Inhibition of cyclooxygenase-2 down-regulates osteoclast and osteoblast differentiation and favours adipocyte formation in vitro

European Journal of Pharmacology 572 (2007) 102 – 110 www.elsevier.com/locate/ejphar

Inhibition of cyclooxygenase-2 down-regulates osteoclast and osteoblast differentiation and favours adipocyte formation in vitro☆ Maarit Kellinsalmi a,⁎, Vilhelmiina Parikka b , Juha Risteli c , Teuvo Hentunen b , Hannu-Ville Leskelä a , Siri Lehtonen a , Katri Selander d , Kalervo Väänänen b , Petri Lehenkari a a

d

Clinical Research Centre, Department of Surgery, University of Oulu, Finland b Department of Anatomy, University of Turku, Finland c Department of Clinical Chemistry, University of Oulu, Finland Department of Medicine, Division of Hematology-Oncology, University of Alabama at Birmingham, AL, USA Received 30 January 2007; received in revised form 8 June 2007; accepted 12 June 2007 Available online 29 June 2007

Abstract Nonsteroidal anti-inflammatory drugs (NSAIDs) inhibit cyclooxygenases (COX) and are widely used for post-trauma musculoskeletal analgesia. In animal models, NSAIDs have been reported to delay fracture healing and cause non-union, possibly due to the drug-induced inhibition of osteoblast recruitment and differentiation. To further investigate the cellular effects of these drugs in the context of bone healing, we examined the effects of COX-1 inhibitor indomethacin and COX-2 inhibitors, parecoxib and NS398 on osteoclast and osteoblast differentiation and activity in vitro. We discovered that all tested COX-inhibitors significantly inhibited osteoclast differentiation, by 93%, 94% and 74% of control for 100 μM indomethacin, 100 μM parecoxib and 3 μM NS398, respectively. Furthermore, inhibition of COX-2 reduced also the resorption activity of mature osteoclasts. All tested COX-inhibitors also significantly inhibited osteoblast differentiation from human mesenchymal stem cells. Simultaneously, the number of adipocytes was significantly increased. The adipocyte covered areas in the cultures with 1 μM indomethacin, 1 μM parecoxib and 3 μM NS398 were 9%, 29% and 24%, respectively, as compared with 6% in the control group. This data suggests that COX-2 inhibition disturbs bone remodelling by inhibiting osteoclast differentiation and diverting stem cell differentiation towards adipocyte lineage instead of osteoblast lineage. In conclusion, our results further suggest cautious use of COX-2 inhibitors after osseous trauma. © 2007 Elsevier B.V. All rights reserved. Keywords: Cyclooxygenase; Human mesenchymal stem cell; Adipocyte; Parecoxib

1. Introduction

☆ Financial support: this study was supported by funds from National Technology Agency of Finland (TEKES), University Hospital of Oulu and Academy of Finland. ⁎ Corresponding author. University Hospital of Oulu, Department of Surgery, P.O. Box 21, Kajaanintie 22, FI-90029 University Hospital of Oulu, Finland. Tel.: +35 8407315689; fax: +35 885376335. E-mail addresses: [email protected] (M. Kellinsalmi), [email protected] (V. Parikka), [email protected] (J. Risteli), [email protected] (T. Hentunen), [email protected] (H.-V. Leskelä), [email protected] (S. Lehtonen), [email protected] (K. Selander), [email protected] (K. Väänänen), [email protected] (P. Lehenkari).

0014-2999/$ - see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.ejphar.2007.06.030

Cyclooxygenase (COX) is the rate-limiting enzyme in the conversion of arachidonic acid to prostaglandins. Two forms of COX enzyme have been characterized extensively thus far: cyclooxygenase-1 (COX-1) and COX-2. COX-1 is involved in the maintenance of physiologic functions like haemostasis and gastric protection. The cyclooxygenase-2 enzyme (COX-2) is produced by a variety of cells in vivo in response to trauma and inflammation (Noor and Gajraj, 2003; Okada et al., 2003; Zhang et al., 2002). COX-3 has only recently been discovered in dogs and rats and its function may be related to paracetamolinduced hypothermia and analgesia (Botting and Ayoub, 2005; Kis et al., 2005).

M. Kellinsalmi et al. / European Journal of Pharmacology 572 (2007) 102–110

Nonsteroidal anti-inflammatory drugs (NSAIDs) inhibit the synthesis of COX-enzymes and hence the production of prostaglandins, which mediate pain. The various NSAIDs exhibit differences in COX isoenzyme inhibition profiles (Gierse et al., 1999). Due to their analgesic and antiinflammatory effects, these drugs are also widely used in the treatment of post-fracture pain and swelling. In addition to their role as mediators of pain, prostaglandins are also potent regulators of bone cell functions (Okada et al., 2003; Rawlinson et al., 2000). Especially prostaglandin E2 (PGE2) can stimulate the differentiation of both osteoblasts and osteoclasts (Liu et al., 2005). The effects of prostaglandins on osteoclast formation are mediated via enhancing the action of receptor activator of nuclear factor-kappa B ligand (RANKL) on osteoclast precursors. This induction is mediated by osteoblasts. The enhanced action of RANKL in osteoclast precursors also stimulates the differentiation of osteoblasts. (Okada et al., 2003; Liu et al., 2005; Choi et al., 2005; Wei et al., 2005). NSAIDs have different capabilities to inhibit different COX isoenzymes (Gierse et al., 1999). Bone repair is a tightly regulated process which involves a highly orchestrated participation of several cell types. During the initial step of fracture healing, a haematoma covers the fracture area and inflammatory cells, such as macrophages and leukocytes, are attracted to the site of trauma (Ozaki et al., 2000; Simon et al., 2002). These cells provoke an inflammatory response, which involves various cytokines, growth factors and arachidonic acid metabolites and activated osteoclasts, all of which contribute to the osteoblastic proliferation and differentiation of stem cells. The final osteosynthesis occurs 3–6 weeks after the fracture. First, the organic extracellular matrix is produced and mineralized after which bone is continuously remodelled by resorption and reformation (Noor and Gajraj, 2003; Barnes et al., 1999; Einhorn 1998). Data from previous animal studies suggests that inhibitors of COX-2 may interfere with the bone healing process (for review see Li et al., 2006). More specifically, Zhang et al. have demonstrated that healing of stabilized tibia fractures was significantly delayed in COX-2−/− mice (Zhang et al., 2002). Goodman et al. (2002) have showed that a COX-2 selective NSAID decreased bone growth in vivo. Giordano et al. reported a significant delay in fracture healing in tenoxicam group and the osseous union was more incomplete the sooner the NSAIDtreatment was initiated (Giordano et al., 2003) So far, there are no prospective clinical studies on COX-2 and fracture healing. The only clinical studies thus far have been retrospective and focused on spinal fusions (Glassman et al., 1998; Reuben and Ekman, 2005). The aim of this study was to further clarify the effects of the COX-inhibitors, with differences in COX-inhibition profiles, on osteoblast and osteoclast differentiation and activity, using various in vitro assays. Our results further suggest that COX-2 inhibition significantly decreases osteoclast differentiation (Sato et al., 1997; Kotake et al., 1999; Okada et al., 2000). We also show that COX-2 inhibitors decrease osteoblast differentiation and promote stem cell commitment towards the adipocyte lineage in human mesenchymal stem cell cultures.

103

2. Materials and methods 2.1. Reagents Indomethacin was purchased from Alpharma OY (Helsinki, Finland), parecoxib from Pfizer OY and NS398 (N-[2-(cyclohexyloxy)-4-nitrophenyl]-methanesulfonamide) from Sigma (St. Louis, MO, USA). α-Modified Essential Medium (α-MEM), phosphate buffered saline, pH 7.2 (PBS), 1 M HEPES Buffer, Trypsin–EDTA solution and penicillin/streptomycin solution (10 000 units/ml penicillin G sodium and 10 000 μg/ml streptomycin sulphate in 0.85% saline) were purchased from Gibco BRL (Paisley, the UK). Peroxidase-conjugated WGA-lectin, dexamethasone, L-ascorbic acid 2-phosphate, β-glyserolphosphate, bovine serum albumin, 4-p-nitrophenylphosphate and TRACP histochemical kit no 386 were purchased from Sigma (St. Louis, MO, USA). O-cresolphthalein-complexone was purchased from Roche Diagnostics (Mannheim, Germany). Oil Red O was obtained from Sigma. Trizol reagent was purchased from Invitrogen Life Technologies (Carlsbad, CA), Prime RNAse inhibitor from Eppendorf (Hamburg, Germany), M-MLV reverse transcriptase from Invitrogen Life Technologies and all primers used were from Oligomer (Helsinki, Finland), and Dynazyme II from Finnzymes (Helsinki, Finland). 2.2. Osteoclast resorption assay The bone slices were prepared from frozen bovine cortical bone shafts and they were cut with a diamond saw to a 200 μm thickness, sterilized with sonication, rinsed briefly in 70% ethanol and then transferred to fresh 37 °C medium into 24-well plates. Osteoclasts were isolated from the long bones (femur, tibia, humerus) of 1- to 2-day old newborn male Sprague– Dawley rats, as previously described in detail. (Boyde et al., 1984; Chambers et al., 1984). Briefly, osteoclasts were scraped from the endosteal surface of split long bones into α-MEM and centrifuged at 200 ×g for 10 min. The cells were then resuspended in the same medium and an aliquot of 50 μl was added to bone slices which were placed on a piece of parafilm. The cells were allowed to attach onto the bone slices for 30 min at 37 °C 5%CO2/95% air. After that, the non-attached cells were rinsed off with sterile PBS. The bone slices with the attached cells were then placed into 24-well plates and cultured in αMEM buffered with 20 mM HEPES, 2 mM L-glutamine, penicillin (100 IU/ml), streptomycin (100 μg/ml) and 10% heatinactivated fetal calf serum, in the presence or absence of 1 μM, 10 μM and 100 μM indomethacin or parecoxib or 0,03 μM, 0,3 μM and 3 μM NS398. The cultures were stopped after 48 h by fixing the cells with 3% paraformaldehyde and 2% sucrose for 5 min. To detect apoptotic osteoclasts, the cells on the bone slices were stained for tartrate resistant acid phosphatase (TRACP) using a histochemical kit according to the manufacturer's instructions (Sigma) and with Hoechst, to detect the nuclei, as described in detail previously (Selander et al., 1996). In some experiments

104

M. Kellinsalmi et al. / European Journal of Pharmacology 572 (2007) 102–110

the bone slices were also stained for 5 minutes with toluidine blue in 1% sodium borate. Osteoclasts were recognized by their multinuclearity. To quantify the resorption pit areas, stainings with peroxidase-conjugated WGA-lectin (diluted 1:40 in PBS) were performed, as previously described in detail by Selander et al. (1994). With this method, the resorption pits are visualised as brown coffee-stain-like areas and can be quantified by combining the 20× light microscopy image (Nikon TMD Diaphot) to a colour-based computer-assisted image analyses system (MCID/M2, Imaging Research Inc., Brock University, Ontario, Canada). 2.3. Osteoclast differentiation assays from mouse bone marrow cultures Cells were harvested from the tibias and femurs of 8– 12 weeks old sacrificed NMRI mice. Briefly, after dissection, the long bones were dipped in 70% ethanol and placed into sterile α-MEM. The bone ends were cut and the marrow was flushed out with α-MEM containing 20 mM HEPES. The obtained cells were pelleted by centrifugation (200 ×g, 10 min) and resuspended in 10 ml of α-MEM supplemented with 10% fetal calf serum, penicillin/streptomycin (100 IU/100 μg/ml) and 20 mM HEPES. Adherent cells were removed by allowing the cells to attach to the tissue culture dish in CO2-incubator for 2 h. Non-adherent cells were collected and the dish was washed with 5 ml of medium. These two fractions were pooled and the cell numbers were counted. The cells were then pipetted onto 24-multiwell plates (1 × 106/ml) and cultured for 6 days in the presence or absence of 1 μM, 10 μM and 100 μM indomethacin or parecoxib and 0.03 μM, 0.3 μM and 3 μM NS398. Half of the media were changed on day 3. At the end of the culture period, media were removed and the cells were fixed with 3% paraformaldehyde in PBS for 20 min, rinsed with PBS and stained for TRACP as above. Supernatants from these cultures were also collected for TRACP measurements (as explained below). 2.4. Mouse spleen cell cultures The spleens were removed from 8–12-weeks old sacrificed NMRI mice and placed in a tissue cell culture dish (diameter 10 cm) containing 5 ml of α-MEM–20 mM HEPES in the laminar hood. Several cuts were made with a scalpel on the spleen, which was then gently rubbed with a syringe piston to release cells into the medium. Cells were collected by centrifugation (10 min, 400 ×g) and the cell pellet was resuspended in 10 ml of α-MEM containing 10% fetal calf serum, penicillin-streptomycin, and 20 mM HEPES. The cells were counted and 5 × 105 cells were added/well in 48-well plates (0.5 ml culture medium/well). Osteoclast formation was induced with the combination of 20 ng/ml of RANKL and 10 ng/ml macrophage colony stimulating factor (M-CSF) and the cells were cultured in the presence or absence of 0,003 μM, 0,03 μM and 0,3 μM NS398. Half of the medium was changed on day 3. On day 7, the culture media were collected from the wells and TRACP 5 b was determined from

the supernatants, as explained below (Alatalo et al., 2000). The animals were maintained in accordance with the guidelines of the Oulu University ethical committee for the use and care of experimental animals. 2.5. Human mesenchymal stem cell derived osteoblast culture The human mesenchymal stem cells were harvested from the femoral bone marrow before the application of the femoral part of hip prosthesis in total endoprosthesis operations. The method for human mesenchymal stem cell derived osteoblast culture has been previously described in detail (Leskelä et al., 2003). The bone marrow samples were cultured in flasks containing 5 ml of culture medium with 0.1 μmol/L dexamethasone at 37 °C in a humidified atmosphere in the presence of 5% CO2. The medium was removed after two days and the flasks were rinsed with Ca2+ and Mg2+-free PBS, fresh culture medium was added and half of the medium was replaced twice a week. Nearconfluence was achieved approximately in 2 weeks and the cells were then washed three times with PBS and detached with trypsin–EDTA digestion solution. After the cells were counted microscopically, they were transferred to 24-well plates at a density of 5000 cells per well. Hereafter the culture medium was supplemented with 0.05 mM ascorbic acid, 100 mM dexamethasone, 10 mM β-glycerophosphate and tested with various, indicated concentrations of indomethacin, parecoxib and NS398. The specific alkaline phosphatase activity was measured at 21 days and expressed as absorbance per protein (mg/ml). The assay buffer containing 50 mmol/L Tris 0.1% Triton X-100 (pH 7.6) and 0.9% NaCl was added to each well and the samples were frozen. After thawing of the samples, the assay buffer containing 0.1 mol/L 4-p-nitrophenylphosphate as substrate in 0.1 mol/L Tris, pH 10 and 1 mmol/L MgCl2 was added to each well. The reaction was stopped with 1.0 mol/L NaOH after incubation for 30 min, after which absorbance was read immediately at 405 nm in a plate reader (Victor2, Wallac Oy, Turku, Finland). Each sample was measured in duplicate. The protein concentrations were measured using the Bio-Rad protein assay kit according to the manufacturer's recommendations. For calcium determination after 5 weeks of culture, the cells were washed three times with PBS and incubated overnight at room temperature in 0.6 M HCl. The extracts were then complexed with o-cresolphthalein-complexone. The colorimetric reaction was read at 570 nm with a plate reader. Each sample was measured in duplicate. In order to determine the absolute calcium concentration, each measurement was compared with a calibrated standard (Roche Diagnostic Corporation, IN, USA). 2.6. Determination of procollagen propeptides (PINP and PIIINP) The concentrations of the amino-terminal propeptide of type I procollagen (PINP) and the amino-terminal propeptide of type III procollagen (PIIINP) in the human mesenchymal stem cell osteoblast cultures were measured with commercially available RIA kits (Orion Diagnostica Ltd, Espoo, Finland) according to

M. Kellinsalmi et al. / European Journal of Pharmacology 572 (2007) 102–110

105

of 40 s at 94 °C, 1 min at 68–72 °C and 90 sec at 72 °C were performed. β-actin expression was analysed by RT-PCR in a similar fashion to assure equal quality of RNA in the same samples. The primers used were 5′-CTCCTTAATGTCACGCACGATTTC-3′ (anti-sense, 24 mer) and 5′-GTGGGGCGCCCCAGGCACCA-3′ (sense, 20 mer). The products were analyzed by agarose gel

Fig. 1. The effects of indomethacin and parecoxib (1 μM, 10 μM, 100 μM) and NS398 (0.03 μM, 3 μM, 30 μM) on the number of osteoclasts (A) and resorption area (B) in rat osteoclast resorption assays. Data is expressed as mean ± S.D. P b 0.05 (⁎), P b 0.01 (⁎⁎), P b 0.001 (⁎⁎⁎) as compared with control.

manufacturer's instructions (Leskelä et al., 2003). Each sample was measured in duplicate. 2.7. Oil Red O staining Adipocyte areas were measured in hMCS cultures after Oil Red O staining solution (60% of Oil Red O stock and 40% of dH2O) (Birk et al., 2006). The stained areas were quantified using the same analysis method as for analysing the resorption pit area (as above). 2.8. RT-PCR for adiponectin measurement Total RNA was isolated from cultured cells by Trizol reagent according to manufacturers' instructions. DNAsetreated RNA (1 μg) was used for the cDNA synthesis, which was performed at 42 °C for 45 min in a reaction mixture containing 10 pmol of each primer for adiponectin (GI 44890057, sense 5′-CTCCGGTTTCACCGATGTCTCCC3-3′, 23 mer), 15 U Prime RNAse inhibitor, 100 U M-MLV reverse transcriptase, 1 × first-strand buffer, 2.5 mM dithiothreitol and 500 μM dNTPs in a total volume of 20 μl. All primers were from Oligomer. An aliquot of 10 μl of the product was taken for the PCR, which contained 1 U Dynazyme II, 1 × Dynazyme II reaction buffer, 500 μM dNTPs and 10 pmol of both anti-sense and sense (5′-CGGTCATGACCAGGAAACCACGAC-3′, 24 mer) primers in a total volume of 25 μl. A total of 30 PCR cycles

Fig. 2. The effects of COX-inhibition on the differentiation of osteoclasts, as measured by A) microscopically counting osteoclast numbers and (B) as a function of culture medium supernatant tartrate resistant acid phosphatase 5b (TRACP 5b) activity values in the presence of indomethacin, parecoxib (1 μM, 10 μM, 100 μM) or NS398 (0.03 μM, 0.3 μM, 3 μM) in mouse bone marrow osteoclast differentiation assays (Data in A and B is presented as % of control, mean ± S.D.). (C) Culture medium supernatant tartrate resistant acid phosphatase 5b (TRACP 5b) activity values in the presence of NS398. The experiment was done using mouse spleen cell cultures as a source for osteoclast precursors (CTRL = control, CTRL Vit-D = vitamin-D added in control medium).

106

M. Kellinsalmi et al. / European Journal of Pharmacology 572 (2007) 102–110

electrophoresis and visualized by ethidium bromide. The expected sizes of the PCR reaction products were 213 base-pairs for adiponectin and 540 base-pairs for β-actin. The products were analyzed by agarose gel electrophoresis, visualized byethidium bromide and quantified by Quantity One 4.5.0. program (BioRadLaboratories Inc).

The enzyme reactions were terminated by the addition of 25 μL of 0.32 mol/L NaOH to the wells, and the absorbance at 405 nm (A405) was measured with a model 2 Victor instrument (EG & G Wallac). Alatalo et al (2000) have shown that the amount of TRACP 5b in the mouse osteoclast differentiation assay correlates with the number of osteoclasts in the culture (r = 0.94).

2.9. TRACP 5b immunoassay 2.10. Statistical analyses The assays were performed as described previously (Alatalo et al., 2000). TRACP 5b rabbit anti-serum at 1:1000 dilution was incubated on anti-rabbit IgG-coated microtiter plates (EG & G Wallac) for 1 h. Culture medium (200 μL) was then added and further incubated in the wells for 1 h. The bound enzyme activity was detected using 8 mmol/L 4-nitrophenyl phosphate as substrate in 0.1 mol/L sodium acetate buffer for 2 h at 37 °C.

All statistical tests were performed using Student's paired t-test and one-way-ANOVA. Values are expressed as mean ± S.D. In the figures, ⁎,⁎⁎ and ⁎⁎⁎ represent the Student's t-test P-values of P b 0.05, P b 0.01 and P b 0.001, respectively, T-tests were only performed if the ANOVA tests showed significant difference between groups (P b 0.05).

Fig. 3. The effects of indomethacin and parecoxib (1 μM, 10 μM, 100 μM) and NS398 (3 μM, 30 μM) on the differentiation and recruitment of osteoblasts measured as a function of PINP (A, B, C) and PIIINP (D, E, F) after 1, 2 and 3 weeks in human mesenchymal stem cell culture.

M. Kellinsalmi et al. / European Journal of Pharmacology 572 (2007) 102–110

3. Results

107

3.2. COX-inhibition decreases osteoblast differentiation and matrix deposition

3.1. COX-inhibition decreases osteoclast differentiation We first investigated what effects the COX-inhibitors may have on the initial phase of bone remodelling, which involves increased osteoclast activity. This was done by using the rat osteoclast resorption assay which utilizes mature osteoclasts. Surprisingly, treatment with all tested COX-inhibitors decreased significantly the number of osteoclasts in this assay (Fig. 1A). Indomethacin decreased osteoclast number in a dosedependent manner, and the levels were 45%, 25% and 3% of control at 1 μM, 10 μM and 100 μM, respectively. The decrease was significant with all three tested concentrations (P b 0.001). Similarly, parecoxib and NS398 decreased the osteoclast numbers significantly (P b 0.01, Fig. 1A). Only the highest concentrations of COX-inhibitors significantly decreased also the resorption area, indomethacin (100 μM) by 95% (P b 0.01) and parecoxib (100 μM) by 84% (P b 0.05) of control, respectively. (Fig. 1B) NS398 was more potent and significantly decreased the bone resorption area by 61% (P b 0.05) of control, already at a very low concentration of 0.3 μM. The volumes of the resorption pits were not quantified with this method. Taken together, these results suggested to us that more than the actual osteoclast activity, COX-inhibitors affect osteoclast differentiation, as the resorption area/cell was unaltered. To further characterize how COX-inhibitors affect osteoclasts, we next tested the effects of these compounds in an osteoclast differentiation assay, where osteoclast precursors are derived from mouse bone marrows. Also in this assay the numbers of osteoclasts were dramatically decreased by all the tested COX-inhibitors. In the presence of 1 μM, 10 μM and 100 μM indomethacin, the number of osteoclasts were 5%, 7% and 1% of control, (P b 0.001) respectively. Treatment with parecoxib (1 μM, 10 μM and 100 μM) also induced a dosedependent decrease in the number of osteoclasts to 33%, 6% and 1% of control (P b 0.001). Similarly, treatment with NS398 had an inhibitory effect on osteoclast differentiation, the osteoclast numbers being 37%, 26% and 14% (P b 0.01) of control at concentrations 0.03 μM, 0.3 μM and 3 μM, respectively (Fig. 2A). The effect on osteoclast differentiation was further confirmed by TRACP 5b activity measurements from the culture supernatants in this assay. Also this parameter decreased significantly with all the three COX-inhibitors, with all tested concentrations. Indomethacin at the concentrations of 1 μM, 10 μM and 100 μM decreased TRACP 5b activity to 20%, 12% and 5% of control (P b 0.001), respectively. In the presence of parecoxib, the supernatant TRACP 5b activities were 44% (P b 0.01), 27% and 7% (P b 0.001) of control and in the presence of NS398 at concentrations 0.03 μM, 0.3 μM and 3 μM, TRACP 5b activities were 12%, 11%, and 8% (P b 0.001) of control respectively (Fig. 2B). Inhibition of osteoclast differentiation by NS398 was also evident when spleen cells were used as a source for osteoclast precursors. As expected, the presence of vitamin-D, which was used as a positive control, increased the supernatant TRACP 5b activity under these culture conditions. (Fig. 2C).

To determine the effect of COX-inhibitors on the second phase in bone remodelling, osteoblast recruitment and differentiation, we used osteoblast differentiation assays. More specifically, we measured the changes of PINP and PIIINP levels in the presence of COX-inhibitors during three weeks of culture of human mesenchymal stem cells under osteoblast differentiation promoting conditions. In the untreated control conditions, collagen synthesis, as measured by PINP quantity in the culture media supernatants, reached the highest level (200 μg/L) at 2 weeks. At all tested time-points one, two and three weeks of culture, the

Fig. 4. The effect of indomethacin and parecoxib (1 μM, 10 μM, 100 μM) and NS398 (0.003 μM, 3 μM, 30 μM) on osteoblast proliferation measured as protein level (A), on osteoblast differentiation measured as alkaline phosphate (ALP) specific activity (B) and on mineralization measured as calcium deposition (C) in human osteoblast cultures.

108

M. Kellinsalmi et al. / European Journal of Pharmacology 572 (2007) 102–110

significant decrease at the PINP levels were detected in the presence of indomethacin, parecoxib (Fig. 3A and B) and NS398 (Fig. 3C). After three weeks of culture, the decrease was significant with all concentrations of all three tested COXinhibitors (Fig. 3C). In the untreated control groups, the production of type III collagen measured by PIIINP levels in the culture media, was highest, 17 μg/L after three weeks of culture. After one week of culture, only minor decreases in PIIINP concentrations were induced by the COX-inhibitors (Fig. 3D). At two weeks of culture, 100 μM Parecoxib increased the PIIINP synthesis while 3 μM and 30 μM NS 398 decreased it (Fig. 3E) Thereafter, at three weeks the two highest parecoxib concentrations (10 μM and 100 μM) induced a statistically significant increase in the PIIINP concentration (157% and 147% of control, respectively). During the same culture periods, treatment with NS398 (3 μM, 30 μM) induced a significant decrease in the PIIINP level (to 80% and 49% of control, P b 0.01 and P b 0.001) (Fig. 3F), thus treatment with parecoxib and NS398 had opposite effects on the PIIINP. Treatment with indomethacin 10 μM and parecoxib 1 μM nearly doubled the protein level (P b 0.01 and P b 0.001) and NS398 3 μM increased the protein level to 150% of control (P b 0.01) (Fig. 4A). We next studied the effects of COXinhibitors on the differentiation and activation of human mesenchymal stem cell derived osteoblasts, by measuring the alkaline phosphatase (ALP) activity and calcium concentrations. A minor decrease in specific ALP activity was observed in the presence of 1 μM indomethacin (P b 0.05). NS398 at 0.003 μM and 3 μM concentrations significantly decreased the specific ALP activity to 52% and 61% of control (P b 0.01 and P b 0.05) Contrary to the effects of indomethacin and NS398 the highest concentration of parecoxib increased the specific ALP activity significantly (P b 0.05) (Fig. 4B). A significant decline of mineral deposition was observed in the presence of the tested COX-inhibitors. Indomethacin had the most evident dose-dependency, and with this treatment the calcium values declined to 54%, 46% and 38% of control (P b 0.001) (Fig. 4C). 3.3. COX-inhibition stimulates adipocytic differentiation of human mesenchymal stem cells

Fig. 5. Microscopic images of human mesenchymal stem cell cultures in the presence of culture medium, containing dexamethasone, indomethacin 100 μM and parecoxib 100 μM. The scale bar is 250μm in all images (A). The effects of indomethacin and parecoxib (1 μM, 10 μM, 100 μM) and NS398 (3 μM, 30 μM) on adipocyte formation were quantified as red oil o stained area (B). RT-PCR for adiponectin and β-actin (C) of cells cultured in the presence of the indomethacin and parecoxib (1 μM, 10 μM) or NS398 (0.3 μM,3 μM). The PCR product for adiponectin was quantified and normalised against β-actin (D) (C=control, C+D=control+dexamethasone).

Unexpectedly, a major increase in the number of adipocytes was observed in the presence of parecoxib and NS398, partly explaining the observed increase in the cell mass. In the nontreated control cultures, only a small proportion of human mesenchymal stem cells exhibited spontaneous differentiation into adipocytes, and addition of both dexamethasone and indomethacin caused only a minor increase in the number of adipocyte islets (Fig. 5A). To quantify the increase of adipocytes in cultures incubated with COX-2 inhibitors, the samples were stained with Oil Red O and the areas of the stained adipocytes were counted. Contrary to the selective COX-2 inhibitors, treatment with indomethacin resulted in only a small increase in the Oil Red O stained proportional area. In the presence of parecoxib an obvious increase in the adipocyte covered area could be detected, the increase compared to control (6.5%) was 29%

M. Kellinsalmi et al. / European Journal of Pharmacology 572 (2007) 102–110

(P b 0.01), 40% (P b 0.01) and 25% (P b 0.001) at concentrations of 1 μM, 10 μM and 100 μM, respectively. NS398 at 3 μM concentration increased significantly (P b 0.001) red oil o stained proportional area (Fig. 5B), but higher concentrations had no such effect. The increment in the number of adipose cells in the cultures was confirmed by measuring the levels of the adiponectin transcript with RT-PCR. (Fig. 5C). Treatment with indomethacin and parecoxib at 1 μM and 100 μM concentrations induced a slight dose-dependent increase in the relative intensity of the adiponectin RT-PCR product. NS398 had only a minor stimulating effect on the expression of adiponectin (Fig. 5D). 4. Discussion The possibility that COX-2 inhibitors could delay fracture healing is frequently under debate. Our data clearly suggests that all tested COX-inhibitors; indomethacin, parecoxib and NS 398, dramatically inhibit osteoclast differentiation, which is the initial step in bone remodelling after injury. The effect on osteoclast formation was rapid because in addition to the long-term cultures, it was also detectable in the short term osteoclast resorption assays, which typically measure the resorption activity of mature osteoclasts but in which also some differentiation can be seen. The drug concentrations used in our study were similar to those described to be the plasma concentrations for indomethacin and parecoxib in humans, suggesting that these in vitro observations may have clinical relevance. Our results agree with previous publications on the topic (Sato et al., 1997; Kotake et al., 1999; Okada et al., 2000, 2003; Liu et al., 2005; Goodman et al., 2002). However, there are also controversial results to these (Im et al., 2004; Brown et al., 2004; Long et al., 2002). These differences could be due to differences in species or experimental setting: in theory the spinal fusion for example might be less dependent on osteoclastic activity. During the initial stage of fracture healing also osteoblast precursors are recruited to the site by molecules released from the resorbed bone (Einhorn, 1998). The pre-osteoblasts are differentiated from the common MSCs, which can also differentiate along the adipocyte lineage. (Gregoire et al., 1998; Li et al., 2003) We show here that inhibition of COX-2 inhibited the expression of osteoblast markers while at the same time promoting differentiation of MSCs towards the adipocyte lineage under culture conditions that normally predominantly stimulate osteoblast differentiation. All COX-inhibitors had a significant stimulatory effect on total protein content of cultures which is most likely a consequence of enhanced cell proliferation. It is notable, that although there were some discrepancies between the effects of COX-inhibitors on the osteoblast markers PINP, PIIINP and ALP, the ultimate manifestation of osteoblast activity, mineral deposition, was unequivocally diminished by all COX-inhibitors. It has also been demonstrated with male Sprague–Dawley rats that COX-2 mRNA levels peak during the first 14 days and returns to basal level until 21 days after bone fracture. Taken together, these observations suggest withholding from COX-2

109

inhibitor use for at least 3 weeks after fracture (Gerstenfeldt et al., 2003). Whether such procedure actually affects bone healing also in the clinical setting requires further investigation. The molecular mechanism of COX-2 inhibition induced adipocyte differentiation from human mesenchymal stem cells is unclear and our observation calls for further investigation. The involvement of the COX-pathway in the regulation of adipogenesis has been, however, suggested also previously (Gregoire et al., 1998; Li et al., 2003). Adiponectin was shown to elevate expression of COX-2 and to induce the release of PGE2 (Yokota et al., 2002). COX-1 and COX-2 isoforms have also been shown to be involved in the negative modulation of adipocyte differentiation in 3T3-L1 adipocyte cell culture (Yan et al., 2003). Furthermore enhanced adipocyte differentiation could have a negative feedback effect on osteoblast differentiation; fatty acids have been demonstrated to block IL-1 induced PGE2 production and also osteoclastogenesis has been slightly reduced by inhibition of RANKL (Ha et al., 2006; Deshimaru et al., 2005). In conclusion, we show here that COX-2 inhibitors can have profound effects on bone cell activities in vitro. These drugs inhibit osteoclast and osteoblast differentiation, while promoting hMCS commitment to the adipocyte lineage. While the in vivo significance of these results needs to be established, they do suggest that the use of COX-2 inhibitors should be carefully considered at a time of recovery when efficient bone healing is critical. Acknowledgement The authors want to express their warmest thanks to Mrs. Savilampi for her tireless technical assistance. References Alatalo, S.L., Halleen, J.M., Hentunen, T.A., Mönkkönen, J., Väänänen, H.K., 2000. Rapid screening method for osteoclast differentiation in vitro that measures tartrate-resistant acid phosphatase 5b activity secreted into the culture medium. Clin. Chem. 46, 1751–1754. Barnes, G.L., Kostenuik, P.J., Gerstenfeld, L.C., Einhorn, T.A., 1999. Growth factor regulation of fracture repair. J. Bone Miner. Res. 14, 1805–1815. Birk, R.Z., Abramovitch-Gottlib, L., Margalit, I., Aviv, M., Forti, E., Geresh, S., Vago, R., 2006. Conversion of adipogenic to osteogenic phenotype using crystalline porous biomatrices of marine origin. Tissue Eng. 12, 21–31. Botting, R., Ayoub, S.S., 2005. COX-3 and the mechanism of action of paracetamol/acetaminophen. Prostaglandins Leukot. Essent. Fat. Acids 72, 85–87. Boyde, A., Ali, N.N., Jones, S.J., 1984. Resorption of dentine by isolated osteoclasts in vitro. Br. Dent. J. 156, 216–220. Brown, K.M., Saunders, M.M., Kirsch, T., Donahue, H.J., Reid, J.S., 2004. Effect of COX-2-specific inhibition on fracture-healing in the rat femur. J. Bone Joint Surg. Am. 86A, 116–123. Chambers, T.J., Thomson, B.M., Fuller, K., 1984. Effect of substrate composition on bone resorption by rabbit osteoclasts. J. Cell. Sci. 70, 61–71. Choi, B.K., Moon, S.Y., Cha, J.H., Kim, K.W., Yoo, Y.J., 2005. Prostaglandin E(2) is a main mediator in receptor activator of nuclear factor-kappaB ligand dependent osteoclastogenesis induced by Porphyromonas gingivalis, Treponema denticola and Treponema socranskii. J. Periodontol. 76, 813–820. Deshimaru, R., Ishitani, K., Makita, K., Horiguchi, F., Nozawa, S., 2005. Analysis of fatty acid composition in human bone marrow aspirates. Keio J. Med. 54, 150–155.

110

M. Kellinsalmi et al. / European Journal of Pharmacology 572 (2007) 102–110

Einhorn, T.A., 1998. The cell and molecular biology of fracture healing. Clin. Orthop. 355, 7–21. Gerstenfeldt, L.C., Thiede, M., Seibert, K., Mielke, C., Phippard, D., Svagr, B., Cullinane, D., Einhorn, T.A., 2003. Differential inhibition of fracture healing by non-selective and cyclooxygenase-2 selective non-steroidal antiinflammatory drugs. J. Orthop. Res. 21, 670–675. Gierse, J.K., Koboldt, C.M., Walker, M.C., Seibert, K., Isakson, P.C., 1999. Kinetic basis for selective inhibition of cyclo-oxygenases. Biochem. J. 339, 607–614. Giordano, V., Giordano, M., Knackfuss, I.G., Apfel, M.I.R., Gomes, R.D.C., 2003. Effect of tenoxicam on fracture healing in rat tibia. Injury 34, 85–94. Glassman, S.D., Rose, S.M., Dimar, J.R., Puno, R.M., Campbell, M.J., Johnson, J.R., 1998. The effect of postoperative nonsteroidal anti-inflammatory drug administration on spinal fusion. Spine 1, 834–838. Goodman, S., Ma, T., Trindade, M., Ikenoue, T., Matsuura, I., Wong, N., Fox, N., Genovese, M., Regula, D., Smith, R.L., 2002. COX-2 selective NSAID decreases bone ingrowth in vivo. J. Orthop. Res. 20, 1164–1169. Gregoire, F.M., Smas, C.M., Sul, H.S., 1998. Understanding adipocyte differentiation. Physiol. Rev. 78, 697–710. Ha, H., Lee, J.H., Kim, H.M., Kwak, H.B., Lee, S., Kim, H.H., Lee, Z.H., 2006. Alpha-lipoic acid inhibits inflammatory bone resorption by suppressing prostaglandin E2 synthesis. J. Immunol. 1, 111–117. Im, G.I., Kwon, B.C., Lee, K.B., 2004. The effect of COX-2 inhibitors on periprosthetic osteolysis. Biomaterials 25, 269–275. Kis, B., Snipes, J.A., Busija, D.W., 2005. Acetaminophen and the cyclooxygenase-3 puzzle: sorting out facts, fictions and uncertainties. J. Pharmacol. Exp. Ther. 315, 1–7. Kotake, S., Udagawa, N., Takahashi, N., Matsuzaki, K., Itoh, K., Ishiyama, S., Saito, S., Inoue, K., Kamatani, N., Gillespie, M.T., Martin, T.J., Suda, T., 1999. IL-17 in synovial fluids from patients with rheumatoid arthritis is a potent stimulator of osteoclastogenesis. J. Clin. Invest. 103 (9), 1345–1352. Leskelä, H.V., Risteli, J., Niskanen, S., Koivunen, J., Ivaska, K.K., Lehenkari, P., 2003. Osteoblast recruitment from stem cells does not decrease by late adulthood. Biochem. Biophys. Res. Commun. 311, 1008–1013. Li, X., Cui, Q., Kao, C., Wang, G.J., Balian, G., 2003. Lovastatin inhibits adipogenic and stimulates osteogenic differentiation by suppressing PPARgamma2 and increasing Cbfa1/Runx2 expression in bone marrow mesenchymal cell cultures. Bone 33, 652–659. Li, L., Pettit, A.R., Gregory, L.S., Forwood, M.R., 2006. Regulation of bone biology by prostaglandin endoperoxide H synthases (PGHS): a rose by any other name 1/4. Cytokine Growth Factor Rev. 17, 203–216. Liu, X.H., Kirschenbaum, A., Yao, S., Levine, A.C., 2005. Cross-talk between the interleukin-6 and prostaglandin E(2) signalin systems results in enhancement of osteoclastogenesis through effects on the osteoprotegerin/ receptor activator of nuclear factor-(kappa)B (RANK) ligand/RANK system. Endocrinology 146, 1991–1998. Long, J., Lewis, S., Kuklo, T., Zhu, Y., Riew, K.D., 2002. The effect of cyclooxygenase-2 inhibitors on spinal fusion. J. Bone Joint Surg. Am. 84A, 1763–1768.

Noor, M., Gajraj, F.R.C.A., 2003. The effect of cyclooxygenase-2 inhibitors on bone healing. Reg. Anesth. Pain Med. 28, 456–465. Okada, Y., Lorenzo, J.A., Freeman, A.M., Tomita, M., Morham, S.G., Raisz, L.G., Pilbeam, C.C., 2000. Prostaglandin G/H synthase-2 is required for maximal formation of osteoclast-like cells in culture. J. Clin. Invest. 105 (6), 823–832. Okada, Y., Pilbeam, C., Raisz, L., Tanaka, Y., 2003. Role of cyclooxygenase-2 in bone resorption. J. UOEH 1, 185–195. Ozaki, A., Tsunoda, M., Kinoshita, S., Saura, R., 2000. Role of fracture haematoma and periosteum during fracture healing in rats: interaction of fracture haematoma and the periosteum in the initial step of the healing process. J. Orthop. Sci. 64–70. Rawlinson, S.C., Wheeler-Jones, C.P., Lanyon, L.E., 2000. Arachinodic acid for loading induced prostacyclin and prostaglandin E(2) release from osteoblasts and osteocytes is derived from the activities of different forms of phospholipase A(2). Bone 27, 241–247. Reuben, S.S., Ekman, E.F., 2005. The effect of cyclooxygenase-2 inhibition on analgesia and spinal fusion. J. Bone Joint Surgery Am. 87, 536–542. Sato, T., Morita, I., Murota, S., 1997. Prostaglandin E2 mediates parathyroid hormone induced osteoclast formation by cyclic AMP independent mechanism. Adv. Exp. Med. Biol. 407, 383–386. Selander, K., Lehenkari, P., Väänänen, H.K., 1994. The effects of bisphosphonates on the resorption cycle of isolated osteoclasts. Calcif. Tissue Int. 55, 368–375. Selander, K.S., Mönkkönen, J., Karhukorpi, E.K., Härkönen, P., Hannuniemi, R., Väänänen, H.K., 1996. Characteristics of clodronate-induced apoptosis in osteoclasts and macrophages. Mol. Pharmacol. 50, 1127–1138. Simon, A.M., Manigrasso, M.B., O'Connor, J.P., 2002. Cyclo-oxygenase 2 function is essential for bone fracture healing. J. Bone Miner. Res. 17, 963–976. Wei, X., Zhang, X., Zuscik, M.J., Drissi, M.H., Schwarz, E.M., O'Keefe, R.J., 2005. Fibroblasts express RANKL and support osteoclastogenesis in a COX-2-dependent manner after stimulation with titanium particles. J. Bone Miner. Res. 20, 1136–1148. Yan, H., Kermouni, A., Abdel-Hafez, M., Lau, D.C.W., 2003. Role of cyclooxygenase COX-1 and COX-2 in modulating adipogenesis in 3T3-L1 cells. J. Lipid. Res. 44, 424–429. Yokota, T., Meka, C.S.R., Medina, K.L., Igarashi, H., Comp, P.C., Takahashi, M., Nishida, M., Oritani, K., Miyagawa, J.-I., Funahashi, T., Tomiyama, Y., Matsuzawa, Y., Kincade, P.W., 2002. Paracrine regulation of fat cell formation in bone marrow cultures via adiponectin and prostaglandins. J. Clin. Invest. 109, 1303–1310. Zhang, X., Schwarz, E.M., Young, D.A., Puzas, J.E., Rosier, R.N., O'Keefe, R.J., 2002. Cyclooxygenase-2 regulates mesenchymal cell differentiation in to the osteoblast lineage and is critically involved in bone repair. J. Clin. Invest. 1, 1405–1415.