Microarray analysis reveals age-related differences in gene expression during the development of osteoarthritis in mice




To better understand the contribution of age to the development of osteoarthritis (OA).


Surgical destabilization of the medial meniscus (DMM) was used to model OA in 12-week-old and 12-month-old male C57BL/6 mice. OA severity was evaluated histologically. RNA used for microarray and real-time polymerase chain reaction analysis was isolated from joint tissue collected from the medial side of the joint, including cartilage, meniscus, subchondral bone, and the joint capsule with synovium. Computational analysis was used to identify patterns of gene expression, and immunohistochemistry was used to evaluate tissue distribution of selected proteins.


OA was more severe in older mice than in young mice. Only 55 genes showed a similar expression with DMM-induced OA in the 2 age groups, while 493 genes showed differential expression, the majority having increased expression in older mice. Functional categories for similarly expressed genes included extracellular matrix– and cell adhesion–related genes; differentially expressed genes included those related to muscle structure and development and immune response genes. Comparison of expression in sham-operated control joints revealed an age-related decrease in matrix gene expression and an increase in immune and defense response gene expression. Interleukin-33 was present in multiple joint tissue cells, while CCL21 was more localized to chondrocytes and meniscal cells. Periostin was found in the extracellular matrix of cartilage and meniscus.


Age affects both the basal pattern of gene expression in joint tissues and the response to surgically induced OA. Examining tissue from the joint beyond only cartilage revealed novel genes and proteins that would be important to consider in OA.

Osteoarthritis (OA) is unusual in young adults but occurs quite commonly in older adults, such that symptomatic OA affects 10–20% of people age >50 years (1). In addition to age, joint injury is a common risk factor, especially for knee OA, with a pooled odds ratio of 3.86 (95% confidence interval 2.6–5.7) for knee OA after joint injury (2). Importantly, aging and joint injury interact. People who experienced a meniscal injury after age 30 years developed radiographic evidence of OA 3 times faster than those who had a similar injury between ages 17 and 30 years (3). The mechanisms by which aging contributes to the development of OA and the ways in which age and joint injury interact are incompletely understood and are the subject of the present work.

Several studies have analyzed gene expression microarray data to discover the genes and pathways that are regulated at the transcriptional level during the development of OA. Studies of human OA have used cartilage removed at the time of joint replacement surgery, which represents end-stage disease, or from autopsies for studies of early lesions (4–7). Gene expression has also been evaluated in animal models of OA, including the rat anterior cruciate ligament transection (ACLT) and meniscal tear models (8, 9). A limitation to these studies is that gene expression was evaluated only in 1 tissue, most often the articular cartilage. It is well accepted that OA is a process that involves the joint as an organ rather than the articular cartilage alone. Microarray studies have been performed using RNA isolated from OA subchondral bone (10) and synovium (11, 12), but, like the cartilage studies, the analysis was limited to those selected tissues.

In the present study, we evaluated and compared gene expression in knee joint tissues from younger and older adult mice after the induction of OA by destabilization of the medial meniscus (DMM). The DMM model is a postinjury model described by Glasson et al (13–15) that has become popular because it involves the meniscus, which is commonly involved in human OA, and because the histologic lesions within the affected joint are similar to those observed in human OA. Most commonly, young male mice (129/SvEv or C57BL/6 strains) in the age range of 8–12 weeks are used in this model. Mice are considered to be skeletally mature at age ∼10 weeks, which is the age generally recommended for studies of surgically induced OA in this species (16). However, a 10-week-old C57BL/6 mouse corresponds approximately to a teenaged human, while a 12-month-old mouse would represent a 40–50-year-old human (17).

Therefore, to study the effects of age on the development of posttraumatic OA, we measured OA severity histologically and analyzed gene expression by microarray in joints from 12-week-old and 12-month-old mice, which we will refer to as young and older adult mice, respectively. RNA was isolated from joint tissue that was removed from the medial (affected) side of the joint, including cartilage, subchondral bone, meniscus, and the joint capsule with synovium, in order to study the joint as an organ. We identified significant histologic differences in OA severity between younger and older mice as well as differences in gene expression that included genes not previously identified in OA that might play an important role in the disease process.



Male C57BL/6 mice were purchased from Charles River (young mice) or from the National Institute on Aging aged rodent colony (older mice). All animals were housed in the same facility and were provided food and water ad libitum on a 12-hour light/dark cycle. A total of 30 animals were included in each age group (12 weeks old or 12 months old at the time of surgery). Half of the animals in each age group underwent DMM surgery and half underwent sham surgery, as described below. Animals were cared for according to institutional animal care and use protocols, and all animal studies were approved by the Wake Forest University School of Medicine Animal Care and Use Committee.

Surgical induction of OA.

The DMM model of OA was established as described by Glasson et al (15), using a protocol and instructions kindly provided by Dr. Glasson. OA was induced by transection of the meniscotibial ligament. The sham-operated controls underwent the same surgery except that the meniscotibial ligament was not cut. Five mice were housed in each cage and allowed free activity. Mice were euthanized 8 weeks after surgery.

Histology and analysis of disease severity.

Six DMM-operated mice and 6 sham-operated mice in each age group were randomly selected for histologic studies. From these, 4 groups of joints were identified: joints that underwent DMM surgery and their respective contralateral control joints, and joints that underwent sham surgery and their respective contralateral control joints. The hind limbs were routinely fixed in 4% paraformaldehyde, decalcified in 10% EDTA, processed, and embedded intact into paraffin for histologic evaluation. The joints were sectioned in a coronal plane, and serial midcoronal 4-μm sections that included the femoral condyles, menisci, and tibial plateaus were cut. Two representative sections from each joint were stained with hematoxylin and eosin (H&E) and Safranin O.

Each medial and lateral tibial plateau and meniscus was graded by an evaluator (MAM) who was blinded with regard to the experimental group, using a recently developed murine OA grading scheme that included semiquantitative grades and quantitative measurements (18). The 2 semiquantitative grades consisted of an articular cartilage structure score and a Safranin O staining score, both ranging in grades from 0 to 12, with 0 being normal and 12 indicating severe OA. The parameters evaluated quantitatively included articular cartilage (thickness and area), chondrocytes (chondrocyte cell death [total area containing ≥2 necrotic chondrocytes], total number of viable chondrocytes, and percentage of chondrocyte cell death in articular cartilage [chondrocyte cell death/articular cartilage area × 100%]), subchondral bone (area and thickness), periarticular bone (e.g., osteophytes, total area if present), and meniscus (total area and area of chondrocyte cell death, if present).


Serial sections immediately adjacent to those stained with H&E and Safranin O were selected from representative joints from 12 animals and were immunostained using antibodies directed against interleukin-33 (IL-33), CCL21, and periostin in order to examine protein tissue distribution. Antigen retrieval was achieved with a citrate buffer (Dako), and endogenous peroxidase was blocked with 3% H2O2. Sections were incubated with universal protein block (Dako) and then for 30 minutes with the appropriate primary antibodies, as follows: anti–IL-33 (1:50 dilution) (catalog no. AF3626; R&D Systems), anti-CCL21 (1:100 dilution) (catalog no. AF457; R&D Systems), or antiperiostin (1:200 dilution) (catalog no. P2195; Invitrogen). Binding of primary antibody was detected with Biocare Goat probe for 15 minutes followed by Goat polymer for 20 minutes (anti–IL-33 and anti-CCL21) or Dako Envision+ anti-rabbit HRP Polymer for 30 minutes (antiperiostin). All sections were developed with 3,3′-diaminobenzidine chromogen and counterstained with Mayer's hematoxylin (Dako). For negative control slides the primary antibody was substituted by negative goat serum (BioGenex) (for anti–IL-33 and anti-CCL21) or rabbit serum (Dako) (for antiperiostin).

RNA isolation and purification.

Nine DMM-operated and 9 sham-operated mice in each age group were used for RNA isolation. Immediately after euthanasia, the knee joints of the operated side were opened and the patella, tendons, and other soft tissues around the medial side of the joint were removed. A coronal section was made through the tibial plateau and femoral condyle tissue, and the medial side of the joint, including the articular cartilage, subchondral bone with any osteophytes, meniscus, and the joint capsule with synovium, was collected for analysis. The tissue was placed in RNAlater (Invitrogen) and stored overnight at 4°C. The next day the RNAlater solution was removed and the tissue was frozen at −80°C until RNA isolation. RNA was extracted from the tissue by homogenization in RNA extraction solution (UltraClean Tissues and Cells RNA isolation kit; Mo Bio Laboratories) using a Precellys 24 tissue homogenizer (Bertin Technologies; purchased from Mo Bio Laboratories) and 1.4-mm ceramic beads. The amount and quality of the RNA were determined using an Agilent 2100 Bioanalyzer. Only samples with an RNA integrity number of ≥8 were used for analysis.


Affymetrix mouse genome 430 2.0 oligonucleotide expression array chips were used. RNA samples isolated from joint tissue from 3 mice were pooled for analysis on each microarray chip, and 3 separate microarray chips were run for each condition (DMM- or sham-operated) from each age group (younger or older) for a total of 12 arrays using RNA from 36 mice. Chip processing and image capturing were performed using Affymetrix AGCC software.

Microarray data analysis.

Details of data normalization, filtering, pattern classification, and functional annotation are provided in the Supplementary Methods (available on the Arthritis & Rheumatism Web site at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)1529-0131). Briefly, raw microarray data were normalized using systemic variation normalization as described (19) (see Supplementary Figure 1, available on the Arthritis & Rheumatism Web site at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)1529-0131). For each gene in each experimental replicate, the signal log ratio was calculated as the log2 ratio of the intensity in the DMM-operated joint to the average intensity in sham-operated joints. Data were then filtered according to the scheme shown in Figure 1. The signal log ratio data associated with the filtered genes were classified into 1 of 3 groups: up, down, and no change. Criteria for classification were as follows: signal log ratio ≥ 0.5, up; signal log ratio ≤ −0.5, down; −0.5 < signal log ratio < 0.5, no change. Once each signal log ratio value was classified, the gene was classified into 1 of 8 patterns (Figure 1). In order to evaluate changes in gene expression due to age, independent of surgically induced OA, differences in expression between the young and old sham-operated groups were also evaluated (see Supplementary Methods, available on the Arthritis & Rheumatism Web site at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)1529-0131).

Figure 1.

Flow chart of the filtering process. As described in Materials and Methods, the 45,101 probe sets on the Affymetrix arrays were filtered, in a stepwise process, to identify genes that responded similarly or differently to destabilization of the medial meniscus (DMM) between younger and older mice. Step 1 identified all genes with a signal that was significantly detected on the chip over the background. Step 2 identified significantly expressed or repressed genes in DMM-operated joints, with respect to sham-operated joints, in ≥2 replicate pools. Step 3 identified genes that either responded in a similar manner between young and older mice or had a significantly different response. These genes were then assigned to an expression pattern in step 4 for further analysis. SLR = signal log ratio; D = down; U = up; NC = no change.

A functional analysis was performed for the genes in each pattern defined in Figure 1 and for the comparison of the young and old sham-operated groups. The Functional Annotation Clustering tool provided by the Database for Annotation, Visualization, and Integrated Discovery (DAVID) (20, 21) was used to obtain gene annotations that were significantly overrepresented in each pattern as compared to the full list of genes on the Affymetrix Mouse 430 2.0 GeneChip (used as background). Ingenuity Pathway Analysis (IPA) (www.ingenuity.com) software version 8.8 (content version 3204, 10/27/2010) was used to identify canonical pathways that were significantly overrepresented. The gene list for each pattern was input into IPA, and a Core Analysis was run on each.

Real-time polymerase chain reaction (PCR).

Samples of RNA from the same pools used for the microarrays were used for real-time PCR. The SYBR Green–based method was used with optimized primer sets obtained in the RT2 qPCR Primer Assays (SABiosciences/Qiagen), with TATA box binding protein used to normalize relative expression as previously described (22).

Statistical analysis.

Statistical analysis methods for the microarray data are provided above and in the Supplementary Methods (available on the Arthritis & Rheumatism Web site at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)1529-0131). Histologic data from both the lateral and medial tibial plateaus were evaluated with analyses of variance (ANOVAs) (intraanimal comparisons) and repeated-measures tests (interanimal comparisons) using SPSS version 15.0 (IBM). The real-time PCR results were analyzed using ANOVA with StatView 5.0 software (SAS Institute).


Development of OA changes in young and older mice.

Both young and older mice exhibited typical histologic features of OA in the DMM-operated knees, including cartilage surface fibrillation and clefting, medial meniscal degeneration, osteophyte formation, and subchondral bone thickening (Figures 2A–D). OA lesions were consistently more severe in the medial tibial plateau than in the lateral tibial plateau and were more severe in the 12-month-old mice than in the 12-week-old mice. The average articular cartilage structure score (medial plus lateral) was ∼2-fold higher in the DMM-operated knees of the older mice than in those of the younger mice (Figure 2E).

Figure 2.

Histologic evaluation of osteoarthritis (OA) in medial knee joint compartments of 12-week-old and 12-month-old mice. Younger and older mice underwent destabilization of the medial meniscus (DMM) or sham surgery, and OA severity was evaluated after 8 weeks. Sections were stained with hematoxylin and eosin (H&E) and Safranin O. A, Twelve-month-old DMM-operated mouse. B, Twelve-month-old sham-operated mouse. C, Twelve-week-old DMM-operated mouse. D, Twelve-week-old sham-operated mouse. Bars = 100 μm. Insets are higher-magnification (20×) images of a central area of the medial tibial plateau of the H&E-stained sections. E, Articular cartilage structure (ACS) scores calculated on a 0–12 scale for the medial and lateral (M + L) sides of the joint. Controls are the contralateral (unoperated) limbs. Values are the mean ± SEM for the medial and lateral tibial plateaus combined. ∗ = P < 0.05 versus young mice, by analysis of variance.

In both the 12-week-old and 12-month-old DMM-operated groups, the medial tibial plateaus had significantly greater areas and thicknesses of subchondral bone (P < 0.001 for both), a higher percentage of chondrocyte cell death in articular cartilage (P = 0.001 for both), a lower total number of viable chondrocytes (P = 0.01 for both), and larger abaxial osteophyte size (P < 0.05 for both) compared with the medial tibial plateaus from the corresponding sham-operated joints (see Supplementary Table 1, available on the Arthritis & Rheumatism Web site at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)1529-0131). In addition, the medial tibial plateaus in DMM-operated joints of the 12-month-old mice had significantly reduced articular cartilage (P < 0.001 for area and thickness) and higher articular cartilage structure scores (P = 0.008) than the lateral tibial plateaus (data not shown). Because the most severe lesions were present medially, only results from the medial tibial plateaus are presented for the surgical treatment group comparisons unless otherwise noted (see Supplementary Table 1, available on the Arthritis & Rheumatism Web site at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)1529-0131). In addition, no significant differences between the contralateral DMM-operated and contralateral sham-operated groups were identified in the younger mice. Therefore, data from these groups were combined into a single “control” group for the young mice.

The sham-operated knees of the 12-week-old mice did not exhibit any significant changes of OA; however, the sham-operated joints in the 12-month-old mice demonstrated mild OA-like changes, suggesting an age-related development of early spontaneous OA (Figure 2B). This was unlikely to be a result of the sham surgery, since similar changes were seen in the contralateral (unoperated) knees of the sham-operated older mice. In addition, the contralateral joints of the DMM-operated older mice exhibited significantly more severe OA than the contralateral joints of the sham-operated older mice, suggesting an effect of the DMM surgery on the contralateral limb in older mice. Additional histologic findings are presented in the Supplementary Results (available on the Arthritis & Rheumatism Web site at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)1529-0131).

Patterns of gene expression in knee joint tissue from young and older mice with surgically induced OA.

A 4-step filtering and classification process (Figure 1) identified those genes that were significantly over- or underexpressed in DMM-operated joints compared to sham-operated joints and classified them into 1 of 8 patterns. The signal log ratio distributions illustrate a key observation: more genes were up-regulated in the RNA from older mice, while more genes were down-regulated in the younger mice. The complete list of genes for each pattern and their DAVID annotations are provided in Supplementary File 1 (available on the Arthritis & Rheumatism Web site at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)1529-0131). The numbers of genes in each of the replicate pools from young and older mice that were up- or down-regulated or showed no change are shown in Supplementary Figure 2 (available on the Arthritis & Rheumatism Web site at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)1529-0131), and heatmaps illustrating levels of expression are shown in Figure 3, along with functional annotations significantly overrepresented in each pattern identified using DAVID (20, 23).

Figure 3.

Heatmaps of the genes identified in each expression pattern. Each group summarized in step 4 (Figure 1) is represented as a heatmap with a summary of annotation terms found to be significantly overrepresented in each pattern by DAVID. Columns represent replicate arrays of young mice (p1, p2, and p3; left) and older mice (p1, p2, and p3; right), with each row representing a single gene. Blue indicates down-regulation, red indicates up–regulation, and white indicates no change (NC) in expression as compared to expression in sham-operated mice. EGF = epidermal growth factor.

Genes regulated similarly in both 12-week-old and 12-month-old DMM-operated mice relative to sham-operated controls.

Fifty-five genes were either up-regulated or down-regulated in both young and older mice. There were only 8 genes in the down–down pattern. IPA identified them as belonging to metabolic pathways (citrate cycle, pyruvate, and linoleic acid metabolism) as well as the complement system (complement factor D) and metabolism of xenobiotics.

The 47 genes in the up–up pattern demonstrated significant functional annotations for extracellular matrix– and cell adhesion–related genes and for carbohydrate binding and epidermal growth factor (EGF)–like proteins. Genes in this pattern included those for asporin, chemokine (CC motif) ligand 21A, several collagen genes including those for types III, VI, and XIV collagen, Dkk-3, HtrA serine peptidase 1, insulin-like growth factor 1 (IGF-1), IL-33, lumican, matrilin 2, matrix metalloproteinases (MMPs) 2 and 3, periostin, and tissue inhibitor of metalloproteinases 1. The canonical pathways identified for these 47 genes included pathways involved in tissue remodeling, including genes also classified as involved in rheumatoid arthritis (RA) (IL33, IGF-1, DKK3, MMP3, MMP2) and several signaling pathways, including genes for IGF-1, IL-17, and hypoxia-inducible factor 1α. The most significant pathway identified was “hepatic fibrosis” (IGF-1, TIMP1, CCL21, MMP2, COL3A1).

Genes with higher expression in older DMM-operated mice compared to young DMM-operated mice.

The young–old patterns of no change–up, down–no change, and down–up included 421 genes for which relative expression was higher in older mice. These 3 patterns had similar functional annotations that included muscle-, sarcomere-, myofibril-, and collagen-related genes. The largest gene pattern, down–no change, included 170 genes exhibiting functional groups dominated by muscle structure and development and carbohydrate metabolism. Both DAVID analysis and IPA indicated that genes involved in ion and metal homeostasis, particularly calcium-signaling genes, were significantly represented. Pathway analysis also indicated that signaling genes were significantly represented, including genes involved in integrin-linked kinase signaling, protein kinase A signaling, actin–cytoskeleton signaling, and cyclin-dependent kinase 5 signaling. Genes in this pattern also included those for 2 chemokine receptors (CCR1 and CCR2) and a chemokine ligand (CXCL10), types XI and IV collagen, 4 interferon-related genes, S100B, glutaredoxin 3, and suppressor of cytokine signaling 6.

The no change–up (young–old) pattern was the second largest with 156 genes. DAVID annotation groups included collagen, extracellular matrix and cell adhesion, signaling (EGF), glycosaminoglycan binding, and bone mineralization. One DAVID annotation group contained genes associated with peroxidase activity and oxidative stress, suggesting perhaps a redox component to this process. Genes in this pattern included those for aggrecan, types II, V, VI, XXII, XXIV, and XXVII collagen, discoidin domain receptor 1, matrilin 2, tumor necrosis factor receptor–associated factor 6, IL-13 receptor (IL-13R), prostaglandin-endoperoxide synthase 2, platelet-derived growth factor, peroxiredoxin 4, F-spondin, syndecan 1, thrombospondin 2 and 4, very low-density lipoprotein receptor, and Wnt inhibitory factor 1.

The down–up (young–old) pattern was the third largest, containing 95 genes. As with the no change–up and down–no change patterns, this gene group exhibited muscle-related annotations and calcium signaling pathways. However, this pattern also included genes for cytochrome c oxidase, type IX collagen, epiphycan, hyaluronan and proteoglycan link protein 1, matrilin 3, titin, and unique cartilage matrix–associated protein. IPA pathways included calcium signaling, actin–cytoskeleton signaling, and protein kinase A signaling. A unique pathway associated with this pattern was oxidative phosphorylation.

Genes with higher expression in young DMM-operated mice compared to older DMM-operated mice.

Seventy-two genes displayed higher expression in young mice relative to the older mice: no change–down, up–down, and up–no change (young–old). These 3 patterns differed from the previous patterns in that the annotations were focused on genes related to the immune response. The largest group in this pattern, up–no change, contained 52 genes that had functional annotations related to immune system development, immune regulation, and leukocyte activation. IPA revealed 2 primary pathway groups: those associated with B cells (including B cell development, various types of signaling in B cells, and altered B cell signaling in RA) and those associated with phospholipid metabolism and signaling. Examples of genes in this category include those for CD19, CD51, CD79, IL-7R, Fas apoptotic inhibitory molecule 3, peroxiredoxin 2, sphingosine kinase 1, and pre–B lymphocyte gene 1.

The 2 smallest patterns were the no change–down (18 genes) and up–down (2 genes) patterns. The no change–down annotations related to immunoglobulin, immune response–related processes, and disulfide bond and transmembrane proteins. Genes of potential interest included those for the chemokine CXCL13, dual-specificity phosphatase 12, phospholipase A2 (group IID), the gene radical S-adenosyl methionine domain–containing 2, and immunoglobulin and histocompatibility genes. Only 1 gene (2 probe sets) was up-regulated in young mice and down-regulated in older mice: the gene for immunoglobulin heavy chain complex.

Age-related differences in gene expression independent of surgically induced OA.

Differences in expression between sham-operated control joints of young mice and those of older mice were examined, revealing significant age-related differences in gene expression. Table 1 provides a list of the 10 most differentially expressed genes in the older sham-operated adult mice compared to the young sham-operated adult mice. A complete list of the 861 differentially expressed genes and their DAVID annotations can be found in Supplementary File 2 (available on the Arthritis & Rheumatism Web site at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)1529-0131). Genes expressed at higher levels in sham-operated joints of older mice included those for 3 chemokines in the top 10: CXCL13 (B lymphocyte chemoattractant), CCL8 (monocyte chemotactic protein 2), and CCL5 (RANTES). DAVID analysis of 430 genes expressed at higher levels in older mice included annotations primarily for immune response and defense response, such as response to wounding and inflammatory response. The increased expression of genes for chemokines, cytokines (e.g., IL-6), and several HLA molecules resulted in the top IPA canonical pathways listed as “altered T cell and B cell signaling in rheumatoid arthritis,” “systemic lupus erythematosus signaling,” and “graft-versus-host disease signaling.”

Table 1. Genes with the highest and lowest SLRs when comparing RNA from sham-operated control joints of older mice to that from sham-operated control joints of younger mice*
 Average SLR
  • *

    Signal log ratios (SLRs) were calculated using the microarray results for sham-operated young and old mice as described in Materials and Methods. Shown are the 11 genes with the highest expression and the 10 genes with the lowest expression in old mice relative to young mice.

Genes with increased expression in sham-operated joints of old mice 
 Hba-a1/2 (hemoglobin alpha, adult chain1/2)5.18
 CXCL13 (B lymphocyte chemoattractant)2.66
 CCL8 (MCP-2)2.51
 Pla2g2d (phospholipase A2, group IID)2.23
 Igh-6 (immunoglobulin heavy chain 6 [heavy chain of IgM])2.2
 Cd51 (alpha V integrin)1.95
 Igh (immunoglobulin heavy chain complex)1.94
 CCL5 (RANTES)1.92
 Ighv14-2 (immunoglobulin heavy variable V14-2)1.72
 H2-Aa (histocompatibility 2, class II antigen A, alpha)1.71
 Faim3 (Fas apoptotic inhibitory molecule 3)1.71
Genes with decreased expression in sham-operated joints of old mice 
 Col9a1 (collagen, type IX, alpha 1)−4.14
 Matn3 (matrilin 3)−3.89
 Lect1 (leukocyte cell-derived chemotaxin 1)−2.83
 Col9a3 (collagen, type IX, alpha 3)−2.82
 Hapln1 (hyaluronan and proteoglycan link protein 1)−2.76
 Col2a1 (collagen, type II, alpha 1)−2.43
 Myh2 (myosin, heavy polypeptide 2, skeletal muscle, adult)−2.43
 Epyc (epiphycan)−2.33
 Mup1 (major urinary protein 1)−2.3
 Prkg2 (protein kinase, cGMP-dependent, type II)−1.95

The 431 genes expressed at lower levels in sham-operated control joints of older mice compared to sham-operated control joints of younger mice were primarily genes for extracellular matrix proteins such as type IX and type II collagen, matrilin 3, and link protein (the gene Hapln1). DAVID analysis of this group revealed annotations for the extracellular matrix, metabolic processes, and tissue development, including cartilage development. Genes in the latter category included those for transforming growth factor β2 (TGFβ2), aggrecan, SOX9, and types II and XI collagen. Other relevant genes decreased with age were those for high mobility group box chromosomal protein 2 (HMGB-2), osteoprotegerin, and cartilage oligomeric matrix protein (COMP). IPA canonical pathways included “calcium signaling” and “RhoA signaling.”

Real-time PCR measurement of selected genes.

We selected 21 genes to measure using real-time PCR based on differences noted in the arrays and/or their potential importance in the OA process and compared expression between sham-operated and DMM-operated joints within an age group as well as between young and old mice (Table 2). Expression of genes for asporin (which inhibits TGFβ and has been implicated in OA) and ADAMTS-5 (aggrecanase 2), increased more in DMM-operated older mice. Other genes with increased expression in DMM-operated joints of old but not young mice included those for aggrecan, CXCR7, HtrA serine peptidase 1, secreted Frizzled-related protein 2, and periostin. Genes increased in DMM-operated joints in both young and old mice included those for CCL21, IGF-1, and MMP-3, while genes that were increased only in young mice and not in old mice were those for CCR7, Dkk-3, and IL-33. Also of interest was that the gene for the cartilage matrix proteoglycan aggrecan was expressed at lower levels in the sham-operated old mice than in the sham-operated young mice, while the genes for CCR7, CXCR2, insulin-degrading enzyme, and IL-33 were expressed at higher levels in the sham-operated old mice than in the sham-operated young mice.

Table 2. Real-time polymerase chain reaction results for selected genes, shown as relative expression levels normalized to expression of TATA box binding protein*
Gene nameSham-operated young miceDMM-operated young miceSham-operated older miceDMM-operated older mice
  • *

    Values are the mean ± SD. DMM = destabilization of the medial meniscus; ADAMTS5 = a disintegrin and metalloproteinase with thrombospondin motif 5; CCL21 = chemokine (CC motif) ligand 21; CCR7 = chemokine (CC motif) receptor 7; CXCR2 = chemokine (CXC motif) receptor 2; CXCR7 = chemokine (CXC motif) receptor 7; COL3 = collagen type III; COL10 = collagen type X; DKK3 = dickkopf homolog 3; DEL1 = epidermal growth factor–like repeats and discoidin I–like domain; Htra1 = HtrA serine peptidase 1; IDE = insulin-degrading enzyme; IGF-1 = insulin-like growth factor 1; IL33 = interleukin-33; MMP3 = matrix metalloproteinase 3; MMP13 = matrix metalloproteinase 13; Nox4 = NADPH oxidase 4; Sfrp2 = secreted Frizzled-related protein 2; TSG6 = tumor necrosis factor α–induced protein 6.

  • P < 0.05 versus young DMM-operated mice.

  • P < 0.05 versus young sham-operated mice.

  • §

    P < 0.05 versus age-matched sham-operated mice.

ADAMTS55.6 ± 0.54.8 ± 0.18.2 ± 311.0 ± 2
Aggrecan1.79 ± 0.21.9 ± 0.30.69 ± 0.11.669 ± 0.6§
Asporin1.0 ± 0.11.6 ± 0.31.3 ± 0.53.22 ± 1§
CCL2149.7 ± 15.6112.8 ± 6.2§62 ± 3.599 ± 11.3§
CCR70.42 ± 0.20.77 ± 0.1§0.82 ± 0.10.84 ± 0.01
CXCR23.0 ± 0.72.6 ± 0.318.6 ± 0.616.7 ± 0.2
CXCR715.4 ± 2.323.3 ± 2.722.2 ± 4.836.5 ± 8.9§
COL353.0 ± 5.8102.1 ± 12.756.6 ± 18.1195.6 ± 91§
COL100.4 ± 0.20.7 ± 0.40.5 ± 0.10.2 ± 0.3
DKK311.3 ± 3.120 ± 6.8§14.8 ± 2.722.5 ± 3.9
DEL16.6 ± 1.45.7 ± 0.395.4 ± 0.357.7 ± 1.6
Htra159.4 ± 13.1107.6 ± 10.180.6 ± 22.1216.8 ± 108§
IDE35.1 ± 4.843.3 ± 6.471.6 ± 9.573.4 ± 13.3
IGF-10.02 ± 0.0050.06 ± 0.007§0.02 ± 0.0040.04 ± 0.006§
IL330.46 ± 0.120.74 ± 0.2§0.76 ± 0.010.81 ± 0.2
MMP319.0 ± 5.844 ± 5.8§24 ± 3.046 ± 8.4§
MMP1375.7 ± 982.1 ± 2076.7 ± 2399.7 ± 18
Nox415 ± 6.612 ± 2.19.5 ± 3.711 ± 3.9
Periostin6.8 ± 0.629.8 ± 1.114.9 ± 3.825.2 ± 8.9§
Sfrp211.1 ± 1.517.3 ± 217.6 ± 6.925.7 ± 2.7§
TSG62.6 ± 0.64.1 ± 0.43.4 ± 0.987.8 ± 4.8

Type X collagen is a marker of chondrocyte hypertrophy, and OA chondrocytes are thought to assume a hypertrophic phenotype, but we did not see a significant difference in gene expression in either younger or older DMM-operated mice, nor did we find increased expression of the gene for MMP-13. However, a previous study examining older C57BL/6 mice with spontaneous OA also did not show an increase in type X collagen (24), and, unlike MMP-3 and ADAMTS-5, MMP-13 expression may not increase until OA is in a more advanced stage (4). Type III collagen RNA was increased in the DMM-operated joints of both young and old mice, but the increase only reached statistical significance in old mice.

Immunohistochemical analysis.

We selected 3 genes from the list validated by real-time PCR in order to examine protein tissue distribution. The negative control sections did not exhibit any immunostaining (not shown). There was very strong cellular staining for IL-33 in chondrocytes in articular cartilage, meniscus, and growth plate, and in synovial cells (Figure 4). IL-33 was variably present in osteocytes, osteoblasts, skeletal muscle, vascular endothelium, and hemopoietic cells. CCL21 had a much more localized distribution, with immunostaining in articular cartilage chondrocytes, meniscal cells, and growth plate chondrocytes. Immunostaining for CCL21 was also present in vascular endothelium, very low numbers of osteocytes, growth plate matrix, and (minimally) in skeletal muscle. Periostin was present in cartilage matrix (articular cartilage, meniscus, growth plate), in low numbers of chondrocytes in these tissues, and in some osteoblasts in the periosteum. We did not examine sufficient numbers of sections from the 4 groups to determine the effects of age and OA on staining intensity, but we plan to do so as part of future studies using these and additional proteins.

Figure 4.

Immunohistochemical staining for selected proteins in mouse knee joints. A and B, Strong positive staining of chondrocytes in articular cartilage (arrows) and meniscus (arrowheads) for interleukin-33 in a young mouse that underwent destabilization of the medial meniscus (DMM) (A) and for CCL21 in an old sham-operated mouse (B). C and D, Positive staining for periostin in articular cartilage matrix (arrows) in a young DMM–operated mouse (C) and in periosteum (arrows) in a young sham–operated mouse (D). Bars = 50 μm.


Similar to findings in humans, we found that age contributes to the development of both spontaneous and injury-induced OA in mice. Mild OA-like pathology was present in the unoperated knees of the 12-month-old mice, suggesting that mice at this age are in the early stages of developing naturally occurring OA, similar to what has been described in pathologic studies of human knees at the equivalent age of ∼40 years (25). The older mice also exhibited more severe histologic OA after a joint injury that destabilized the medial meniscus, which is also consistent with studies on the development of OA after joint injury in humans (3) and supports the use of this model to study age-related differences. More severe OA lesions were also noted in the contralateral (unoperated) knee of older DMM-operated mice relative to the contralateral (unoperated) knee of older sham-operated mice, which could have been due to either gait changes or systemic factors released from the DMM-operated joint, suggesting that contralateral joints should not be used as controls in mouse models of surgically induced OA.

Annotations of genes expressed at lower levels in the older sham-operated control animals included extracellular matrix, metabolic processes, and tissue development, with a predominance of cartilage extracellular matrix genes including those for types II, IX, and XI collagen, aggrecan, COMP, and link protein, as well as for the transcription factor SOX9 and the growth factor TGFβ2. These findings are consistent with those of studies demonstrating an age-related decline in matrix production when chondrocytes are stimulated with anabolic factors (for review, see ref.26) and of a study showing reduced immunostaining for TGFβ2 in the articular cartilage of older mice (27). We also noted a decrease in expression of the gene for HMGB-2, consistent with a previous study in mice that showed that decreased levels may contribute to chondrocyte death in the superficial zone (28). Genes expressed at higher levels in older sham-operated controls included chemokine and HLA genes and had annotations for immune and defense response. Expression of the gene for CXCR2, which serves as an IL-8 receptor and has been found to play an important role in promoting cell senescence (29), was also increased in the sham-operated joints of older mice.

In the joints with surgically induced OA, more genes were significantly up-regulated in older mice (n = 421) compared to younger mice (n = 72), potentially indicating a more active disease process. Genes with increased expression in OA joints of older mice included extracellular matrix genes, such as those for aggrecan and type II collagen, the expression of which decreased with age in the sham-operated controls. DAVID analysis and IPA of the relatively smaller set of genes that were up-regulated in younger (but not older) mice with OA identified significant annotations for immune response genes and B cell signaling, while these annotations were found in genes expressed in the sham-operated joints of older mice. Genes with muscle-related annotations and genes involved in calcium signaling were more often down-regulated in the younger mice with OA, while these genes were up-regulated in DMM-operated joints of older mice. Because we did not include muscle in the tissue used for RNA isolation, muscle-related genes are most likely being expressed by other tissues including cartilage, as previous studies have noted chondrocyte expression of genes such as that for α-smooth muscle actin (30).

A unique aspect of the present study was that the RNA isolated for microarray analysis was extracted from the multiple tissues that make up the joint rather than from a single tissue. Although this approach might be less sensitive in detecting genes that were up- or down-regulated in a single tissue and might limit the ability to determine which particular tissue contributed to expression of a specific gene, it has the advantage of allowing discovery of genes that are more globally involved in the OA process. Despite the potential limitations, the genes for aggrecan, link protein, and type II collagen, primarily expressed in articular cartilage and in the inner zone of the meniscus, were detected on the arrays as being up-regulated in DMM-operated joints of older mice, even though cartilage and meniscus loss was greater in these mice than in the younger mice.

Periostin was up-regulated in DMM-operated joints of older but not younger mice. Periostin is a vitamin K–dependent (Gla-containing) protein produced by osteoblasts and also found in the periosteum (31). We confirmed its presence in periosteum and osteoblasts and in addition found that it was present in the cartilage matrix and in low numbers of chondrocytes. Periostin appears to play a role in diverse processes including tooth development (32), cancer metastasis (33), and tissue repair after injury, such as repair of heart tissue after a myocardial infarction (34). Likewise, secreted Frizzled-related protein 2, which can augment Wnt3a signaling (35) and whose gene expression was increased in DMM-operated joints from older mice, has also been shown to be involved in myocardial repair (36). The increased expression of these and other genes involved in extracellular matrix formation and tissue repair indicates an active repair response in the knee joints of the older mice.

IL-33 was found to be up-regulated in the DMM-operated joints of younger but not older mice, although its expression was higher in sham-operated joints of older mice than in sham-operated joints of younger mice. IL-33 is a member of the IL-1 superfamily and is thought to serve as an alarmin in a number of inflammatory diseases, including RA (for review, see ref.37), but, to our knowledge, it has not been reported to be up-regulated in OA. In RA, IL-33 was found in synovial tissue; however, it can also be expressed by osteoblasts, where it is thought to inhibit bone resorption (38). We found that although IL-33 is fairly widely distributed in joint tissues, there was strong and consistent positive immunostaining for IL-33 in articular and growth plate chondrocytes and meniscal cells, suggesting that it may function in these tissues as well.

The 55 genes that were similarly expressed in DMM-operated knees relative to sham-operated knees of younger and older mice were involved in extracellular matrix remodeling and included up-regulated genes involved in matrix degradation, such as those for MMP-2, MMP-3, and HtrA serine peptidase 1. Although not as extensively studied as the MMPs, several serine proteases, including HtrA serine peptidase 1, have been shown to be increased in human OA cartilage (39). HtrA serine peptidase 1 has been implicated in degradation of fibronectin (40) and aggrecan (21) and was also found by others to be increased in the cartilage of mice 8 weeks after DMM surgery (41). In addition, HtrA serine peptidase 1 was a prominent protease found in a proteomic analysis of human OA cartilage (42). CCL21 is a novel chemokine gene found to have increased expression in DMM-operated joints of both young and old mice; it was found by immunostaining to be localized to chondrocytes and meniscal cells and the growth plate matrix. CCL21 is a ligand for CCR7, the gene for which was also increased in DMM-operated joints of young mice and in sham-operated control joints of older mice. A recent study demonstrated elevated levels of CCL21 in both RA and OA synovial fluid, with higher levels in RA synovial fluid, compared to those in normal controls (43).

In a recently reported cartilage microarray study that used the rat meniscal tear model of OA, Wei et al (9) integrated their findings with those of previous microarray studies of differential gene expression in the rat ACLT model (8) and in human OA cartilage (5) and generated a list of 20 OA genes that were in common in the human study and at least 1 rat model. We searched for those 20 genes in our list of 548 genes that were differentially expressed in either younger or older DMM-operated mice and in the list of 861 genes differentially regulated by age in sham-operated joints of young and older mice. We found 7 genes in common in the list of genes differentially expressed in DMM-operated mice (COL3A1, COL6A2, lumican, MMP3, NDRG2, PCOLCE, and TIMP1); 6 were differentially regulated in the older DMM-operated mice and 5 in the young DMM-operated mice. We also found 4 genes in the list of genes in sham-operated mice that were differentially regulated with age in the same direction as differential regulation in the OA list (LTBP2 [the gene for latent TGFβ binding protein 2], NDRG2, SERPINA1, and TIMP1). Given the importance of TGFβ in joint tissues, the increase in LTBP2 expression in OA and with age may be particularly important.

We also compared our list of genes differentially expressed in DMM-operated mice with a list of 150 genes reported by Hopwood et al to be differentially expressed in human OA subchondral bone (10), and we found 4 genes in common (CCR2, crystallin alpha B, synuclein alpha, and tyrosyl–transfer RNA [tRNA] synthetase). Of these, the expression patterns matched for DMM-operated joints of young but not old mice for crystallin alpha B and tyrosyl-tRNA synthetase (down-regulated) and synuclein alpha (up-regulated). However, we found more genes (n = 10) in common in a comparison of the bone list with the list of genes differentially expressed in sham-operated young versus old mice, with 8 genes having the same expression pattern (crystallin alpha B, guanine nucleotide binding protein alpha Z, glycoprotein V, lymphotoxin beta, matrix extracellular phosphoglycoprotein with ASARM motif, RAB27B, selectin P, and tubulin, beta 1).

Finally, we made a similar comparison between the mouse joint microarray gene lists and a list of 260 genes found to be ≥2-fold differentially expressed between inflamed and uninflamed biopsy specimens of human synovial tissue obtained from subjects undergoing meniscectomy for meniscal injuries (12). A total of 30 genes were present in both the synovial gene list and the mouse joint lists, with expression of 8 genes regulated in the same direction in synovium and the lists of genes differentially regulated in DMM-operated and sham-operated mice (see Supplementary Table 2, available on the Arthritis & Rheumatism Web site at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)1529-0131). One of those was the gene for CCL21, which our immunohistochemical staining showed to be present in articular cartilage, suggesting that this gene is up-regulated in more than one tissue in joints with meniscal damage. The comparison of genes regulated in the same direction between the sham-operated joints of old–young mice and the human synovial samples showed 20 genes in common, demonstrating that a significant number of genes that are up-regulated with age in the joint are also found in inflamed synovia from meniscal injuries. This included the gene for IL-7R, which our group had previously shown to be expressed by human chondrocytes (44).

In summary, the analysis of gene expression in joint tissues in a meniscal injury model of OA found OA-related genes that had been previously reported in other animal models and in human OA, but also revealed novel genes and pathways that could be important in the OA process. The results also demonstrated clear age-related transcriptional differences in both the sham-operated control joints and the DMM-operated joints. These findings demonstrate the importance of age when considering the results of animal model studies of OA. Most studies that have used transgenic mice to study the role of specific genes in OA have used animals at an age similar to that in our younger group. Genes and pathways important in the OA process may be missed if only young animals are used in such studies. Age at onset of joint injury clearly affects the way in which the cells within joint tissues respond. The older animals exhibited a very active response to joint injury, including the up-regulation of matrix genes, chemokines, and matrix-degrading enzymes, consistent with the concept that OA is not a degenerative disease but rather a condition that activates remodeling of joint tissues. Further studies of the differences in the transcriptional response between joints of younger and older mice should help elucidate the mechanisms underlying the contribution of age to the development of OA.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Loeser 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 conception and design. Loeser, Carlson, Callahan, Leng, Fetrow.

Acquisition of data. Loeser, McNulty, Carlson, Callahan, Ferguson, Chou.

Analysis and interpretation of data. Loeser, Olex, McNulty, Carlson, Chou, Leng, Fetrow.


The authors thank Mary Zhao for technical assistance and the Microarray Core Facility of the Comprehensive Cancer Center of Wake Forest University for performing the microarrays. We also thank Dr. Sonya Glasson for instructions on the DMM model, Drs. Carla Scanzello and Blair Hopwood for providing microarray data sets from synovium and bone, respectively, and Dr. John Williams for helpful discussions.