify the bone marrow cavity including trabecular bone (TB). CB₅₅ (55% of length) was digitally isolated. The total volume was determined, including CB, TB, and masked voxels. The TMD parameters were determined in individual TMD histograms of typical (b) whole bone, (c) CB, and (d) TB for sham and OVX groups.

Fig. 2.

Steps of segmentation for (a) cortical bone (CB₅₅, 55% of length) layers from periosteum to endosteum to obtain TMD (b) mean, (c) SD, (d) Low₅, and (e) High₅. The OVX group had significantly lower values of all TMD parameters at each distance compared to the sham group (p<0.046), except SD between 350 μm and 540 μm, and Low₅ between 675 μm and 810 μm (p>0.051).

2.3. Dynamic mechanical analysis (DMA) and static fracture testing

After the non-destructive scanning, a series of mechanical testing was performed as described in previous studies (Amorosa et al. 2013; Kim et al. 2017; Kim et al. 2018; Kim et al. 2016). The specimen was kept wet during the whole mounting process on an electromagnetic loading machine (Bose Corporation, Framingham, MA, USA) with a displacement transducer with 15 nm resolution (Fig. 3a). The anterior-posterior loading location was determined at 55% of the femur length from the femoral head. A preload of 10 N was applied to confirm contact between the specimen and loading jig. The K and W were assessed by applying a static displacement up to 0.01 mm (Fig. 3b). The DMA used non-destructive oscillatory bending displacement with a mean level of 0.01 mm and amplitude of 0.005 mm at the range of 0.5 to 3 Hz (Fig. 3c). A dynamic complex stiffness (K*) was computed from elastic (storage) (K′) and viscous (loss) (K′′) stiffness with an equation of K*= K′+iK′′. Tangent delta (tan δ), which accounts for ability of loading energy dissipation, was computed by K′′/K′. The phase shift (δ) was detected in cyclic load and displacement curve. Following the non-destructive DMA testing, static fracture testing was conducted with a displacement rate of 0.5 mm/second up to fracture. Maximum force and displacement at the fracture point were used to determine strength (F_max) and displacement (d_max), respectively (Fig. 3d). Toughness (U) was obtained by computing area under the load-displacement curve up to the fracture point.

Fig. 3.

(a) 3-point bending at 55% of femur length from the femoral head, (b) hysteresis (W) computed using a non-destructive load-displacement up to 0.01 mm, (c) non-destructive oscillatory bending displacement using a mean level of 0.01 mm and amplitude of 0.005 mm at the range of 0.5 to 3 Hz of DMA, and (d) static fracture tests for typical femurs of sham and OVX groups. Red dashed lines: nanoindentation locations

2.4. Nanoindentation

After fracture testing, the non-damaged regions of femur outside the jig were dissected and the surface of specimens was polished for nanoindentation (Fig. 4a,b). A pyramidal Berkovich tip for the nanoindenter was used to probe the specimen up to 500 nm deep with a displacement rate of 10 nm/sec (Fig. 4c). The plastic hardness (H) was measured at a peak indenting load (P_max). The viscosity (η) and normalized creep (Creep/P_max) were assessed during a 30-second hold period under peak load. Tan δ was also assessed by continuous stiffness measurement (CSM) using a phase shift under 45 Hz oscillatory force corresponding to 2 nm of displacement during the 30-second holding period (Kim et al. 2015). Tan δ is computed by $\delta ={tan}^{-1}(\frac{\omega \text{C}}{\text{K}-\text{m}{\omega }^2})$ where δ is a phase angle between the force and displacement signals, ω is the oscillation frequency, C is the damping coefficient of indentation contact, K is the spring constant of the contact, and m is the indenter mass. Finally, the elastic modulus (E_b) was measured during the unloading period of nanoindentation. As results, the five parameters (E_b, H, η, Creep/P_max and tan δ) of elastic, plastic, and viscoelastic mechanical properties could be assessed at the same site of bone matrix using a cycle of nanoindentation.

Fig. 4.

(a) Nanoindentation sites in a femur. Although this figure shows a bigger size with 1000 nm depth of indentation to clearly illustrate the locations, the real size of indentation is smaller (5 to 6 μm width) with 500 nm depth for 3×3 array sites for periosteal, core, and endosteal regions of the cortical bone. (b) Elastic modulus (E_b), plastic hardness (H), static viscoelastic normalized creep (Creep/P_max) and viscosity (η), and dynamic viscoelastic tangent delta (tan δ) are provided using a cycle of nanoindentation at the same site of bone matrix.

2.5. Statistical analysis

One specimen of the OVX group was lost, one specimen of the sham group had experimental errors during mechanical testing, and distal condyles obtained from 3 specimens of OVX group had been broken at the process of dissecting. As a result, femurs of 19 sham and 16 OVX rats were analyzed. However, 20 sham surgery and 19 OVX rats were included for obtaining CB₅₅ parameters. For assessing DMA and static mechanical parameters, 19 femurs of each group were analyzed.

The current study measured 54 parameters as listed in Table 1. A Student’s t-test was utilized to compare between sham and OVX groups for each parameter except nanoindentation parameters that were analyzed using a linear mixed effect model with individual animal-specific random intercept to account for the intra-specimen correlation. A total of 1396 nanoindentation sites (734 for sham group and 662 for OVX group) was analyzed for each parameter. To test if the within-specimen variance of the nanoindentation parameters is different between sham and OVX groups, a likelihood ratio test was performed based on a mixed effect model assuming heterogeneous residual variance between the two groups, and a mixed effect model assuming homogeneous residual variance. A paired t-test was used to compare between CB₅₅ layers of sham and OVX groups. Since the number of tests is fairly large, false discovery rate (FDR) was examined. FDR is the proportion of significant tests that are false positive. We used the Benjamini-Hochberg procedure (Benjamini and Hochberg 1995) to calculate the q-value for each test. The q value of a given test can be interpreted as that, if one rejects all null hypotheses with p values less than that of the given test, on average 100*q% of those null hypotheses will be falsely rejected. Significance was set at p<0.05.

Table 1.

Comparison of measured parameters (mean±standard deviation) between sham and OVX groups. Significantly different parameters between sham and OVX groups are highlighted in bold (p<0.05). Var: within-specimen variance, heterogeneity.

Parameters			sham	OVX	p value
Volumetric	BV (mm3)		354.470±38.271	325.143±23.213	0.012
	TV (mm3)		508.259±32.689	501.825±29.235	0.547
	BV/TV		0.698±0.064	0.648±0.020	0.005
	BVCB (mm3)		273.720±34.937	257.477±19.623	0.108
	BVTB (mm3)		80.750±7.327	67.666±6.831	0.001
Mineral Density	TMC (mg HA)		533.580±70.226	464.748±32.114	0.001
	BMD (mg HA/cm3)		1050.334±126.482	925.918±28.058	0.001
	Mean (mg HA/cm3)		1502.244±52.102	1429.209±9.998	0.001
	SD (mg HA/cm3)		119.738±22.575	98.821±4.446	0.001
	COV		0.079±0.012	0.069±0.003	0.003
	Low5 (mg HA/cm3)		1296.442±23.095	1254.363±11.298	0.001
	High5 (mg HA/cm3)		1672.154±86.215	1564.623±14.133	0.001
	MeanCB (mg HA/ cm3)		1531.334±55.486	1454.175±9.931	0.001
	SDCB (mg HA/ cm3)		112.378±24.170	88.630±4.381	0.001
	COVCB		0.073±0.013	0.061±0.003	0.001
	Low5CB (mg HA/cm3)		1315.444±24.533	1281.287±13.521	0.001
	High5CB (mg HA/cm3)		1678.632±86.971	1569.751±14.251	0.001
	MeanTB (mg HA/ cm3)		1403.102±33.124	1336.105±12.643	0.001
	SDTB (mg HA/ cm3)		83.502±6.864	76.974±3.790	0.002
	COVTB		0.059±0.004	0.058±0.003	0.141
	Low5TB (mg HA/cm3)		1272.258±30.482	1214.363±14.668	0.001
	High5TB (mg HA/cm3)		1545.189±45.187	1465.647±12.876	0.001
Morphology	Length (mm)		36.383±0.674	36.676±1.224	0.377
	BV55 (mm3)		8.574±0.789	8.394±0.474	0.432
	TV55 (mm3)		13.594±1.006	13.476±0.757	0.701
	Ct.Th (mm)		0.710±0.072	0.705±0.037	0.836
	DAP_outer (mm)		3.171±0.129	3.179±0.092	0.831
	DML_outer (mm)		3.993±0.179	3.974±0.125	0.730
	AP/ML		0.795±0.023	0.800±0.014	0.411
	DAP_inner (mm)		1.837±0.245	1.855±0.144	0.801
	DML_inner (mm)		2.495±0.284	2.487±0.161	0.924
	Imax (mm4)		8.541±1.245	8.229±0.786	0.353
	Imin (mm4)		5.687±0.741	5.644±0.616	0.853
	Periosteal Perimeter55 (mm)		12.080±0.449	12.011±0.331	0.618
	Endosteal Perimeter55 (mm)		8.789±0.601	8.861±0.361	0.654
	BV/TVTB		0.299±0.17	0.069±0.051	0.001
	BS/BV (mm−1)		31.872±4.965	43.336±5.191	0.001
	Tb.N (mm−1)		2.453±1.278	0.670±0.478	0.001
	Tb.Th (mm)		0.118±0.011	0.096±0.008	0.001
	Tb.Sp (mm)		0.303±0.178	0.552±0.157	0.001
DMA	K* (N/mm)		517.644±38.899	538.723±28.82	0.068
	K′ (N/mm)		517.499±38.905	538.602±29.599	0.068
	K′′ (N/mm)		11.839±2.378	11.098±1.789	0.285
	tan δ		0.023±0.005	0.021±0.003	0.073
Static Mechanical	W (Nmm)		0.002±0.001	0.001±0.001	0.093
	K (N/mm)		529.114±75.33	555.026±52.427	0.226
	Fmax (N)		144.804±24.857	131.814±21.511	0.094
	dmax (mm)		0.374±0.072	0.316±0.072	0.019
	U (Nmm)		43.341±40.72	26.915±14.984	0.108
Nano-indentation	Eb (GPa)	Mean	20.693±4.586	19.949±3.953	0.524
		Var	11.228	10.009	0.135
	H (GPa)	Mean	0.735±0.158	0.722±0.136	0.632
		Var	0.018	0.016	0.166
	η (GPa·S)	Mean	46475.28±14095.79	45003.7±11100.94	0.579
		Var	1.37E+08	1.03E+08	0.001
	Creep/Pmax (nm/mN)	Mean	8.723±3.412	8.597±2.606	0.725
		Var	7.470	5.996	0.004
	tan δ	Mean	0.048±0.028	0.046±0.021	0.557
		Var	0.000539	0.000409	0.001

3. RESULTS

The OVX group had significantly lower values of BV, BV/TV, BMD, whole and regional TMD, BV/TV_TB, Tb.N, and Tb.Th than the sham group (p<0.005) while it had significantly higher BS/BV and Tb.Sp (p<0.001). For mechanical testing, the OVX femur group had significantly less fracture displacement and heterogeneity (within-specimen variance) of viscoelastic nanoindentation parameters (η, Creep/P_max, and tan δ) than the sham femur group (p<0.03). All other parameters were not significantly different between sham and OVX groups (p>0.063) (Table 1). The values of CB₅₅ TMD were significant lower for the OVX group than the sham group at most of the layers (p<0.042) (Fig. 2b). The q values of all significant tests based on the original p values were below 0.05. Therefore, we expect less than 5% of the significant tests to be false positive.

4. DISCUSSION

Previous studies using OVX rat models showed that BMD, trabecular morphology, static elastic stiffness and strength, and nanoindentation modulus of femoral bone were altered due to estrogen deficiency (Chen et al. 2015; Comelekoglu et al. 2007; Fox et al. 2006; Liu et al. 2015). However, these properties were partly determined for different rats from several independent studies. As such, it was hard to correlate those previous results obtained under different experimental conditions and draw conclusions to elucidate the effects of estrogen deficiency on bone properties. The multiscale characterization of bone used in the current study was able to determine parameters of TMD distribution and viscoelastic characteristics responsible for controlling mechanical behavior of bone at the multiscale (10⁻² to 10⁻⁷ m) of the same femur as well as those traditional parameters.

Consistent with previous studies (Chen et al. 2015; Comelekoglu et al. 2007; Fox et al. 2006; Liu et al. 2015), the current study found that OVX deteriorates bone quantity and quality of femur, resulting in reduction of its load bearing capacity. The bone quantity is represented by bone mass (TMC and BMD) and bone quality includes regional variations of volume and mineral density distribution, morphology, dynamic and static mechanical responses, and elastic and viscoelastic properties of femur. Previous studies also observed multiscale characterization of bone, indicating that the microscale material behavior plays a critical role in determining skeletal load-bearing ability (Bart et al. 2014; Donnelly et al. 2010; Shah et al. 2018). For example, a previous study (Shah et al. 2018) investigated properties of bone matrix at the nano- and micro-level using the OVX rat model and found that the OVX group diminished BMD but increased bone mass through periosteal apposition compared with the wild-type control rats. However, the authors did not find differences in nanomechanical behavior and mineralized collagen fibrils, indicating that rats ovariectomized at 3 months undergo simultaneous bone loss and growth, resulting in the effects of OVX being less obvious. In contrast, the study presented here used rats at 6 months of age that are fully matured and have reached social maturity (Adams and Boice 1983; Johnston and Ward 2015), showing the best osteoporotic response by OVX (Francisco et al. 2011). In addition, the effects of estrogen deficiency on bone were observed starting at 2 months following OVX (Lelovas et al. 2008; Wronski et al. 1988). As a result, we also found that the nanoindentation properties were not different between sham and OVX groups, but we show that the distribution of TMD and viscoelastic properties at the tissue level of bone can affect mechanical behaviors at the organ level.

Estrogen deficiency caused by menopause or OVX results in high bone turnover with increased activity of both osteoclasts and osteoblasts. However, bone resorption outpaces formation, leading to net bone loss (Eastell et al., 2016; Seeman E. 2003a; Wronski et al., 1988). This results in smaller bone volume (BV) and fraction (BV/TV) of OVX animals than those of the sham control group, as observed in the current study. The BMD, which was computed by multiplying the mean value of TMD by BV/TV, also decreased because both TMD and BV/TV were reduced in the OVX group. The lower mean TMD of the OVX group resulted in part from decreases in both Low₅ and High₅ values that shifted the TMD histograms of OVX leftward (Table 1 and Figure 1). TMD Low₅ values measure mineralization of the 5% least mineralized bone and therefore reflect new bone matrix that is not yet fully mineralized. In contrast, High₅ values measure mineralization of the 5% most mineralized bone and therefore reflect mature bone (Kim et al. 2012; Kim et al., 2015). As such, the decreased Low₅ value of the OVX group indicates an increase in less mineralized, newly formed bone, while the decreased High₅ value indicates greater active resorption of more fully mineralized pre-existing bone tissues. These results likely arose because estrogen deficiency induced rapid bone turnover and increased activity of both osteoblasts and osteoclasts (Eastell et al. 2016; Evans et al. 1994). These patterns of TMD alteration due to OVX were observed regardless of whether cortical or trabecular bone was observed. Therefore, the results demonstrate that analyses of TMD distribution can depict biological activities of bone cells.

A limitation of the current study should be mentioned. We did not aim to clarify the detailed mechanisms of how viscoelastic properties are interrelated with crack initiation and propagation at the bone matrix and development of bone fracture at the organ level. It is accepted that the mineral contents mainly control elastic and plastic responses, while collagen and water are more responsible for determining viscoelastic responses of bone. As such, the viscoelastic behavior of bone likely results from interactions between these components responding to loading. Nair et al. (Nair et al. 2013) simulated that the energy dissipation capacity of the mineral-collagen interactions enables a stick–slip deformation process activated at large deformation. The current study observed that the brittle characteristics of OVX bone could result from decreased heterogeneity of TMD and viscoelastic properties. We speculate that uniform properties of bone matrix could limit local viscoelastic energy dissipation more than heterogeneous properties, thus increasing brittleness. Further studies are certainly needed to integrate these components in understanding the multiscale bone fracture.

Our multiscale characterization included innovative analyses for TMD distribution and viscoelastic characteristics in addition to the traditional BMD, morphology, elastic, and fracture parameters. The current findings indicate that OVX reduces bone quantity (mass and BMD) and quality (TMD distribution, TB morphology, mechanical properties). As the cortical bone morphology at the bending point was not different between sham and OVX groups, the more brittle property of OVX femur at the organ level likely resulted from its decreased ability to resist deformation at the tissue level compared to the sham control femur. These results are consistent with previous studies that speculated postmenopausal bone becomes brittle, resulting in micro-damage accumulation in the bone tissue and eventual reduction of bone integrity at the organ level (McNamara 2010; Schaffler 2003). The current study investigated direct connections between the multiscale characteristics at the tissue and organ levels using the same age and strain of OVX animals. Therefore, we demonstrate that the brittle characteristics of OVX bone could result from decreased heterogeneity of TMD and viscoelastic properties. Combinations of these characteristics likely increase the risk of micro-crack initiation and propagation at the tissue level under loading at the organ level. Therefore, these results validate that multiscale characterization of the same bone specimen can provide more comprehensive insights to understand how the bone components contributed to mechanical behavior at different levels. These results also allow speculation that if clinical CT analysis can assess BMD distribution in a manner similar to the methodology presented here, these data will suggest additional information on diagnosis of the fracture risk of postmenopausal patients.