A novel model of a biomechanically induced osteoarthritis‐like cartilage for pharmacological in vitro studies

Abstract Excessive pressure or overload induces and aggravates osteoarthritic changes in articular cartilage, but the underlying biomechanical forces are largely ignored in existing pharmacological in vitro models that are used to investigate drugs against osteoarthritis (OA). Here, we introduce a novel in vitro model to perform pathophysiological and pharmacological investigations, in which cartilage explants are subjected to intermittent cyclic pressure, and characterize its ability to mimic OA‐like tissue reactivity. Mechanical loading time‐dependently increased the biosynthesis, content and retention of fibronectin (Fn), whereas collagen metabolism remained unchanged. This protocol upregulated the production and release of proteoglycans (PGs). The release of PGs from explants was significantly inhibited by a matrix metalloproteinase (MMP) inhibitor, suggesting the involvement of such proteinases in the destruction of the model tissue, similar to what is observed in human OA cartilage. In conclusion, the metabolic alterations in our new biomechanical in vitro model are similar to those of early human OA cartilage, and our pharmacological prevalidation with an MMP‐inhibitor supports its value for further in vitro drug studies.

tissue of joints. 21 Chondrocytes sense their mechanical surroundings and respond accordingly by producing an extracellular matrix that is well able to dissipate mechanical loads, thereby protecting cartilage against injury. 9,[22][23][24] Mechanosensitive ion channels, primary cilia and integrin receptors in the cell membrane sense changes in the streaming potentials, interstitial fluid, and hydrostatic and osmotic pressure that are induced by dynamic and static mechanical forces. These physical stimuli activate mechanotransduction pathways, which ultimately regulate transcription in chondrocytes. 9,22,23 Using animal models is one way to practically examine the mechanisms involved in the initiation and progression of OA. With in vivo models, morphological changes of cartilage and joints can be reliably evaluated, but the cellular mechanisms that control these alterations are difficult to study and remain mostly unknown. 16,25,26 25 As such, cartilage explant cultures offer a more realistic method of investigating mechanical signals being transmitted through the sophisticated architecture of the tissue and the responses and functions of cells in their natural milieu. 16,25,28 However, most organotypic cartilage explants are cultured at normal atmospheric pressure, whereby the chondrocytes are not exposed to compressive pressure or tensile strains. Since hyaline articular cartilage is exposed to compressive stimuli, especially in joints, some in vitro studies have already been carried out. 13,14 The effects of mechanical compression on cell proliferation, hypertrophy and apoptosis, as well as the biosynthetic response of the chondrocytes with regard to the extracellular matrix and catabolic enzymes were investigated. 13,29,30 The mechanical devices used differ depending on the mechanical stimuli to be examined, that is to what extent static or cyclic forces, sliding forces or strains are exerted (ie 12,13,[31][32][33]. The mechanical bioreactor developed by us is able to exert constantly or intermittently a large number of static and cyclic compressive forces with different load levels and frequencies. 33 One major finding is that the metabolism of chondrocytes is clearly controlled by the mechanical conditions, wherein the magnitude of the metabolic response is subject to the load frequency, strain and type of loading (eg compression, sliding, continuous, intermittent and static). 7,11-14,28-32, [34][35][36] In vitro research on OA has been undertaken to screen for drugs, cytokines and growth factors that promote cartilage repair and reduce cartilage destruction. Most in vitro pharmacological models have used IL-1 to induce cartilage degradation. [37][38][39][40] However, the similarity of OA-like changes in these in vitro models to those in humans is limited and most valid with respect to the loss of proteoglycan (PG). Thus, the overall goal of our studies was, using mechanical forces in initially healthy cartilage explants, to cause a disrupted metabolism that strongly resembles the phenotype in OA for pharmacological research. We have developed and described a mechanical loading protocol 29,30,34,35 that induces OA-like conditions in initially healthy articular cartilage explants. In the current study, we perform comprehensive biochemical analyses of the time-dependent alterations and the first pharmacological validation of this new in vitro model using a synthetic inhibitor of matrix metalloproteinases (MMPs).

| Mechanical loading of cartilage explants
To be able to culture and mechanically stress cartilage discs for a longer time period in a sterile surrounding, an apparatus was specially designed and manufactured for this purpose (Figure 1), as described previously. 33 Briefly, the load amplitude, frequency and load function were controlled by a computerized unit positioned outside of the incubator. Figure 1A shows the six load chambers being mounted in a hermetically sealed case containing a servo-controlled, electropneumatic valve that regulated the amplitude and frequency of air inlet into the spring-controlled air bellows. A piston with a porous glass load platen was connected to the dynamic air bellows and mechanically compressed the cartilage explant cultured within a specimen holder ( Figure 1B). The specimen holder and piston were made of surgical-grade titanium. The degree of dynamic cartilage compression was continuously monitored during loading with a load displacement transducer system in the air bellow and the sealed blue case ( Figure 1A,B). Eight filter-controlled air ports per chamber exist to enable under sterile conditions the exchange with the incubator air.
In our study, we used a laminar flow cabinet to obtain two cartilage explants, which were lying side by side and macroscopically healthy, at full-thickness from the load-bearing area of a metacarpophalangeal condyle belonging to a slaughtered 18-to 24-months-old bull. A biopsy punch was used to collect 7 mm diameter cartilage discs. We measured the thickness in the middle of the cartilage explants twice using a digital calliper gauge so that the amount of compression during mechanical loading could be determined. 29,30,34,35 The discs were washed and centrally located within the specimen holder thereby placing the cutting edge of discs to the bottom of the specimen holders. Articular explants were cultured in 2.5 ml Ham's F-12 nutrient medium containing a serum substitute and additional supplements as already reported. 29,30,34,35 The loading device was then equilibrated for 2 h in a normal bench top incubator at 37°C, 5% CO 2 and 95% humidity ( Figure 1A). A peak stress of 0.5 MPa was used to apply uniaxial intermittent cyclic loading for 1, 3, 6 or 9 days. The cyclic loads were applied using a sinusoidal waveform of 0.1 Hz frequency for 10 s, which was followed by a 100-s period of unloading ( Figure 1C).
The air inlet was reduced by 10 N during the unloaded period so that the stamp detached from the cartilage surface, enabling the explant to reswell. The cartilage explants serving as controls were cultivated in identically constructed chambers without the air bellows located above. Media were changed on Days 3 and 6 during experiments lasting 6 or 9 days; obtained media received 10% protease inhibitor mixture (vol/vol) and were frozen at −20°C until analysed. The cultured discs were radioactively labelled with 10 µCi/ml  The mechanical device with six load chambers connected to a sealed blue case (bottom) and six control chambers (top) housed within a standard CO 2 incubator, as described previously. 33 (B) Schematic of one load chamber, composed primarily of surgical-grade titanium and stainless steel, with parts thereof being coated with gold-coloured TiN to make the steel corrosion-resistant. (C) Schematic representation of the loading protocol. Briefly, explants were intermittent mechanically loaded, in that the loads were cyclically applied in an approximately sinusoidal waveform of 0.1 Hz, followed by a period of unloading lasting 100 s during which the load platen was lifted from the articular surface. This loading sequence was continuously repeated x times until termination of the experiments Darmstadt, Germany), followed by liquid scintillation counting, as described previously. 30,34

| Quantitation of proteoglycans and DNA
The sulphated PGs of the papain-digested extracts and media were quantified spectrophotometrically at 523 nm after chemical action with 0.25 ml of dimethylmethylene blue using a microplate reader.
We used as standard CS type A of bovine trachea (Sigma GmbH, Deisenhofen, Germany). 30,34 The bisbenzimidazole dye Hoechst 33258 was applied to fluorometrically determine the quantity of DNA in the papain-digested discs. 34

Sulfation of endogenous and [ 3 H]-glucosamine-labelled CS was
monitored in papain-digested cartilage explants, as described previously. 30,41 Briefly, papain-digested tissue samples were centrifuged to remove tissue debris, and GAGs were pelleted, washed, dried, resuspended and digested with chondroitinases ABC and ACII (Seikagaku Corp., Tokyo, Japan). Enzymatically digested CS, containing the unsaturated disaccharides of the internal CS chains, and the nonreducing terminal mono-and disaccharides were concentrated using the MicroCon ® 3 filters, reduced with NaBH 4 and injected onto a CarboPac PA1 column that was connected to a guard column (Dionex GmbH, Idstein, Germany). Specimens were eluted by a step gradient and scanned on a HPAEC-PAD system (model DX 500; Dionex GmbH, Idstein, Germany) to quantify the unlabelled, unsaturated disaccharides ΔDi6S, ΔDi4S and ΔDi0S, obtained from the digested interior part of the CS chains. To quantify [ 3 H]-radiolabelled mono-and disaccharides, 1-ml fractions of the eluant were collected and the radioactivity was quantified by liquid scintillation counting.
We calculated the mean length of the newly synthesized CS chains (ie the number of internal Δdisaccharides per chain). For this purpose, the ratio of the radioactively labelled internal unsaturated disaccharides to the terminal saccharides was computed. 41,42

| Fibronectin and protein synthesis
The wet weight of explants was determined after weight equilibrium of thawed discs had been reached. Cartilage and load platen were extracted three times over 2 d for Fn with 4 M urea in 0.05 M phosphate buffer, pH 7.3 and proteinase inhibitors. 35 [ 3 H]-Phenylalaninelabelled Fn was isolated from nutrient media and extracts by gelatin Sepharose 4B affinity chromatography. 35 In addition to Fn, a large number of cartilage proteins were extracted by urea. We determined the total [ 3 H]-phenylalanine incorporated into proteins within the media and urea extracts by precipitation with an identical volume of ice-cold 1N perchloric acid. After 1 h, the precipitated radioactively labelled proteins were concentrated, washed, and dried and the radioactivity was then determined by liquid scintillation counting. 35 We determined the Fn content in the extracts and nutrient media of the explants using ELISA. 35  The difference between wet and dry weight was calculated, and the percentage of water content was then calculated for both unloaded controls and loaded cartilage discs.

| Chondrocyte viability
The same culture conditions as described above were applied to cartilage discs that were cultured exclusively to determine the viability of chondrocytes in each cartilage zone separately using the vital stains fluorescein diacetate and propidium iodide. 34

| Statistical analysis
Our experiments were replicated four to five times on different days, each time using joints from different steers (N = 5-6). We

| Proteoglycan metabolism of loaded explants
Intermittent cyclic mechanical loading increased the incorporation of a radioactive precursor into PGs of cartilage explants compared with the case in unloaded controls ( Figure 2B). The elevated incorporation appeared after 3 (p ≤ 0.001) and 6 days, returning to preload levels by Day 9, indicating a period of maximum mechanical stimulation of PG synthesis. Elevated amounts of endogenous PGs were released into the nutrient media during the 3-, 6-and 9-day experiments, whereas the total PG content (data not shown) and loss of newly synthesized PGs remained unchanged.

| Fine structure of chondroitin sulphate in cartilage explants
The glycosaminoglycan chain chondroitin sulphate (CS) of aggrecan, the main proteoglycan of the cartilage extracellular matrix, is composed out of repeating disaccharides, which themselves consist out of glucuronic acid (glcUA) and ß1,3N-acetylgalactosamine (galNAc). Analysis of the CS composition being newly synthesized by unloaded explants revealed that culture for up to 9 days did not Less galNAc4,6S was synthesized in cartilage discs that were loaded for 6 days, which was the only terminal CS chain that was significantly altered compared with the unloaded controls ( Figure 2D). Furthermore, the internal disaccharides of newly produced CS chains and the elongation of these chains remained unaffected by the applied loads.

| Fibronectin and protein metabolism of loaded cartilage
The applied loading protocol significantly elevated the incorporation of [ 3 H]-phenylalanine into Fn on Days 3 and 6 compared with that of the unloaded control explants ( Figure 3A), again demonstrating a window in which the biosynthetic activity of chondrocytes was enhanced. Protein synthesis, however, remained unchanged; thus, on Days 3 and 6, only the biosynthesis of Fn was specifically stimulated, as the proportion of Fn relative to total newly synthesized protein was increased (data not shown). As shown in Figure 3A, the total endogenous Fn content as found in explants, media and extracts of load platen was also markedly higher, as determined with ELISA. Furthermore, endogenous Fn was less released from loaded explants compared to unloaded controls ( Figure 3B).

| Collagen synthesis of loaded cartilage
During the first 3 days of cyclic mechanical stimulation, the incorporation of [ 14 C]-proline, respectively, the specific collagen synthesis rate adapted to the mechanical environment and subsequently remained constant (Figure 4). However, mechanical loading did not modulate the biosynthesis of non-collagenous proteins (NCP) throughout the 9-day culture. Furthermore, release of newly synthesized proteins from the loaded cartilage discs into media remained unaffected (data not shown).

| DNA and water content of loaded explants
As shown in Table 1, the content of DNA and water from loaded cartilage discs did not differ markedly from values obtained from unloaded explants used as controls.

| Load induced compression
The extent of the maximum compression of the mechanically loaded cartilage discs remained constant throughout the experiments ( Table 1). As shown in Table 1, the range of standard deviations was ± 7.3%. This implies that the on average 31.6% load-induced compression of cartilage discs, which were originally 0.54 ± 0.05 mm thick (N = 141), was reproducible under the respective conditions. were not affected (data not shown).

| First pharmacological validation of the in vitro model
The suitability of the in vitro model for pharmacological research was tested in that the effect of the MMP-inhibitor A976157 on PG and Fn metabolism of cartilage explants was examined at 10 −5 M in a 6-day experiment.
The MMP inhibitor led to significant reversal of the mechanically induced release of endogenous PGs to control levels ( Figure 5A). However, the agent did not modify the biosynthesis or total content of PGs, DNA content, the maximum amplitude of compression of explants or the release of newly synthesized PGs (data not shown). Figure 5B shows that the load-induced increases in the biosynthesis and content of Fn and protein synthesis were not significantly affected by A976157. However, as expected, mechanical loading of drug-treated explants increased the synthesis of Fn and protein as well as Fn content. In addition, the release of Fn and proteins remained unaffected (data not shown).
Cartilage sections were dyed with the vital stains propidium iodide and fluorescein diacetate, revealing that A976157 did not impair chondrocyte viability (data not shown).

| DISCUSS ION
In humans, secondary OA has a largely mechanical aetiology, and in many animal models, OA can be artificially induced by mechanical instability. [15][16][17]19 This in vitro study revealed the time dependence of the mechanically induced changes of our new in vitro model, developed for the preclinical testing of drugs against OA, as well as the first pharmacological validation of it.
The loading device that we used 33 44 The pressure in our study was set to 0.5 MPa, which is below the stresses experienced by the human knee joints while walking (0.8-6.3 MPa). 45 Although this pressure is low, the strains in our experiments were above those that are expected to occur in vivo because the cartilage discs were removed from the subchondral bone and its enclosing cartilage. Moreover, cartilage discs were obtained from bovine metacarpophalangeal joints and not from human knee or hip joints. However, our previous studies identified the frequency of cyclic loading as the chief mechanical factor that influences cartilage metabolism. 29,30,34,35 In this study, we determined a significantly enhanced loss of endogenous PGs from cartilage discs after 3 days of loading. A distinctive feature of OA is the depletion of PGs from damaged joint cartilage. 46,47 Our findings are corroborated by earlier in vitro studies also reporting mechanical load-induced damage of the collagenous network, leading to an enhanced loss of PGs. 48 Consistent with earlier results, 49-51 we observed greater release of PGs only after a latent period, perhaps due to the production and activation of proteoglycanolytic enzymes, which could have dis- Our loading protocol also significantly increased Fn synthesis after a latent period of 3 and 6 days, whereas protein synthesis remained unchanged. Several studies have confirmed that OA cartilage synthesizes up to five times more Fn than healthy cartilage, that Fn levels are 10-fold to20-fold higher in early OA cartilage lesions and that OA cartilage explants release less newly synthesized Fn into media. [58][59][60] We also observed significantly greater retention of endogenous and newly synthesized Fn in loaded explants. Our results are supported by previous studies that demonstrated higher synthesis and concentration of Fn in canine cartilage discs ten days post intermittent impact pressure. 61,62 Any load-induced increases in Fn synthesis and content remained unaffected by the addition of the MMP inhibitor A976157.
MMP-3 and MMP-8 were reported to cleave Fn, 63 and a single blunt impact on bovine osteochondral explants was found to promote Fn fragmentation. 64  Mechanical loading of explants did not alter collagen synthesis for up to 9 days. After skeletal growth, the biosynthesis of collagen type II decreased significantly, and by in situ hybridization and immunohistochemistry, collagen type II was reported to be not TA B L E 1 Effect of loading duration on the DNA and water content and the magnitude of cartilage compression. Note: Cartilage discs were cyclically mechanically loaded with a maximum pressure of 0.5 MPa, applied in a sinusoidal waveform with a frequency of 0.1 Hz. Discs were subjected to intermittent loading with 10 s of cyclic loading followed by a 100-s unloading phase repeated x times for 1, 3, 6 or 9 days. Data are shown as mean values ± S.D. The numbers in parentheses indicate the numbers of specimens used to quantify the degree of compression.  Furthermore, in vitro studies revealed that the water content of cartilage explants increases after impact loading. 61,62,71 These findings suggest irreversible weakening of the retaining collagen network, wherein hydrophilic PGs swell by hydration to a greater degree than normal. Again, our data indicate that the collagen network of explants is not damaged severely enough for an increase in water content to be quantifiable.

F I G U R E 5
During OA, the articular cartilage surface loses its homogeneous nature and is disrupted and fragmented by pitting, clefts and ulcerations. 72,73 In our experiments, the porous load platen slightly damaged the cartilage surface and thus contributed to the induction of OA-like changes. A porous load platen was chosen to ensure that the explants were adequately supplied with nutrients.
In conclusion, a notable feature of our new in vitro model is the induction of an early-stage OA-like appearance in response to purely mechanical stimuli, which, when in excess, is a major pathogenic factor of human OA. Our first pharmacological prevalidation already showed that this new biomechanical in vitro model is suitable to screen potentially chondroprotective agents. Compared with animal models of OA, nearly every biochemical, histological or biomechanical method can be applied in our novel in vitro model within a reasonable period. Thus, our novel biomechanical in vitro model should deepen our understanding of the early stages of degenerating cartilage and its specific treatment.

ACK N OWLED G EM ENTS
The authors would like to thank O. Jähn, I. Pelzer and W. Ortmann for excellent technical assistance. This study was supported by grants of the German Federal Ministry of Education and Research (BMBF no. 0311058) and of the SET Foundation (no. 022).

CO N FLI C T O F I NTE R E S T
The authors declare no competing or financial interests. Manfred Schudok is a Sanofi employee and may hold shares and/or stock options in the company.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data supporting the findings of this study are available from the corresponding author upon reasonable request.