Osteoarthritis (OA) is a common joint disorder affecting elderly populations. The prevalence of radiographic knee OA in adults was reported as 37.4% in the USA and 61.9% in Japan, and it increased with age [1] [2] . Indirect and nonmedical costs for OA have significant implications, given its high prevalence and an upcoming aging society [3] .

One of the characteristic changes in OA is wear and tear of articular cartilage. Chondrocytes are the only cellular component in articular cartilage, and they maintain homeostasis of articular cartilage synthesizing matrix molecules such as collagen and proteoglycan. In OA, cartilage homeostasis is disrupted by the loss of balance between cartilage anabolism and catabolism [4] . Cytokines modulate the degradation of articular cartilage. Interleukin-1 (IL-1) plays a central role in the pathogenesis of OA [5] . IL-1 is present in OA chondrocytes [6] , and it stimulates chondrocytes to produce chemical mediators such as prostaglandin E₂, nitric oxide and metalloproteinases [7] -[9] . These mediators have catabolic effects on cartilage [10] .

There are no evident treatments that cure OA or delay the disease progression as yet. Thus, the current goal of treatment is to modify symptoms of OA. Some guidelines recommend NSAIDs including selective COX-2 inhibitors [11] [12] . However, conventional NSAIDs have serious adverse effects and degrade articular cartilage [13] -[15] . In our in vitro experiments, both diclofenac and celecoxib induced apoptosis in chondrocytes in a dose-dependent manner, however the latter induced it to a lesser degree [16] . Recently, celecoxib is suggested to have potential disease-modifying properties in OA [17] . Thus, celecoxib might be preferable to use for the treatment of OA. The aim of this study is to identify potent candidate proteins that are targets of celecoxib in OA chondrocytes.

To investigate cellular phenotypes under different conditions, a method of comprehensive analysis of proteins has been developed [18] [19] . In the present study, chondrocytes from OA patients were prepared under four different conditions: 1) unstimulated (control), 2) treated with IL-1β, 3) treated with IL-1β and celecoxib, and 4) treated with celecoxib. To simulate in vivo conditions of OA, chondrocytes were stimulated with IL-1β, and celecoxib was added to simulate the clinical effects of celecoxib [16] . Moreover, stimulation by celecoxib alone was considered to reflect the effects of celecoxib on senescent chondrocytes.

To investigate whether celecoxib affects protein profiles of chondrocytes, and also whether celecoxib modulates the effects of IL-1β on chondrocytes, we applied 2D-DIGE to chondrocyte cellular proteins. Then, we separated proteins extracted from unstimulated chondrocytes (control) and the stimulated chondrocytes (treated with Il-1β, celecoxib, Il-β + celecoxib) by 2DE (Figure 1). As a result, more than 1100 protein spots were detected per gel image, of which 86 protein spots showed significantly different intensity among the 4 groups.

Several comprehensive proteomic analyses of cancer cells treated with celecoxib have been studied in lung can-

^*Intensity was resumed by addition of celecoxib.

cer, oral squamous cell carcinoma and colorectal carcinoma [22] -[24] . Another study used cardiomyocytes to investigate the cardiovascular toxicity of NSAIDs including celecoxib [25] . The present study is believed to be the first one to investigate the effects of celecoxib on OA chondrocytes.

With the proteomic analysis, no less than 86 proteins were found to be affected by celecoxib and/or IL-1β in this study. Out of these 86 proteins, we could identify 5 decreased proteins and 6 enhanced proteins in chondrocytes treated with celecoxib. COX-2 was not implicated in this condition, as chondrocytes from OA as well as normal cartilage expressed extremely low levels of PGE₂ and COX-2 mRNA when incubated in an unstimulated condition [26] . Thus, the modification of these proteins was mediated by mechanisms unrelated to COX-2. Dimethyl-celecoxib, which is a close structural analogue of celecoxib, lacks the inhibitory function of COX-2, and has antitumor effects through activation of endoplasmic stress in tumor cells [27] .

Two proteins out of the 5 decreased ones were cytoskeletons, actin and tubulin. The remaining proteins were ATPase family AAA domain-containing protein 3C, serpin H1 and cAMP-dependent protein kinase type II-beta regulatory subunit. The AAA+ (ATPases associated with various cellular activities) family is a large and func-

tionally diverse group of enzymes that are able to induce conformational changes in a wide range of substrate proteins [28] . Serpin H1 is one of the serine protease inhibitors and is commonly known as heat shock protein 47 (HSP47). HSP47/Serpin H1 is a molecular chaperon and maintains collagen biosynthesis. Antibodies against it are found in autoimmune diseases such as rheumatoid arthritis and systemic lupus erythematodes [29] . HSP47 is a stress protein that is induced under some stress in endoplasmic reticulum (ER). In early OA in a mouse model, HSP47 mRNA was up-regulated associated with IL-1β, IL-6 and TNFα, suggesting that celecoxib may suppress the stress in OA chondrocytes [30] . The significance of cAMP-dependent protein kinase and AAA+ family protein on osteoarthritis is unclear as far as we have found in the literature.

The 6 enhanced proteins included apolipoprotein A-V, 78 kDa glucose-regulated protein (GRP78), glutamate carboxy peptidase 2, mitochondrial stress-70 protein and sorting nexin-9. GRP78 was identified from 2 different spots. Apolipoprotein A-V was first found in 2001, and it maintains plasma levels of triglyceride [31] . Sorting nexin 9 is widely expressed and plays a role in endocytosis [32] . An alternative name for mitochondrial stress- 70 protein is 75Kd glucose regulated protein (GRP75). It is located in mitochondria and causes cell proliferation [33] . No implication of chondrocytes in OA with these proteins has been reported as yet. On the other hand, GRP78 was physiologically induced by starvation [34] , and interestingly, also by dimethyl celecoxib, the analog of celecoxib that lacks COX-2 activity [35] , which suggested that GRP78 induction was COX-2 independent. Proteomic analysis of human chondrocytes has shown that GRP78 as well as GRP94 increased in late stage OA [36] , and advanced glycation end products (AGEs) induced GRP78 in chondrocytes [37] . GRP78 is an ER chaperon and is induced by stress to refold unfolded proteins, however, sustained ER stress leads to an inflammatory response and apoptosis [38] . Under stress, GRP78 is expressed and its specific overexpression was reported to reduce apoptosis in Chinese hamster ovary cells [39] . Thus, appropriate induction of GRP78 by celecoxib may have a chondroprotective effect, in part.

We identified 2 proteins whose IL-1β-induced expression was neutralized by celecoxib; CD200 receptor increased and moesin decreased with IL-1β stimulation, and both returned to baseline by addition of celecoxib. CD200 receptor (CD200R) is a ligand of CD200, also known as OX-2. Expression of CD200R is found in the macrophage lineage cells and delivers an inhibitory signaling [40] . Activation of CD200R by CD200-Fc ameliorates collagen-induced arthritis in mice [41] . Expression of CD200R in chondrocytes has not been studied yet, and further research is needed for the function of CD200R in chondrocytes of OA. Moesin together with Erzin and Radixin serves as a nexus between cell membrane and underlying cytoskeleton. It regulates signaling pathways by linking transmembrane receptors and downstream signaling components [42] . Though there is very little knowledge regarding the effects of moesin on chondrocytes, increased phosphorylation of moesin has been observed on stretched chondrocytes [43] .

Some limitations of our study should be mentioned. First, the clinical serum concentration of celecoxib was not more than 1/10 of the concentration used in this study. As the aim of this study was to identify candidate proteins, we used higher concentration to detect as much potent proteins as possible. Validation of each protein should be necessary in the next step. Second, a number of patients who provided chondrocytes was only 3. As a larger number of patients are required for representing OA, validation of each protein will be necessary using an appropriate number of patients. Even though, it should be emphasized that differences of protein expression reached statistical significance in this study. Lastly, chondrocytes from healthy subjects were not evaluated as a control. Our previous study showed that responses of normal chondrocytes by IL-1 stimulation were different from those of OA [26] and the results might be different. While celecoxib is used for patients with symptom of OA, normal chondrocytes should also be evaluated as a scientific interest.

In conclusion, the protein profile of OA chondrocytes was found to change after the administration of celecoxib. AAA+ protein, HSP47/Serpin, cAMP-dependent protein kinase type II-beta regulatory subunit, alpha-actin-4 and tubulin decreased with administration of celecoxib, and apoplipoprotein A-V, glutamate carboxipeptide 2, mitochondrial stress-70 protein, sorting nexin-9 and GRP78 increased with addition of celecoxib. Of these, GRP78 is a stress protein, and overexpression of it may have a chondroprotective effect by inhibiting apoptosis. In the present study, celecoxib modulated IL-1β-stimulated chondrocytes, and CD200R and moesin were identified as two such resulting proteins. Further investigation is needed to elucidate the function of eachprotein in OA.

HN received a research grant from Pfizer Inc.