Effect of Rosiglitazone on Bone Quality in a Rat Model of Insulin Resistance and Osteoporosis

A total of 61 female ZF rats (fa/fa), aged 16 weeks, were purchased from Charles River (Senneville, Quebec, Canada). All rats were housed singly on cornmeal bedding in plastic cages that contained a hide-away tube. The rats were fed and watered ad lib and aged for 9 weeks prior to treatment. During this period, all rats received a 1 ml piece of artificially sweetened strawberry Jell-O 5 days a week. Select groups were ovariectomized at age 16 weeks (Table 1). All procedures were carried out in accordance with the Canadian Council of Animal Care guidelines and were approved by the University of Toronto Animal Care Committee.

The rats were divided into treatment groups described in Table 1. RSG was administered in the Jell-O 5 days a week for 12 weeks. ALN was administered via subcutaneous injection once a week for 12 weeks. At 12 days and 2 days prior to killing, all rats were injected with Calcein Green, a bone mineralization marker (32). After 12 weeks of treatment, all rats were killed. The right and left femora and the L5 and L6 lumbar vertebrae were dissected and kept moist with saline-soaked gauze. These specimens were stored at −20°C and were thawed at room temperature prior to testing. The right tibiae were dissected and immediately fixed in 70% ethanol in preparation for embedding for histomorphometry.

Dual-energy X-ray absorptiometry (DEXA) was performed on all right femora and L5 and L6 lumbar vertebrae to determine BMD (33). This method generates a two-dimensional measurement of BMD using a small animal densitometer (PIXImus; GE, Mississauga, Ontario, Canada), which was calibrated with an aluminum/leucite phantom (33).

Microcomputed tomography (MicroCT) was performed on all right femora and L6 lumbar vertebrae providing a three-dimensional measurement of BMD and bone architecture (34). Analysis of L6 vertebrae yielded measurements of percent trabecular bone volume (BV), trabecular thickness, trabecular number, and trabecular separation. Analysis of right femora yielded measurements for anterior/posterior diameter, medial/lateral diameter, polar moment of inertia (mm⁴), cross-sectional area (mm²), and elliptical moment of inertia (mm⁴) (34). These femoral parameters were used to normalize the results from three-point bending and torsion testing. MicroCT was also used to assess the total porosity (percent) for all right femora.

To evaluate the mechanical properties of cortical bone, right femora were tested in three-point bending, and the diaphyses of left femora were tested in torsion. To evaluate the mechanical properties of trabecular bone, all proximal halves of right femora were tested in femoral neck fracture and all L6 vertebrae were tested in vertebral compression. For three-point bending, a preload between 1 and 2 N was applied to the midpoint of the right diaphysis, and the bone was loaded in bending until failure at a rate of 1 mm/min using an Instron 4465 (Instron, Norwood, MA) with a 1,000-N load cell as previously described (35). For femoral neck fracture testing, a preload between 0.5 and 1 N was applied and the femoral head was loaded until failure at 2.5 mm/min (36). For torsion testing, left femur diaphyses were tested on a custom built machine with a 20–lb-in load cell until failure at a speed of 1.5 degrees/sec as previously described (35). L6 lumbar vertebrae were tested in compression using an Instron 4465 as previously described (37). A preload of 5 N was applied, and the vertebra was loaded until failure at a rate of 1 mm/min.

Load-displacement curves were generated using LabView 5.0 software (National Instruments, Austin, TX) to calculate the bone structural properties (38). The ultimate load represents the maximum load that the specimen sustains. The stiffness is determined from the slope of the elastic region of the curve and represents the extrinsic stiffness or the rigidity of the specimen. The failure displacement measures the extent of deformation of the bone at failure, and the energy to failure represents the amount of energy that the specimen can absorb prior to breaking (38).

Three-point bending, torsion, and vertebral compression data were normalized using the geometrical measurements to generate a stress-strain curve and to determine the material properties (ultimate stress, modulus, and toughness) of the bones (39,40). Femoral neck fracture data cannot be normalized as a result of the complex geometry of the femoral head/neck.

After dissection, all right tibiae were cut in half, and each proximal half was cut coronally using an Isomet low speed saw (Buehler, Whitby, Ontario, Canada). Proximal cranial halves of right tibiae were then fixed in 70% ethanol for 1 week, followed by dehydration in acetone and infiltration with Spurr resin. Specimens were then embedded in Spurr resin and polymerized. The Spurr blocks containing the tibia samples were sectioned into 5-μm sections (Leica RM2265 microtome; Leica Microsystems, Richmond Hill, Ontario, Canada) for static histomorphometry and 7-μm sections for dynamic histomorphometry. The 5-μm sections were stained with Goldner’s Trichrome (41). The region of interest was located 1 mm distal to the tibial growth plate to ensure that only trabecular bone was being analyzed. Using a magnification of ×100, four fields were analyzed using an image analysis system (Bioquant Nova Prime, version 6.50.10; Bioquant, Nashville, TN) to obtain the following parameters: osteoid surface/bone surface (OS/BS), osteoid volume/BV (OV/BV), and ES.

Dynamic histomorphometry was performed on unstained 7-μm sections using the same tibial region of interest as in static histomorphometry. Sections were analyzed under ultraviolet light for Calcein Green labeling (41). Using a magnification of ×100, four fields were analyzed using Bioquant image analysis system to obtain bone formation parameters, including mineralized surface/BS (MS/BS) and bone formation rate (BFR). All measurements were completed following the nomenclature and guidelines from the American Society for Bone and Mineral Research (42).

SPSS version 17.0 software (SPSS, Inc., Chicago, IL) was used to perform statistical analysis. Student t test was used to evaluate differences between the two female Sham groups and between the female Sham control and OVX control groups. For the female OVX study, a two-way ANOVA was used to determine if there were any differences in means among the two treatments (RSG and ALN) and to determine if there was any interaction between the two drugs. Multiple comparisons were performed using Fisher least significant differences post hoc test. P < 0.05 was accepted as significant. Data are presented as mean ± SD.

Results from DEXA and MicroCT analysis can be found in Table 2. DEXA and MicroCT analysis both confirmed the effect of OVX in the ZF rat model. OVX controls showed significantly lower vertebral aerial and volumetric BMD than Sham controls (P < 0.05) as expected after OVX. The OVX RSG group demonstrated significantly lower vertebral aerial and volumetric BMD (P < 0.05) compared with OVX controls and the OVX ALN group. Both OVX ALN–treated groups showed significantly increased vertebral aerial and volumetric BMD compared with OVX controls and OVX RSG groups (P < 0.05). In the case of vertebral volumetric BMD, ALN may be attenuating the effects of RSG because there was only a 5% decrease in BMD from the OVX ALN to the OVX RSG plus ALN group compared with a 15% decrease in BMD from OVX controls to the OVX RSG group.

OVX controls showed significantly lower femoral aerial BMD than Sham controls (P < 0.05). The OVX RSG group demonstrated significantly lower femoral aerial BMD (P < 0.05) compared with OVX controls. Both OVX ALN–treated groups showed significantly increased femoral aerial and volumetric BMD compared with OVX controls and OVX RSG groups (P < 0.05). However, there were no differences when comparing femoral volumetric BMD of the OVX RSG group to the other OVX groups.

Trabecular architecture properties, as analyzed by MicroCT, can be found in Table 3. The OVX control group demonstrated significantly decreased trabecular percent BV and trabecular number and increased trabecular separation compared with Sham controls. RSG treatment resulted in further decreased percent BV, trabecular thickness, and trabecular number and increased trabecular separation compared with the OVX control group (P < 0.05). The OVX ALN groups, with or without RSG, showed the opposite result with increased percent BV and increased trabecular number compared with the OVX controls and OVX RSG groups (P < 0.05).

The results from porosity measurements via MicroCT can be found in Table 4 and Fig. 1. The OVX RSG group demonstrated significantly higher femoral total porosity than OVX controls (P = 0.023). Both OVX ALN groups had significantly lower total porosity than the OVX RSG group (P < 0.05). However, it should be noted that RSG did increase cortical porosity in the presence and absence of ALN.

After mechanical testing of cortical bone, there were no significant differences in any parameters between the Sham controls and OVX controls. Results from all mechanical testing can be found in Table 5. In three-point bending, ALN treatment increased mechanical strength relative to RSG-treated groups. The OVX RSG plus ALN group showed significantly higher ultimate load and failure load compared with the OVX RSG group (P < 0.05) (Table 5). Once data were normalized, the OVX RSG plus ALN group demonstrated significantly higher ultimate stress and toughness than the OVX RSG group (P < 0.05) (Supplementary Table 1). Also, these values for ultimate stress and toughness were not decreased compared with the OVX ALN group, suggesting that ALN is protecting against the effects of RSG in three-point bending testing.

After torsion testing of left femora, the OVX RSG group demonstrated significantly lower failure torque and torsional stiffness compared with OVX controls (P < 0.05) (Table 5). Of interest, the OVX RSG plus ALN group also demonstrated lower failure torque and stiffness compared with OVX controls and the OVX ALN group (P < 0.05) (Table 5), suggesting that ALN cannot protect against the effects of RSG in torsion. Once data were normalized, the OVX RSG group showed significantly decreased shear stress and shear modulus compared with OVX controls (P < 0.05) (Supplementary Table 1). The OVX RSG plus ALN group demonstrated decreased shear stress compared with the OVX ALN group (P < 0.05) (Supplementary Table 1), again suggesting that the presence of ALN does not protect against the effects of RSG in this parameter.

There were no significant differences between the Sham controls and OVX controls for any of the parameters from vertebral compression testing of all L6 vertebrae (Supplementary Table 1). The OVX RSG group showed significantly decreased ultimate load compared with OVX controls (P < 0.05) (Table 5). Both OVX ALN groups had significantly higher ultimate load compared with OVX controls. The OVX ALN also had significantly increased stiffness compared with OVX controls (Table 5). Once normalized, the OVX RSG group demonstrated significantly decreased ultimate stress and toughness compared with OVX controls (P < 0.05) (Supplementary Table 1). The OVX RSG plus ALN group had significantly increased ultimate stress compared with OVX RSG groups (P < 0.05) (Supplementary Table 1).

After femoral neck fracture testing on all proximal femora, the OVX RSG group showed significantly decreased ultimate load compared with OVX controls (P < 0.05) (Table 5). The OVX RSG plus ALN group showed significantly increased ultimate load and energy to failure compared with the OVX RSG group (P < 0.05) (Table 5). In this case, ALN treatment largely reverses the adverse effects of RSG. In the presence of ALN, RSG does not cause the decrease in ultimate load that is observed in the absence of ALN.

Results from static and dynamic histomorphometry can be found in Table 6. Static histomorphometry analysis revealed significantly lower percent OV/BV and percent OS/BS for the OVX RSG group compared with OVX controls (P < 0.05), suggesting impaired bone formation with RSG treatment. In addition, both OVX ALN groups showed decreased percent OV/BV and percent OS/BS compared with OVX controls (P < 0.05), which is expected after chronic usage of bisphosphonates. We also observed an increase in ES, an indicator of increased bone resorption, which was significant in the OVX RSG group compared with OVX controls (P < 0.05). RSG treatment also appeared to result in increased adiposity (Fig. 2), which was consistently observed in all RSG-treated tibiae.

After dynamic histomorphometry, OVX controls showed significantly increased percent MS compared with Sham controls (P < 0.05). The OVX RSG group showed significantly decreased MS compared with OVX controls (P < 0.05). This group also demonstrated significantly lower BFR compared with OVX controls (P < 0.05). All bisphosphonates, including ALN, will eventually reduce bone formation to a very large extent (from 70 to 90%) as a result of their suppression of bone resorption, and this is likely the reason why we were not able to measure BFR (labeling) in the ALN-treated rats (43–45).

This study examines the effects of RSG on bone quality in an obese, hyperglycemic, and hyperinsulinemic animal model of type 2 diabetes and osteoporosis. Previous studies examine the effect of RSG on bone loss and bone formation/resorption, but none relate bone structure and composition to bone quality determined by bone mechanical properties.

In women, but not men, limb fractures have been reported in those taking RSG (5). Our results support this idea of compromised cortical bone strength. In our female ZF rat model, the negative effects of RSG were specific to OVX rats. No effect of RSG was observed in intact rats. Total femoral porosity was increased by >100% in the OVX RSG–treated group compared with OVX controls. This is especially important because even small increases in porosity can decrease bone strength substantially (39), and this cortical porosity is not normally seen in OVX rats. Consequently, after torsion testing of left femora, we observed decreased cortical mechanical properties (failure torque, stiffness, shear stress, and modulus) after RSG treatment. ALN treatment resulted in significantly decreased porosity as expected, owing to its antiresorptive properties. This decreased porosity in the ALN groups was reflected in the improved mechanical properties. However, these parameters did not improve to the levels seen in OVX controls, suggesting that ALN treatment does not completely prevent the decrease in mechanical properties caused by RSG treatment. Decreased bone formation may be to blame for the detrimental effects of RSG on mechanical properties in the cases where ALN is unable to completely protect against these RSG effects. In cases where ALN does exert a protective effect (for example, femoral neck fracture and three-point bending testing), an increase in bone resorption may be the cause of decreased mechanical properties after RSG treatment.

As expected, both cortical and trabecular BMD and bone mineral content (BMC) were decreased in the OVX model after RSG treatment. In a similar manner, Sottile et al. (21) performed a study with OVX, female Wistar rats and saw enhanced bone loss with RSG treatment in tibia, femur, and lumbar spine. Sorocéanu et al. (19) also observed decreased vertebral BMD and lower trabecular BV after RSG treatment in mice. In addition to our BMD analysis, our mechanical testing results suggest decreased vertebral bone strength. RSG treatment resulted in decreased ultimate load, ultimate stress, and energy to failure in the vertebrae of our female OVX model. We speculate that RSG is causing decreased strength and increased brittleness in trabecular bone of the female OVX model, resulting in an increased susceptibility to fracture. ALN treatment maintained cortical and trabecular BMD, suggesting that ALN may prevent the loss of bone mass associated with RSG treatment.

While TZDs such as RSG are a commonly prescribed treatment for type 2 diabetic patients, there are conflicting studies regarding the influence of RSG on bone turnover. Experimentally, in vitro, decreased bone formation after RSG treatment has been reported. PPAR-γ is a known positive adipocyte differentiation regulator; thus, activation of PPAR-γ by RSG could reduce bone formation through promotion of the adipocyte phenotype. RSG administration has been shown to decrease bone formation while also increasing adipogenesis (10,16). Our results support this evidence because histomorphometric analysis revealed decreases in bone formation parameters (OV/BV, OS/BS, and BFR) of trabecular bone in the female OVX model after RSG administration.

Recent in vivo studies suggest increased osteoclastogenesis due to RSG treatment. Previous studies suggest that the effect of PPAR-γ activation is specific to osteoclast genes that function along the osteoclast differentiation pathway. Wan et al. (20) discovered a pro-osteoclastogenic effect of PPAR-γ after RSG administration. Sottile et al. (21) also found evidence supporting increased osteoclastogenesis by reporting both an increase in marrow adipogenesis and increased bone resorption in Wistar rats after activation of PPAR-γ with RSG. Our results are consistent with these findings; we observed significantly increased ES in our model, which further suggests an increase in bone resorption. This, along with ALN’s antiresorptive ability to prevent losses of trabecular architecture and bone mass, leads us to conclude that RSG significantly influences bone resorption in our rat model.

Activation of PPAR-γ can regulate c-fos levels, part of the receptor activator of nuclear factor-κB ligand signaling pathway, which can lead to increased osteoclastogenesis (19). Together, our data indicate that bone formation and bone resorption are both being affected by RSG in the female ZF OVX model. Giaginis et al. (23) also demonstrated a decrease in bone formation and increased osteoclast generation in OVX Wistar rats, whereas they saw no change in intact non-OVX rats after RSG administration.

This rat model of obesity and type 2 diabetes in combination with OVX is especially relevant because in ADOPT, increased limb fractures were reported specifically in women with type 2 diabetes. Using ALN in combination with RSG and OVX not only allowed us to examine a possible countermeasure to the deleterious effects of RSG on the skeleton but also provided insight into the mechanism by which RSG is inducing bone loss in this in vivo system. Estrogen’s protection against the negative effects of RSG on bone suggests a potential mechanism by which RSG leads to bone loss. RSG activation of PPAR-γ may be interacting with receptor activator of nuclear factor-κB ligand signaling resulting in increased osteoclastogenesis and increased bone resorption. ALN, however, does not completely protect against the effects of RSG on the skeleton. In the case of three-point bending and femoral neck fracture, we have observed the ability of ALN to alleviate the effects of RSG in the female ZF rat model in vivo, whereas in torsion testing, we have not. The lack of a detectable effect in three-point bending, contrary to what we found in the torsion data, may be due to the distribution of extra porosity and the difference in the distribution of the peak stresses in the two test methods. Three-point bending causes a very small region of peak bending moment (and, therefore, stress) at the midspan of the specimen on one face of the bone. Torsion testing causes constant torque through the length of the specimen and, therefore, a much larger region of bone is equivalently stressed. This much larger region of stress results in a much higher probability of achieving failure stress at a defect (such as a pore) acting as a stress concentrator. This explains the smaller difference in means in the three-point bending data between OVX CON and OVX RSG. Since ALN is known to enhance the mechanical strength of bone through decreasing resorption and increasing mineralization, we suggest that RSG’s seemingly detrimental effect on bone may be due to excess resorption in addition to decreased formation. It is important that these changes in bone remodeling have a direct effect on the mechanical properties whereby strength was decreased in both cortical and trabecular bone treated with RSG (Supplementary Fig. 1). This finding has significant implications for women taking this drug.

While this study was carried out in the ZF fa/fa rat model of obesity, associated with insulin resistance and glucose intolerance, it should be noted that the diabetes state itself has been documented to have adverse effects on bone characteristics (11). We examined the effects of RSG in this context since the TZD drugs are administered to human subjects with type 2 diabetes. While some rodent studies of RSG effects in nondiabetic rats (14) suggest similar effects, a direct comparison would be interesting to explore the interacting effects of RSG and diabetes on bone.

One other potential caveat is that RSG treatment may improve the diabetes state. Since, as noted above, diabetes may adversely affect bone, such an effect of RSG would underestimate its adverse effects. We have measured random morning-fed glucose values in the different groups of rats at baseline and over the 3 months of treatment (Supplementary Table 2). By two-way repeated-measures ANOVA, there were no significant differences in any of the glucose values and no interaction detected between treatment groups and time. There was a trend toward lower mean glucose in the OVX group treated with RSG from baseline to 3 months, which was not seen in the combination RSG plus ALN-treated group. This was not likely significant and if so, would have caused an underestimation of the benefit of ALN. However, there is no evidence in human subjects with diabetes treated with bisphosphonates that glycemic control is altered (46,47). Taken together, these data indicate that the antidiabetes effects of RSG and the diabetes state itself are not significant confounders of the RSG actions on bone.

In conclusion, we found that RSG treatment enhanced the effects of OVX on bone loss in cortical and trabecular bone in the female OVX ZF model. Congruent with this bone loss were decreased mechanical properties in both cortical and trabecular bone after RSG treatment. ALN appeared to prevent the bone loss caused by RSG and partially prevented decreases in mechanical strength of cortical and trabecular bone. RSG treatment appeared to cause decreased bone formation as well as large increases in bone resorption, suggesting a possible mechanism by which RSG induces bone loss and inferior bone quality in our OVX female model. Especially interesting is the increased long bone porosity after RSG treatment, which may explain the increased susceptibility to limb fractures experienced by postmenopausal women taking RSG. Further examination into the proresorptive capabilities of RSG should be carried out to confirm the influence that PPAR-γ activation by RSG is having on bone resorption versus bone formation.

L.D.S. has received funding from the Canadian Institutes of Health Research (CIHR) and the Banting and Best Diabetes Center at the University of Toronto. I.G.F. and M.D.G. have received funding from the CIHR and GlaxoSmithKline. No other potential conflicts of interest relevant to this article were reported.

L.D.S. conducted experimental work, analyzed data, and wrote the manuscript. R.R. contributed to the animal experiment. T.L.W. contributed to experimental work (mechanical testing and MicroCT) and wrote the manuscript. I.G.F. contributed to the design of the experiment and wrote the manuscript. M.D.G. designed and supervised the experiment and wrote the manuscript.

The authors thank Mr. Richard Cheung, Samuel Lunenfeld Research Institute, Mount Sinai Hospital; Dr. David Cole, Womens College, Women's College Hospital, Toronto, Ontario, Canada; Dr. Reinhold Vieth, Mount Sinai Hospital; Mr. Doug Holmyard, Mount Sinai Hospital; and Dr. Mircea Dumitriu, Samuel Lunenfeld Research Institute, Mount Sinai Hospital, for their help.