Ex vivo culture platform for assessment of cartilage repair treatment strategies

• A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website. • The final author version and the galley proof are versions of the publication after peer review. • The final published version features the final layout of the paper including the volume, issue and page numbers.

sheep, goat, dog and pig), whose use is accompanied by high costs for animal caring and ethical issues (Cook et al., 2014;Hurtig et al., 2011).
In order to refine and reduce the number of animal studies performed in the field of cartilage repair, standardized and representative ex vivo models that allow for a whole array of simultaneous tests are valuable tools (de Vries-van Melle et al., 2012;Vinardell et al., 2009;Pretzel et al., 2013). These ex vivo models need to fulfill the following requirements: The model has to be stable, without loss of physiological properties, viability and metabolic activity, for a relevant culture period, optimally 8 weeks, as this is the typical duration of in vivo animal studies on cartilage regeneration therapies (Sakata et al., 2015). Additionally, the model should provide the possibility to monitor functional parameters, such as metabolic activity, viability and, preferably, also mechanical properties. Existing models to study cartilage, to test biomaterials or therapies for cartilage repair can be classified into three groups: cell seeded biomaterials under free swelling conditions, cartilage-only models, and osteochondral models. They differ in the presence and type of often result in the formation of fibrotic tissue, which consists mainly of collagen I and is therefore mechanically inferior to native cartilage (Wakitani et al., 1994;Dewan et al., 2014;Caldwell and Wang, 2015). With regard to the long-term outcome of existing strategies there is still a need to develop new, respectively improved treatment strategies that lead to proper cartilage regeneration.
The development of new treatments and the refinement of existing treatments to repair cartilage defects is typically studied in animal models, such as rodent, sheep, goat, pig and horse. They represent the complexity of the in vivo situation. It has to be taken into account that cartilage defect models in rodents and mature rabbits show spontaneous self-repair (Cook et al., 2014). However, this phenomenon of intrinsic repair is limited in humans and other species (Fox et al., 2009). Further disadvantages of animal models are the limited control over physiological parameters and the limited possibilities for monitoring and controlling the healing progress from a biological and biomechanical point of view. With respect to joint size, the only clinically relevant models are large animal models (horse,

Model
Cell-laden 3D biomaterials under free swelling conditions 1. Cartilage derived matrix scaffolds (Benders et al., 2014) 2. Natural hydrogels (Sheehy et al., 2015) 3. Synthetic scaffolds (Endres et al., 2012) Cartilage-only models 1. Mature human cartilage discs (ø 5 mm) from OA femoral condyles (Wardale et al., 2015) 2. Mature bovine cartilage discs (ø 6 mm) from lateral facets of trochlea/patella groove (Pretzel et al., 2013) 3. Immature porcine chondral explants (ø 6mm) from femoropatellar joints (Vinardell et al., 2009)  which is more representative of the in vivo situation than the models summarized in Table 1. Furthermore, this design allows adding specific media and factors to the different tissues and allows us to analyze secretion of soluble factors by bone and cartilage separately. The overall aim was to obtain an ex vivo osteochondral culture model that is highly suitable to, at least partly, replace animal testing by serving as a screening tool prior to animal studies. We investigated in study A whether separation of the cartilage and bone compartments and/or culture media results in the long-term preservation of viability, structural and functional properties of cartilage tissues. In study B, we evaluated for how long we can maintain matrix stability of osteochondral explants during long-term culture. Finally, in study C, we determined the optimal defect size in which self-healing does not occur during 28 days of culture.
Cell-seeded 3D biomaterials under free swelling conditions and cartilage-only models are useful for gaining insight into cell-material interactions and to characterize the stimulatory effect of different materials on cell phenotype and metabolic activity, but the absence of bone is an inherent limitation for testing of cartilage repair treatments. It has been reported before that the presence of subchondral bone is critical for the success of microfracture procedures (Frisbie et al., 2003(Frisbie et al., , 2006. Moreover, de Vries-van Melle et al. (2012, 2014b showed that cartilage-related genes were more highly expressed in their ex vivo cultured osteochondral explant compared to cartilage-only explants. Their osteochondral model is a promising approach to studying cartilage biology as it simulates an in vivo joint-like environment. However, the model is not suitable for long-term ex vivo cultures since the expression of cartilage-related genes decreases over 28 days of culture. We hypothesized that this decrease may be caused by factors secreted from the subchondral bone that are released and concentrated in the culture medium, which also serves as nutrient supply for the cartilage of the osteochondral explants. In vivo, the crosstalk between the tissues is restricted to the bone cartilage interface. Factors secreted by osteoblasts isolated from sclerotic bone, but also mononucleated cells that are present in the bone marrow, have been shown to influence chondrocyte metabolism indicated by a reduction of cartilage matrix synthesis in a co-culture model (Sanchez et al., 2005a,b;Iwai et al., 2011). Therefore, we thought that if the direct communication of all factors released from the subchondral bone, including the bone marrow, with the cartilage of osteochondral explants can be restricted to crosstalk via the cartilage bone interface, cartilage matrix composition may be maintained.
Biopsies were cultured for 14 or 28 days at 37°C and 5% CO 2 . Medium was changed every 3-4 days. At each time point, four biopsies per group were used for analyses. Biopsies were cut in half, cartilage and bone tissue were separated, and the cartilage tissues of one set of halves was analyzed for their biochemical content. The other set of halves was used for histology (n = 2) and evaluation of cell viability (n = 2). Four further biopsies were used to create time point zero for all measurements.

Study B: Evaluation of long-term preservation of matrix stability in osteochondral explants cultured with tissue-specific medium in the osteochondral platform
Osteochondral biopsies (n = 28) were cultured for 0, 14, 28, 42, 56, 70 and 84 days at 37°C and 5% CO 2 in the platform with cartilage medium in the cartilage compartment and bone medium in the bone compartment (based on results of study A; Fig. 1). Medium was changed every 3-4 days. At each time point, n = 4 biopsies were used to determine the equilibrium compressive modulus. Afterwards, these biopsies were cut in half, cartilage and bone were separated, and the cartilage tissues of one self of halves were analyzed for the biochemical content (n = 4). The other set of halves was used for histology (n = 2) and evaluation of cell viability (n = 2).

Study C: Define critical size of chondral defects ex vivo
Full thickness cartilage defects with a diameter of 1 mm, 2 mm or 4 mm (each n = 3) were created with a biopsy punch (PFM medical AG, Cologne, GER) (Fig. 2). Residual debris on the calcified layer was removed with a sharp spoon (MF Dental, Mantel, GER). The osteochondral biopsies with inflicted defects were cultured for 28 days in the osteochondral platform with 3 ml cartilage medium in the upper compartment and 3 ml bone medium in the bone compartment (Fig. 1, group 3). Medium was changed every 3-4 days and after 28 days of culture, samples underwent plastic embedding. 5 mm in total height (consequently 3.5-4 mm mean bone height), were isolated from 10 medial femoral condyles of 5-7-month-old landrace pigs (retrieved from a local slaughter house) as described above.

Ex vivo osteochondral culture platform
The ex vivo osteochondral culture platform is developed and owned by LifeTec Group BV. It consists of custom inserts that are suspended into an in-house developed six-well plate (Fig.  1). The osteochondral explants are mounted into the inserts using an O-ring positioned exactly at the interface between cartilage and bone, creating two isolated culture compartments: the upper compartment containing the cartilage and the lower compartment containing the bone, each with their respective media. Complete separation of the two compartments has been confirmed (data not shown).

Experimental outline
Study A: Effect of separation of the cartilage and bone compartments and of culture with tissue-specific media Osteochondral biopsies were divided into three experimental groups, each n = 8 (Fig. 1). Group 1 consisted of biopsies that were placed in a standard six-well plate filled with 2% lowgelling agarose (Sigma, Zwijndrecht, NL) in PBS (Sigma, Zwijndrecht, NL), such that the subchondral bone was surrounded by agarose and the cartilage was above the agarose surface as previously described (de Vries-van Melle et al., 2012). Biopsies were cultured with 5.5 ml cartilage medium consisting of DMEM high glucose (42430, Gibco, LifeTechnologies, Bleiswijk, NL) supplemented with 100 U/ml penicillin, 100 µg/ml streptomycin, 2.5 µg/ml fungizone, 1 mM sodium pyruvate (Gibco, LifeTechnologies, Bleiswijk, NL), 40 µg/ ml L-proline (Sigma-Aldrich, Zwijndrecht, NL), 50 µg/ml Lascorbic acid-2-phosphate (Sigma-Aldrich, Zwijndrecht, NL), 1% ITS+ Premix (Corning, Fisher Scientific, Landsmeer, NL) and 100 nM dexamethasone (Sigma-Aldrich, Zwijndrecht, NL). The biopsies of group 2 and 3 were positioned in our platform such that the cartilage was separated from the bone. The inserts with biopsies of group 2 were placed in custom 6-wells plates and the upper and lower compartment was filled with 2.5 ml and 3 ml cartilage medium respectively (Fig. 1). In group 3, biopsies were cultured in the same setup as group 2, but with Amsterdam, NL), and orthohydroxyproline content was assessed as described previously using an assay modified from Huszar et al. (1980), and a trans-4-hydroxyproline reference (Sigma, Zwijndrecht, NL).

Unconfined compression testing
Mechanical testing of samples of study B was performed at every time point in unconfined compression between impermeable platens using a custom-designed biomechanical testing apparatus containing a voice coil system (SMC Pneumatics, Amsterdam, NL) to determine the compressive equilibrium modulus (Eq) as previously described by Olivera et al. (2015). The cartilage thickness of each sample was measured. Then the samples were submerged in PBS at room temperature and equilibrated in creep to a tare load of 1 N for 120 s. From this offset, stress relaxation tests were performed with a ramp speed of 1 μm/s until reaching 10% strain. After equilibrium was reached, Eq was calculated from the ratio of the equilibrium stress to the applied strain.

Statistical analysis
Statistical analysis was performed with GraphPad Prism 7.00. In study A, the effects of treatment and time as well as their interaction were examined for significant differences by two-way ANO-VA with GAG, DNA and HYP content as dependent variables. Differences were assumed statistically significant for p < 0.05 of the corrected values by Tukey HSD post hoc test. For study B, the effect of culture time on Eq, GAG, DNA and HYP was analyzed by one-way ANOVA with Fisher's LSD post-hoc test.

Study A: Effect of separation of the cartilage and bone compartments and of culture with tissue-specific media
MTT staining of explants at day 0, 14 and 28 of ex vivo culture revealed preserved metabolic activity over the full depth of the cartilage tissue without apparent differences between time points and groups (Fig. 3). Metabolic activity of cells residing

Analyses
Viability assay 3-(4,5-dimethylthiazolyl-2)-2,5-diphenyltetrazoliumbromide (MTT; Sigma-Aldrich, Zwijndrecht, NL) was used to evaluate viability of the cells in explants. Osteochondral biopsies were washed three times with PBS and incubated with 1 mg/ml MTT solution in DMEM for 90 min at 37°C. After washing with PBS, macroscopic pictures were taken. Freshly isolated samples served as positive control and samples that were incubated in dimethylsulfoxide (DMSO; Sigma-Aldrich, Zwijndrecht, NL) for 24 h served as negative control (data not shown).

Safranin-O/Fast Green staining
Samples from each group (study A and B) were fixed in 3.7% formalin (Sigma, Zwijndrecht, NL), decalcified in 10% EDTA (Amresco, USA) in deionized water, and processed for paraffin embedding. In study C, samples were fixed with 4.0% formalin (Roti ® -Histofix, Carl Roth, Karlsruhe, GER), washed in deionized water, dehydrated and incubated in pre-infiltration solution (Heraeus Kulzer, Hanau, GER) before polymerized for plastic embedding. All samples were cut in 4 µm sections. Slides were stained with Safranin-O and Fast Green (Sigma, Zwijndrecht, NL) to visualize the distribution of proteoglycans and proteins, respectively, in the tissue. Hematoxylin (Sigma, Zwijndrecht, NL) was used as nuclear counterstaining. Slides of all three studies were imaged by light microscopy (study A: Zeiss Axio Observer Z1; study B,C: Keyence BZ-9000, Biorevo, Neu-Isenburg, GER).

Biochemical content in cartilage tissue
Cartilage samples were lyophilized and digested in a 140 µg/ ml papain solution (Sigma Aldrich, Zwijndrecht, NL) for 16 h at 60°C. DNA content was determined using the Hoechst dye method with a reference curve of calf thymus DNA (Sigma, Zwijndrecht, NL). The glycosaminoglycan (GAGs) content was determined using a modification of the DMMB assay described by Farndale et al. (1982) and a shark cartilage chondroitin sulfate reference (Sigma, Zwijndrecht, NL). For determination of the collagen content (HYP), digested tissue samples were hydrolyzed in 6 M hydrochloric acid (Merck,

Fig. 3: Vitality staining (MTT in blue) of crosssections of osteochondral explants cultured for 14 and 28 days ex vivo in study A
No differences in metabolic activity were observed in explants between groups (group 1: agarose embedding, group 2: cartilage medium in both compartments, group 3: tissue-specific medium). 28 showed significantly higher GAG content compared to group 1 (p < 0.0001) and group 2 (p = 0.0417).
Safranin-O staining confirmed the quantitative findings and demonstrated that GAG loss in group 1 and 2 was most pronounced in the superficial zone of the cartilage tissue (Fig. 4A). In contrast to the GAG content, the HYP and DNA content of the samples did not show significant differences among groups or time points (Fig. 4C, D).

Study B: Long-term culture of osteochondral explants with tissue-specific culture media
No changes in MTT staining were observed over time in the cartilage tissue during 84 days of culture (Fig. 5). MTT staining in the subchondral bone seemed to decrease during the first 14 days of culture, beyond which no apparent differences were observed.
The intensity of Safranin-O staining was maintained in explants up to day 42 (Fig. 6). A slight reduction in red staining intensity was observed in the superficial zone of the cartilage in the subchondral bone appeared slightly decreased over time in all groups.
Osteochondral samples embedded in agarose that were cultured only in presence of cartilage medium (group 1) demonstrated a significant decrease in GAG content in culture (Fig.  4B). At day 14, GAG content of group 1 samples was significantly reduced to 68.9% of the value at day 0 (p = 0.0231) and GAG content decreased further to 39.3% at day 28 (p < 0.0001). Samples of group 2, cultured in our osteochondral platform with cartilage medium in both compartments, revealed that culturing in separated compartments reduced GAG loss (group 2, 3). After 14 days of culture 77.1% of the day 0 GAG content was still present in group 2 and no further decrease was observed between days 14 and 28. At day 28, GAG content in group 2 was significantly higher compared to group 1 at day 28 (p = 0.0035).
Remarkably, full preservation of the native GAG content during 28 days of culture was observed in osteochondral explants cultured in separated compartments with specific medium added to the respective tissues (group 3). GAG content in group 3 at day Values are displayed as mean ± SD (n = 3-4). *p < 0.05, **p < 0.005, ****p < 0.0001 compared between groups and time points. tent (Fig. 7D). After day 56, the equilibrium modulus decreased albeit not statistically significantly.

Study C: Defining ex vivo critical size defects
Full thickness cartilage defects with diameters of 1, 2 and 4 mm did not heal spontaneously during 28 days of ex vivo culture (Fig. 8B-D). Neo-tissue formation was mostly observed in defects with a diameter of 2 and 4 mm (8C'-D'). Safranin-O staining revealed deposition of GAG-containing matrix on the surface of the 2 and 4 mm cartilage defects at day 28 (indicated by arrows in Fig. 8C' and D', respectively), which was not tissue at day 56. From then on, staining intensity further decreased, also becoming noticeable in the middle and deep zones of the tissue at day 70 and 84. Quantitative measurements of the amount of GAG confirmed the histological staining results (Fig. 7A). Native GAG content was preserved during 56 days of culture and then decreased (day 70 vs. day 28: p = 0.0252).
The collagen content in the cartilage remained stable during the entire culture period of 84 days (Fig. 7B). DNA content was higher at day 28, day 42 and day 70 compared to day 0 (Fig. 7C). The equilibrium modulus of the native cartilage tissue was preserved during 56 days of culture, similar to GAG con-    2002; Urban et al., 2012). In contrast to chondrocytes, bone cells require β-glycerophosphate, dexamethasone and serum, providing growth factors (Urban et al., 2012;Langenbach and Handschel, 2013).
Unlike the cartilage, MTT staining intensity of the subchondral bone decreased during culture. There are a number of cell types present in the bone, such as bone cells and blood cells. The bone cells are osteoclasts, osteoblasts and osteocytes, which are responsible for bone remodeling, but make up only a small percentage of the total cells present in the bone (Taichman, 2005). Due to the low concentration of bone cells in trabecular matrix, we believe that they are hardly visible in the MTT staining. We therefore assume that mainly dying blood cells are the cause of the decreased viability staining of the bone. Because of the interrupted blood circulation after explanting the biopsies, the blood cells and platelets no longer have a vital function and therefore likely undergo apoptosis. In future studies, we aim to gain more insight into the process of active bone remodeling in our explants by staining or quantification of alkaline phosphatase activity or by visualization of osteoclast activity by means of tartrate-resistant acid phosphatase staining.
We demonstrated that we are able to maintain native GAG content and mechanical properties for 56 days with tissue-specific medium, which, as far as we know, has not been shown in any other ex vivo osteochondral culture model before. Culture duration of 56 days is a relevant period for testing repair strategies and the model can therefore, at least partly, replace animal testing or serve as a screening tool before testing in animals. An array of tests with dozens of samples can be performed simultaneously to study the effects of material composition and mechanical properties, compound concentrations, drug dose responses, etc.
We hypothesize that incorporating mechanical loading into the system will allow for even longer culture periods, as we postulate that the decrease in GAGs and reduction in equilibrium modulus in cartilage after 56 days of ex vivo culture is caused by the absence of mechanical triggers. Numerous groups have shown the importance of mechanical loading on cartilage metabolism (Guilak et al., 2004;Vinardell et al., 2012;Kock et al., 2012Kock et al., , 2013. As a next step in the development of our model we are now integrating a mechanical loading module into the platform while maintaining a closed system. In study C, all three defect sizes tested in our platform proved to be critical size defects. No explant cultured ex vivo over 28 days attained complete healing of full thickness cartilage defects. Since larger defect diameters are more feasible for use with implant materials, we will use 4 mm as the optimal ex vivo critical size defect diameter of full thickness cartilage defects. This is smaller than the typically described critical sizes of in vivo cartilage defects in animal studies, such as for example in adult miniature pigs in which critical defect diameters range from 4.5 mm to 6.0 mm (Harman et al., 2006;Chang et al., 2014). However, compared to animal studies, it is much easier to create large numbers of samples with cartilage defects with different diameters and depths (chondral, subchondral, osteochondral) in a reproducible and controlled manner in our model. visible in samples with a 1 mm defect (Fig. 8B-B'). Migration of cells into the defect site was only observed in 2 mm and 4 mm defects. Metabolic activity was similar to results shown for study A and B (not shown).

Discussion
In this study, we developed a culture platform that allows ex vivo long term culture over at least 56 days of osteochondral biopsies under controlled conditions, while maintaining cartilage tissue content, structure and mechanical properties. It is a predictive platform suitable for preclinical assessment of cartilage repair strategies in critical size defects.
The platform is characterized by its two separated media compartments, which allows culture of cartilage and bone with tissue-specific nutrients. To our knowledge, this unique feature of two separated media compartments differs from existing ex vivo osteochondral culture models. The platform mimics the in vivo situation, in which diffusion of signaling molecules between the cartilage and bone is only possible through the subchondral bone plate at the bone-cartilage interface. In contrast to the previously described agarose model, there is no external communication via the surrounding culture medium (de Vriesvan Melle et al., 2012).
We have shown that separation of compartments already partly reduced the unwanted GAG loss observed in group 1 and previously by de Vries-van Melle et al. (2014a). We postulate that this can be explained by the fact that cartilage tissue in our platform is not exposed to factors secreted by bone cells or other cells residing in the bone marrow, such as cytokines and other stress factors via the culture media. Such factors are naturally secreted by cells in the bone during apoptosis or inflammation, which can inhibit cartilage metabolism or stimulate degradation thereof. Sanchez et al. (2005a,b) have shown that sclerotic osteoblasts inhibit aggrecan production on protein and gene expression level and lead to a reduction in chondrogenic transcription factor SOX9 and collagen II on mRNA level compared to non-sclerotic osteoblasts. Furthermore, Tetlow et al. (2001) have demonstrated that the presence of inflammatory cytokines such as IL-1β and TNF-1α reduced cartilage matrix synthesis and caused proteoglycan loss.
Astonishingly, supplying the osteochondral explants with tissue-specific culture medium resulted in the complete preservation of the native ECM composition during at least 56 days of culture. It is difficult to explain which exact mechanism is responsible for maintaining cartilage tissue content and properties in case of separation of bone and cartilage tissues and using tissue-specific medium. It has been shown before that cartilage homeostasis and function strongly depend on the crosstalk between cartilage and bone (Mariani et al., 2014). It is likely that cartilage and bone cells perform optimally when they receive the specific factors they require for their metabolism, which in turn can lead to positive signaling between the two tissues. The supplementation of cartilage medium with ascorbic acid, insulin, transferrin and selenic acid is crucial to maintain the cartilaginous matrix production (Clark et al., Therefore, the defect model represents a tool for studying clinically relevant cartilage repair therapies, namely preclinical testing of cartilage biomaterials for their regenerative potential, evaluation of drug and cell therapies, or to study basic biological mechanisms and pathways during cartilage regeneration, which will certainly reduce the number of animals needed for in vivo testing.

Conclusion
To conclude, our ex vivo culture platform with the unique feature of separated compartments for cartilage and bone allows for long-term culture of osteochondral explants while maintaining cartilage tissue content, structure and mechanical properties. As such, our model offers a highly relevant new platform for assessment of cartilage repair strategies in critical size defects.