One Year of Transgenic Overexpression of Osteoprotegerin in Rats Suppressed Bone Resorption and Increased Vertebral Bone Volume, Density, and Strength


  • The authors state that they have no conflict of interest

  • Published online on February 16, 2009


RANKL is an essential mediator of bone resorption, and its activity is inhibited by osteoprotegerin (OPG). Transgenic (Tg) rats were engineered to continuously overexpress OPG to study the effects of continuous long-term RANKL inhibition on bone volume, density, and strength. Lumbar vertebrae, femurs, and blood were obtained from 1-yr-old female OPG-Tg rats (n = 32) and from age-matched wildtype (WT) controls (n = 23). OPG-Tg rats had significantly greater serum OPG (up to 260-fold) and significantly lower serum TRACP5b and osteocalcin compared with WT controls. Vertebral histomorphometry showed significant reductions in osteoclasts and bone turnover parameters in OPG-Tg rats versus WT controls, and these reductions were associated with significantly greater peak load in vertebrae tested through compression. No apparent differences in bone material properties were observed in OPG-Tg rat vertebrae, based on their unchanged intrinsic strength parameters and their normal linear relationship between vertebral bone mass and strength. Femurs from OPG-Tg rats were of normal length but showed mild osteopetrotic changes, including reduced periosteal perimeter (−6%) and an associated reduction in bending strength. Serum OPG levels in WT rats showed no correlations with any measured parameter of bone turnover, mass, or strength, whereas the supraphysiological serum OPG levels in OPG-Tg rats correlated negatively with bone turnover parameters and positively with vertebral bone mass and strength parameters. In summary, low bone turnover after 1 yr of OPG overexpression in rats was associated with increased vertebral bone mass and proportional increases in bone strength, with no evidence for deleterious effects on vertebral material properties.


Osteoprotegerin (OPG) is a soluble decoy receptor that binds to RANKL. RANKL is a member of the TNF superfamily that plays an important role in osteoclast formation, activation, attachment to bone, and survival.(1) OPG prevents each of these processes by preventing RANKL from binding and activating its cognate receptor, RANK, which is found on osteoclasts and osteoclast precursors.(2,3) Important roles for OPG and for RANKL in bone metabolism were established by the profound skeletal changes observed in knockout and transgenic (Tg) mice. RANKL knockout mice were virtually devoid of osteoclasts and had extremely high bone mass.(4) OPG knockout mice had increased bone remodeling, reduced bone density, and reductions in bone strength that led to spontaneous fragility fractures.(5,6) RANKL-Tg mice had markedly increased bone resorption and significant reductions in bone volume, density, and strength.(7) OPG-Tg mice had low levels of bone remodeling and increased bone volume and density, but bone geometry and strength parameters were not examined.(8)

RANKL inhibition has been shown to markedly reduce bone resorption and to significantly increase bone mass and bone strength in normal healthy animals(9) and in various animal models of bone disease.(10–12) However, the longest duration of RANKL inhibition in these studies was 24 wk.(10) The consequences of continuous life-long RANKL inhibition on whole bone strength and on the material properties of bone tissue are of interest when considering RANKL inhibition as a therapeutic strategy for chronic bone loss conditions such as osteoporosis. We therefore created OPG-Tg rats that overexpress the same full-length rat OPG construct first used to identify the skeletal role of OPG in Tg mice.(8) These OPG-Tg rats have advantages over the mice, including syngeneic OPG overexpression, greater blood volume for biochemical markers of bone turnover, and larger bones for studying trabecular microarchitecture and vertebral bone strength.

Although the ability of OPG to inhibit bone resorption and to maintain or increase bone mass is well validated, serum OPG concentrations frequently fail to correlate with bone turnover markers or BMD in clinical observational studies.(1,13,14) As such, the extent to which circulating levels of OPG regulate bone remodeling and bone mass in humans is unclear. In pursuit of a rodent model to study such relationships, we identified a murine OPG assay that cross-reacted with a recombinant rat OPG protein standard. This assay was used to quantify the levels of native OPG in rat blood and bone for the first time and to examine how these levels related to parameters of bone turnover, density, and strength in wildtype (WT) and OPG-Tg rats.


Generation of OPG transgenic rats

Single-cell embryos from outbred Sprague-Dawley–derived CD rats (Charles River) were microinjected with the full-length rat OPG transgene driven by the human apolipoprotein E (ApoE) promoter and liver-specific enhancer(8) and transferred into pseudopregnant recipient CD rats. OPG-Tg offspring were identified by PCR screening for the ApoE-OPG transgene in tail DNA.

Biochemical and blood analyses

Blood was collected at necropsy by cardiac puncture. Serum osteocalcin, total (free and bound) RANKL, and OPG were measured with commercially available microbead kits (Linco Research). The murine OPG assay showed >95% cross-reactivity with recombinant rat OPG standards, whereas the murine RANKL assay showed ∼40% cross-reactivity with recombinant rat RANKL standards (Amgen; data not shown). All OPG and RANKL data were generated using these recombinant rat protein standard curves. TRACP5b levels were evaluated with a rat ELISA kit (IDS). Clinical chemistry analytes were measured with a Hitachi 717 Automatic Chemistry Analyzer (Roche Diagnostics). A tibia was removed at necropsy, flash-frozen in liquid nitrogen, and pulverized. Total protein was extracted with a digestion buffer (50 mM Tris buffer, pH 7.4, containing 0.1 M sodium chloride and 0.1% Triton X-100), as previously described.(15) OPG and TRACP5b were measured from these protein extracts as described above, and levels were standardized to total protein concentrations that were determined by a BCA kit (Pierce). Red blood cell count, hemoglobin, and hematocrit were determined from EDTA-treated whole blood obtained from 13-mo-old female OPG-Tg rats (n = 15) and WT controls (n = 18), using an Advia 120 analyzer (Bayer).

Vertebral bone densitometry (DXA) and μCT

Areal BMD (aBMD) of the lumbar spine (L1–L5 vertebrae) was determined before death using DXA (QDR 4500a; Hologic). Volumetric BMD (vBMD) and trabecular microarchitecture were determined using a desktop μCT system (GE eXplore Locus SP Specimen Scanner; GE Healthcare). The fourth lumbar vertebra (L4) was scanned at 0.5° rotations for 200° (80 kVp, 80 μA). Images were calibrated to milligram per cubic centimeter mineral using a hydroxyapatite phantom and reconstructed to yield images with an 18-μm isotropic voxel size. All 3D image manipulations and analyses were performed within the GE MicroView software (v.2.1). For the central 70% of the vertebral axial height, contours were drawn outlining the vertebral body and within the cortical shell using a semiautomated contouring algorithm. The whole vertebral body was analyzed for total volumetric BMC and BMD (no threshold), total bone area (B.Ar), and total cross-sectional area. The cancellous region was analyzed for region vBMD (no threshold), trabecular bone volume per tissue volume (Tb.BVF), trabecular thickness (Tb.Th), trabecular spacing (Tb.Sp), trabecular number (Tb.N), structure model index (SMI), connectivity density, and degree of anisotropy. Single pixels were removed before analysis within the software using a purification algorithm. For the cortical shell, cortical area, cortical vBMD, and cortical thickness (Ct.Th) were calculated. Tb.Th, Tb.Sp, and Ct.Th were assessed using a sphere-packing algorithm, and trabecular structures <2 pixels (36 μm) thick were excluded. The segmentation threshold used for μCT analysis was determined for each specimen using an algorithm that selected a value based on the bimodal density distribution of marrow and bone.(16)

Vertebral histomorphometry

Rats were injected with tetracycline (25 mg/kg) on days −28 and −27 and calcein (10 mg/kg) on days −4 and −3 relative to termination. L2 vertebrae were embedded undecalcified in polymethylmethacrylate. Static and dynamic histomorphometry were performed on 4- and 8-μm-thick sections, respectively, using an Osteomeasure workstation (OsteoMetrics). Static measurements included trabecular bone volume as a percentage of tissue volume (BV/TV) and osteoclast and osteoblast surfaces as percentages of total surface (Oc.S/BS, Ob.S/BS). Total osteocyte lacunae were also counted, as was the percentage of lacunae occupied by osteocytes with apparently intact nuclei. Dynamic parameters included mineralizing surface (MS/BS), mineral apposition rate (MAR), and bone formation rate (BFR/BS). A zero was assigned to MAR for those samples in which single labels but not double labels were present.

Compression testing of lumbar vertebrae

Processes, laminae, and endplates were removed from the fifth lumbar vertebrae (L5) using a diamond wire saw (Model 3241; Well Diamond Wire Saws). The resulting vertebral body specimens (axial height = 4.75 mm) were compressed to failure using an MTS 858 Mini-Bionix II servohydraulic test system (FlexTest MPT software v3.5B). Load and displacement data were recorded by the test system, and the extrinsic strength parameters, peak load (Fmax), stiffness (S), and energy to failure (W), were determined. Intrinsic strength parameters were calculated from the whole bone strength results, total bone area (B.Ar) from μCT analysis, and trimmed vertebral body height (H) as follows: ultimate strength = Fmax/B.Ar; elastic modulus = S/(B.Ar/H); and toughness = W/(B.Ar × H).

Femoral μCT and bone strength analyses

Right femurs were scanned using the desktop μCT scanner, and images were reconstructed to a 32-μm isotropic voxel size. For the central 10% of the femur height, contours were drawn at the periosteal and endocortical surfaces using a semiautomated contouring algorithm. This integrated diaphyseal segment was analyzed for cross-sectional moment of inertia (CSMI) and periosteal perimeter, whereas the endocortical region of OPG-Tg rats was also analyzed for trabecular bone volume per tissue volume (Tb.BVF). Periosteal perimeter was assessed in a single distal femur metaphyseal cross-section located at 17% of the bone length from the distal end. A threshold of 845 mg/cm3 was used for these analyses. The femurs were tested to failure in three-point bending (MTS Mini-Bionix II, lower span 14 mm, displacement rate 3 mm/min). Intrinsic properties were calculated from data on whole bone extrinsic properties and cortical geometry.

Statistical analyses

Transgenic rats were assigned to one of three groups based on circulating OPG levels (as described in the Results section), and comparisons to WT littermates were performed using an ANOVA with Dunnett's posthoc analysis. Regression analyses were performed between biomarker, densitometry, histomorphometry, and strength endpoints for WT and OPG-Tg rats separately. OPG-Tg rats were grouped together for regression analyses. Log transformation provided the highest correlation of determination (R2) values for serum OPG and TRACP5b across the Tg groups. The slopes for linear regressions were tested for significant deviations from zero. When regressions for both WT and OPG-Tg rats were significant, the lines were tested for group differences. All statistical analyses were performed within GraphPad Prism v.4.02.


General phenotype of OPG-Tg rats

Transgenic overexpression of OPG for 1 yr resulted in radiographic evidence of increased BMD throughout the skeleton. Clinical chemistry analysis showed no significant differences between OPG-Tg and WT groups for serum calcium, phosphorus, total alkaline phosphatase, total bilirubin, cholesterol, triglyceride, blood urea nitrogen, creatinine, glucose, total protein, albumin, globulin, or the liver enzymes aspartate and alanine aminotransferase (AST and ALT). OPG-Tg rats had no evidence of anemia (normal red blood count, hematocrit, and hemoglobin levels), despite μCT evidence of reduced bone marrow volume in femurs (–53%) and L2 vertebrae (−31%; both p < 0.05 versus WT controls; data not shown). The lack of anemia in OPG-Tg rats was probably related to the mild to moderate extramedullary hematopoiesis observed by histologic assessment of the spleen. OPG-Tg rats did not exhibit impaired tooth eruption, had no spontaneous skeletal fractures, and did not differ in body weight from WT controls. The immune phenotype of adult OPG-Tg rats was previously described, which indicated no significant alterations in the structure or histology of lymph nodes or thymus, no significant differences in circulating cytokines or complete blood cell differentials, and no observed alterations in innate or humoral responses to defined immune challenges.(17)

Protein analyses

The range of serum OPG in Tg rats was 220 to 67,800 pg/ml (mean ± SE = 18,560 ± 3720 pg/ml), whereas the range in their WT littermates was 182 to 528 pg/ml (mean = 257 ± 16 pg/ml; Fig. 1A). For statistical analyses, OPG-Tg rats were allocated to one of three groups based on serum OPG concentrations. Tg-Low (n = 10), Tg-Med (n = 7), and Tg-High (n = 15) expressers had circulating OPG levels that were, respectively, <5, 5- to 50-, and 50- to 260-fold greater than the average concentration of WT littermate controls (Fig. 1A). There were no significant differences in OPG concentrations in tibial protein extracts for any OPG-Tg group compared with WT controls (Fig. 1B), and these tibial OPG levels did not correlate with serum OPG concentrations in either WT or OPG-Tg rats (R2 = 0.01-0.02; regressions not shown). Serum TRACP5b was significantly lower in Tg-Med and Tg-High groups compared with WT controls (−47 and −65%, respectively; Fig. 1C). Serum osteocalcin was significantly lower than WT controls only in the Tg-High group (−23%; Fig. 1D). Serum OPG levels in WT controls showed no significant correlations with either serum TRACP5b or serum osteocalcin (Figs. 1E and 1F). In contrast, serum OPG in OPG-Tg rats showed significant inverse correlations with both TRACP5b and osteocalcin (Figs. 1E and 1F). The average concentration of serum RANKL in WT controls was 47.6 ± 2.3 (SE) pg/ml, and the average serum RANKL:OPG ratio in WT controls was 0.096 ± 0.005. Serum analyses from these WT animals showed no significant correlations between RANKL versus TRACP5b (R2 = 0.006), RANKL versus osteocalcin (R2 = 0.008), RANKL:OPG ratio versus TRACP5b (R2 = 0.044), or RANKL:OPG ratio versus osteocalcin (R2 = 0.008; regressions not shown). Measurements of serum RANKL and serum RANKL:OPG ratios in OPG-Tg rats showed very low values (data not shown). The validity of these data is unclear because high levels of recombinant OPG (1-3 ng/ml) were observed to greatly reduce the ability of the RANKL assay to detect RANKL in vitro.

Figure Figure 1.

Biochemical markers of bone turnover. Serum was collected from 1-yr-old WT (n = 23) and OPG-Tg rats (n = 32) at necropsy. Transgenic groups were binned by serum OPG level (low, medium, and high). A tibia was harvested for analysis of OPG protein concentration. (A) Increasing levels of serum OPG in OPG-Tg rats. (B) No difference in OPG concentrations in whole bone protein extracts from OPG-Tg rats vs. WT controls. (C and D) OPG-Tg rats had reduced serum levels of the osteoclast marker TRACP5b and the formation marker osteocalcin. Data represent means ± SE. *Significantly different from WT littermates, p < 0.05. (E and F) Serum OPG concentrations in OPG-Tg rats, but not in WT controls, showed significant inverse correlations with serum TRACP5b and with serum osteocalcin. Regression lines represent OPG-Tg rats only. *Significant correlation, p < 0.01.

Vertebral DXA and μCT

μCT analyses indicated that the external geometry (cross-sectional area and height) of lumbar vertebrae was similar in OPG-Tg rats and WT controls (Table 1). DXA analyses showed that aBMD of the lumbar vertebrae was significantly increased in all OPG-Tg rats compared with WT controls (Fig. 2A). Serum OPG had a significant positive correlation with aBMD in OPG-Tg rats, whereas there was no significant correlation in WT controls (Fig. 2B). μCT analysis of whole vertebral L5 bodies showed increased vBMC and vBMD in all of the Tg groups (Table 1). Cortical area and thickness were significantly increased in OPG-Tg rats compared with WT controls (Table 1). In vertebral trabecular regions, Tb.BVF, Tb.vBMD, Tb.N, connectivity density, and degree of anisotropy were significantly greater in Tg-High rats, whereas Tb.Sp and SMI were significantly lower. Trabecular thickness was similar across all groups (Table 1). The autothresholding algorithm resulted in values for BVF and Tb.Sp that were highly correlated with those from histomorphometry across WT and OPG-Tg rats (R2 = 0.88 and 0.90, respectively). Representative μCT images of vertebrae are shown in Fig. 3. Trabecular architectural changes were generally proportional to the level of OPG overexpression in OPG-Tg rats. Significant positive correlations were found in OPG-Tg rats between serum OPG and total bone area, total vBMD, trabecular BVF, trabecular number (Fig. 4), and connectivity density (R2 = 0.86; regression not shown). In WT rats, no significant correlations were found between serum OPG and any μCT endpoints.

Table Table 1.. μCT Data for L5 Vertebra From 1-yr-Old OPG-Tg and Wildtype (WT) Rats
original image
Figure Figure 2.

Areal BMD and peak load of the lumbar spine. (A) aBMD of L1–L5 was analyzed by DXA before the necropsy of 1-yr-old WT (n = 23) and OPG-Tg rats (n = 32). Transgenic groups were binned by serum OPG level. Areal BMD increased concentration dependently with increasing levels of serum OPG (*p < 0.05). (B) Serum OPG correlated positively with lumbar aBMD in OPG-Tg rats but not in WT controls. Regression line represents OPG-Tg rats only (*p < 0.05). (C) Destructive compressive testing of L5 showed concentration-dependent increases in peak load with increasing levels of serum OPG. *Significantly different from WT littermates, p < 0.05. (D) Serum OPG correlated positively with lumbar peak load in OPG-Tg rats but not in WT controls. Regression line represents OPG-Tg rats only (*p < 0.05).

Figure Figure 3.

Representative μCT images of fifth lumbar vertebrae (L5). Transgenic rats were binned by serum OPG levels. Representative surface rendered images were selected based on the median values for whole vertebra vBMC within each group. The frontal 2D sections (18 μm thick) were thresholded. The axial 3D isosurface-rendered images consist of 20 slices (a 360-μm-thick region) from the center of the vertebral body and were thresholded.

Figure Figure 4.

Correlations of serum OPG levels with μCT parameters. Serum OPG levels in OPG-Tg rats correlated positively with (A) total bone area, (B) total volumetric BMD (vBMD), (C) trabecular bone volume fraction (BVF), and (D) trabecular number. R2 values are presented for individual regressions of WT and OPG-Tg rats. The regression lines shown in each panel are for OPG-Tg rats only (*all p < 0.001), because no significant correlations were found in wildtype (WT) controls.

Vertebral histomorphometry

The lumbar vertebrae of OPG-Tg rats had significantly greater trabecular bone volume, with significant differences noted in the Tg-Med and Tg-High groups (41% and 77% greater, respectively, than WT controls; Table 2). Increased bone volume was related to the reduced osteoclast surface, which was significantly lower (−68%) in the Tg-High group versus WT controls. Osteoblast surface was statistically similar in OPG-Tg groups compared with WT controls (Table 2). Bone formation was suppressed in OPG-Tg rats, as shown by reductions in mineralizing surface and bone formation rate. MAR was not significantly affected by OPG overexpression (Table 2). BFR was negatively correlated with serum OPG level in Tg rats (R2 = 0.51) but not in WT controls (R2 = 0.08; regressions not shown). The occupancy of trabecular osteocyte lacunae was analyzed in the Tg-High group and the WT controls, and there was a slight but statistically significant increase in the percentage of lacunae occupied by osteocytes in the Tg-High group (93.1 ± 0.8% versus 90.7 ± 0.7%, p = 0.034).

Table Table 2.. Histomorphometry Data for L2 Vertebra From 1-yr-Old OPG-Tg and Wildtype (WT) Rats
original image

Vertebral bone strength

Compression testing of L5 showed significant increases in strength parameters for OPG-Tg rats (Tg-Med and Tg-High groups) versus WT controls (Table 3; Fig. 2C). L5 from Tg-High rats had 56% greater peak load, 62% greater stiffness, and 53% greater energy absorption compared with WT controls. To approximate material properties, strength parameters were normalized to μCT-derived total bone area and height. Toughness and elastic modulus were similar between WT and Tg groups, whereas ultimate strength was significantly higher in the Tg-High group (+12%; p < 0.01 versus WT). No significant differences in ultimate displacement were found between WT and OPG-Tg rats (data not shown). Regression analyses showed that serum OPG was significantly and positively correlated with L5 peak load in OPG-Tg rats, whereas no association was found in WT rats (Fig. 2D). DXA-derived aBMD proved to be a strong predictor of vertebral peak load in both WT and OPG-Tg rats, with regression lines that were superimposed (Fig. 5A). μCT-derived total bone area (trabecular plus cortical) and trabecular BVF also showed similarly strong correlations with bone strength in WT and OPG-Tg rats (Figs. 5B and 5C). Peak load correlated positively with μCT-derived total vBMC (OPG-Tg: R2 = 0.71; WT: R2 = 0.57) and cortical thickness (OPG-Tg: R2 = 0.45; WT: R2 = 0.38; regressions not shown). No significant differences were found between OPG-Tg and WT groups for any of these regression lines. Peak load was inversely correlated with parameters of bone turnover in OPG-Tg rats, including BFR/BS (Fig. 5D), serum osteocalcin (R2 = 0.21; regression not shown), and serum TRACP5b (R2 = 0.32; regression not shown). There were no significant correlations between these bone turnover parameters and peak load in WT controls (data not shown).

Table Table 3.. Compressive Strength Data for L5 Vertebral Body From 1-yr-Old OPG-Tg and Wildtype (WT) Rats
original image
Figure Figure 5.

Correlations of L5 vertebral strength with bone turnover and density parameters. For A–C, black regression lines represent OPG-Tg rats (n = 32) and gray dashed lines represent WT controls (n = 23). Each of these regressions was statistically significant (p < 0.05), and OPG-Tg and WT were not significantly different. (A) Peak load of L5 from OPG-Tg rats and WT controls showed similar and strong positive correlations with DXA-derived lumbar aBMD. (B and C) L5 peak load also correlated positively with μCT-derived total (trabecular and cortical) bone area and with trabecular bone volume fraction (BVF) in OPG-Tg rats and in WT controls. (D) L5 peak load correlated negatively with histology-derived trabecular bone formation rate in OPG-Tg rats, whereas there was no significant relationship in WT controls. The black regression line in D represents OPG-Tg rats only.

Femur geometry, density, and strength

μCT showed that OPG-Tg rat femurs were of normal length (Table 4). The marrow cavity of OPG-Tg rat femur diaphyses contained ectopic trabecular bone in proportion to the level of OPG overexpression (Fig. 6A; Table 4). As a result, total BMC of the femur midshaft was 11% higher in Tg-High versus WT rats (Table 4). Periosteal perimeter was 6% lower in the Tg-High group (Table 4), and periosteal perimeter was inversely related to trabecular bone volume in the femur midshaft of OPG-Tg rats (Fig. 6B). The maximum perimeter of the distal femoral metaphysis was modestly (+4%) but significantly greater in Tg-High rats compared with WT controls, and the ratio of metaphyseal perimeter versus diaphyseal perimeter increased in proportion to the level of OPG overexpression (Table 4). Femurs from Tg-High rats exhibited a 21% lower CSMI, and 9% lower maximum load in a three-point bending test (Table 4). Regression analysis of CSMI versus peak load showed similar positive linear correlations in OPG-Tg rats and in WT controls (Fig. 6C). No significant differences were found in femur stiffness between any groups. Energy to failure, and the associated intrinsic property toughness, were significantly lower in femurs from Tg-Med and Tg-High animals (−25% and −35%, respectively; Table 4). The intrinsic material property ultimate strength was similar across all groups, whereas elastic modulus was significantly higher in Tg-High rats (+16%, p < 0.05 versus WT controls).

Table Table 4.. Femur Geometry and Strength in 1-yr-Old OPG-Tg and Wildtype (WT) Rats
original image
Figure Figure 6.

Femur geometry and strength relationships. (A) Representative cross-sectional μCT images of the femur mid-diaphysis (top panels), longitudinal images of distal femurs (middle panels), and cross-sectional images of the distal femur metaphysis from 1-yr-old WT and OPG-Tg rats. The top of each longitudinal image represents the mid-femur plane of the cross-sectional diaphysis images, and the dotted line represents the plane of distal femur cross-sectional images and data analysis. Retention of trabecular bone is evident in the diaphysis of OPG-Tg rats, in proportion to the level of OPG overexpression (low, med, and high). A narrower diaphysis is apparent in the Tg-High example shown. For B and C, black regression lines represent OPG-Tg rats (n = 32) and gray dashed lines represent WT controls (n = 18). (B) Regression analysis of trabecular bone volume fraction (BVF) vs. periosteal perimeter of the femur mid-diaphysis of OPG-Tg rats showed a significant inverse correlation (p < 0.05). (C) Regression analysis of femur mid-diaphysis CSMI vs. maximum bending load showed similar significant positive linear correlations for OPG-Tg rats and WT controls (p < 0.05 for each).


RANKL is an essential mediator of bone resorption, and the magnitude of bone turnover suppression that can be achieved through RANKL inhibition seems to exceed that associated with bisphosphonate therapy.(18–20) It is therefore important from a safety and efficacy perspective to understand the potential long-term effects of RANKL inhibition on bone strength and bone quality. Marked suppression of bone resorption and formation was evident in the vertebrae of 1-yr-old OPG-Tg rats, with reductions in osteoclast surface of up to 68% and reductions in bone formation rate of up to 90%. Analysis of similar OPG-Tg rats at 28 wk of age showed a 99% reduction in osteoclast surface versus WT controls.(21) Based on these data and their high bone mass phenotype at birth,(17) it can be assumed that OPG-Tg rats had very low bone turnover throughout their first year of life (half of their natural lifespan). Suppression of bone resorption in OPG-Tg rats was accompanied by significant increases in cortical and trabecular bone mass in lumbar vertebrae, which was achieved despite the occurrence of coupling-related suppression of bone formation. Increased cortical and trabecular bone volume in OPG-Tg rat vertebrae indicated that bone resorption was suppressed to a greater degree than bone formation. BFR was inversely correlated with vertebral bone strength in OPG-Tg rats across a broad range of supraphysiological OPG levels, indicating that the lowest levels of bone turnover were associated with the greatest vertebral strength. In fact, although Tg-High rats had markedly lower BFRs than the other groups, the maximum load-bearing capacity of vertebra from all 15 Tg-High rats (range = 471-835 N) was higher than the group mean (416 N) or median value (397 N) of the WT control group. These results are consistent with a meta-analysis of clinical trials of antiresorptive agents, wherein the greatest suppression of bone formation markers was associated with the greatest reduction in fracture risk.(22) These results indicate that the suppression of bone formation does not reduce bone strength when coupled with appropriate reductions in bone resorption, at least for the durations covered by the aforementioned studies. Biomechanical testing of lumbar vertebrae from OPG-Tg rats showed bone strength parameters that were proportional to vertebral bone mass parameters and to the level of OPG overexpression. The strong natural relationship between bone mass and bone strength, which was clearly evident in WT control rats, was not altered in OPG-Tg rat vertebrae apart from their shift to higher bone mass and strength.

Although the purpose of bone remodeling in any species is unclear, it has been hypothesized that the remodeling of human cancellous bone might serve to maintain osteocyte viability.(23) Older interstitial human bone, which had not been recently remodeled, was shown to have a greater proportion of empty osteocyte lacunae.(24) However, in human bone, the prevalence of empty osteocyte lacunae was shown to increase with age,(24) and increased bone remodeling in ovariectomized rats was associated with reductions in osteocyte density and bone strength.(25) These data suggest that, regardless of its evolutionary role, bone remodeling is an inefficient mechanism for the replacement of dead osteocytes and for the maintenance of skeletal integrity. In OPG-Tg rats, the occupancy of trabecular osteocyte lacunae was modestly but significantly higher than in WT controls. It seems unlikely that the transgenic OPG would have directly influenced osteocyte viability, because osteocytes are found only in bone and the overall skeletal concentration of OPG was not increased in OPG-Tg rats. The mechanism for this result is therefore unclear, and these preliminary analyses require corroboration in other models of RANKL inhibition. Nonetheless, greater osteocyte lacunar occupancy in OPG-Tg rats suggests that bone remodeling might not serve to maintain osteocyte viability in rat vertebrae.

The ability to modulate serum OPG levels through the ApoE transgene, and our recently validated method for measuring rat OPG(17,26) allowed us to examine the extent to which serum OPG might regulate bone resorption in rats. Our findings in 1-yr-old WT rats and previous data from 6-mo-old WT rats(21) showed that serum OPG levels failed to correlate with any measured parameter of bone turnover, mass, or strength. These observations, which are consistent with results from many clinical observational studies,(27–30) suggest that other factors besides circulating OPG contribute to the regulation of bone resorption. It is noteworthy, however, that high serum OPG levels in OPG-Tg rats were invariably associated with low levels of bone resorption and high bone mass. This suggests that circulating OPG is capable of regulating bone resorption but perhaps only at supraphysiological or pharmacological(31) levels. Circulating OPG is well positioned to regulate the differentiation of circulating osteoclast precursors,(32) which have been shown by parabiosis experiments to represent a key source of osteoclasts.(33) The high serum OPG levels in OPG-Tg rats might have constrained peripheral osteoclastogenesis to the extent that other normally important osteoclast regulatory factors became somewhat redundant and therefore less confounding to the relationship between serum OPG and bone resorption.

A potentially related finding was the total lack of correlation between serum and tibial OPG concentrations in both WT and OPG-Tg rats. The majority of OPG measured in tibial extracts was probably matrix derived, because the flushing of bone marrow from rat long bones only modestly reduced the yield of OPG levels (data not shown). Tibial OPG levels in OPG-Tg rats were apparently unaffected by their high circulating OPG levels, suggesting that bone matrix OPG did not originate from blood. Osteocytes(34) and osteoblasts(35) produce OPG and are well positioned to be potential sources of OPG found in bone matrix. Native OPG has been identified within bone matrix by immunohistochemistry,(36–38) and we corroborated these findings in rat tibias (data not shown). Interestingly, similar immunohistochemistry approaches showed no matrix localization with recombinant OPG-Fc.(39,40) Native OPG contains a heparin-binding domain,(41) and the absence of this domain in OPG-Fc might account for its lack of uptake in bone matrix. Although the biological role of matrix OPG is unknown, our data suggest that matrix OPG levels are unlikely to be reflected in serum measurements. OPG concentrations in arthritic rat paw extracts dropped precipitously with the progression of focal bone erosions, whereas serum OPG levels increased.(42) These results suggest that systemic and local regulation of osteoclasts by OPG might be independent processes, which could also help to explain why serum OPG alone tends to explain little of the variation in bone resorption in physiologic or pathologic states.

The OPG transgene used in these rats was previously shown to be overexpressed during midgestation, and OPG-Tg rat pups were born with evidence of retained trabeculae within their femur diaphyses.(17) We now show that these diaphyseal trabeculae are retained through 1 yr of age and that this retention was associated with geometric changes and a reduction in femur bending strength. The significantly decreased bending strength of Tg-High rat femurs could be attributed to a reduced midshaft periosteal perimeter and a resulting decrease in CSMI, which is an important attribute of long bone bending strength.(43) Periosteal perimeter was inversely related to trabecular bone volume within the femur diaphysis of Tg rats, suggesting that the inability of their osteoclasts to resorb primary spongiosa during bone elongation might have contributed to reduced periosteal expansion of the diaphysis. Ectopic trabeculae in the diaphysis of Tg rats might reduce the mechanical strain associated with normal weight bearing, thereby diminishing the stimulus for periosteal apposition in accordance with mechanostat theory.(44) Such interactions would be more appropriately studied in younger OPG-Tg rats, because reduced periosteal perimeter was already established in 6-mo-old OPG-Tg rat femurs (data not shown). This hypothetical periosteal adaptation might have served to optimize compressive resistance with the minimal amount of cortical bone. However, OPG-Tg rat femurs were not as well suited as normal femurs for resisting the bending forces exerted during strength testing. These observations suggest that during long bone growth, the mechanostat system is less capable of optimizing bone mass and geometry when osteoclasts are dysfunctional (in osteopetroses) or restrained by overwhelming levels of OPG.

There are several features of OPG-Tg rats that differ from classic osteopetrosis phenotypes,(3,4,45) including normal tooth eruption, lack of anemia, and lack of runting. Unlike RANKL knockout mice,(4) femur length in OPG-Tg rats was normal despite the presence of large amounts of unresorbed trabecular bone in the diaphyseal marrow. This finding suggests that these two features of osteopetrosis might not be directly related. Tg-high rats exhibited modest widening of the distal femoral metaphysis, and the ratio of metaphyseal:diaphyseal perimeters increased in proportion to the level of OPG overexpression. These findings might be analogous to changes in the “metaphyseal index” described by Ward et al.(46) in children treated with bisphosphonates. An important question is whether these geometric abnormalities explain the reduction in femur bending strength in these OPG-Tg rats or whether strength reductions are a signal for reduced “bone quality.” In favor of a geometric explanation, CSMI correlated positively with femur bending strength in OPG-Tg rats, and this relationship was indistinguishable from that observed in WT controls. Furthermore, in three other rat osteopetrosis models characterized by unfavorable long bone geometry, reduced femur bending strength was not associated with changes in measured bone material properties.(47,48) The lack of evidence for reduced bone quality or impaired material properties in OPG-Tg rat vertebrae, which had normal external dimensions, provides indirect support for a role of altered long bone geometry in the reduction of femur bending strength.

It is unclear whether these findings in OPG-Tg rat femurs would translate into clinical pediatric risks, but we have no a priori reason to predict that osteoclast inhibition through recombinant RANKL inhibitors would carry less risk to the young growing skeleton than bisphosphonate therapy. While bisphosphonates have been shown to increase BMD and to reduce the risk of vertebral fractures in children with osteogenesis imperfecta,(49) the effects of bisphosphonates on long bone geometry and fracture have not been established in pediatric clinical trials.(50) Furthermore, osteopetrotic long bone changes have been documented in a bisphosphonate-treated child.(51) It is important to determine the age range wherein osteoclast inhibition influences long bone geometry, which is likely to vary by species. The OPG transgene was overexpressed from mid-gestation in rat embryos(17) to at least 1 yr of age, making it difficult to determine an age range or the minimal duration of RANKL inhibition responsible for the changes observed in these OPG-Tg rats. Pharmacology data provide some insights, in that recombinant OPG had neutral or beneficial effects on femur bending strength and/or cortical geometry in rats as young as 6 wk of age,(9) in mice as young as 7.5 wk of age,(52) and in cynomolgus monkeys as young as 2–3 yr of age.(53) Bisphosphonate treatment of rats as young as 5 wk of age was also shown to improve femur bending strength.(54) Preliminary data from normal rats treated from 2–8 wk of age with recombinant OPG-Fc (1–10 mg/kg/wk) or with alendronate (0.1–1.0 mg/kg/wk) did not indicate any deficit in femur peak load or reduced periosteal circumference of the mid-diaphysis (data not shown). However, the duration of treatment in these pharmacology studies represented a small proportion of those species' life spans, and it remains possible that prolonged OPG administration, in young or neonatal animals, could recapitulate the long bone changes observed in these OPG-Tg rats.

This study has several limitations, including the fact that RANKL was inhibited throughout growth and development. This inhibition led to larger gains in vertebral bone mass and strength than would be expected had OPG overexpression begun during adulthood. Indeed, when aged (15 mo old) ovariectomized rats were treated for 24 wk with recombinant OPG,(10) the improvements in vertebral bone mass and density were significantly less robust than the results reported here. Nonetheless, marked bone turnover suppression for 24 and 52 wk provided no evidence of impaired vertebral bone quality in either study. Another limitation of this study is that the bone volume fraction of rat vertebrae greatly exceeds that of aged human vertebrae. As such, the relationships reported for trabecular architecture vs. bone strength might not be consistent between the two species. This limitation may have been exacerbated in the Tg animals to the extent that trabecular spacing and connectivity were well beyond normal ranges for rats.

In summary, 1 yr of RANKL inhibition through continuous overexpression of OPG suppressed bone resorption and resulted in altered geometry of long bones but not vertebrae. Long bones in OPG-Tg rats were of normal length, but they exhibited unresorbed trabeculae in the diaphyseal marrow, reduced periosteal perimeter, and an associated reduction in bending strength. In contrast, lumbar vertebrae from OPG-Tg rats had normal external dimensions and marked increases in trabecular bone volume, density, and strength. The strength of lumbar vertebrae was directly related to bone mass and inversely related to bone turnover parameters in both OPG-Tg and WT rats. These results suggest that the vertebral bone material properties were not significantly influenced by long-term RANKL inhibition and that the extrinsic bone strength improvements in OPG-Tg rat vertebrae were related primarily to increased bone mass and improvements in trabecular microarchitecture.


We thank Larry Ross (Amgen) for clinical chemistry analyses. Holly Brenza Zoog and Michelle N. Bradley assisted in the preparation of this manuscript. This work was sponsored by Amgen.