To analyze genome-wide changes in chondrocyte gene expression in a surgically induced model of early osteoarthritis (OA) in rats, to assess the similarity of this model to human OA, and to identify genes and mechanisms leading to OA pathogenesis.
OA was surgically induced in 5 rats by anterior cruciate ligament transection and partial medial meniscectomy. Sham surgery was performed in 5 additional animals, which were used as controls. Both groups underwent 4 weeks of forced mobilization, 3 times per week. RNA was extracted directly from articular chondrocytes in the OA (operated), contralateral, and sham-operated knees. Affymetrix GeneChip expression arrays were used to assess genome-wide changes in gene expression. Expression patterns of selected dysregulated genes, including Col2a1, Mmp13, Adamts5, Ctsc, Ptges, and Cxcr4, were validated by real-time polymerase chain reaction, immunofluorescence, or immunohistochemistry 2, 4, and 8 weeks after surgery.
After normalization, comparison of OA and sham-operated samples showed 1,619 differentially expressed probe sets with changes in their levels of expression ≥1.5-fold, 722 with changes ≥2-fold, 135 with changes ≥4-fold, and 20 with changes of 8-fold. Dysregulated genes known to be involved in human OA included Mmp13, Adamts5, and Ptgs2, among others. Several dysregulated genes (e.g., Reln, Phex, and Ltbp2) had been identified in our earlier microarray study of hypertrophic chondrocyte differentiation. Other genes involved in cytokine and chemokine signaling, including Cxcr4 and Ccl2, were identified. Changes in gene expression were also observed in the contralateral knee, validating the sham operation as the appropriate control.
Our results demonstrate that the animal model mimics gene expression changes seen in human OA, supporting the relevance of newly identified genes and pathways to early human OA. We propose new avenues for OA pathogenesis research and potential targets for novel OA treatments, including cathepsins and cytokine, chemokine, and growth factor signaling pathways, in addition to factors controlling the progression of chondrocyte differentiation.
Osteoarthritis (OA) is the most common degenerative joint disease in the world. Although many environmental and behavioral factors have been correlated with the onset of OA, its etiology is largely unknown. Our understanding is confounded by the fact that OA appears to be affected differently by various case-specific influences. Secondary OA arises from an initial trauma (such as a ligament tear), resulting in joint instability, unnatural articulation, and eventual development of histopathologic abnormality (1, 2), and commonly affects people in their thirties and forties (3). Less obtrusive causes, including genetic predisposition, precipitate idiopathic or primary OA (4, 5).
OA is characterized by articular cartilage degradation. The cells of articular cartilage (chondrocytes) produce and maintain their surrounding extracellular matrix (6, 7). Cartilage homeostasis relies on the controlled catabolism of matrix proteins, such as type II collagen and aggrecan, by proteolytic enzymes of the matrix metalloproteinase (MMP) and aggrecanase families (8, 9), and subsequent replacement with new protein synthesized by chondrocytes (10, 11), resulting in a balance between anabolism and catabolism. With the onset of OA, the balance shifts toward degradation.
One possible mechanism that might precipitate this shift in cartilage homeostasis involves a change in chondrocyte phenotype. Articular chondrocytes are normally prevented from progressing to hypertrophy by mechanisms that are not well understood. It has been proposed that loss of this inhibitory control may mediate the degradation observed in OA (12, 13). During chondrocyte hypertrophy, the expression levels of genes such as MMP-13 and alkaline phosphatase are increased (14). These genes promote cartilage degradation and calcification. While hypertrophic differentiation facilitates development via endochondral ossification at the growth plate, it becomes pathologic when ectopically recapitulated in articular cartilage.
Thus, it is necessary to understand the collection of factors that are disrupted in OA in order to elucidate which of them promote chondrocyte hypertrophy or other mechanisms involved in cartilage breakdown. Of particular interest is the identification of molecular changes during the onset of OA, since this would allow the development of strategies for early detection and intervention. Genome-wide analyses of dysregulated genes in early OA are a step toward achieving this understanding. Although recent reports have described genome-wide microarray analyses of human OA samples (15, 16), including some changes during early to moderate OA, human samples from early stages of OA are largely inaccessible and are characterized by substantial heterogeneity. This complicates data analyses and increases the risk of missing crucial genes. In addition, putative therapies require preclinical investigation in models of early human OA. Accordingly, gene expression studies also need to be performed in appropriate animal models.
Numerous animal models of OA have been described, each with advantages and disadvantages (17–19). Due to the heterogeneity of human OA, each of these models likely reflects only a subset of cases. For example, anterior cruciate ligament transection (ACLT), with or without partial meniscectomy, mimics posttraumatic (secondary) OA but is less similar to other forms. In contrast to spontaneous OA (observed in certain strains of mice and guinea pigs), however, surgical induction of OA allows tight control of disease onset and progression, facilitating the identification of temporal developments. We previously demonstrated this in a pilot study of the surgical model used here, in which we identified stages of OA (20).
Rodent models are particularly useful for genetic studies of OA pathogenesis because of advanced genome annotation, availability of genetic tools (e.g., whole-genome microarrays), and low housing costs. The rat, in particular, provides the additional advantage of a body size that facilitates surgical manipulation and yields sufficient joint tissue for molecular analyses. Moreover, recent studies have shown that articular chondrocyte gene expression profiles in intact areas are largely similar to those in damaged areas (16), indicating that analysis of total knee cartilage (required for genetic studies in small animals) is an accurate method of identifying genome-wide changes in articular chondrocyte gene expression.
Several reports have described the effects of ACLT in rats. While those authors have investigated histopathologic abnormalities using either ACLT alone (19, 21), ACLT with partial meniscectomy (22, 23), or ACLT with medial collateral ligament transection (24, 25), none have investigated genome-wide changes in gene expression in early or late stages of OA in a rat model.
Recently, we completed a study characterizing OA pathogenesis in the ACLT with partial meniscectomy model, which was used in the present study, and assessed the effects of forced mobilization (20). We concluded that this model more closely resembles human OA when forced mobilization is used, and that 4 weeks after surgery this model is reminiscent of early human OA. Thus, in the present study, we made use of this recently developed model of OA, in combination with microarray technology, with the goal of identifying genes and mechanisms involved in degradation of articular cartilage. In addition, we compared gene expression profiles with our earlier microarray data sets on hypertrophic chondrocyte differentiation (14) in order to examine similarities between OA and cartilage development. This study was undertaken to expand our current understanding of OA pathogenesis in surgical models of OA and to provide relevant insight into the factors and mechanisms involved in early OA pathogenesis.
MATERIALS AND METHODS
Surgery was performed on the right knees of male Sprague-Dawley rats weighing 300–325 gm (Charles River, St. Constant, Quebec, Canada). Anesthetic (50% ketamine [100 mg/ml], 25% xylazine [20 mg/ml], 10% acepromazine [10 mg/ml], and 15% saline [0.9% solution]) and Trisbrissen antibiotic (Schering Canada, Pointe Claire, Quebec, Canada) were both administered at a dose of 100 μl/100 gm body weight. The animals were randomly assigned to 2 groups. The first group underwent ACLT and partial medial meniscectomy via an incision on the medial aspect of the right knee joint capsule, anterior to the medial collateral ligament. Surgical treatment induced pathologic changes of OA in the ipsilateral joint; the left knee was the contralateral joint. The second group underwent a sham operation, via a similar incision in the right joint capsule, without ACLT or partial medial meniscectomy. All animals underwent 30 minutes of forced mobilization as previously described (20), 3 times per week. Cartilage for microarray analysis was harvested from these 10 animals 28 days after surgery. Additional animals underwent the same surgical protocol and were killed 2, 4, and 8 weeks after surgery for histologic analysis and 2 weeks and 4 weeks after surgery for RNA isolation. This study was approved by the Animal Care and Use Committee at the University of Western Ontario.
RNA isolation and preparation.
Articular cartilage was dissected from each knee, cleaned of noncartilaginous tissue, and immersed in QIAzol (Qiagen, Mississauga, Ontario, Canada). It was necessary to pool cartilage from the femoral condyles and tibial plateaus to obtain 15 mg of tissue per sample, yielding enough RNA for subsequent analyses (500–800 ng per sample). After homogenization, total RNA was isolated using the Lipid Tissues Mini Kit, according to the instructions of the manufacturer (Qiagen). RNA quantity was assessed using the RiboGreen Assay (Molecular Probes, Burlington, Ontario, Canada), and RNA quality was confirmed using an Agilent 2100 Bioanalyzer (Agilent, Palo Alto, CA).
Total RNA (100 ng per sample) from the articular cartilage of 5 ipsilateral, 5 contralateral, and 5 sham-operated knee joints was subjected to 2 rounds of amplification, followed by labeling and hybridization to Affymetrix RAE230_2.0 GeneChips (Affymetrix, Santa Clara, CA) containing 31,099 probes. Each joint comprised one sample that was hybridized to a separate chip (n = 5 per condition). Sample amplification, labeling, hybridization, and detection were carried out, according to the instructions of the manufacturer (http://www.affymetrix.com/support/technical/manuals.affx), at the London Regional Genomics Center (London, Ontario, Canada).
Raw data gene expression files from Affymetrix GeneChips were imported into GeneSpring, version 7.2, software (Silicon Genetics, Redwood City, CA). GC-robust multichip analysis preprocessing was performed. Raw data transformation set values <0.01 at 0.01, per-chip normalization was set at the 50th percentile, and per-gene normalization was set to the median and to values in control/sham samples. Data sets from sham-operated samples were assigned to the “normal” treatment group, and thus defined baseline expression for each probe. The remaining data sets (samples from contralateral and ipsilateral joints) were assigned to “diseased” treatment groups, averaged, and used in subsequent analysis. All data were interpreted using the log-ratio setting. From the starting list of 31,099 probes, 17,597 probes were determined to have a reliable signal, using the GeneSpring 7.2 SG1a-1 signal intensity quality control script. The script was adjusted to require signal intensity above a threshold of 50 in ≥2 of the 3 conditions. The data were then passed through a parametric Welch one-way analysis of variance (ANOVA) script, with P values less than 0.05 considered significant, which reduced the list to 3,877 probes. Post hoc Bonferroni multiple comparison testing was performed to identify statistically significant changes in expression between sham-operated samples and contralateral samples, and between sham-operated samples and ipsilateral samples, with P values less than 0.05 considered significant. Differentially expressed genes exhibited similar changes in samples from the same experimental condition. GeneSpring 7.2 software was used to create gene lists based on fold change.
Principal components analysis (PCA).
GeneSpring 7.2 software was used to perform PCA of conditions, using the list of 3,877 probe sets demonstrating significant signal intensity in each sample, as well as the list of 1,619 probe sets that demonstrated a change of 1.5-fold or higher between OA samples and control samples.
Raw data CEL files were imported into BRB ArrayTools (http://linus.nci.nih.gov.proxy1.lib.uwo.ca:2048/BRB-ArrayTools.html) developed by Dr. Richard Simon and Amy Peng Lam. After base 2 logarithm transformations, data were normalized by centering each array using the robust multichip average algorithm. Intersection of the secondary lists resulted in 1,437 probe sets that demonstrated a consistently strong signal in each sample. These sets were used for agglomerative hierarchical clustering analysis in BRB ArrayTools, using centered metric correlation and average linkage.
After probe sets were filtered using the criterion of a minimum 1.5-fold change in differential gene expression between sham-operated and ipsilateral cartilage, the resulting list containing 1,619 probes was used in gene ontology analysis. Categorized lists were generated based on Rattus norvegicus annotations for biologic process, cellular component, or molecular function gene ontologies using FatiGO software (26, 27). The percentage of the total number of annotated probes attributed to each category was calculated.
To quantitatively determine relative gene expression in cartilage RNA samples, real-time PCR was carried out, as previously described (28). The reactions were prepared with the TaqMan One-Step Master Mix kit (Applied Biosystems, Foster City, CA). TaqMan GAPDH (forward primer 5′-GAAGGTGAAGGTCGGAGTC-3′, reverse primer 5′-GAAGATGGTGATGGGATTTC-3′, probe JOECAAGCTTCCCGTTCTCAGCC-TAMRA) control reagents were used as the internal control. The target primer/probe sets for all tested genes were purchased as TaqMan gene expression assays (Applied Biosystems). An ABI Prism 7900 HT Real-Time PCR system (PerkinElmer, Emeryville, CA) was used to detect amplification over 40 cycles for these experiments. Five independent RNA samples (from animals other than those used in the array experiments) were assayed from each treatment 2 weeks and 4 weeks after surgery, each in quadruplicate. In each experiment, a reaction without template RNA was used as a negative control. All relative expression values were calculated using the ΔCt method, normalized to GAPDH expression, and expressed in arbitrary units relative to the expression values in the sham-operated group (set at 1). One-way ANOVA was performed to determine the statistical significance of the differences between the means for each treatment type. Tukey's post hoc comparison was performed to compare the means for all treatment types. Expression values are shown as the mean ± SEM. P values less than 0.05 were considered significant. Analyses were carried out using GraphPad Prism, version 4 (GraphPad Software, San Diego, CA).
Immunofluorescence and immunohistochemistry.
Knee joints were obtained from rats 2, 4, and 8 weeks after surgery. Tissue samples were fixed via intracardial perfusion with 4% paraformaldehyde and dissected. The joints were demineralized in 1% EDTA/glycerol for 4–5 weeks and embedded in paraffin. Sagittal sectioning of the decalcified knees was performed at the Robarts Research Institute Molecular Pathology Laboratory (London, Ontario, Canada). Six-micrometer sections from the medial compartment of each joint were used for immunohistochemical analyses. Primary antibody against MMP-13 (Cedarlane, Hornby, Ontario, Canada) and fluorescein isothiocyanate–conjugated secondary antibody were used to detect MMP-13 within articular cartilage of ipsilateral, contralateral, and sham-operated knee joints 4 weeks after surgery. TOTO-3 iodide (Molecular Probes) was used as a nuclear counterstain. Image development and confocal microscopy were performed using a Zeiss LSM 510 META microscope and software (Carl Zeiss, Toronto, Ontario, Canada). These experiments were repeated 3 times with similar results; no signal was detected in samples without primary antibody. For immunohistochemistry, sections from 2, 4, and 8 weeks postsurgery were probed with anti–cathepsin C or anti-CXCR4 primary antibodies (Abcam, Cambridge, MA) and horseradish peroxidase–conjugated secondary antibodies. Colorimetric detection with diaminobenzidine substrate (Dako, Carpinteria, CA) was carried out for equal time periods for each section. Experiments for each protein were carried out on sections from ≥3 different animals, with reproducible results.
Assessment of microarray gene expression profiles in articular cartilage.
Genome-wide changes in gene expression were analyzed 4 weeks postsurgery because the ipsilateral knee showed indications of early OA at this time point. For example, surface abrasions, edema of the superficial zone, and shallow fissures were present in the ipsilateral joint but not in the sham-operated or contralateral joints (Figure 1A). RNA from the articular cartilage of sham-operated (control) knees, contralateral knees, and ipsilateral (OA) knees was hybridized to Affymetrix GeneChips. Data analysis with GeneSpring 7.2 resulted in a set of 3,877 statistically significant differentially expressed probes (P ≤ 0.05). To identify which condition(s) were responsible for variations in gene expression, we performed PCA using GeneSpring 7.2 and identified 3 clusters. The ipsilateral cluster accounted for 62.83% of the variation, while the contralateral and sham-operated clusters accounted for 8.92% and 6.07% of the variation, respectively. These results indicated that all 3 conditions had distinct gene expression profiles and that most of the differences in gene expression occurred in the ipsilateral condition.
Unsupervised clustering was used to determine whether gene expression profiles distinguished samples in one group from those in the other groups (Figure 1B). Clustering initially resulted in 2 classes of samples. All ipsilateral samples clustered in one class, and all sham-operated samples in the other. Interestingly, 3 of the 5 contralateral samples clustered with ipsilateral samples, while the remaining 2 contralateral samples clustered with sham-operated samples. The first subcluster distinguished all ipsilateral samples from all contralateral samples, while the 5 sham-operated and 2 contralateral samples were not distinguished from one another. Surgery was the most important factor in determining where each sample clustered. These data clearly demonstrated that ipsilateral gene expression profiles were different from gene expression profiles in sham-operated samples, while expression profiles in contralateral samples were heterogeneous. Three of the 5 contralateral samples were identified as having expression profiles different from those in sham-operated samples, which suggested that contralateral joints were not an appropriate control.
Dysregulation of genes involved in OA.
We next looked for dysregulation of genes known to be dysregulated in human OA (Figure 1C). The expression levels of proteases, including Adamts5 (29, 30), Mmp2 (31), and Mmp13 (32–34), were increased in ipsilateral OA cartilage, similar to other OA-related factors, including Chi3l1 (encoding cartilage glycoprotein 39) (35), prostaglandin E synthase (Ptges) (36), prostaglandin-endoperoxide synthase 2 (Ptgs2, encoding cyclooxygenase 2), which was increased but not at a significant level (37), and Tgfb2 (38).
Altered production of extracellular matrix (ECM) components is common in OA cartilage (39, 40) and was observed in ipsilateral cartilage samples (Figure 1D). This included increases in the levels of several types of collagen, many of which (e.g., Col1a1) are not expressed at high levels in healthy cartilage (41, 42) but were increased 2–4-fold in ipsilateral samples. Versican (Cspg2), lumican (Lum), and syndecan 1 (Sdc1) were other up-regulated ECM genes (Figure 1D). Interestingly, type II collagen (Col2a1) was not significantly dysregulated (Figure 1D). These profiles suggest that OA-like changes occurred at the gene expression level in this model.
Verification of changes in gene expression by real-time PCR.
Additional animals were used to confirm selected changes in gene expression by alternative approaches. Real-time PCR confirmed the increased expression levels of known OA genes, such as Mmp13, Adamts5, Ptgs2, Ptges, and Ccl2, and of novel genes, including Ednra, which mediates endothelin 1 signaling (43), and kit ligand (Kitl), which activates the c-kit protooncogene (44), in ipsilateral cartilage (Figure 2). Microarray analyses also demonstrated changes in contralateral joint chondrocyte expression of some genes compared with expression in sham-operated controls. Although similar changes were observed with real-time PCR, no statistically significant differences were seen with the probes tested, perhaps due to the heterogeneity of contralateral samples suggested by unsupervised clustering. We analyzed the expression of Col2a1, although it was not included in the probe list due to substantial variation between samples. Real-time PCR revealed similar variability in levels of Col2a1 between samples, without statistically significant differences between groups (Figure 2). Overall, these real-time PCR data indicated that our microarray data accurately reflected gene expression patterns.
Confirmation of gene expression at the protein level.
To validate the microarray gene expression data at the protein level, we examined the expression of MMP-13 in tissue sections from sham-operated, contralateral, and ipsilateral knees by immunofluorescence (Figure 3). The level of MMP-13 protein expression was markedly increased in ipsilateral cartilage, compared with that in contralateral or sham-operated cartilage. MMP-13 expression in ipsilateral cartilage was highly concentrated at the major articular surface defect (fibrillated matrix in the superficial and upper mid-zones) and was also detected in the cartilage territorial matrix and chondrocyte lacunae.
We also investigated the spatial and temporal expression of 2 novel genes identified by microarray analysis. Increased levels of Ctsc and Cxcr4 messenger RNA (mRNA) expression (at 4 weeks) were confirmed by real-time PCR (Figures 4A and C). Analysis of RNA samples isolated 2 weeks postsurgery determined that levels of Ctsc, but not Cxcr4, expression were increased (Figures 4A and C). Immunohistochemistry corroborated mRNA expression data for both genes, and demonstrated that the level of cathepsin C expression was increased in ipsilateral cartilage at 2, 4, and 8 weeks postsurgery compared with sham-operated controls (Figure 4B). However, the CXCR4 expression level was not increased until 4 weeks postsurgery, and was increased further at 8 weeks (Figure 4D).
Interestingly, CXCR4 was also expressed in hypertrophic growth plate chondrocytes in all animals (Figure 4E), and has been shown to increase in chondrocyte differentiation (14). Cathepsin C was not detected in the growth plate (results not shown). Similar up-regulation of cathepsin C, CXCR4, and MMP-13 was seen in tibial plateau and in femoral condyle cartilage. These results further validate our microarray data, and describe differential temporal expression of different genes identified by microarray. The coexpression of CXCR4 in degrading cartilage and hypertrophic growth plate chondrocytes suggests a role of chondrocyte hypertrophy in early OA pathogenesis.
Analysis of categories of dysregulated genes.
Having confirmed known OA gene expression patterns in this model, and identified individual novel genes, we analyzed our data sets in more detail. Groups of differentially expressed transcripts in each treatment group were separated according to fold change. Comparison of ipsilateral cartilage with sham-operated cartilage revealed 1,619 differentially expressed probe sets with changes in their levels of expression ≥1.5-fold, 722 with changes ≥2-fold, 135 with changes ≥4-fold, and 20 with changes of 8-fold (Figure 5A). Comparison of contralateral cartilage with sham-operated cartilage showed 398 differentially expressed probes with changes ≥1.5-fold, 91 with changes ≥2-fold, and 10 with changes ≥4-fold. No gene was changed >8-fold (Figure 5A).
We then inspected the dysregulated contralateral and ipsilateral gene lists for overlap. Interestingly, 354 of the 398 genes that were differentially regulated in contralateral cartilage were also dysregulated in ipsilateral OA cartilage (Figure 5B), the vast majority of which changed in the same direction (either increased in both contralateral cartilage and OA cartilage or decreased in both contralateral cartilage and OA cartilage). The only exceptions were the levels of Ptgs2 and mannose-binding lectin–associated serine protease 1 (Masp1), both of which were decreased in contralateral cartilage and increased in ipsilateral cartilage. For the full list of differentially expressed genes in both conditions, see Supplementary Table 1 (available on the Arthritis & Rheumatism Web site at http://www.mrw.interscience.wiley.com/suppmat/0004-3591/suppmat/).
Gene ontology analysis was performed to gain perspective on which classes of genes were differentially regulated in ipsilateral cartilage. Of the dysregulated genes that had annotations, 52% were involved in metabolism, 25% in cell communication, 12% in cell differentiation, and 11% in transcription (Figure 5C). Specific classifications were then assessed, demonstrating that differential regulation of genes encoding ECM proteins, cytoskeletal components, receptors, ligands, growth factors, cytokines, cell cycle proteins, proteolytic enzymes, and apoptosis factors occurs in early experimental OA (Figure 5D). Interestingly, expression levels of most of the dysregulated genes found in the categories of ECM molecules, ligands, cytokines, growth factors, and proteolytic enzymes were increased, rather than decreased, correlating with expected profiles in OA. For example, Supplementary Table 2 (available on the Arthritis & Rheumatism Web site at http://www.mrw.interscience.wiley.com/suppmat/0004-3591/suppmat) shows a detailed list of differentially regulated cytokines and growth factors in OA cartilage. The distribution of dysregulated genes in OA cartilage across different functional classes implicates anabolic responses, proteolytic enzyme production, and cytokine, chemokine, and growth factor signaling in early OA pathogenesis in this model.
Similarities of phenotypic modulation of OA chondrocytes to chondrocyte differentiation.
Phenotypic alterations in articular chondrocytes at various stages of OA have been described in several previous reports (45, 46). Such alterations are often reminiscent of chondrocyte differentiation to hypertrophy in the growth plate (47). To assess the similarities of early OA pathogenesis to chondrocyte differentiation, we compared the list of differentially expressed genes in OA cartilage from this model (≥2-fold change in gene expression level in ipsilateral cartilage versus that in sham-operated cartilage) with our previous microarray analyses of hypertrophic chondrocyte differentiation in 3-dimensional micromass culture (≥5-fold change in late versus early differentiation) (14) (Table 1).
Table 1. Genes dysregulated both in ipsilateral cartilage (determined by comparison with sham-operated cartilage) and in differentiating chondrocytes (determined by comparison of early versus late differentiation)
α1-actin, skeletal muscle
Bone marrow stromal cell antigen 1
Complement component 1, s subcomponent
Chitinase 3-like 1
Chimerin (chimaerin) 2
CXC chemokine receptor 4
Cytochrome b-245, β polypeptide
D site albumin promoter binding protein
Double cortin and calcium/calmodulin-dependent protein kinase–like 1
Extracellular matrix protein 1
Coagulation factor III
Fc receptor, IgG, low-affinity Iib
Growth arrest and DNA damage–induced 45α
Growth associated protein 43
Growth arrest specific 6
Guanylate binding protein 2
Insulin-like growth factor binding protein 6
Immunoglobulin superfamily, member 6
Interleukin-2 receptor, γ-chain
Lipopolysaccharide binding protein
Latent transforming growth factor β binding protein 2
Phosphate regulating gene with homologies to endopeptidases on the X chromosome
Protein tyrosine phosphatase, receptor type C
Protein tyrosine phosphatase, receptor type O
Regulator of G protein signaling 5
Serine (or cysteine) proteinase inhibitor, clade G, member 1
Transglutaminase 2, C polypeptide
Toll-like receptor 2
Wnt1-induced signaling protein 2
Of the 46 genes common to both lists, 37 (80%) changed in the same direction (2 genes decreased in both arrays and 35 increased in both). Only 9 genes displayed opposing trends in the 2 settings. Genes up-regulated in both scenarios included the proteases cathepsin S (Ctss), reelin (Reln), and Mmp13, as well as other genes previously implicated in chondrocyte differentiation, including transglutaminase 2 (Tgm2) (48), extracellular matrix protein 1 (Ecm1) (49), Fc receptor, IgG, low affinity III (Fcgr3) (50), latent transforming growth factor β binding protein 2 (Ltbp2) (51), and phosphate regulating endopeptidase homolog, X-linked (Phex) (52).
The similarities of chondrocyte gene expression during physiologic hypertrophy to gene expression in early OA are also supported by the results of our analyses of CXCR4 expression. The level of Cxcr4 gene expression was increased in micromass chondrocyte differentiation (Table 1) and in ipsilateral cartilage 4 weeks after surgery (Figures 4C and D), as well as in hypertrophic chondrocytes of the growth plate (Figure 4E). While some hypertrophic markers (e.g., type X collagen) were not present in the OA list, it is likely that not all hypertrophic markers are expressed at this early stage. Indeed, our previous data show that type X collagen expression is up-regulated at later stages in the model used in the present study (20). These data indicate that mechanisms of early OA pathogenesis in this model are related to chondrocyte differentiation during development.
In this study we used microarray technology and a recently developed rat model of OA (20, 53) to obtain a global view of early changes in gene expression in degrading cartilage. Identification of differentially expressed genes will expand our understanding of mechanisms involved in early OA pathogenesis. We chose ACLT with partial meniscectomy in rats for a number of reasons. First, the rat genome is well annotated, and powerful genomic tools are available for study of this species. Second, the size of the rat allows surgical manipulation and isolation of sufficient biologic material (e.g., chondrocyte RNA). Third, housing of rats is relatively inexpensive, allowing functional studies of multiple compounds that might modulate OA progression. Fourth, the ACLT with partial meniscectomy model allows controlled initiation and predictable progression of OA. This is in contrast to models of spontaneous disease that show greater variability, complicating identification of early events in OA development.
Our studies have demonstrated that 4 weeks after surgery this model reproducibly reflects many aspects of early human OA, making it suitable for investigating early changes in cartilage gene expression in this disease (20). Our analyses confirmed that established molecular changes in OA were already under way at this time and identified novel changes. Thus, this animal model and the time point chosen are appropriate for identifying previously unknown factors in early OA pathogenesis.
However, as with every animal model, some limitations apply. First, cartilage degradation was surgically induced, making this study most relevant to posttraumatic OA. Due to the heterogeneity of human OA, this limitation is not unique to our model, and many of the changes we observed may also occur in other forms of OA. Second, for practical and economic reasons we have, so far, restricted our microarray analyses to a single time point (4 weeks after surgery). Therefore, molecular changes at earlier and later time points were missed in this study. Finally, although several of the genes relevant to OA identified by microarray in this study were validated by real-time PCR, many more OA-relevant genes were not.
Thus, interpretation of changes in the nonvalidated genes should be made with caution, since microarray data almost always contain some false-positive findings. However, several of our recent studies, as well as the data presented here, indicate that Affymetrix microarray data are consistently validated by real-time PCR, in situ hybridization, immunohistochemistry, and other techniques (14, 54, 55). Nevertheless, validation of microarray data should be performed as a first step in subsequent studies of the regulation or function of specific genes.
Some molecular changes identified in the ipsilateral joint merely represent adaptive responses to altered joint loading. To demonstrate this, we assessed RNA and protein expression of 2 factors, cathepsin C and CXCR4, at additional time points. Whereas levels of cathepsin C expression were increased 2 weeks postsurgery, CXCR4 levels were not increased until 4 weeks postsurgery but were further increased at 8 weeks. While early induction of cathepsin C might be due to immediate adaptation to an altered mechanical environment, delayed induction of CXCR4 expression is coincidental with changes in chondrocyte phenotype and not with immediate responses to altered mechanical factors. Thus, different genes respond with different dynamics in this model, and future study should examine these individual responses in detail. Moreover, OA has been correlated with mechanical alterations in joint biomechanics (56). Thus, at early stages, molecular changes in response to mechanical stimuli are likely etiologic for OA, leading to subsequent pathologic events.
One of the important findings of our study is that ACLT with partial meniscectomy with forced mobilization results in dysregulation of gene expression in contralateral joint chondrocytes. Several of the genes found to be dysregulated in contralateral cartilage were also dysregulated in the same direction in ipsilateral cartilage, although the changes were milder in the contralateral joint. Unsupervised clustering demonstrated that while contralateral samples were different from sham-operated samples, heterogeneity among contralateral samples was present, with some animals showing greater dysregulation than others. Differences in gene expression in the contralateral and sham-operated samples could be due to altered biomechanics (e.g., altered loading of the contralateral joint) or to systemic factors (e.g., mediators of inflammation) released as OA progresses in the ipsilateral joint (57). Further, differences in compensatory contralateral joint loading between individuals may account for heterogeneity among contralateral samples. Overall, we conclude that contralateral joints are not suitable controls in gene expression studies.
After comparing the results of the present study with those of an earlier study of differentially expressed genes in chondrocyte differentiation (14), we found similarities between gene expression profiles in early OA pathogenesis in this model and differentiating chondrocytes. Further, chondrocyte phenotype modulations in OA in the form of dedifferentiation (58, 59), proliferation (60), apoptosis (61), and hypertrophy (46, 62) have been proposed. We found that many genes (e.g., Ltbp2, Phex, and Reln) were regulated similarly in both degrading cartilage and differentiating chondrocytes, positioning mechanisms leading to hypertrophic differentiation of chondrocytes in early OA. This finding is supported by the results of other studies, which have led investigators to propose that recapitulation of chondrocyte differentiation to hypertrophy may be responsible for cartilage loss in human OA (12, 13, 46, 63).
As mentioned above, an abnormal mechanical environment is a risk factor for OA (56), and this may be mediated by stimulating chondrocyte phenotype modification. However, CXCR4 expression was not induced in degrading articular cartilage until 4 weeks postsurgery, but was expressed at all time points assessed in hypertrophic growth plate chondrocytes and was increased during chondrocyte differentiation (14). Thus, it is not likely that CXCR4 induction in early OA is directly due to mechanical alterations in this model, which occur immediately following surgery. Instead, CXCR4 induction may be stimulated by other factors that are induced by mechanical changes. These changes may cause a gradual shift in chondrocyte phenotype to hypertrophy, thereby increasing CXCR4 expression in parallel with cartilage degradation. Indeed, increased signaling from the CXCR4 receptor (either by increased levels of its ligand CXCL12 or by increased receptor expression) has been shown to increase the level of expression of the catabolic enzyme MMP-13 in chondrocytes (64).
Induction of chondrocyte hypertrophy in early OA also leads to the production of factors that calcify matrix (e.g., alkaline phosphatase). Induction of articular chondrocyte hypertrophy is therefore detrimental, since it impairs maintenance and promotes degradation of articular cartilage, thus contributing to OA pathogenesis.
Array data analysis and confirmatory studies for known mediators of OA at the RNA and protein levels suggest that the model of ACLT with partial meniscectomy in rats mimics established molecular events in human OA. Among these, we observed increased levels of expression of the cartilage glycoprotein Chi3l1, mediators of inflammation, such as Ptges, Ptgs2, and tumor necrosis factor α–induced protein 6 (Tnfaip6) (65), proteases (e.g., Mmp2, Mmp13, and Adamts5), and anabolic response genes (e.g., collagens, Cspg2, and Lum). Many other groups of genes were dysregulated, such as members of the insulin-like growth factor (IGF) axis (Supplementary Tables 1 and 2). This included increased levels of expression of IGF binding proteins, which sequester IGFs and prevent their anabolic influences in chondrocytes. In addition, modulation of oxidative defense genes (e.g. Sod2, encoding superoxide dismutase 2) implicates oxidative defense mechanisms in the development of early OA (Supplementary Table 1), defined in a recent study of multiple stages of human OA (15). Findings of dysregulation of many genes known to be involved in OA provide strong evidence that newly identified genes are also involved in OA progression.
Interestingly, gene ontology analyses demonstrated that many cytokine and growth factor signaling genes were up-regulated early in degrading cartilage. For example, the dramatic change we observed in the expression of Ccl2 (19-fold increase in ipsilateral cartilage) distinguished it as a potential factor in OA. Ccl2 has been implicated in OA by several in vitro studies and has been detected in human OA chondrocytes (66, 67). Our study also implicates additional novel chemokine signaling factors, including Cxcr4, Cklf1, and Cx3cl1, in cartilage degradation. Chemokines produced by chondrocytes may contribute to OA pathogenesis by recruiting inflammatory cells in the synovium, or by changing chondrocyte gene expression and phenotype via autocrine/paracrine signaling (68), through chemokine receptors expressed by chondrocytes (66). Our data, along with findings of others who have identified chemokine signaling as a key player in OA pathogenesis (56, 69), strongly suggest the need for in-depth studies on the role of chemokine signaling in OA. However, chemokines represent just one example of new candidate regulators of OA progression identified in our study. Additional examples include members of the cathepsin family of cysteine proteases; our study demonstrated up-regulation of cathepsins, including Cfsc, in degrading cartilage. It is noteworthy that other cathepsin family members (e.g., B and K) have previously been implicated in OA (68, 70).
In summary, it is necessary to understand the dysregulation of gene expression in OA chondrocytes for better comprehension of pathogenesis, identification of causes, and determination of potential drug targets. Our findings in the present study, which is among the first investigations of genome-wide gene expression in an animal model of OA, expand the current understanding of genetic mechanisms involved in OA pathogenesis in such models. Further, the data suggest that this animal model is comparable with human OA, since we found dysregulation of many of the same OA susceptibility genes and genes that facilitate OA pathology reported in human OA. Therefore, these data provide a basis for future studies on the function of numerous newly identified genes (e.g., Cxcr4 and Ctsc) in OA pathogenesis and on their suitability as drug targets in OA.
Dr. Beier 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.