Prostaglandins (PGs) are synthesized from arachidonic acid (AA) that is converted into PGH2 by the rate-limiting enzymatic reaction carried out by cyclooxygenase (COX) [1, 2, 3, 4]. Prostaglandin H2 is then isomerized to form a number of biologically active prostanoids that include PGE2 and PGI2 [2]. The two COX isozymes, COX-1 and COX-2, are each encoded by a different gene that is subject to unique regulation in vivo. COX-1 is constitutively expressed throughout the body but COX-2 is usually not present in most tissues and its expression is induced by pro-inflammatory agents, and in case of bone, by mechanical stimulus [1, 4, 5, 6].

Prostaglandins have been shown to regulate bone metabolism in vivo and in vitro [7]. Exogeneously added PGE2 (i) stimulates periosteal bone formation to increase total tissue area; (ii) enhances endocortical bone formation that accompanies reduced area of bone marrow; (iii) stimulates trabecular bone formation; and (iv) increases cortical porosity by activating intra-cortical bone remodeling [8]. PGE2 was also reported to increase the mechanical strength of bones in rapidly growing and mature male rats and could completely restore the deficit of cancellous bone in ovariectomized rats with established osteopenia [9].

Since both COX isozymes are present in the bone, the precise contribution of COX-1 and COX-2 in the turnover and remodeling of bone is unclear [7]. It has been suggested that COX-2 plays a unique role in bone remodeling in the adult that is different from its action on de novo formation of skeleton during development. Induction of COX-2 following injury and inflammation and elaboration of PGs is thought to directly regulate bone remodeling. Consistent with this hypothesis, COX-2−/− mice were shown to have decreased bone density compared with their heterozygous littermates; COX-2 was needed for optimal differentiation of mesenchymal cells into osteoclasts and osteoblasls that contributed to both intra-membranous and endochondral bone formation during bone repair [5, 6]. Since the abnormal bone repair in COX-2−/− mice could be reversed by exogenous PGE2, the authors concluded that PGs produced via the action of COX-2 directly affected bone remodeling [5, 6].

The effects of COX-2 on the mechanical properties of bone have not been directly investigated. Based on the previous observations, we hypothesized that the bones of COX-2−/− mice may have lower mechanical properties. Therefore we compared bulk material (elastic modulus and ultimate stress) and structural (stiffness and breaking force) properties by three-point bending, bone matrix properties (elastic modulus and hardness) by nanoindentation, and bone geometries (moment of inertia, cortical thickness, cortical area, and cortical diameter) in the wild type, COX-1−/− and COX-2−/− mice. We specifically tested the bone matrix properties by nanoindentation methods with a goal to test whether the altered bulk mechanical properties of bone were due to its structure and geometry or to changes in the bone matrix.

Materials and Methods

C57B6/DBA COX-1−/− (PGHS-1 null), COX-2−/− (PGHS-2 null) and their wild type littermates (C57BL6/DBA) were developed at the Department of Veteran’s Affairs Medical Center, Memphis, TN [10, 11] from mating pairs that originated at the University of North Carolina [3, 12]. Prior to experiments mice were housed in plexiglass cages at 25°C and kept on a 12-hr light/12-hr dark cycle in a virus-free environment. Food and water were available ad libitum. Due to fertility problems of the PGHS-null females, homozygous males were bred with heterozygous females to obtain homozygous COX-1 and COX-2-ablated mice used in these experiments. The progeny of heterozygous mating, containing both wild-type COX alleles, was used as control.

The genotypes of mice were determined by PCR [11]. DNA from tails was extracted using the Dneasy Tissue Kit from Qiagen (Valencia, CA). Tail samples were lysed overnight in a buffer containing proteinase K, and loaded onto a mini-column. After the column was washed, DNA was eluted in water and used for PCR. Three primers were used in the same PCR reactions for identification of the PGHS-1 or PGHS-2 allele. The WT allele 5′ primer (PGHS1-5′) AGGAGATGGCTGCTGAGTTGG, the mutant allele 5′ primer (PGHS1-neo) GCAGCCTCTGTTCCACATACAC, and the 3′ primer (PGHS1-3′) AATCTGACTTTCTGAGTTGCC were used to yield a fragment of 600 or 700 bp for the PGHS-1 WT or mutant allele, respectively. The WT allele 5′ primer (PGHS-2 5′-ACACATCTCACTGGCACC, the mutant allele 5′ primer (NeoPro) ACGCGTCACCTTAATATGCG, and the 3′ primer (TGC2-3) GTACGGTTCAGGGAGAA yield a fragment of 600 or 800 bp for the PGHS-2 WT or mutant allele, respectively. Three month-old mice were sacrificed by CO2 suffocation. The right femur from each mouse was excised and cleaned of soft tissue.

Femora were subjected to biomechanical testing in three-point bending (Instron 8500, Canton, MA) to measure bulk bone properties. Each femur was placed on two supports at a span length of 8.5 mm. Force was applied vertically at the mid-shaft of the bone on the anterior surface such that the anterior surface was under compression while the posterior surface was under tension. The press head as well as the two support points were rounded to avoid shear load and cutting [13, 14, 15]. Each bone was compressed at a constant rate of 3.6 mm/min until fracture [13]. Data were collected using a custom-written LabVIEW (National Instruments, Austin, TX) program and were converted into a load versus displacement curve. Stiffness (bulk structural property) was calculated as the slope of the linear (elastic) part of the load-displacement curve. Ultimate stress σ and elastic modulus E (bulk material properties) were calculated according to the previously reported formulas [14, 16, 17] as follows:

where F is the breaking force, L is the span of the support points, c is the half-diameter of the mid-shaft in the load direction, d is the displacement, and I is the principal moment of inertia.

The proximal part of each femur used in the three-point bending tests was embedded without vacuum in epoxy resin (EPO-THIN™ low viscosity epoxy, Buehler, Lake Bluff, IL) at room temperature to measure bone matrix properties by nanoindentation. Embedded samples were metallographically polished to expose the bone material to be measured and produce smooth surfaces needed for nanoindentation. Specimens were ground with silicon carbide abrasive papers of decreasing grit size (400, 600, 800, and 1200 grits) under de-ionized water, followed by micro-cloths (TEXMET, Buehler, Lake Bluff, IL) with alumina powder of 0.05 µm grit. After the polishing, the specimens were ultrasonically cleaned to remove surface debris.

All nanoindentation tests were performed using the Tribolndenter (Hysitron Inc., St. Paul, MN) in load control mode with a pyramid-shaped diamond tip. Fused silica was used prior to the measurements to calibrate the tip shape function and machine compliance. Four indents were made in each target area to reduce the effects of variations. To minimize the effects of viscoelasticity and creep on property measurements, a relatively long constant load hold period was conducted before the final unloading to diminish visco-elastic deformation to a negligible rate. The maximum load of the nanoindentation tests was 3 mN, with a loading/unloading rate of 200 µN/s, which produced impressions in bones 500 nm deep.

The Oliver-Pharr method [18] for determining the indentation modulus and hardness has been well documented and will not be discussed at length. Measurements of load and displacement were used to determine the contact stiffness. The reduced modulus E r was determined from the contact stiffness. The equations used to calculate the hardness (H) and the reduced modulus (E r) were:

and

where P max is the maximum force, A c is the contact area, and S the contact stiffness. The elastic modulus of specimen was derived from:

where ν is Poisson’s ratio. The Poisson’s ratio and elastic modulus of the diamond indenter tip, νtip and E tip, were 0.07 and 1140 GPa, respectively.

Bone cross-sectional geometry was evaluated at one femoral site adjacent to the fracture site. A transverse section about 1 mm thick was cut from each femur using Isomet 1000 (Bulhler, Lake Bluff, IL). Cross-sectional pictures were obtained using a digitizing optical microscope. A custom-written Visual C++ program was used to measure the geometric parameters by digitizing the periosteal and endocortical boundary [19].

The bone cross-sectional area was further examined by histological sections that were stained with hematoxylin and eosin (H&E). Specimens were demineralized and embedded in paraffin. Sections 5 µm thick were cut and mounted on slides to be assessed by light microscopy.

To compare the bulk mechanical properties, bone matrix properties, and geometric properties for different groups, one-way analysis of variance (ANOVA) was used (P < 0.05 for significance). Schéffe test (significance level: 5%) was employed to determine if specific pair-wise values were significantly different. All statistical analyses were performed with StatView 5.0.1 (SAS Institute Inc., USA).

Results

There were statistically significant differences in bulk material properties of C57BL6/DBA COX-2−/− mice when compared with either C57BL6/DBA COX-1−/− or wild-type mice (P < 0.05). Bulk material properties of COX-2−/− mice were 29% lower than those of the wild-type mice and 21% lower than those of the C57BL6/DBA COX-1−/− mice. In contrast, there were no significant differences in bulk material properties between COX-1−/− and their heterozygous littermates. The bulk structural properties (i.e., stiffness and breaking force) of bone among the three groups of mice, regardless of their COX genotype, were also similar (Table 1). Finally, as judged by the nanoindenation tests, no significant differences in bone matrix properties could be demonstrated among the wild type COX-1−/− and COX-2−/− mice (Table 2).

Table 1 Bulk mechanical properties of mouse femora measured by three-point bending test
Full size table
Table 2 Bone matrix properties measured with nanoindentation test (mean ± standard deviation)
Full size table

As shown in Table 3, no significant differences in cortical thickness and cortical cross-sectional area could be measured among the three groups of mice. This is in contrast to the measurements of the moment of inertia that was 11% higher in COX-2−/− mice compared with either COX-1−/− or wild-type mice (P < 0.001). There was slightly larger pericortical diameter of bone in COX-2−/− mice (7%) compared with either wild-type or COX-1−/− mice but this apparent difference in pericortical diameter was not statistically significant. In contrast, the endocortical diameter of the femora of COX-2−/− mice was larger than that of either COX-1−/− or wild-type mice (P < 0.001). In general, COX-1−/− mice had the smallest bones accompanied by the thickest cortical bone. The bones of COX-2−/− mice had significantly larger cortical diameter. As depicted in the representative cross-section of H&E-stained bone (Fig. 1), the cortices of COX2−/− were thinner and contained 2–3 times more intracortical pores compared with COX-1−/− or wild-type mice. The histological appearances of bones from heterozygous C57BL6/DBA COX-2−/+ mice were similar to those of COX1−/−, COX1−/+ or WT (unpublished data).

Table 3 Bone geometric parameters of mouse femora (mean ± standard deviation)
Full size table
Figure 1
figure 1

Histological sections stained with hematoxylin and eosin of femoral cortical bones from (A) COX-1−/−, (B) COX-2−/−, and (C) wild-type (heterozygous littermates) mice. Cortical bones of COX-2−/− mice were thinner and contained many more intra-cortical pores compared with the bones of the other two groups. Scale bar indicates 500 µm.

Full size image

Discussion

We report that the bulk material properties of C57BL6/DBA COX-2−/− bones are lower than those of the wild-type or COX-1−/− mice. The mechanical tests, combined with the results of geometry, histology, and nanoindentation tests, indicate that the decreased bulk material properties of the COX-2−/− mice are mainly due to the decreased thickness of cortical bone containing larger numbers of small intracortical pores (Fig. 1). It appears therefore that the material properties of bones were compensated for by differences in the geometries of bones from COX-1−/− and COX-2−/− mice and their heterozygous littermates. The cellular and molecular mechanisms by which numbers of intracortical pores are increased in the COX-2−/− mice needs to be investigated. How the lack of a functional COX-2 causes the observed changes in the bones of these mice remains unclear.

Since the regulation of bone metabolism and remodeling by PGs has been established both in vivo and in vitro, we hypothesize that lack of a functional COX-2 may potentially lead to the altered metabolism of prostaglandins in the bone. Exogenous PGE2 stimulated bone formation at the periosteal, endocortical, and marrow trabecular surfaces; PGs have also been reported to inhibit endocortical bone resorption with a consequent increase in the cortical bone mass [7, 8, 9, 20, 21, 22]. Although bone is capable of expressing COX-1 and COX-2, experiments with selective inhibitors of COX isozymes implicate COX-2 as the main regulator of bone remodeling. These experiments are largely corroborated by observations of COX-ablated mice, COX-1 and COX-2 knockout mice have revealed unique as well as overlapping roles of the two COX isozymes in various patho-physiological processes, including bone remodeling [23, 24, 25]. Since COX-2−/− mice are born with renal abnormalities [26, 27], it may be argued that the abnormalities in their bones may indirectly emanate from renal dysfunction. We believe that such a scenario is unlikely. The severity of renal pathology and lifespan of COX-2−/− mice appears to be strain-dependent; COX-2−/− mice with a mixed genetic background (DBA/B6) used in our study do not die of renal failure, unlike their B6/COX-2 counterparts [27, 29].

Furthermore, two recent studies provided direct evidence for the involvement of COX-2 in bone remodeling [5, 6]. Bone marrow cells from COX-2−/− mice cultured in vitro elicited decreased osteoclastogenesis and osteoblastogenesis and bone nodule formation. Both of these deficits in osteoclast and osteoblast formation in vitro could be overcome by exogenous PGE2 [5, 6]. Okada et al. [5] also showed that bone marrow cultures from COX-2 heterozygous mice had an intermediate deficit in osteoclastogenesis. In contrast, our COX2 heterozygous mice were similar to wild type with respect to bulk material properties, perhaps reflecting a different genetic background of our strain [10]. These differences not withstanding, we believe that the lower bulk material properties of COX-2−/− mice are not caused by renal complications but are a direct result of a lack of congenital COX-2.

We can only speculate if slightly larger cortical bones of COX-2−/− mice reflect a compensatory mechanical function (i.e., greater moment of inertia) in these animals. Both COX-1 and COX-2 catalyze the same reaction in the synthesis of PGs and it is conceivable that the lack of one type of COX may induce a compensatory change in the remaining isozyme. We have shown earlier that compensatory changes in the metabolism of PGs in fact occur in a number of tissues of COX-2−/− mice both in vivo and in vitro [10, 28, 30].

Nanoindentation is a widely used tool for measuring bone matrix properties [31, 32]. It allows us to probe a surface and map its properties with high resolution [31] and therefore it is a suitable method for assessing whether changes in bulk material properties of bone are due to changes in the bone matrix or geometry [33]. Our nanoindentation data showed that absence of either COX-1 or COX-2 gene had a significant effect on the bone matrix property. We are aware that bulk material properties may not represent true material properties because of a number of assumptions in the interpretation of nanoindentation data that may be too simplistic. For example, the cross-sectional geometry of a long bone varies along its length and is also poorly represented by a hollow circular cross-section, although the bulk material properties derived from whole bone tests assume both geometric and bone matrix properties to be constant [34, 35].

A number of COX-2-selectivc nonsteroidal antiinflammatory drugs (NSAIDs) have been developed in recent years [7, 36]. Although it is well appreciated that non-COX selective NSAIDs inhibit healing of bone-factures, the effects of COX-2 selective NSAIDs on bone remodeling and mechanical properties have not been investigated in detail. The long-term administration of COX-2 selective NSAIDs may decrease the bulk material properties in patients because of increased intracortical porosity and altered bone remodeling. We should point out, however, that the treatment with COX-2-selective NSAIDs in the adults and congenital ablation of COX-2 may not have similar consequences in the bone.