SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES

Objective

The biologic changes associated with osteoarthritis (OA) are incompletely understood. The aim of this study was to elucidate the molecular mechanisms underlying OA progression in an STR/Ort murine model of spontaneous disease.

Methods

Global patterns of gene expression were assessed using microarray analysis of articular cartilage/subchondral bone from the tibial plateaus of STR/Ort mice at 3, 9, and 12 months of age. The age-dependent severity of osteophyte formation and extent of cartilage damage were determined in the corresponding femurs using microfocal computed tomography and the Mankin histologic scoring system. Pathway analysis was used to identify the functions of genes associated with OA progression, and changes in gene expression were confirmed using immunohistochemistry.

Results

Six hundred twenty-one genes were associated with both osteophyte formation and cartilage damage in the STR/Ort joints. Genes involved in the development/function of connective tissue and in lipid metabolism were most significantly enriched and regulated during disease progression. Genes directly interacting with peroxisome proliferator–activated receptor α (PPARα)/PPARγ were down-regulated, whereas those genes involved with connective tissue remodeling were up-regulated during disease progression. Associations of down-regulation of myotubularin-related phosphatase 1 (a phosphoinositide 3-phosphatase involved in lipid signaling) and up-regulation of biglycan (a member of the small leucine-rich protein family known to modulate osteoblast differentiation and matrix mineralization) with OA progression were confirmed by immunohistochemistry.

Conclusion

Since adipogenesis and osteogenesis are inversely related in the developing skeletal tissue, these results suggest that a shift in the differentiation of mesenchymal cells from adipogenesis toward osteogenesis is a component of the OA pathophysiologic processes occurring in the tibial plateau joints of STR/Ort mice.

Osteoarthritis (OA), a disease associated with reduced synovial joint function and increased pain, is a major cause of disability in humans (1). There are no consistently effective methods for preventing OA or slowing its progression, and symptomatic treatments provide limited benefit for many patients. Gross changes in the structure and content of articular cartilage, subchondral bone, synovial membrane, joint ligaments, and tendons have been described for many years in patients with OA (2). However, the molecular changes associated with OA in these tissues have only recently begun to be elucidated (for review, see refs.3–8).

Important features of OA include the degradation of articular cartilage and remodeling of subchondral bone (9). Cartilage damage is thought to be mediated through excess synthesis and release of catabolic factors including proinflammatory cytokines, matrix metalloproteinases (MMPs), and nitric oxide, as well as a reduced synthesis of anabolic factors such as insulin-like growth factor 1 (IGF-1) (10, 11). However, other studies indicate that OA is likely to be a systemic disease, involving stromal cell differentiation and lipid metabolism (12). Indeed, generalized changes in many tissues of the joints have been observed, including increased adiposity (13), muscle weakness (14), and weakening of the anterior cruciate ligament (ACL) (15). Because adipocytes share a common mesenchymal cell precursor with chondrocytes, tenocytes, and osteoblasts (16), a biologic link between lipid metabolism and connective tissue remodeling may be a central component of OA.

In support of this hypothesis, recent studies have shown that up-regulation of peroxisome proliferator–activated receptor γ (PPARγ) signaling is therapeutic in surgically induced OA or type II collagen–induced arthritis in animal models (17, 18). Although PPARγ was originally identified as a key regulator of lipid metabolism and adipocyte differentiation (19, 20), more recent evidence has suggested that activation of PPARγ can also regulate inflammatory responses. For example, PPARγ has been shown to be a negative regulator of macrophage activation (21). Moreover, PPARγ activation inhibits the production of interleukin-1β (IL-1β), tumor necrosis factor α (TNFα), and IL-6 in monocytes (22). In addition, PPARγ agonists can inhibit the production of MMP-13 in human chondrocytes and MMP-1 in human synovial fibroblasts (23, 24).

It has therefore been suggested that PPARγ agonists exert their therapeutic effects on surgically induced or collagen-induced arthritis through the suppression of these inflammatory mediators (17, 18). However, these animal models involve an external mechanism of inflammation in the development of OA, and therefore the disease in these models may not be representative of generalized OA in humans. As such, the current level of understanding has not yielded a satisfactory hypothesis as to how spontaneous OA is initiated or what major signaling mechanisms are involved in the progression of spontaneous, idiopathic OA.

In order to identify the molecular mechanisms underlying the progression of spontaneous OA, we analyzed the changes in gene expression that occur in affected joints during OA progression in the STR/Ort mouse, a strain derived from the common inbred strain STR/1N. STR/Ort animals spontaneously develop histologic lesions resembling those of human OA, with ∼85% of male STR/Ort mice developing the disease in the medial tibial plateau at age 1 year (25). Using genome-wide expression profiling and functional analysis, development and function of the connective tissue and lipid metabolism were shown to be the biologic functions that are most significantly up-regulated and down-regulated, respectively, during OA progression. Furthermore, genes regulated by PPARα and PPARγ were down-regulated in a coordinated manner during disease progression. These results suggest that PPAR signaling is down-regulated during the progression of OA in STR/Ort animals, and that a shift away from adipocyte formation and toward osteoblast differentiation in mesenchymal precursor cells is an important component in this spontaneous, idiopathic model of OA in mice.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES

Collection of joint samples and tissue preparation.

A colony of STR/Ort mice was established from 3 original breeding pairs obtained from Dr. R. M. Mason (Imperial College of Medicine, London, UK) (25). The joints from both hind limbs were collected from 3 groups of male STR/Ort mice at 3, 9, and 12 months of age. After careful disarticulation, the femur and corresponding tibia from each joint were collected for histology and RNA preparation, respectively. The femur was fixed in 4% paraformaldehyde (Fisher Scientific, Fair Lawn, NJ) in phosphate buffered saline (PBS). The corresponding tibia was carefully cleaned of attached ligaments and muscle.

Tibial articular cartilage and subchondral bone were microdissected together by collecting a 200-μm–thick section measured from the cartilage surface of the tibial plateau. From each group of mice, the cartilage/bone sections from the left and the right tibia were quickly frozen and stored in liquid nitrogen until processing for RNA. In addition, the dorsal segments of the ribs, from approximately equal lengths (2 mm) of cartilage and bone, were collected from each mouse. The rib tissues were pooled for RNA preparation.

Mankin histologic scores of cartilage changes.

Semiquantitative histopathologic grading was performed as described previously (9), according to a modified Mankin scoring system established for grading OA changes (26–28). The 5 subcategories of the Mankin score evaluated were structure, chondrocyte number, chondrocyte clustering, proteoglycan content, and subchondral plate and/or tidemark changes. After being scanned by microfocal computed tomography (micro-CT), the femurs were decalcified in 0.5M EDTA, pH 7.6, and embedded in paraffin in the same orientation.

OA develops focally in STR/Ort mice, and it is therefore very difficult to compare the histopathologic characteristics of OA cartilage between mouse joints. This limitation is frequently attributable to the technical inability to collect histologic sections at the same depth of tissue in the joints. We therefore collected 20 sections, each of 5 μm in thickness and 50 μm apart, from each femur. Three toluidine blue–stained sections were carefully selected from each sample on the basis of whether they were of comparable orientation and tissue depth. After microscopic evaluation of each section of cartilage, the modified Mankin score was assessed, as previously described (29). Briefly, scores were assessed for the severity of cartilage surface structural damage (scores 0–10), changes in cellularity (scores 0–4), cell clustering (scores 0–4), pericellular staining (scores 0–4), and matrix proteoglycan staining (scores 0–4), and the maximum score was 26 (28).

Immunohistochemistry.

Tissue sections were deparaffinized in xylene, hydrated in graded ethanol, and then treated with 500 units/ml testicular hyaluronidase (Sigma, St. Louis, MO) at 37°C for 15 minutes. Endogenous biotin and biotin binding activity were blocked with an avidin–biotin blocking kit (Zymed, Burlingame, CA) followed by serum blocking. Tissue sections were then incubated with either anti-human biglycan antibodies (R&D Systems, Minneapolis, MN) or anti-human myotubularin-related phosphatase 1 (MTMR-1) antibodies (Abgent, Bioggio-Lugano, Switzerland), overnight at 4°C.

For immunostaining to detect MTMR1, sections were rinsed in PBS with 0.3% Tween 20 and then incubated with biotin-conjugated anti-rabbit antibodies (Vector, Burlingame, CA) for 30 minutes, followed by streptavidin–horseradish peroxidase conjugate (Zymed) for 10 minutes. These sections were again rinsed with PBS, developed using the aminoethylcarbazole chromogen of the Histostain SP kit (Zymed), and counterstained with hematoxylin. For immunostaining to detect biglycan, sections were rinsed in PBS with 0.3% Tween 20 and then incubated with biotin-conjugated anti-goat antibodies (Vector) for 30 minutes, followed by high-sensitivity streptavidin–horseradish peroxidase (R&D Systems) for 30 minutes. Sections were rinsed and developed to a brown color using 2.5% 3,3′-diaminobenzidine. For controls, the same procedures were carried out in the absence of the primary antibodies.

Osteophyte score.

The left and right femurs were fixed in 4% paraformaldehyde for 24 hours, and immersed in 70% ethanol. The distal regions were scanned by micro-CT (μCT-40; Scanco Medical, Bassersdorf, Switzerland) at a resolution of 12 μm. Osteophytes were scored by 2 independent observers (MP and GAW). A grading system was developed whereby joints were assigned an osteophyte score of 0–5 depending on the number and size of the osteophytes observed.

Gene expression profiling.

Total RNA was isolated from mouse tissues and converted to fluorescently labeled complementary RNA (cRNA) that was hybridized to DNA oligonucleotide microarrays as described previously (30, 31). Briefly, 4 μg of total RNA from each tissue sample was used to synthesize double-stranded DNA through reverse transcription. The cRNA was produced by in vitro transcription and labeled postsynthetically with Cy3 or Cy5. The cRNA derived from individual tibial plateaus (experimental sample) were hybridized against pools of cRNA from the rib tissue derived from the same individuals (reference sample). Separate reference pools were constructed for each time point to account for age-induced changes in gene expression. This hybridization scheme was used to 1) remove the majority of age-related genes, 2) enrich for genes locally regulated in weight-bearing joints (the knee), and 3) normalize the reference values for different rates of disease progression between the left and the right knee joints and between animals within the same age group.

Two hybridizations were done with each pair of cRNA samples, using a fluorescent dye reversal strategy. The microarrays contained 23,564 probes that were representative of genes or expressed sequence tags. Probe sequences were chosen to maximize gene specificity and minimize the 3′-replication bias inherent in reverse transcription of messenger RNA (mRNA). In addition, the microarrays contained ∼100 control probes for quality control purposes. All probes on the microarrays were synthesized in situ with inkjet technology (Agilent Technologies, Palo Alto, CA) (30). After hybridization, arrays were scanned, and fluorescence intensities for each probe were recorded. Ratios of transcript abundance (experimental to control) were obtained following normalization and correction of the array intensity data. Gene expression data were analyzed using Rosetta Resolver gene expression analysis software (version 5.1; Rosetta Biosoftware, Seattle, WA).

Identification of genes associated with disease phenotypes.

In order to identify genes associated with the osteophyte score, joints were placed into 2 groups (group 1 comprising samples with osteophyte scores ≤1; group 2 comprising samples with osteophyte scores ≥4), and an analysis of variance (ANOVA) calculation was performed to identify probes differentially expressed between groups. Correlation analysis was used to identify genes associated with the Mankin score. An estimated false discovery rate (FDR) of 5%, based on 500 permutations of the data, was used to determine the thresholds for significant values in the ANOVA and correlation analyses. The significance of the size of overlap between genes associated with each disease phenotype was calculated using the hypergeometric distribution.

Gene function and network analysis.

Genes identified as being positively or negatively associated with the Mankin score and the osteophyte score were used for network and gene function analyses. These genes comprised the seed set. Locus identification numbers were imported into the Ingenuity Pathway Analysis (IPA) system (Ingenuity Systems, Mountain View, CA), and genes were then categorized based on the published findings regarding biochemical, biologic, or molecular functions. Calculations of the P value for enrichment of gene functions were based on the hypergeometric distribution.

The identified genes were also mapped to interaction networks as described previously (32). Briefly, the construction of interaction networks involves 1) overlay of genes identified as significant from the experimental data onto the IPA interactome, 2) determination of the specificity of connections between genes by calculating the percentage of each gene's connections to other significant genes (networks are grown from genes with the highest specificity connections), and 3) assessment of the significance of the identified networks by determining the probability that a collection of genes with a sample size equal to or greater than the number in the network could be achieved by chance alone. The resulting networks are ranked by score, with a score of 3 indicating that there is a lower than 1 in 1,000 chance that the focus genes are in a network due to random chance. Networks with a ranking score >3 were considered significant.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES

Gross morphologic and histologic features of individual joints.

Male STR/Ort mice were killed for morphology and histology. OA-like pathologic features were surveyed by histologic evaluation of the individual joints from mice at 3, 9, and 12 months of age (n = 15–18 samples per group). Typically, the normal femoral condyles (Figure 1A, panel a) and tibial plateaus (results not shown) of the joints of 3-month-old STR/Ort mice had intact cartilage integrity (Figure 1A, panel a, inset), occasionally accompanied by limited penetration of calcified tissue from the subchondral surface (Figure 1A, panel a, arrows). In contrast, OA-associated lesions could be readily observed in joints from mice as early as 3 months of age (Figure 1A, panel b, inset), including a loss of cartilage cellularity and proteoglycan staining, increase in focal surface damage, bone growth into the cartilage, and growth of small osteophytes (Figure 1A, panel b, arrows). As the animals aged, surface erosion became more severe, accompanied by increases in proteoglycan loss, osteophyte formation, and subchondral bone sclerosis (Figure 1A, panel c). By 12 months of age, complete loss of cartilage and bone eburnation could be observed in the animals (Figure 1A, panel c).

thumbnail image

Figure 1. Evaluation of the progression of osteoarthritis (OA) in STR/Ort mice. A, The pathogenesis of age-dependent OA progression was assessed histologically in normal femoral condyles of 3-month-old STR/Ort animals, revealing typical intact cartilage integrity (Mankin scores 0–1) accompanied by limited penetration of calcified tissue into the articular cartilage (arrows) (a). Inset, Higher magnification view of the boxed area in a. As early as 3 months of age, joints developed more advanced disease, with focal cartilage surface damage (Mankin score 8) and small chondro-osteophytes (arrows) (b). Inset, Higher magnification view of the boxed area in b. At 12 months of age, many joints had significant cartilage thinning (arrows) and osteophytosis, or developed either severe bone sclerosis associated with reductions in bone marrow (BM) space or complete bone (B) eburnation (Mankin score 22) (c). Original magnification × 4 in A (a–c); × 20 in insets of a and b. B, Osteophyte growth (arrows) was assessed by microfocal computed tomography of the coronal (a and c) and axial (b and d) planes of the femur of a 12-month-old CBA mouse as control (a and b) or an age-matched STR/Ort mouse (c and d). The osteophyte score (range 0–5) was assigned based on the overall size and extent of the osteophytes as assessed on all axial plane images for each femur. Asterisks in d indicate subchondral sclerosis.

Download figure to PowerPoint

At each time point, joint damage was assessed by measuring the presence and severity of osteophytes and extent of cartilage damage. Micro-CT scans were performed on individual joints, and presence/severity of osteophytes was scored by 2 independent observers. A grading system was developed whereby joints were assigned a score of 0–5 depending on the number and size of the observed osteophytes. Figure 1B shows micro-CT images of the coronal and axial planes of a typical normal femur from a 12-month-old CBA mouse (Figure 1B, panels a and b) compared with those from an age-matched STR/Ort mouse, the latter of which showed severe osteophyte development (Figures 1B, panels c and d).

Cartilage damage was assessed using the modified Mankin scores for various joint histologic features (as described in Figure 1 and in Materials and Methods). As illustrated in Figures 2A and B, the osteophyte scores and Mankin scores of individual joints showed an age-dependent increase, and the disease progression in individual joints from the same animal appeared to behave in an independent manner. Two-way ANOVA calculations revealed that both the osteophyte score and the Mankin score of joints from mice at 9 months of age and at 12 months of age were significantly increased compared with these scores at 3 months of age (Figures 2A and B).

thumbnail image

Figure 2. Assessment of osteoarthritis-related pathologic markers as a function of increasing age in STR/Ort mice. Osteophyte scores determined by microfocal computed tomography (μCT) (A) and Mankin scores of histologic changes to the cartilage (B) were calculated as described in Materials and Methods. Scores for individual joints are shown separately for the left (L) and right (R) femurs. Broken lines connect data points (open circles) from the same animal. Solid circles with bars represent the mean and SEM for each age group. Triangles represent the scores for animals in which only 1 joint was phenotyped. Raw P values (by analysis of variance) for differences between age groups were calculated as follows: in A, P = 0.004, P = 6 × 10−4, and P = 0.037 for age 3 months versus 9 months, age 3 months versus 12 months, and age 9 months versus 12 months, respectively; in B, P = 0.012, P = 9 × 10−5, and P = 0.044 for age 3 months versus 9 months, age 3 months versus 12 months, and age 9 months versus 12 months, respectively.

Download figure to PowerPoint

Interestingly, disease progression was not observed in all of the joints assessed. No osteophytes were observed in 4 of 18 joints from 3-month-old animals, 2 of 18 joints from 9-month-old animals, and 1 of 15 joints from 12-month-old animals (Figure 2A). This observation is consistent with previous reports of incomplete disease penetration at 1 year of age (25). Whereas 4 of the 18 joints from 3-month-old animals showed absence of osteophyte development, all 18 joints from the 3-month-old animals showed Mankin scores ≥1 (Figure 2B), indicating that mild cartilage damage precedes osteophyte development in the STR/Ort murine disease model. This is consistent with the observations in the rat ACL transection model of OA, in which cartilage damage has also been shown to precede the development of osteophytes (29).

Joint anabolic and catabolic factors.

Data on gene expression were generated from the same joints used to assess disease phenotypes. Joint tissue, comprising articular cartilage and subchondral bone, was obtained from the mouse tibial plateaus for profiling on Agilent Technologies oligonucleotide-based DNA microarrays. Due to the small size of rodent joints and the inherent difficulties in separating articular cartilage from bone, the tibial plateaus that contained both articular cartilage and subchondral bone tissue were microdissected together in a 200-μm section. The tibial tissue section from each animal was processed for profiling studies.

To assess the catabolic and anabolic processes that occur in STR/Ort joints, we first analyzed the expression of candidate genes involved in biologic functions thought to be important for the initiation or progression of OA. We chose candidate genes that are commonly described in the literature, from 4 general categories: 1) cartilage components (Col2a1, Agc1, Hapln1, and Comp), 2) cartilage catabolism (Adamts5, Il1b, and Il6), 3) osteoclast function (Ctsk and Mmp9), and 4) bone anabolism (Igf1 and Tgfb1). Figure 3 shows the regulation of these genes in the STR/Ort joints by age groups.

thumbnail image

Figure 3. Regulation of candidate genes involved in murine osteoarthritis. Genes were assigned to 1 of 4 categories: 1) cartilage components (Col2a1, Agc1, Hapln1, and Comp), 2) cartilage catabolism (Adamts5, Il1b, and Il6), 3) osteoclast function (Ctsk and Mmp9), and 4) bone anabolism (Igf1 and Tgfb1). Data represent the log10 of the ratio of expression in each joint sample relative to the pooled reference sample from mouse rib tissue. Genes that were up-regulated relative to the reference sample are shown in magenta, while genes that were down-regulated relative to the reference sample are shown in cyan. Asterisks denote genes whose average expression in the joints from mice at age 3 months was significantly higher than that in mice at age 12 months (P < 0.05 by analysis of variance).

Download figure to PowerPoint

The expression levels of cartilage components, cartilage catabolic factors, and genes indicative of osteoclast function were significantly higher in the joints of mice at 3 months of age compared with 12 months of age (P < 0.05). This is again consistent with the observations in the rat ACL transection model of OA, in which subchondral bone resorption and cartilage loss are early events during disease progression (9, 29). Igf1 and Tgfb1, which are genes that have been previously implicated in osteophyte development in experimentally induced OA in the mouse and rat (9, 33), were not up-regulated with increasing age in the STR/Ort joints. Since these joints clearly showed an increase in the presence and severity of osteophytes at 12 months of age relative to the findings at 3 months of age (Figure 2A), these data suggest that other molecular mechanisms are contributing to osteophyte development in STR/Ort animals.

Genes significantly associated with joint pathologic changes.

We then performed a genome-wide analysis to identify the genes associated with disease progression in STR/Ort joints. In order to identify the genes associated with osteophyte formation, joints were placed into either group 1 (osteophyte score ≤1) or group 2 (osteophyte score ≥4). An ANOVA calculation was performed, and 2,214 genes were identified as being significantly differentially expressed between groups (P < 0.005); this significance threshold was selected so that the FDR in 500 random permutations was lower than 5%. In order to identify the sequences associated with cartilage damage, we performed a correlation analysis, and 3,624 genes were identified as being significantly correlated with the Mankin histologic score (Pearson's r >0.45, FDR <5%).

Six hundred twenty-one genes were identified as being associated with both the osteophyte score and the Mankin score (P for overlap < 0.001). The expression levels of 331 genes were up-regulated with increasing age, while 290 genes were down-regulated with increasing age (results not shown). These findings suggest that the development of OA in STR/Ort animals involves both the up-regulation and the down-regulation, coinciding with increasing age, of similar numbers of genes associated with disease pathology.

Functions of genes in relation to disease progression.

In order to gain insights into the biologic processes and signaling networks involved in OA progression, we performed analyses of the biologic functions and contributory pathways in the STR/Ort joints, in order to uncover the relationships among genes associated with disease progression in this model. Instead of focusing on individual genes or a group of candidate genes that were previously demonstrated to change with disease progression, we utilized a different approach to allow a more unbiased assessment of the biologic processes underlying disease progression in the STR/Ort model. This approach involved use of the IPA software tool as described in Materials and Methods (see the Ingenuity Systems Web site at http://www.ingenuity.com), which enabled us to identify the gene functions that were statistically significantly enriched among the 621 genes described as being associated with the osteophyte and Mankin histologic scores. Note that others have previously used the IPA tool to identify biologic networks involved in complex processes, including inflammation, glucocorticoid receptor signaling, and cancer (32, 34, 35).

The general function most significantly enriched among the 331 genes up-regulated during disease progression was the development and function of the connective tissue (P = 9.32 × 10−5–4.55 × 10−2). The specific function most significantly associated with the group of up-regulated genes was patterning of bone (P = 9.32 × 10−5 for the genes Bgn, Cdx1, Enpp1, Ggt1, Nog, Ptn, Ptprv, Src, and Wnt3a), which is consistent with the development of osteophytes with increasing age in STR/Ort animals (Figure 2A). The general function most significantly enriched among the group of 290 down-regulated genes was lipid metabolism (P = 8.08 × 10−7–4.93 × 10−2), and modification of fatty acid was the most significantly enriched specific function (P = 8.08 × 10−7 for the genes Acas2, Acox1, Amacr, Ech1, Gpam, Hadha, Hadhb, Phyh, and Slc27a2). These findings suggest that a general down-regulation of genes involved in lipid metabolism may play a key role in OA pathogenesis. The inverse time-dependent regulation of genes identified as being involved in the development and function of the connective tissue and in lipid metabolism is summarized in Figure 4.

thumbnail image

Figure 4. Inverse age-dependent regulation of murine genes involved in connective tissue development/function and lipid metabolism during osteoarthritis progression. Data represent the log10 of the ratio of expression in each tibial sample relative to the pooled reference sample of rib tissue derived from age-matched individual mice. Genes that were up-regulated relative to the reference sample are shown in magenta, while genes that were down-regulated relative to the reference sample are shown in cyan.

Download figure to PowerPoint

Interaction networks.

In order to further assess the biologic processes associated with lipid metabolism during OA progression, we investigated, with the use of the IPA platform, interaction networks formed by the above-described 290 genes identified as being down-regulated in association with disease progression. This system identifies biologic interaction networks among genes of interest by mining published findings regarding biochemical, biologic, or molecular interactions (see Materials and Methods).

Interestingly, 2 networks that contain PPAR transcription factors as central nodes were observed to be highly significant. As shown in Figure 5A, PPARα was identified as a central node of the most significant interaction network identified among the down-regulated genes (significance score of 23). This network includes 3 transcriptional targets of PPARα (Acox1, Fgg, and Acas2, as highlighted in Figure 5A). In addition, a second significant interaction network (significance score of 14) contained PPARγ as a central node (Figure 5B). This network includes 4 transcriptional targets of PPARγ (Fasn, Vnn1, Cyp4b1, and Tpm2, as highlighted in Figure 5B) and 1 gene known to physically interact with PPARγ (Smarcd3). These results suggest that reduction of PPAR signaling is an important mechanism underlying the progression of OA in the STR/Ort mouse. This is supported by the results of previous studies that have suggested a therapeutic role for PPARγ activation in animal models of surgically induced OA (17).

thumbnail image

Figure 5. Networks involving peroxisome proliferator–activated receptor α (PPARα) (A) and PPARγ (B) formed by genes down-regulated during osteoarthritis progression in mice. Genes indicated in green were identified by microarray analysis (comprising the seed set as described in Materials and Methods), while all other genes were brought into the network based on their known interactions with genes in the seed set. The intensity of the color represents the mean level of down-regulation in mice at age 12 months relative to that in mice at age 3 months. Direct PPARα or PPARγ interactions are denoted by blue arrows.

Download figure to PowerPoint

Relative protein expression of MTMR-1 and biglycan.

To provide an additional level of validation of these changes in mRNA expression, we performed a qualitative assessment of the protein levels in the STR/Ort joints by immunohistochemistry. We chose to evaluate protein expression of the signature genes for which immunohistochemistry reagents are available for use in a murine system. Among the signature genes that correlated with OA progression in the STR/Ort mice, the protein expression of MTMR-1 and biglycan, as assessed by immunohistochemical methods, was further verified (Figure 6).

thumbnail image

Figure 6. Immunohistochemical analysis of the relative protein expression levels of myotubularin-related phosphatase 1 (MTMR-1) and biglycan, to illustrate the inverse relationship between adipogenesis and bone matrix formation in mice. A and B, Expression of MTMR-1, a protein involved in lipid signaling, was found to be higher in cartilage of a typical 3-month-old mouse (A) compared with the levels in cartilage of a typical 12-month-old mouse (B), suggesting a down-regulation in lipid signaling. MTMR-1 expression was primarily found in articular cartilage. D and E, Expression of biglycan, a marker of bone formation, increased with age-dependent disease progression, and higher levels of biglycan were detected in the joints of a typical 12-month-old mouse (E) compared with levels in the joints of a typical 3-month-old mouse (D). Biglycan was primarily found in articular cartilage and in cells lining the articular cartilage and the subchondral trabecular bone surface. C and F, Controls for analyses of MTMR-1 (C) and biglycan (F) comprised nonspecific staining in the absence of primary antibodies.

Download figure to PowerPoint

MTMR-1, a phosphoinositide 3-phosphatase involved in lipid signaling, was found to be highly expressed in the joints of 3-month-old mice compared with 12-month-old mice (Figures 6A and B, brown stain). In the joints of 3-month-old mice, MTMR-1 appeared to be highly expressed in the hypertrophic chondrocytes and, to a lesser extent, in chondrocytes near the articular surface (results not shown). MTMR-1 expression was not detected in the ligaments and muscles surrounding the joints. The regulation of MTMR-1 protein expression was consistent with our findings of the general down-regulation of genes involved in lipid metabolism occurring in conjunction with OA progression (as shown in Figure 4).

Conversely, biglycan, a small leucine-rich proteoglycan that plays a critical role in the formation of collagen fibrils, was highly expressed in the joints of 12-month-old mice compared with 3-month-old mice (Figures 6D and E). In the joints of 12-month-old mice, we detected high levels of biglycan protein in the hypertrophic cartilage/bone interphase and in cells lining the surfaces of articular cartilage and subchondral trabecular bone. This increase in protein expression of biglycan supports our observations of the up-regulation of genes involved in connective tissue development and function occurring in parallel with OA progression (also shown in Figure 4).

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES

OA is currently considered to be a complex joint disease in which all tissues in the joints play an important role in disease initiation and/or progression. It has long been suggested that the progression of articular cartilage degeneration is concomitant with intense remodeling of the subchondral bone and increased bone stiffness, leading to abnormal mechanical stress across the overlying cartilage (36, 37). Indeed, increased subchondral bone activity in OA patients, as determined by enhanced uptake of scintigraphic technetium-labeled disphosphonate, has been shown to precede detectable cartilage loss (38), and the bone formation marker osteocalcin was reported to be higher in synovial fluid from patients with severe scintigraphic signals compared with that from patients with mild alterations on knee scans (39). Unlike patients with osteoporosis, OA patients tend to have a high body mass index together with an elevated rate of bone turnover, resulting in increased bone density (40). This suggests that the process of new bone synthesis exceeds degradation in susceptible individuals.

Studying the pathogenesis of OA in humans is hampered by inherent difficulties such as a lack of availability of normal and diseased tissue at early stages of the disease. Therefore, animal models of OA are essential for understanding disease etiology and for the development of effective therapies. Although numerous animal models have been developed and characterized (for review, see ref.41), many of these models involve surgical intervention or inflammatory stimuli to induce disease. In models of OA involving surgically induced joint instability in several species, both bone and cartilage changes occur concomitantly (29). In contrast, in animal models in which a spontaneous OA-like disease develops (resembling human disease), increased bone density and osteoid volume are often more severe than cartilage changes. For example, Dunkin-Hartley guinea pigs (42) and cynomolgous macaques (43) have age-related changes in bone that precede those in cartilage. Evidence demonstrating that subchondral bone remodeling is linked to cartilage destruction in both humans and animals is well accepted; however, the mechanisms by which the changes in subchondral bone influence articular cartilage are incompletely understood.

In this study, we sought to elucidate the molecular mechanisms underlying disease progression in the spontaneous STR/Ort mouse model. STR/Ort mice share physical characteristics similar to those in human subjects with OA, such as high body weight and high bone mineral density. In addition, a number of biochemical features of OA in STR/Ort mice are similar to the changes observed in human disease, such as matrix proteoglycan depletion by MMPs and aggrecanase (25). Due to recent progress in mouse genomics, the STR/Ort model provides a unique opportunity to investigate the events associated with the initiation and progression of spontaneous OA.

We used gene expression profiling to understand how biologic pathways change with disease progression in the STR/Ort model. By associating gene expression changes with cartilage damage and micro-CT scores in SRT/Ort joints, we were able to identify the genes that were up- or down-regulated with disease progression. Rather than focusing on individual candidate genes, we analyzed biologic functions in an unbiased manner by leveraging informatics tools to identify ascribed biologic functions that were statistically significantly associated with the disease. In this way, we can build a more pathway-centric view of disease biology that leverages information across many genes and is not reliant on individual genes that are chosen for analysis based on preexisting knowledge of other disease models.

This study is the first to provide evidence that altered lipid metabolism through a reduction in PPAR signaling is an important component of spontaneous, idiopathic OA. While PPARγ and PPARα themselves were not found to be significantly altered at the mRNA level, multiple direct targets of the PPARs were altered. We speculate that there is a non–mRNA-based means of down-regulating the activity of the PPARs, for example, a posttranslational modification or a binding to other coactivators or corepressors that alters the activity of PPAR signaling without actually altering the mRNA for PPARs. The PPARγ agonist rosiglitazone has recently been shown to be therapeutic for surgically induced OA in guinea pigs (17), an effect hypothesized to be mediated through the inhibition of proinflammatory signals in the affected joint.

While no direct link between PPARα and OA has been reported, PPARα has also been shown to have antiinflammatory properties (44, 45). Therefore, one possible explanation for the relationship between PPAR signaling and OA progression is that decreased PPAR signaling leads to an elevation of inflammation in the joint microenvironment that favors catabolic signals over anabolic signals, resulting in the observed cartilage damage and osteophyte formation. An alternative explanation involves the developmental link between adipogenesis, chondrogenesis, and osteogenesis. These cells share a common mesenchymal cell precursor that can be induced to differentiate into one of these cell types in vitro by adjustment of the culture microenvironment (46, 47). It is thus possible that an early event in the initiation and progression of OA is a preferential shift toward osteoblastogenesis resulting from the down-regulation of PPAR signaling. This is supported by the observed increase in subchondral bone formation and sclerosis associated with disease progression in humans and animal models of OA.

Cumulative evidence suggests that the dysregulation of subchondral bone metabolism is different in OA, possibly due to an altered osteoblast phenotype. Indeed, osteoblasts isolated from the subchondral bone of patients with OA demonstrated altered phenotypes (48, 49). In comparison with normal osteoblasts, OA osteoblasts produce more alkaline phosphatase, osteocalcin, IGF-1, urokinase plasminogen activator, cytokines, and eicosanoids, including IL-1β, IL-6, prostaglandin E2, and leukotrienes. All of these factors could promote subchondral bone remodeling and are also involved in deposition and turnover of matrix. Because of the development of microcracks, vascular channels or neovascularization may provide a link between subchondral bone tissue and cartilage, potentially enabling these factors to influence the abnormal metabolism of articular chondrocytes and remodeling of OA cartilage.

Currently, no single affector responsible for osteoblast-induced cartilage degradation has been identified. Unlike in other animal models of experimentally induced OA, IGF-1 and transforming growth factor β1 (TGFβ1) were not associated with disease phenotypes in our STR/Ort model, as determined by microarray profiling and reverse transcription–polymerase chain reaction analysis on the same RNA samples used in profiling studies (results not shown). This suggests that alternative mechanisms are important for the development of spontaneous OA in the STR/Ort model. Nevertheless, even with the lack of observed changes in IGF-1 and TGFβ1 mRNA levels during disease progression, we cannot rule out the possibility that posttranscriptional modifications or storage of these growth factors in the extracellular matrix and release via osteoclastic bone resorption during subchondral bone remodeling are alternative mechanisms through which IGF-1 and TGFβ1 might affect OA. However, their lack of association with disease progression is consistent with recent reports that IGF-1 and TGFβ1 mRNA levels are not different between OA and non-OA human bone (50).

In summary, the involvement of bone formation in OA initiation and progression has been recognized for many years (36), and our previous work in the rat ACL transection model of OA supports an early role of bone remodeling in disease development (9, 29). Based on the data presented herein, we hypothesize that a shift of mesenchymal cell differentiation from adipogenesis toward osteogenesis in the subchondral region is an important component of the pathogenesis of spontaneous OA. Due to the observed inverse relationship between lipid metabolism and matrix remodeling in this spontaneous disease model, it is likely that up-regulation of PPAR signaling abrogates OA progression by inhibiting early bone formation, which constitutes a therapeutic strategy that could be applicable to human OA.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES

Dr. Duong had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Watters, Cheng, Zhuo, Hayami, Phillips, Duong.

Acquisition of data. Watters, Pickarski, Wesolowski, Zhuo, Hayami, Wang.

Analysis and interpretation of data. Watters, Cheng, Pickarski, Wesolowski, Zhuo, Hayami, Wang, Szumiloski, Phillips.

Manuscript preparation. Watters, Pickarski, Duong.

Statistical analysis. Watters, Cheng, Szumiloski, Phillips.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES