Subchondral bone microarchitecture and mineral density in human osteoarthritis and osteoporosis: A regional and compartmental analysis

Osteoarthritis (OA) and osteoporosis (OP) are historically considered to be inversely correlated but there may be an overlap between the pathophysiology of the two diseases. This study aimed to investigate the subchondral bone microarchitecture and matrix mineralization, and the association between them in OA and OP in relation to the degree of cartilage degeneration. Fifty‐six osteochondral plugs were collected from 16 OA femoral heads. They were graded on a regional basis according to the stages of cartilage degeneration, as evaluated by a new macroscopic and a modified microscopic grading system. Twenty‐one plugs were collected from seven femoral heads with OP. Plugs were scanned by microcomputed tomography and the microarchitectural and mineral properties were obtained for both subchondral plate and trabecular bone. Microarchitecture and material and apparent densities of subchondral bone in OP were similar to regions with early cartilage degeneration but different from regions with advanced cartilage degradation in OA femoral heads. Subchondral trabecular bone was more mineralized than subchondral plate in both OP and OA, and this compartmental difference varied by severity of cartilage degradation. Furthermore, the relationship among trabecular bone volume fraction, tissue mineral density, and apparent bone density was similar in OP and different stages of OA. Subchondral bone microarchitecture and mineral properties in OP are different from OA in a regionalized manner in relation to stages of cartilage degeneration. Both regional and compartmental differences at structural, material, and cellular levels need to be studied to understand the transition of OA subchondral bone from being osteoporotic to sclerotic.


| INTRODUCTION
Osteoarthritis (OA) and osteoporosis (OP) are two of the most common skeletal diseases in the ageing population. 1 The relationship between them has been a topic of discussion for decades. OA is characterized by cartilage degradation, subchondral bone sclerosis, synovial inflammation, and osteophyte formation at diarthrotic joints. 2 OP is characterized by systemic loss of bone due to unbalanced bone resorption, leading to an increased risk of fracture. 3 Mechanism(s) of initiation and progression of OA remains unclear, but the interactions between cartilage and subchondral bone have been suggested to play a significant role. 4 In the 1970s, Radin et al. 5 hypothesized that sclerotic and stiff subchondral bone increases mechanical stress in cartilage and may be the cause of cartilage degradation. However, many studies reported that increased bone turnover and resorptive bone changes, rather than sclerosis, characterize early stage OA and/or promote disease progression. 2,[6][7][8][9] Furthermore, the hardness of subchondral bone at tissue level has been reported to be reduced in OA compared with normal 10,11 and even with OP. 12 Such reduction has been attributed to increased bone remodeling and compromised matrix mineralization. 4,10,[13][14][15] Historically, OA and OP are thought to be inversely related as they rarely occur in the same patient, 16 and the presence of OA is generally associated with higher systemic bone mineral density. 1 One hypothesis is that more compliant subchondral support due to bone loss in OP protects the overlying cartilage. 14,15,[17][18][19] However, the fact that increased remodeling and bone loss are also present in OA, as mentioned above, suggests a possible overlap between the pathophysiology of the two diseases. 20 This overlap provides the rationale for the use of bone-targeting agents as a potential disease modifying OA drugs (DMOADs). 20 But, various established antiresorptive OP drugs, such as bisphosphonates and estrogen, have failed to show solid clinical evidence of efficacy for OA. 21 These paradoxes point to a more complicated situation when studying the relationship between bone remodeling in OA and OP.
First, OP is a systemic skeletal condition whereas bone changes in OA are mainly within the subchondral and adjacent epiphyseal/metaphyseal regions. 3,22 Bone remodeling in OA and OP need to be investigated in these areas rather than vertebral, iliac, or diaphyseal bone, to provide more relevant information. Second, subchondral bone properties are closely associated with the condition of the overlying cartilage in OA joint. [23][24][25][26][27] Such spatial variation is important as it may indicate the temporal change in OA development. 23,24,27,27,28 Third, microstructure and matrix mineralization of subchondral bone need to be studied simultaneously as the biomechanical properties of bone are determined at both apparent and material levels. 10,12,29 Moreover, the subchondral plate and trabecular bone compartments need to be studied concurrently as they are biologically and mechanically distinct and may have different impacts on the overlying cartilage. 9,30 In our review of the scientific literature, we identified studies investigating one or more aspects of this broad topic. 17,24,31-33 They provided invaluable data contrasting subchondral bone remodeling in OP and OA, but none of them comprehensively addressed the relationship between OP and OA covering the whole criteria discussed above. Therefore, the purpose of this study was to investigate the differences in subchondral bone between OP and OA by comparing the microarchitecture, mineralization, and the correlation between microarchitecture and mineralization in a regionalized and compartmentalized manner, using a coupled macroscopic and microscopic sampling procedure. Seven femoral heads were collected from patients undergoing THA for low-energy fracture of femoral neck, who were diagnosed with OP according to established clinical guidelines (4 male and 3 female, mean age 69.7 ± 5.9 years). 34,35 Patients with secondary OP due to prolonged use of corticosteroids, or on medications that affect bone metabolism were excluded. Inclusion criteria also involved a macroscopic inspection of specimens, as detailed below.

| Macroscopic evaluation and sampling
A macroscopic grading system, based on previously described methods 31,36,37 and our observations, was developed for this study to visually evaluate the severity of cartilage degeneration on the articular surface (Table 1 and Figure 1). For OA group, articular surface of each femoral head was divided and graded on a regional basis from Grade 1 to Grade 5. For the OP group, femoral heads from patients with hip fracture were included in the study only when their articular surface was normal or comparatively normal (Grade 1). After macroscopic evaluation, a steel hollow punch with 4 mm inner diameter was used to extract osteochondral plugs. OA plugs (N = 56) were collected from regions with different macroscopic Grades (one plug from each Grade) ( Figure 1B-E). Macroscopic Grade 5 was not included in the study as the integrity of subchondral plate is compromised when bone is exposed. 38 Regions with bone cysts were also avoided. OP plugs (N = 21) were collected from three anatomical sites: anterior, posterior, and superior ( Figure 1A). Plugs were kept frozen until micro-CT scanning. Before micro-CT scanning the macroscopic grade of each plug was re-evaluated to investigate intra-observer variability.

| Microscopic evaluation and histology
After micro-CT scanning, plugs were fixed in formalin, decalcified, and embedded in paraffin. Osteochondral tissue sections with 7 µm thickness were cut. Four sections from each plug were randomly picked and stained with safranin O-fast green. The microscopic severity of cartilage degeneration was evaluated using the OARSI histopathology grading system, 36 with modifications to suit the localized sampling procedure (Table 2 and Figure 1). Two examiners, blinded to sample origins and macroscopic grades, scored sections independently. The scoring was carried out using the advanced grading with subgrade of 0.5 as shown in Table 2. Re-evaluation was accomplished with 8-week interval for intra-observer variability analysis. There were no samples with Grade 5-6.5 since regions showing complete loss of cartilage and osteophytes were excluded.
Sections were viewed and photographed with DM5500 microscope and digital camera (Leica Microsystems).

| Mineral density analysis
Two calcium hydroxyapatite phantoms of known mineral density (0.25 and 0.75 g/cm 3 ) were scanned and reconstructed under the same conditions described above. The X-ray attenuation coefficient values of these phantoms were recorded and a linear calibration equation was used to calculate the mineral density of samples. The apparent density, or volumetric bone mineral density (BMD), was defined as mineral density over the total volume of the ROI, including bone and marrow space. 39 It was measured for trabecular bone only. The material density, or tissue mineral density (TMD), was measured as mineral density over the volume of bone only. 39 TMD reflects the mean degree of bone matrix mineralization 25,40 and was measured for both subchondral plate and trabecular bone. For measurement of TMD, the outer layer (1 voxel in thickness) was removed from the bone surface to correct for partial volume effect.

| Statistical analysis
Inter-and intra-observer variability were evaluated by intraclass correlation coefficient (ICC). 41   For cartilage changes in OA, the generalized Grades 1 and 2 (advanced Grade 1-2.5) are considered as early degeneration; generalized Grades 3 and 4 (advanced Grades 3-4.5) are considered as advanced degeneration; generalized Grades 5 and 6 (advanced Grades 5-6.5) are considered as end-stage degeneration.
For detailed nomenclature of histologic descriptions, refer to the Appendix A in Pritzker et al. 36 microscopic grading was evaluated by Spearman's rank test. 41  3 | RESULTS

| Macroscopic and microscopic evaluation
Macroscopic grading was carried out by one observer (Li) Figure S1).
In the following comparisons of bone microarchitecture and mineral properties, for simplicity and due to limited sample size, the advanced grading of OA plugs was generalized to Grades 1, 2, 3, and 4 as in Table 2. In addition, the advanced grading of OP plugs ranged from 0-1.5, and 1-1.5 were treated as acceptable age-related minor degeneration, accordingly the data from the OP group were pooled without further differentiation.

| Mineral densities
The TMD and results of analysis of variance are reported in Table 3.
Results of post-hoc multi-comparisons are depicted in Figure 2A.  Note: Values are mean ± SD. Analysis was first made between OA grades to examine the regional variations in relation to the severity of OA. Then the tests were repeated to include the OP group to compare OP with OA grades. One-way ANOVA was used for parametric data.
Abbreviations: ANOVA, analysis of variance; BMD, bone mineral density; BS/BV, bone surface/bone volume; BV/TV, bone volume/total volume; OA, osteoarthritis; OP, osteoporosis. a Kruskal-Wallis test was used for non-parametric data.
For compartmental comparisons, trabecular bone had a significantly higher TMD than subchondral plate in all study groups (Figure 2A,C).
The BMD of trabecular bone in OA increased with histopathological grading (p < .001; Table 3 and Figure 2B). OA Grades 3 and 4 had a value significantly higher than Grade 1 and/or Grade 2 ( Figure 2B). The trabecular BMD of OP showed no statistically significant difference with OA Grades 1 and 2 but was lower than that of OA Grades 3 and 4 ( Figure 2B). whereas SMI and BS/BV of OP were significantly higher than those of OA Grade 3 and/or Grade 4 ( Figure 3). An exception was Tb.Sp which showed a downward trend with increased OA grades, but this was not statistically significant (p = .052. Table 3). Tb.Sp showed no significant difference between OP and OA grades ( Figure 3).

| Microarchitecture
Subchondral plate thickness (Pl.Th) in OA increased significantly with worsening cartilage histopathology (p < .001; Table 3 and Figure 3). The OP group had a significantly thinner plate compared to OA Grades 2, 3, and 4 (p = .017, p < .001, p < .001, respectively; Figure 3). No statistically significant differences were found between all study groups in terms of Pl.Po (Table 3 and Figure 3).

| Correlation between BV/TV and mineral density
The results of the linear regression analyses among BV/TV, TMD, and BMD are summarized in Table 4 and depicted in  Table 4).

| DISCUSSION
In this study, we used a combined macroscopic and microscopic sampling procedure which permitted a better representation of the material density or TMD, decreased in OA compared to healthy and OP. 13,15,17 More recently it was shown that trabecular bone was more hypo-mineralized in regions with complete cartilage loss compared to those still covered by cartilage in OA specimens. 24,25 Our results are consistent with these studies and additionally showed that the changes were stagewise in terms of cartilage degeneration and already significant before exposure of subchondral bone, thus providing further evidence that subchondral bone mineralization and cartilage degradation are intrinsically related in OA.
The matrix mineralization of trabeculae in OP was shown to be decreased, 42 increased, 43 or unchanged 15 compared to control by different studies. The conflicting results are likely to be due to site of tissue sampling (spine, ilium, femur, etc.) and the stage of disease. 44,45 In this study, we focused on the subchondral bone of hip joint in established OP and OA and showed that trabecular mineralization of OP was similar with that of regions with early cartilage degradation in OA, and significantly higher than that of regions with advanced degradation. This cartilage degradation based regionalized observation may partly explain why the tissue hardness of trabeculae, measured by nanoindentation, was found to be different between OP and OA by one study 12 but indifferent by another study. 17 In contrast to trabecular bone, there were subtle variations in the TMD of subchondral plate in relation to local severity of cartilage degradation. This is in line with Aspden et al. 14 who showed that there was no site-related variation in the material density of subchondral plate, without accounting for condition of the overlying cartilage. In another study 15 they also reported values for trabecular bone but a comparison with subchondral plate was missing. Our study provided new data showing that the TMD of the subchondral plate was significantly lower than that of trabecular bone in OP and OA, and this compartmental difference varied significantly between regions with varying severity of cartilage degradation. The compartmental difference of TMD is consistent with Cox et al. 25 who showed decreased subchondral trabecular bone mineralization toward articular surface. Our results also indicate that such depthrelated difference is not OA-specific, but also exists in OP. However, Cox et al. reported that the cartilage degradation-related regional difference of trabecular TMD was larger toward articular surface, whereas the regional difference of subchondral plate TMD in our study was not significant. Taken together, our results support the concept that the subchondral plate and trabecular bone are biologically and mechanically distinct, 2,38 and provide further evidence for the view that they respond differently during the progression of OA. 9,30 The regional comparisons of microarchitecture in OA knees in relation to the stage of cartilage degeneration have been reported by several previous studies. 26,27 Overall, our data from hip OA are consistent with these studies and provide further information regarding how these stagewise differences are related to the properties of subchondral bone in OP. OP and early cartilage degeneration regions in OA showed similarly reduced trabecular bone volume and deteriorated trabecular structure compared to advanced degeneration regions. The difference in subchondral plate thickness between OP and OA was already significant from OA Grade 2, which seems to agree with that in non-traumatic type of OA the change in subchondral plate precedes the change in trabecular bone and couples more closely with cartilage degeneration. 46,47 Our regionalized analysis of subchondral bone may suggest that before a comparatively normal region progresses to end-stage OA with degraded cartilage and subchondral sclerosis, there is a period when subchondral bone is osteoporosis-like in terms of both microarchitecture and matrix mineralization, in both subchondral plate and trabecular compartment ( Figure 2C). This supports the hypothesis that subchondral bone remodeling in OA is a biphasic procedure with a transition from favoring resorption at early stage to favoring formation at late stage. 9,21 We suggest that this transition may be caused by increased shear/tensile stress concentrated at the junction between sclerotic and porotic regions. The elevated mechanical stress perceived by osteocytes and osteoblasts in this area may trigger the phenotype change described by previous studies, [48][49][50] switching osteoblasts from pro-resorption to pro-formation. In the meantime, abnormal production of collagen homotrimers and changes in mineralization-related proteins lead to disrupted mineral deposition and consequently hypo-mineralization. 49,50 This theory parallels the hypothesis that the elevated shear stress, rather than compressive stress, is the cause of cartilage deformation and degradation. 4,9 The biomechanical properties of bone are characterized by a series of parameters at both microstructural and tissue material levels. 51 Of these parameters, BV/TV and TMD are considered the most important for trabecular bone. 10,15,32 They are the determinants of apparent density which is often clinically measured by Dual Energy X-ray Absorptiometry (DXA) or Quantitative CT (QCT) as areal or volumetric BMD, 39,51 and are assumed to have a mutually adaptive relationship to define the overall stiffness of bone. 29,45 Compared to clinical imaging techniques, micro-CT benefits from higher resolution and is able to provide assessment of volumetric BMD, TMD, and more accurate measurement of bone microarchitecture simultaneously, enabling investigation of the relationships among them. Our study showed that there was an inverse correlation between BV/TV and TMD and between BMD and TMD of trabecular bone in OA. Together they indicate that in OA the decreased matrix mineralization can be over-compensated by the increased bone volume and lead to increased apparent density. This is consistent with previous studies showing that decreased mineralization in OA trabecular bone only compromised but did not completely abolish the increase in bone strength. 1,15 The result is also consistent with the report that a 4%-6% decrease in TMD in sclerotic bone samples was responsible for only a 4%-9% increase in BV/TV, much less than the actual change (69%) in BV/TV. 25 Another interesting observation from our study is that the pattern of the mutual correlations among BMD, BV/TV, and TMD in OA remained the same when analyzed separately by the microscopic grades. This finding was unexpected as bone remodeling in each region was expected to be different, as discussed above and in. 23,50 Also, a study reported a nonsignificant correlation between BV/TV and TMD in end-stage OA subchondral trabecular bone with cysts and no cartilage coverage. 52 One possible explanation for our finding is that the mineralization of bone is affected by a complex mechanism involving not just remodeling rate, but also remodeling balance and mineralization kinetics. 53 We have also shown that the correlations among BV/TV, TMD, and BMD discussed above also exist in the OP group. These correlations may suggest that loss of trabecular bone in OP is associated with, but not compensated by the increased mineral content, leading to a lower apparent density. This is consistent with the findings that the decreased trabecular bone apparent density in OP was accompanied by stronger and denser trabeculae. 54 The increasing TMD in OP trabecular bone can be a result of active response of osteocytes and osteoblasts to counteract the decreasing stiffness caused by the reducing bone volume, 29,45 or a nonspecific phenomenon as the bone surviving resorption was inner, older, and thus better mineralized trabecular laminae. 29 A limitation of this study is that measurements were made at a single time point, representing spatial differences rather than temporal changes. However, the progression of OA can be indicated by the regionalized progression of cartilage degeneration. [23][24][25]28 Further studies of specimens from young and age-matched healthy controls, and subjects with early diseases would be required to provide reliable data on disease progression. Another limitation of this type of this study is the assumption of independence of samples, 15,24,25 when multiple plugs were sampled from the same specimen. We made this assumption because the plug collection was based on the condition of the overlying cartilage, and the histological evaluation confirmed that our macroscopic sampling procedure well represented and differentiated between varying degrees of cartilage degeneration. This assumption is supported by the intra-sample variation of bone parameters in relation to regional severity of cartilage degradation observed in this and other previous studies. [23][24][25][26] Moreover, the anatomical distribution of cartilage degeneration and subchondral bone properties across an OA femoral head varied between our specimens, and this is consistent with a previous study showing that there is no anatomical site in the OA specimens for which the bone properties are systematically different. 55 Accordingly, OP plugs were collected from three representative anatomical sites of each femoral head to compare with OA.
It should also be noted that the TMD calculated from a desktop micro-CT is subjected to beam-hardening artifact and only represents the mean degree of mineralization. 25,39,56 In this regard, techniques such as gravimetric measurement, compositional analysis, quantitative backscatter scanning electron microscopy and calibrated monochromatic synchrotron micro-CT imaging can be used in the future to confirm our results. In addition, the current study focused on the microstructural and mineral properties of subchondral bone, further studies involving mechanical testing should be conducted to explore the relationships between these parameters and bone biomechanical features (stiffness, elastic modulus, etc.) in different disease conditions. Finally, the region of interest in this study was confined to the most subchondral areas, whereas the changes in epiphyseal/metaphyseal bone may also contribute to OA pathogenesis and should be investigated in future studies.
In conclusion, the results of our study suggest that the comparison of subchondral bone between OP and OA should take both regional and compartmental differences into account, at both structural and material level. This may be particularly relevant when studying the mechanical interactions between cartilage and subchondral bone in future studies. Longitudinal and cross-sectional studies are needed to better understand the mechanism of the transition of OA subchondral bone from being porous to being sclerotic, which may hold the key to the development of DMOADs targeting bone.