Background

Osteoarthritis (OA) is a heterogeneous and multifactorial degenerative joint disease characterized by gradual loss of articular cartilage, formation of osteophytes, and synovial inflammation. Current non-surgical treatments for OA, such as non-steroid anti-inflammatory drugs and steroid injections, only relieves pain, inflammation, and effusion. There is a need for the development of disease-modifying drugs that can not only relieve pain and inflammation, but also inhibit cartilage degeneration. The lack of progress in the development of disease-modifying drugs is largely due to our limited understanding of the pathogenesis of OA and insufficient knowledge regarding the molecular targets for therapeutic intervention.

The biochemical events involved in OA are poorly understood. Many extracellular matrix degrading enzymes and inflammatory cytokines, including matrix metollproteinase-13 (MMP-13), ADAM metallopeptidase with thrombospondin type 1 motif 5 (ADAMTS5), interleukin-1 (IL-1), and cyclooxygenase-2 (Cox-2) have been implicated in OA [1–3]. Pathological calcification has also been implicated. Basic calcium phosphate crystals and calcium pyrophosphate dihydrate crystals are the two most common articular calcium-containing crystals. The presence of these crystals within the knee joints of end-stage OA patients is well recognized [4–9]. Injection of these crystals into the knee joints of dogs and mice induce a severe inflammatory response [10, 11]. These crystals also induced cell mitogenesis and stimulated the production of matrix metalloproteinases (MMPs), nitric oxide, and inflammatory cytokines [12–15], suggesting that crystals may play a role in the development or progression of OA.

Phosphocitrate (PC) is a naturally occurring small molecule originally identified in rat liver mitochondrial extract [16]. Since its original identification, PC has been shown to be a powerful calcification inhibitor [17, 18]. PC prevented soft tissue calcification and didn’t produce any significant toxic side effect in rats in doses up to 150 μmol/kg/day [19]. In addition, PC inhibited crystal-induced mitogenesis, expression of MMPs, and cell death [20–22]. Based on these findings, a hypothesis that PC is a disease-modifying drug for calcification-induced OA therapy was postulated [23]. A subsequent study demonstrated that PC inhibited meniscal calcification, and that a decrease in meniscal calcification was accompanied with reduced cartilage degeneration in Hartley guinea pigs (calcification-induced OA), but had no significant effect on cartilage degeneration in partial meniscectomy-induced OA in rabbit (posttraumatic OA or non-calcification induced OA) [24]. The investigators concluded that PC is a disease-modifying drug for calcification-induced OA therapy but not for non-calcification-induced OA. It was believed that PC exerted its OA disease-modifying activity by inhibiting the formation of articular calcium crystals and the detrimental interaction between these crystals and joint cells (crystal-dependent action) [23, 24]. Although this theory was well received at the time, doubt emerged upon the findings that bisphosphonates, which were potent calcification inhibitors [25, 26], failed to inhibit cartilage degeneration in animal models of OA, including Hartley guinea pig model of calcification-induced OA [27, 28].

We recently found that PC downregulated the expression of many genes classified in cell proliferation, angiogenesis, and inflammatory response, while upregulating the expressions of many genes classified in skeletal system development in the absence of calcium crystals [29–31]. These newer findings suggest that crystal-dependent action of PC may not be a sole action underlying the OA disease-modifying effect of PC. It is likely that PC exerts its OA disease-modifying activity through two independent actions: i) inhibiting the formation of crystals and crystal-induced expressions of MMPs (a crystal-dependent action), and ii) modulating the expressions of genes implicated in OA (a crystal-independent action).

PC-β-ethyl ester (PC-E) is a PC analogue where a β-carboxyl group is replaced by an ethyl ester group (Fig. 1). We are interested in PC-E not only because compared to PC, PC-E has less negative charges, therefore it may be more easily absorbed in the intestine if administered through the oral route, but also because novel new PC analogues may be prepared by linking other active group(s) to this carboxyl group. In this study, we sought to examine disease-modifying activity of PC and PC-E on posttraumatic OA and investigate the structure-activity relationship. The results of this study may not only provide information valuable for the design and development of new PC analogues as disease-modifying drug for OA therapy, but also for a better understanding of pathogenesis of OA and the molecular mechanism underlying the disease-modifying activity of PC.

Fig. 1
figure 1

Molecular structures of PC and PC-E

Full size image

Methods

Dulbecco’s minimum essential medium (DMEM), fetal bovine serum, stock antibiotic/antimycotic mixture were obtained from Invitrogen (Carlsbad, CA). 45Calcium was obtained from Perkin-Elmer (Boston, MA). Antibody specific to MMP-13 (Lifespan Biosciences, Seattle, WA), ADAMTS5 (Santa Cruz Biotechnology, Dallas, TX), Cox-2 (Santa Cruz Biotechnology, Dallas, TX), and CCL-5 (Bioss, Woburn, MA) were obtained from the commercial sources indicated. Safranin-O, fast green, picrosirius red, and alcian blue were obtained from Polysciences (Warrington, PA). PC and PC-E were prepared according to the method described [32]. All other chemicals were obtained from Sigma-Aldrich (St. Louis, MO).

Calcification assay

OA fibroblast-like synoviocytes (FLSs), similar to OA chondrocytes, play a role in the formation of articular crystal [9, 33, 34]. To compare the calcification-inhibitory activity of PC and PC-E, we performed an ATP-induced calcification assay using telomerase immortalized human OA fibroblast-like synoviocytes (FLSs), hTERT-OA 13A FLSs [9]. Briefly, hTERT-OA 13A FLSs were plated in a 24 well cluster plate at 95 % confluence. On the second day, DMEM with 10 % serum was changed to DMEM containing 0.5 % serum. On the third day, after cells became quiescent, DMEM containing 1 mM ATP and trace-labeled with 1 μCi/ml 45calcium was added. Immediately, increasing amounts of PC, PC-E, or citrate were added into the wells. Forty-eight hours later, cells were washed with cold Hank’s balanced salt solution five times and lysed with 0.1 N NaOH. Radioactivity of lysate in each well was quantified using liquid scintigraphy. Calcification-inhibitory activity of disodium ethane-1-hydroxy-1, 1-diphosphonate (EHDP) was also examined. Results were presented as the mean ± SD of five independent experiments.

Proliferation assay

Proliferation-inhibitory activities of PC-E and PC were examined as described [29]. Briefly, hTERT-OA 13A FLSs (4×104) were plated in six well cluster plates. On the second day, DMEM containing 10 % serum and PC-E (0.5 mM) or PC (0.5 mM) was added into the top three wells. DMEM containing 10 % serum without PC-E or PC was added into the bottom three wells as a control. DMEM was changed every three days until the cells in the bottom wells reached 85 % confluence (cultured for about 12 to 14 days). All cells were then harvested and the cell number in each well was counted. Results were presented as the mean ± SD of three independent experiments.

Experimental animals

This study was performed according to the guidelines set forth by the Institutional Animal Care & Use Committee of Carolinas Medical Center, which approved the animal protocol. Male Hartley guinea pigs at three weeks of age were obtained from Charles River Laboratories (Wilmington, MA) and individually housed in 29 × 21 × 10 inch solid bottom cages. Guinea Pig Chow (No. 5025; Ralston Purina, Richmond, Indiana) and water were available ad libitum. The first group of guinea pigs (n = 5) received intraperitoneal injection of PC (40 mg/kg) twice per week, second group (n = 5) received PC-E (40 mg/kg) and the last group (n = 5) received physiological saline. Two months later, partial medial meniscectomy was performed on the right knee of all guinea pigs to induce posttraumatic OA. A week after the surgery, injection of PC, PC-E, or saline was resumed. Five months later, these guinea pigs were euthanized by the administration of Euthasol (Virbac Animal Health, Ft. Worth, Texas). Hind limbs were collected, fixed in 10 % formalin, and transferred to 70 % ethanol until use. Structural changes in the articular cartilage of Hartley guinea pigs were not observed until three months of age [35], therefore pretreatment of the young guinea pigs with PC or PC-E for two months before partial meniscectomy surgery will not result in detectable structural changes in the articular cartilage.

Radiographic, microscopic, and histological examinations

Radiographs of knee joints were obtained with a digital radiography system (piXaray 100, Bioptics Inc., Tucson, AZ). After dissection of the knee joints, radiographs of medial meniscus were obtained. All tibia plateaus were first stained with Indian ink as described [36]. The tibia plateaus were then decalcified in Cal-Ex II solution (Fisher Scientific, Fairlawn, NJ) and cut coronally in the center to produce two equal portions. The posterior portion was embedded in paraffin and sectioned with a Leica RM2025 microtome (Nussloch, Germany) to obtain 4 μm sections. Three non-consecutive sets of sections (three consecutive sections in each set) obtained at 400 μm intervals were stained with safranin-O and counter stained with fast green. These sections (nine sections for each cartilage specimen) were graded according to standard Mankin criteria with minor modifications [37]. Two sections in each cartilage specimens were also stained with picrosirius red and counter stained with alcian blue.

Cartilage thickness

Central portion of safranin-stained sections (the most degenerative area) was photographed and the area of cartilage (same length for each section) was measured using the measuring tool in Adobe acrobat software (San Jose, CA). Briefly, the image file was opened with Adobe acrobat and the cartilage was traced continuously with the “pen” in the measuring tool along the irregular surface of the articular cartilage (numerous very short straight lines were connected together to form an irregular line) and the border between cartilage and subchondral bone. After tracing the cartilage was finished, the area was automatically calculated. Cartilage thickness is obtained by dividing the area with the length of the cartilage measured.

Immunohistochemistry

Two sections in each tibia plateau were deparaffinized with xylene and rehydrated with graded ethanol. Endogenous peroxidase activity was blocked by incubation with deionized water containing 3 % H2O2 for five minutes. Non-specific binding was blocked by incubation with 100 μl of 10 % normal horse serum diluted in base solution (4 % BSA and 5 % non-fat dry milk in PBS) for 20 min. These sections were incubated with primary antibodies (1:100 dilution) for one hour, followed with secondary reagent for 30 min (Immpress reagent kit, Vector, Inc., Burlingame, CA). Negative control was performed using mouse IgG. Slides were rinsed in phosphate buffered saline three times and visualized with 3, 3′-diaminobenzidin. Slides were counterstained with light green, dehydrated and mounted with resinous mounting media. These slides were graded on a scale of 0–4, where 0 = very weak staining; 1 = weak staining; 2 = moderate staining; 3 = strong staining; 4 = very strong staining as described [38].

Microarray

Briefly, hTERT-OA 13A FLSs [9] were plated in two 100 mm plates at 90 % confluence. On the second day, medium was changed to medium containing 1 % serum. On the third day, medium containing 1 % serum and PC-E (0.6 mM) was added to a plate, and medium containing 1 % serum but without PC-E was added to the other plate. Twenty-four hour later, total RNA was extracted using Trizol reagent (Invitrogen, Carlsbad, CA) and purified using Oligotex kit (Qiagen, Valencia, CA). These RNA samples were used for microarray as described [29]. Microarray analysis of PC on gene expressions has been performed previously [29].

Statistical analyses

Results of calcification assay, proliferation assay, and cartilage thickness measurements (variables measured on continuous interval) are presented as the mean ± SD. The differences between the results in 2 groups were analyzed using Student’s t Test. Scores of histological staining and immunostaining (variables presented as ordinal data) were presented as the mean ± SD. The differences between the scores in two groups were analyzed using Wilcoxon rank sum test. Statistical analysis was performed using the statistical analysis tool in the Sigma Plot software, version 12 (Systat Software, Inc., San Jose, CA).

Results

Calcification- and proliferation-inhibitory activities

PC, PC-E, and EHDP, but not citrate, inhibited ATP-induced calcium deposition in a dose dependent manner (Fig. 2a). PC is the most potent calcification inhibitor, with PC-E being 19 % and EHDP being 13 % less powerful than PC in the inhibition of OA FLSs-mediated calcium deposition at the concentration of 0.25 mM (p < 0.01). PC-E, similar to PC [29], also inhibited the proliferation of human OA FLSs (Fig. 2b). There were about 66 % and 64 % fewer human OA FLSs in the PC and PC-E treated wells, respectively, compared to controls (p < 0.01). PC and PC-E had no effect on cell viability up to the concentration of 10 mM whereas EHDP caused cell death when its concentration was higher than 1.5 mM (not shown).

Fig. 2
figure 2

Calcification- and proliferation-inhibitory activities. a PC, PC-E, and EHDP inhibit OA FLSs-mediated calcification. CPM (count per minute) in control, PC-E, PC and EHDP treated cells are 22,784 (±2,156), 6,452 (±1,095), 2,466 (±352), and 5,124 (±1114), respectively, at the concentration of 0.25 mM. b PC and PC-E inhibit OA FLSs proliferation. Cell numbers in control and PC treated wells are 845,236 (±89,521) and 285,706 (±34,132). Cell numbers in control and PC-E treated wells are 924,484 (±98,718) and 331,096 (±48,694). *= p < 0.01, versus control

Full size image

Radiographic examinations

Representative radiographs of the knee joints and medial menisci were provided in Fig. 3. As expected, calcified medial meniscus was observed in the non-operated left knee, but not in the meniscectomied right knee (Fig. 3a). Meniscal calcification predominantly occurred in the anterior horn of the medial meniscus (Fig. 3b). The severely calcified anterior horn of the medial meniscus was absent in the meniscectomied right knee. As shown in Fig. 3a, meniscus in the non-operated left knee of PC and PC-E treated guinea pigs were slightly less calcified compared to untreated guinea pigs, indicating that PC and PC-E reduced meniscal calcification. However, meniscal calcification in PC and PC-E treated guinea pigs appeared similar. No signs of cartilage calcification were observed.

Fig. 3
figure 3

Radiographs of knees and medial menisci. a Representative radiographs of meniscectomied right knee and non-operated left knee of untreated, PC-treated, and PC-E treated guinea pigs. Calcified medial menisci are indicated by arrows. b Radiographs of the meniscectomied right knee medial meniscus and the non-operated left knee medial meniscus of an untreated guinea pig. Severely calcified anterior horn of the medial meniscus (arrow) is present in the left knee, but not in the right knee

Full size image

Microscopic examinations

Indian ink staining was used to examine the tibia plateaus of all guinea pigs. Representative Indian ink stained tibia plateaus are provided in Fig. 4. As shown, cartilage damage, visualized with the help of Indian ink, spanned a larger area in the medial tibia plateaus of the meniscectomied right knee than the cartilage damage in the medial tibia plateaus of the non-operated left knee (Fig. 4a). In the non-operated left knee, cartilage damage was confined in the central area of the medial tibia plateau (Green arrow) whereas in the meniscectomied right knee, cartilage damage was not only present in the central area (Green arrow) but also present in the peripheral area that was originally covered with meniscus (Red arrow). These results demonstrated that meniscal injury and joint instability resulted in increased cartilage degeneration, especially cartilage degeneration in the peripheral area, which was not observed in the non-operated left knee. As shown in Fig. 4b and c, there was less cartilage damage in the medial tibia plateau of the meniscectomied right knee, especially in the peripheral area, in PC and PC-E treated guinea pigs compared to untreated controls, indicating that PC and PC-E inhibited cartilage degeneration induced by meniscal injury and joint instability. There was also less cartilage damage in the medial tibia plateau of the non-operated left knee in PC and PC-E treated guinea pigs compared to untreated controls.

Fig. 4
figure 4

Representative images of Indian ink staining. a Indian ink-stained tibia plateaus of the meniscectomied right knee and the non-operated left knee of an untreated guinea pig. b Indian ink-stained tibia plateaus of the meniscectomied right knee and the non-operated left knee of a PC-treated guinea pig. c Indian ink-stained tibia plateaus of the meniscectomied right knee and the non-operated left knee of a PC-E-treated guinea pig. Green arrow - central area of medial tibia plateau. Red arrow - peripheral area of medial tibia plateau

Full size image

Histological examinations

Representative three non-consecutive safranin-O stained sections from the meniscectomied right knee in untreated and PC treated guinea pigs were provided in Fig. 5. Consistent with Indian ink staining, safranin-O staining revealed that severe cartilage damage occurred in both the central and peripheral areas of medial tibia plateau in untreated guinea pigs. Cartilage damage and proteoglycan loss, in many cases, extended into deep and calcified zones. Cartilage damage and proteoglycan loss were significantly reduced in PC treated guinea pigs. In the central area of medial tibia plateaus, mild to moderate cartilage damage and proteoglycan loss were observed, whereas in the peripheral area, only mild cartilage damage and proteoglycan loss were observed. Consistent with Indian ink staining, cartilage damage and proteoglycan loss were also reduced in PC-E treated guinea pigs (photos not shown).

Fig. 5
figure 5

Safranin-O stained sections of medial tibia plateaus of meniscectomied right knee. a A section in the central area of medial tibia plateau of an untreated and a PC treated guinea pig. b Second section, 400 μm away from first section. c Third section, 800 μm away from first section. Green arrows - central area of medial tibia plateau. Red arrows - peripheral area

Full size image

In addition, cartilage in untreated guinea pigs appeared thinner than the cartilage in PC treated guinea pigs (Fig. 5). We measured the cartilage thickness of all guinea pigs as described. As shown in Fig. 6a, cartilage in PC treated guinea pigs was 31 % thicker than the cartilage in untreated guinea pigs (p = 0.01). Similarly, PC-E also inhibited cartilage thinning. Cartilage in PC-E treated guinea pigs was 18 % thicker than the cartilage in untreated guinea pigs (p = 0.02).

Fig. 6
figure 6

Cartilage thickness and histological scores of medial tibia plateaus. a Normalized cartilage thickness in untreated, PC-treated, and PC-E treated guinea pigs are 1 (±0.11), 1.31 (±0.13), and 1.18 (±0.10).* p < 0.01, versus controls. b Histological scores of medial tibia plateau in the meniscectomied right knee in untreated, PC, and PC-E treated guinea pigs are 11.95 (±1.05), 6.43 (±0.93), and 8.24 (±0.1.37). Histological scores of medial tibia plateau in the non-operated left knee in untreated, PC, and PC-E treated guinea pigs are 8.11 (±1.39), 3.72 (±0.91), and 5.74 (±0.12). * p < 0.05, versus controls. c Histological scores of the peripheral area of medial tibia plateau in the non-operated left knee and the meniscectomied right knee of untreated guinea pigs are 2.92 (±0.45) and 8.67 (±0.74). Histological scores of the peripheral area of medial tibia plateau in the meniscectomied right knee of PC and PC-E treated guinea pigs are 4.59 (±0.81) and 6.42 (±0.0.86). * p < 0.05, versus controls

Full size image

These safranin-O stained sections (nine sections for each tibia plateau specimen) were graded. As shown in Fig. 6b, PC and PC-E significantly reduced the histologic score of medial tibia plateau (black bars), resulting in a 46  and 30 % reduction in the histological scores, respectively (p < 0.05). For comparison, safranin-O stained sections in the non-operated left knee medial tibia plateau were graded. The histological score of the medial tibia plateau in the non-operated left knee was much lower than the histological score of the medial tibia plateau in the meniscectomied right knee (p < 0.01), confirming that meniscal injury and joint instability resulted in significantly increased cartilage damage. PC and PC-E reduced the histologic score of the medial tibia plateau in the non-operated left knee (grey bars), resulting in a 54  and 28 % reduction in the histological scores, respectively (p < 0.05).

Due to severe cartilage damage being observed in the peripheral area of the medial tibia plateau in the meniscectomied right knee but not in the non-operated left knee, we decided to grade the peripheral area specifically. As shown in Fig. 6c, meniscal injury and joint instability in the meniscectomied right knee resulted in severe cartilage degeneration in the peripheral area compared to the non-operated left knee. Cartilage damage in the peripheral area of the medial tibia plateau in the meniscectomied right knee was significantly reduced in PC and PC-E treated guinea pigs, indicating again that PC and its analogue inhibits injury and joint instability-induced cartilage degeneration.

Two sections in each tibia plateau were also stained with picrosirius red and counter-stained with alcian blue. Representative images of picrosirius red stained sections are provided in Fig. 7. As shown, severe cartilage damage was observed in the untreated guinea pigs and in many cases, cartilage lesions extended into the middle, deep, and calcified zones. In contrast, cartilage damage was much less prominent in PC treated guinea pigs. Surprisingly, unlike the severe loss of safranin-O staining, there was no severe loss of picrosirius red staining in untreated guinea pigs compared to PC treated guinea pigs. These findings indicate that OA cartilage in the untreated Hartley guinea pigs was characterized by breakdown of collagen fibers, not by collagen loss. Interestingly, alcian blue staining was observed in some chondrocytes residing in the middle zone of articular cartilage in the untreated guinea pig, which was not observed in the cartilage of PC-treated guinea pigs. One explanation for this finding is that some chondrocytes in OA cartilage synthesize more proteoglycans in response to matrix degeneration. Consistent with this explanation, the intensity of safranin-O staining in the nucleus and cytoplasm of some chondrocytes in the untreated guinea pigs appeared much higher than the intensity of safranin-O staining in the nucleus and cytoplasm of most chondrocytes in the PC-treated guinea pigs (Fig. 5). Severe loss of safranin-O staining in the articular cartilage of untreated guinea pigs mainly occurred in the extracellular matrix space (Fig. 5).

Fig. 7
figure 7

Picrosirius red stained sections. Left: section at the central area of the medial tibia plateau from two untreated guinea pigs. Right: section at the central area of the medial tibia plateau from two PC treated guinea pigs

Full size image

Immunohistochemistry

Representative images of immunostaining are provided in Fig. 8a. In untreated guinea pigs, high levels of ADAMTS-5 protein was present in superficial, middle, and deep zones, whereas high level of MMP-13 protein was present in the middle and deep zones. The level of ADAMTS5 protein was reduced in PC and PC-E treated guinea pigs. The level of MMP-13 protein was also reduced, especially in the middle zone in PC treated guinea pigs. However, the level of MMP-13 protein appeared only slightly reduced in PC-E treated guinea pigs. In untreated guinea pigs, CCL-5 protein was present in superficial, middle, and deep zones, whereas Cox-2 protein was present in superficial and middle zones. It was clear that the level of CCL-5 protein was reduced in PC and PC-E treated guinea pigs. However, the level of Cox-2 protein appeared only slightly reduced in PC and PC-E treated guinea pigs compared to the untreated controls.

Fig. 8
figure 8

Immunostaining. a Immunostaining of ADAMTS5, MMP-13, CCL-5, and Cox-2. No staining was noted in negative control. b Enlarged images of immunostaining of ADAMTS5, MMP-13, CCL-5, and Cox-2 in untreated guinea pigs. ADAMTS staining was mainly found in the pericellular area and extracellular matrix, whereas MMP-13 staining was found in the nucleus, cytoplasm, pericellular area, and extracellular matrix. CCL-5 staining was found in the nucleus, cytoplasm, and pericellular area. Cox-2 staining was mainly found in the nucleus and cytoplasm

Full size image

The scores of these immunostainings are provided in Fig. 9. As shown, PC and PC-E treatments resulted in 42 % and 38 % reductions in the immunostaining score of ADAMTS5, respectively (p < 0.05). PC treatment also resulted in 44 % reduction in the immunostaining score of MMP-13 (p < 0.05). Although PC-E treatment resulted in reduction in the immunostaining score of MMP-13 (16 %), the difference did not reach statistical significance. PC and PC-E treatment resulted in 35 % and 40 % reductions in the immunostaining score of CCL-5, respectively (p < 0.05). Although PC and PC-E treatments resulted in reductions in the immunostaining score of Cox-2 (about 10-14 % reductions), the differences did not reach statistical significance.

Fig. 9
figure 9

Scores of immunostaining. Scores for ADAMTS5 immunostaining in untreated, PC, and PC-E treated guinea pigs are 2.40 (±0.61), 1.40 (±0.49), and 1.50 (±0.45); Scores for MMP-13 immunostaining are 3.20 (±0.71), 1.80 (±0.64), and 2.70 (±0.55); Scores for CCL-5 immunostaining are 2.00 (±0.55), 1.30 (±0.35), and 1.20 (±0.33); Scores for Cox-2 immunostaining are 1.40 (±0.40), 1.20 (±0.55), and 1.20 (±0.49). *p < 0.05, versus controls

Full size image

The effect of PC-E on gene expressions

We have shown that PC downregulated the expression of numerous genes classified in cell proliferation, angiogenesis, and inflammatory response while upregulating the expression of many genes classified in skeletal system development in the absence of calcium crystals [29]. To examine the molecular mechanisms underlying the decreased disease-modifying activity of PC-E, we compared the effect of PC-E on gene expressions with the effect of PC on gene expressions [29]. As shown in Table 1, PC-E downregulated almost all of the PC-downregulated genes classified in cell proliferation, suggesting that PC-E may inhibit cell proliferation as effectively as PC. However, PC-E had little effect on the expression of most of the PC downregulated genes classified in angiogenesis and inflammatory response, including prostaglandin-endoperoxide synthase 2 (PTGS2/Cox-2). PC-E also had little effect on the expressions of most of the PC upregulated genes classified in muscle tissue and skeletal system development (Table 2).

Table 1 Differentially expressed genes in PC-treated and PC-E-treated cells compared with untreated cells
Full size table
Table 2 Differentially expressed genes in PC-treated and PC-E-treated cells compared with untreated cells
Full size table

Discussion

PC, PC-E, and EHDP are powerful calcification inhibitors, suggesting that the phosphate group within these molecules play a key role in their calcification-inhibitory activity. These findings cast some doubt about the role of the calcification-inhibitory activity of these molecules in their OA disease-modifying activity because bisphosphonates displayed little disease-modifying effect on the animal model of OA [27, 28]. PC, consistent with previous findings [18, 26], is more powerful than EHDP as a calcification inhibitor, suggesting that the three carboxyl groups also play a key role in the calcification-inhibitory activity of PC. However, replacement of the β-carboxyl group with a β-ester group only resulted in a moderate 19 % reduction in the calcification-inhibitory activity of PC, indicating that three carboxyl groups are required for the strong calcification-inhibitory activity and that a single carboxyl group only plays a moderate role in the calcification-inhibitory activity of PC.

PC-E, similar to PC, inhibited the proliferation of OA FLSs. Consistent with this finding, PC-E downregulated the expressions of almost all of the PC-downregulated genes classified in cell proliferation as effectively as PC (Table 1). These findings indicate that the β-carboxyl group plays little role in the proliferation-inhibitory activity of PC and that the phosphate group within these molecules plays a key role in the proliferation-inhibitory activity [29]. Because PC-E is much less powerful than PC as an OA disease-modifying drug, it suggests that the proliferation-inhibitory activity of PC plays little role in its OA disease-modifying activity and that those PC downregulated genes classified in cell proliferation are unlikely key OA disease candidate genes.

Medial meniscal calcification is absent in the meniscectomied right knee and severe articular cartilage calcification is not observed in the Hartley guinea pigs [24]; therefore, OA in the meniscectomied right knee was mainly induced by meniscal injury and joint instability, and had less to do with pathological calcification. In this model of posttraumatic OA, PC significantly reduced cartilage damage and proteoglycan loss, demonstrating that PC is a disease-modifying drug for posttraumatic OA therapy. We should point out that our finding and conclusion contradict with a previous study. Cheung et al. found that PC had no significant effect on cartilage degeneration in partial meniscectomied rabbits, and concluded that PC is potentially a disease-modifying drug for calcification-induced OA, but not for non-calcification-induced OA [24]. One possible explanation for these contradicting findings is that different doses of PC were used in the two studies. A single weekly injection (40 mg/kg) was used in the previous study whereas two weekly injections of PC (40 mg/kg) were used in this study. It is worth noting that PC treatment did reduce the histological score of the medial tibia in the rabbit (from 9.8 ± 1.7 to 8.1 ± 2.2) [24]. If higher doses or more injections of PC per week was used, statistically significant difference might have been observed.

The molecular mechanisms underlying the disease-modifying effect of PC remain poorly understood. If PC exerted its disease-modifying effect solely by inhibiting the formation of articular crystals, PC should have displayed little disease-modifying effect on posttraumatic OA. Moreover, PC-E should have displayed only moderately decreased disease-modifying effect on posttraumatic OA compared to PC because PC-E is still a powerful calcification inhibitor. However, we found that PC significantly inhibited cartilage degeneration and that PC-E was 46 % less powerful than PC in the inhibition of cartilage degeneration in the meniscectomied right knee. These findings indicate that crystal-dependent action is unlikely the sole action underlying the disease-modifying effect of PC or PC-E. PC and PC-E likely exert their disease-modifying activity through both a crystal-dependent action and a crystal-independent action. Consistent with this mechanism, PC-E had no effect on the expressions of PC downregulated genes classified in angiogenesis and inflammatory response and PC upregulated genes classified in skeletal system development; therefore, resulting in decreased OA disease-modifying activity. Taken together, it suggests that the gene expression-modulatory activity of PC may play an important role in its OA disease-modifying activity.

We demonstrated that PC reduced the levels of ADAMTS5 and MMP-13 proteins. It is worth noting that the level of MMP-13 protein is much higher in the middle zone than in the superficial and calcified zones (Fig. 7). In human OA articular cartilage, calcium crystals are detected in the superficial and calcified zones [39–41]. If calcium crystals are also present in the superficial or calcified zones of articular cartilage in the guinea pigs and those crystals played a key role in the induction of MMP-13 expression [42], high level of MMP-13 protein should have been observed in the superficial zone or calcified zone. Similarly, if the reduction in the level of MMP-13 protein in PC treated guinea pigs was due to the inhibition of PC on the interaction between calcium crystals and chondrocytes [23], significant reduction in the level of MMP-13 protein should have been observed in the superficial zone or calcified zone, but not in the middle zone. However, these were not what we observed, indicating that calcium-containing crystals is unlikely a key inducer for the production of MMP-13 in the articular cartilage and that PC exerts its inhibitory effect on the production of MMP-13 through a crystal-independent action.

Our findings do suggest that crystal-dependent action of PC plays a role. For example, PC treatment resulted in a 46 % reduction in the histological scores of cartilage in meniscectomied right knee but a 54 % reduction in non-operated left knee. One explanation for this difference (46 % reduction via 54 % reduction) is that pathological meniscal calcification plays a role in the non-operated left knee OA [7, 43]. It is likely that PC exerts its disease-modifying effect on the posttraumatic OA in the right knee through a crystal-independent action whereas PC exerts its disease-modifying effect on the primary OA in the left knee through both a crystal-dependent action and a crystal-independent action. It is conceivable that inhibition of cartilage degeneration through a single action (inhibiting non-crystal-dependent disease pathway) is less effective than through 2 actions (inhibiting both crystal-dependent disease pathway and non-crystal-dependent disease pathway).

PC-E was 46 % less powerful than PC in the inhibition of cartilage degeneration in the non-operated left knee but was 33 % less powerful than PC in the inhibition of cartilage degeneration in the meniscectomied right knee. Again, this difference (46 % reduction via 33 % reduction) can be explained similarly. The replacement of β-carboxyl group with an ester group resulted in partial impairment of both crystal-dependent action and crystal-independent action of PC. It is conceivable that partial impairment of both actions would result in a greater loss in the disease-modifying effect on the primary OA in non-operated left knee than on the posttraumatic OA in meniscectomied right knee because crystal-dependent disease pathway was absent in the right knee.

Our study has limitations. One limitation is that the OA in the meniscectomied right knee is not absolutely a non-calcification-induced OA because crystals must be present in the calcified zone. However, the increased severity of cartilage damage in the meniscectomied right knee compared to the non-operated left knee indicate that the most severe cartilage lesions in the right knee medial tibia plateau, especially the cartilage lesions in the peripheral area, were caused by meniscal injury and joint instability and has little to do with pathological calcification. In addition, a previous study found that PC had no significant effect on crystal formation in cartilage [16]. Taken together, it indicates that PC inhibits cartilage degeneration, at least in part, through a crystal-independent action.

Conclusions

Posttraumatic OA in Hartley guinea pigs are characterized by breakdown of collagen fibers and proteoglycan loss. PC is not only potentially a disease-modifying drug for calcification-induced OA therapy but also potentially a disease-modifying drug for posttraumatic OA therapy. PC exerts its disease-modifying activity on OA through two independent actions, a crystal-dependent action and a crystal-independent action. The β-carboxyl group plays little role in the proliferation-inhibitory activity and the modulatory effect on the expressions of genes classified in cell proliferation, but a major role in the OA disease-modifying activity and the modulatory effect on the expressions of genes classified in angiogenesis, inflammatory response, and skeletal system development of PC. The β-carboxyl group is not a group that should be used to link other active group(s) to create new PC analogues as OA disease-modifying drugs.