Co-culture with macrophages reflects the interaction of the type B cells in the synovial lining with cartilage. Upon addition of a LPS-activated murine macrophage cell line to human OA chondrocytes cultured in poly-(ethylene glycol)-diacrylate (PEGDA) using culture inserts, a significant increase of MMP-1, MMP-3, MMP-9, MMP-13, IL-1β, TNF-α, IL-6, IL-8, and IFN-γ compared to OA chondrocytes in monoculture was observed (Samavedi et al., 2017). Conversely, in macrophages, IL-1β and Arginase-1, a macrophage inflammatory marker, were significantly increased by the presence of the chondrocytes. This suggests that changes induced by OA in either cell type intensified response in the neighboring cell type, highlighting the use of such cultures to study the crosstalk of macrophages and chondrocytes in OA (Samavedi et al., 2017). The catabolic marker panel found was similar to that observed in vitro OA models, indicating the validity of the model (Blasioli et al., 2014; Samavedi et al., 2017). Whether the effect of OA chondrocytes on activated macrophages was specific for the OA state is not clear, as healthy chondrocytes were not included in the study. Addition of LPS- activated and non-activated murine macrophages to gelatin-embedded porcine chondrocytes increased chondrocyte expression of MMP-1, MMP-3 after 1 week of co-culture (Peck et al., 2014). Activated macrophages also increased cell proliferation as well as collagen II and aggrecan expression, which was claimed to mimic anabolism in early OA. However, at the protein level total GAG and collagen II content were significantly reduced upon co-culture with both unstimulated and LPS-stimulated macrophages. Celecoxib treatment of the co-culture with LPS-activated macrophages was able to significantly reduce MMP-1 and MMP-3 expression after 3 days. PGE₂ levels were significantly reduced after 7 days of treatment indicating that the model might be applicable for drug testing (Peck et al., 2014). The inhibition of chondrocyte ECM synthesis by non-stimulated macrophages however, and the combination of murine with porcine cells should be viewed critically.

Focusing on the interaction between cartilage and bone, several models incorporating bone cells have been described, many of these based on alginate-encapsulated chondrocytes in culture inserts and monolayers of osteoblasts (Lin et al., 2010; Sanchez et al., 2005a; Sanchez et al., 2005b; Sanchez et al., 2015) (Figure 2). Using healthy tissue as source, porcine osteoblasts induced chondrocyte hypertrophy as shown by decreased collagen II and aggrecan expression and increased expression of collagen X and bone sialoprotein (Lin et al., 2010), factors involved in OA pathology (Pesesse et al., 2014). If osteoblasts were subjected to cyclic tensile stress, an even more distinct shift towards hypertrophy and additional increase of MMP-1, MMP-3 and MMP-13 expression in chondrocytes was observed, supporting the view that mechanical stress of bone could induce degenerative changes in cartilage. In vivo, mechanical stress of bone lead to increased TGFβ signaling, which in turn could promote OA (Zhen and Cao, 2014), matching the findings in this in vitro model. TGFβ expression was also increased in the stressed osteoblasts in the model (Lin et al., 2010). Still, the observation that healthy osteoblasts can also induce OA-like changes may indicate less validity of this model.

Human osteoblasts also decreased SOX-9 and collagen II gene expression of human chondrocytes, which was more pronounced by addition of sclerotic compared to non-sclerotic osteoblasts (Sanchez et al., 2005a). Fibroblasts did not induce these changes in chondrocytes, indicating that this influence might be specific to the communication of chondrocytes and osteoblasts (Sanchez et al., 2005a). Pretreating non-sclerotic osteoblasts with an inflammatory stimulus (IL-1β + IL-6 + soluble IL-6 receptor or oncostatin M (OSM)) at levels found in OA synovial fluid upregulated MMP expression and downregulated aggrecan expression to similar extent as the sclerotic osteoblasts (Sanchez et al., 2005b). Although also here healthy cells negatively affected chondrocyte behavior, the more pronounced effects of diseased osteoblasts, in line with in vivo observations, suggest its applicability as culture model.

Also in a microsystem bioreactor using microfluidics (Lozito et al., 2013) (Figure 3A), the influence of IL-1β stimulation on osseous and cartilage-like tissue components was studied. This model demonstrated that stimulating hBMSC-generated bone-like tissue with IL-1β resulted in a greater inflammatory response (e.g., increased expression of MMPs) in the adjoining cartilage-like component than by stimulating the cartilage component directly, suggesting communication between both joint compartments (Lin et al., 2014).

A bioengineered co-culture model of chondrocytes and synovial fibroblasts was used to investigate whether overexpression of glutamine fructose-6-phosphate amidotransferase (GFAT), an enzyme involved in glucosamine production, can influence matrix production. Rat synovial fibroblasts were adenovirally transduced with GFAT cDNA, co-cultured with alginate-encapsulated chondrocytes and stimulated with IL-1β to investigate the influence of GFAT on the response to IL-1β (Gouze et al., 2004). While in the non-transduced co-culture control, a decrease of proteoglycan production in chondrocytes and a simultaneous increase of nitric oxide and PGE₂ in the medium were observed after IL-1β stimulation, GFAT overexpression in synoviocytes prevented these changes (Gouze et al., 2004). Thereby the study demonstrated that a co-culture model can be utilized to study how gene therapy in one tissue can affect an adjoining tissue in OA.

Complex systems can nowadays easily be reconstructed in organ-on-a-chip approaches. Monocyte extravasation in response to chemokines or the synovial fluid in OA was studied in a joint-on-a-chip-model including OA patient-derived synovial fluid, fibrin hydrogel-embedded OA chondrocytes and OA synovial fibroblasts, as well as perfusable endothelialized channels for monocyte injection (Mondadori et al., 2021) (Figure 3B). To mimic shear stress present in the channel, laminar flow of the medium was applied. This induced a shift towards more physiological expression levels of endothelial markers VCAM and ICAM, demonstrating the ability to mimic the in vivo situation to a certain extent. Additional TNFα treatment of the endothelial cells synergistically increased ICAM expression. Addition of a chemokine mix (CCL 2, CCL 3, CCL 4, and CCL 5) to the synovial fluid-mimicking compartment further stimulated monocyte extravasation. OA synovial fluid was similarly able to increase monocyte extravasation significantly, suggesting that synovial fluid might be relevant for monocyte extravasation in vivo (Mondadori et al., 2021). This model was used to test a CCR2 chemokine receptor antagonist and an antagonist for chemokine receptors CCR2 and CCR5. The antagonists reduced monocyte extravasation, showing that the model could be utilized for drug testing. The effects on the cartilage compartment was unfortunately not further investigated, nor was the added value of any of the other compartments investigated.

In order to create an OA model for subchondral bone and its vasculature, a photo cross-linked gelatin methacrylate (GelMA)-based co-culture of a mix of immortalized mesenchymal stromal cell line (MSOD) and HUVEC cells was utilized. MSOD cells underwent osteogenic differentiation by the presence of the HUVEC cells (Pirosa et al., 2021) MSOD mono and co-culture showed signs of mineralization. Cytokine stimulation (IL-1β, TNF-α and IL-6, all at concentrations found in synovial fluid of OA patients) increased endothelial network formation, similar to what is found in OA, in both HUVEC and MSOD-HUVEC culture. However, addition of the MSOD was required to maintain the network. Stimulation also induced demineralization and increase of collagen synthesis similar to changes in OA bone (Pirosa et al., 2021). Moreover, the cytokine-induced increased expression of alkaline phosphatase (ALP) and vascular endothelial growth factor (VEGF), markers for osteogenesis and angiogenesis, respectively, was more pronounced in co-culture. The co-culture was also stimulated with conditioned medium of a cartilage-on-a-chip model which consisted of chondrocytes that were embedded in a PEG-based hydrogel loaded onto a compressible PDMS device (Pirosa et al., 2021). Cells in the chip were supra-physiologically compressed to induce an osteoarthritic phenotype (Occhetta et al., 2019). Stimulation of the MSOD-HUVEC co-culture with the conditioned medium induced demineralization but prevented endothelial network formation, suggesting the pathways triggered were distinctly different from those induced by the abovementioned cytokine panel (Pirosa et al., 2021). Taken together co-culture could clearly demonstrate crosstalk between both cell types and also indicates that bone-like cells influence the phenotype of endothelial cells and vice versa. However, to what extent the MSODs really differentiated into bone cells was not clear.

Upon exposure to mechanically injured equine cartilage, primary equine healthy synovial fibroblasts reduced their ADAMTS-4 and 5 expression while increasing expression of MMP-1 as compared to co-culture with normal cartilage (Lee et al., 2013). Vice versa the fibroblasts induced higher collagen II expression in the injured cartilage and also histologically the progression towards an OA phenotype in injured cartilage seemed inhibited, evidenced by decreased cell clusters and focal cell loss (Lee et al., 2013). Also in IL-1β-stimulated equine cartilage, synoviocytes reduced GAG loss and diminished downregulation of aggrecan, although MMP-3 in synoviocytes was upregulated (Gregg et al., 2006). Altogether, synovial cells seem to exert a protective effect on cartilage. To what extent this is found in vivo is not clear.

In an elegant study where healthy rat chondrocytes were co-cultured with synovium harvested at different time points after OA induction, addition of synovium initially showed higher aggrecan production of the chondrocytes, suggesting injury induced a transient anabolic effect. With increased OA stage of the isolated synovium, however, this changed into an increase of monocyte chemoattractant protein 1 (MCP-1), a chemokine initiating inflammation (Xu et al., 2015; Lai-Zhao et al., 2021). The effects did not seem to be related to the OA induction surgery, but rather the injury of the joint, as similar results were observed using synovium from sham surgery controls (Lai-Zhao et al., 2021). A similar approach of determining the effects on joint tissue interaction of early pathological processes involved in OA development was taken by adding human ACL explants from joints with acute and chronic ACL damage to chondrocytes culture. Both types of explants increased human chondrocyte periostin and ADAMTS-4 expression, chronic remnants reduced Il-1, and increased collagen II and MMP-13 expression, whereas explants from acutely injured ACL decreased collagen II expression. It must be noted that only effects at the mRNA level were studied, and hence may not entirely reflect the metabolic state of the chondrocytes (Chinzei et al., 2018). Although applied as part of the search for cartilage regenerative strategies, a co-culture model of pellets of human chondrocytes from preserved areas in OA cartilage and periosteum explants separated by a culture insert can also be used to mimic the paracrine interaction of these tissues in the joint. Addition of chondrocytes significantly increased TGF-β and COL1A1 gene expression and induced expression of IL-6, MMP-2, -7, and -13 expression in periosteum. The periosteum induced collagen I deposition in pellets (Grässel et al., 2010; Rickert et al., 2010). In contrast, the addition of healthy bovine periosteum increased proliferation and decreased collagen II matrix deposition as well as aggrecan synthesis and release in chondrocyte monolayer, suggesting an overall inhibitory effect of periosteum (Steinhagen et al., 2012).

Osteochondral explants consist of cartilage attached to the underlying subchondral bone (Figure 4A). Although in this culture system it is not possible to point out the added value of one or the other tissue in the model, it is a frequently used model that has shown to reflect many aspects of in vivo disease (Houtman et al., 2021a; Houtman et al., 2021b). A transcriptome wide analysis of mechanically stressed human osteochondral explants demonstrated an increase of MMP-13 gene expression (Dunn et al., 2016; Houtman et al., 2021a). Moreover, increased expression of senescence markers such as FOXO and MYC1 were found after mechanical stimulation, which have also been found to be dysregulated in OA chondrocytes in vivo (Matsuzaki et al., 2018; Wu et al., 2018) indicating that the model harbors features of OA. In a study on finding a suitable culture system for OA, conventional stimulation of osteochondral explants by IL-1β was compared to triiodothyronine (T3) stimulation and mechanical stress (Houtman et al., 2021b). On the one hand all three stimuli induced an increase of HIF-2A and MMP-13 and no change in ADAMTS-5 gene expression. On the other hand only IL-1β reduced COL2A1 expression, while T3 stimulation increased COL1A1 expression and hypertrophy, while the mechanical loading only altered mechanical properties of the explant. Thus, depending on the stimulus, but most likely also its dose, different changes in the osteochondral explant can be induced that mimic the different disease aspects of OA (Houtman et al., 2021b).

The effect of Toll-like receptor 4 (TLR-4) agonist lipopolysaccharide (LPS) was tested on human osteochondral explants to see whether it could be used to create an OA model for drug testing. LPS activates TLR-4 in a similar fashion as damage-associated molecular patterns (DAMP) which are thought to play a key role in the inflammatory processes in OA (Geurts et al., 2018; Lambert et al., 2021). In the model, LPS induction increased secretion of the inflammatory markers IL-6 and MCP-1 (Geurts et al., 2018). A similar increase of IL-6 and MCP-1 was also found in LPS-stimulated bone explants, suggesting that the effects seen in the osteochondral explants involved the osseous compartment of the osteochondral explant. Treatment of the LPS stimulated osteochondral explant with SB-505124, a TGF-β receptor type I inhibitor that was shown to attenuate cartilage degradation in vivo (Zhang et al., 2018), could reduce IL-6 secretion but increased MCP-1 secretion into the medium (Geurts et al., 2018). Unfortunately, the effects on cartilage integrity were not studied to verify whether these coincided with previous in vivo findings. Of note, in the described models the osteochondral explants were cultured in the same culture medium. However, bone cells require other culture conditions than chondrocytes. Using either chondrogenic or osteogenic medium for osteochondral explants precludes optimal culture of both tissue compartments (Schwab et al., 2017). Moreover, the bone tissue may secrete cytokines and other substances into the medium at non-physiological levels (Sanchez et al., 2005a; Iwai et al., 2011). A co-culture platform for osteochondral explants consisting of two separated media compartments was developed to overcome this drawback (Figure 4B). A porcine model proved that the platform enabled supply of tissue specific medium and factors. Here, osteochondral explants could be cultured for at least 8 weeks maintaining matrix content and structural and mechanical properties without decreasing viability (Schwab et al., 2017). Combining this culture set up with the aforementioned stimuli could lead to a better OA model where both bone and cartilage could be investigated in a setup closer to the in vivo situation.

Also the crosstalk between cartilaginous tissue and synovium, where the release of inflammatory cytokines by the inflamed synovium has been considered a major driver of the production of matrix degrading enzymes by the cartilage, was addressed in co-culture systems ( Cook et al., 2007; Lee et al., 2009; Beekhuizen et al., 2011; Topoluk et al., 2018; Favero et al., 2019; Lai-Zhao et al., 2021). Adding healthy synovial tissue to canine cartilage explants (Figure 4C) maintained Toluidine blue staining indicative of proteoglycan content to in vivo levels, whereas in cartilage monoculture staining was more faint, suggesting proteoglycan loss (Cook et al., 2007). IL-1β stimulation induced a significant increase of both nitric oxide as well as prostaglandin E₂ (PGE₂) in culture medium of co-cultures, but not in cartilage monoculture. Moreover, MMP-13 secretion was increased to a higher extent after IL-1β stimulation in co-culture compared to cartilage monoculture, in line with the postulated role of the inflamed synovial lining in vivo. Gene expression of collagens, aggrecan and catabolic enzymes (MMPs, ADAMTS-4 and 5) of canine OA cartilage in co-culture with OA synovial tissue was more in line with tissue directly derived from OA patients than of cartilage in monoculture (Cook et al., 2007). Also in cultures of human OA cartilage and synovial tissue, a series of cytokines were produced that matched the expression profile seen in synovial fluid of OA patients (Zhang et al., 2018). Combining human OA synovium with OA cartilage explant culture increased cartilage expression of catabolic markers such as MMP-13, and cartilage matrix degradation. Moreover, the addition of human OA synovium to cartilage reduced viability in chondrocytes and progressively reduced GAG content (Topoluk et al., 2018), or decreased GAG production in cartilage. However, in the latter study GAG release was not affected (Beekhuizen et al., 2011). Although this does not match the classical concept of inflammation-induced matrix degradation, it is in line with several in vivo studies showing that inhibition of synovial inflammation in OA does not improve cartilage integrity (Tellegen et al., 2018; Rudnik-Jansen et al., 2019). However, using the same type of co-culture in another study the presence of OA synovium did induce an increase of GAG release (Hardy et al., 2002). The catabolic/anti-anabolic cues of OA synovial tissue may be related to the presence of macrophages, as their depletion from human OA synovial tissue not only reduced levels of IL-1β in co-culture, but also reduced catabolic responses such as MMP-13 increase (Topoluk et al., 2018).

Using human OA tissue, also a differential effect of the corticosteroid triamcinolone acetonide (TAA) was shown in co-culture. TAA prevented the decrease in GAG production induced by the presence of synovium, while in cartilage monoculture it decreased GAG production (Beekhuizen et al., 2011). Using such an OA tissue based co-culture system, IL-1β-induced cartilage proteoglycan degradation was shown to correlate with induction of COX-2 expression and PGE₂ production in the synovium. A selective COX-2 inhibitor, SC-236, blocked this degradation in the co-culture, and its effect was reversed by exogenous PGE₂. (Hardy et al., 2002). The relevance of this finding to OA, however, was not entirely clear, as COX-2 protein could not be detected in unstimulated co-culture. Generally, it should be questioned whether additional inflammatory stimulation with IL-1β would be meaningful, as such co-cultures are fully based on OA tissue. Also meniscal tissue was affected by addition of synovium from patients with early stage OA, resulting in higher expression of inflammatory markers such as IL-6 and IL-8, catabolic markers such as MMP-3 and MMP-10 and GAG release (Favero et al., 2019). Thus, co-culture of cartilage or meniscus with synovium demonstrated that co-culturing influences the degenerative and inflammatory state of the cartilaginous tissues involved, mimicking OA in vivo. However, using bovine tissue, also healthy synovium induced GAG release and collagen II degradation in healthy cartilage, which suggests that all of these models are equally valid (Siebuhr et al., 2020).

Addition of healthy equine synovium to healthy equine osteochondral explant (Figure 4D) culture increased collagen II expression in the cartilage (Haltmayer et al., 2019). While IL-1β stimulation increased secretion of TNF-α and MMP-13 by cartilage in both monoculture and co-culture, the increase was attenuated in the co-culture, suggesting a protective effect of the synovium tissue (Byron and Trahan, 2017). In contrast, a set of stimuli comprising mechanical injury, IL-1β and TNF-α addition induced a significantly higher increase in MMP-1 gene expression in co-culture with healthy equine synovium compared to equine osteochondral explants alone (Haltmayer et al., 2019). Moreover, these stimuli induced a shift towards the more inflammatory M1 type in synovial macrophages, which mimicked the in vivo situation closer than the corresponding model of osteochondral explants alone (Haltmayer et al., 2019). The lack of protection by the healthy synovium in the latter study may be explained by the impact of the combined stimuli that may have been too large to overcome.

Also the joint capsule with the fibrous outer layer has been used in co-culture (Lee et al., 2009; Swärd et al., 2017). Adding healthy bovine synovial capsule tissue to healthy bovine cartilage culture increased MMP-13 and ADAMTS-4 expression in cartilage. Mechanical injury of the cartilage additionally increased ADAMTS-5 and MMP-3 expression as well as ADAMTS- and MMP- mediated digestion of aggrecan in the cartilage (Lee et al., 2009; Swärd et al., 2017) only in the presence of the synovial capsule tissue.

Unfortunately, using osteochondral tissue makes it difficult to pinpoint the role of bone and cartilage. Investigating the interplay and assigning changes with processes in one of the tissues may help revealing the pathological mechanisms involved in OA. However, taking into account that the joint is comprised of multiple tissues, it could also be argued that the more tissue types are included the closer is the model to the in vivo situation.

Apart from bone, cartilage and synovium, other joint structures such as infrapatellar fat pad or innervating tissue are known to influence joint homeostasis, OA onset and progression (Clements et al., 2009; Cai et al., 2015; Kim et al., 2016). While no effect was observed of adding bovine healthy fat pad tissue to bovine meniscus explant culture, it did increase GAG release in cartilage explant culture (Nishimuta et al., 2017). As net GAG content within the cartilage was not altered, GAG release may have been attributed to increased GAG production, suggesting that the infrapatellar fat may also have beneficial effects (Nishimuta et al., 2017). However, addition of OA IPFP-derived conditioned medium induced proteoglycan and collagen II loss in preserved cartilage from OA patients and an increase of MMP-3 and COX-2 positive cells (Zhou et al., 2020). Additionally, conditioned medium from OA-derived IPFP increased p38MAPK and ERK1/2 signaling in human chondrocytes significantly inducing the upregulation of MMPs, ADAMTS-4 as well as IL-1β, IL-6 and COX-2 in chondrocytes. (Zhou et al., 2020). These results indicate the role of IPFP as a causative factor in OA. This is in line with an in vivo study in which lipodystrophic mice on a high-fat diet were protected from OA development and lost this protection upon fat transplantation (Collins et al., 2021). Possibly early pathological changes convert the positive effect of the IPFP in a healthy joint to a negative influence.

Clinically, pain is a major burden in OA, but the mechanisms involved in the generation of pain are still poorly understood. To investigate the influence of inflamed synovium on nervous tissue, synovial tissue from healthy human donors or osteoarthritic patients was co-cultured with dorsal root ganglia (DRG) obtained from healthy rats (Li et al., 2011). In co-culture with human osteoarthritic synovium, rat DRGs showed increased expression of the neurokinin substance P and its receptors NK1 and NK2, indicative of increased nerve stimulation (Li et al., 2011). Moreover, expression of neuropeptide Y receptor, which has been shown to be associated with chronic pain (Upadhya et al., 2009), was increased upon co-culture with OA synovium, but not healthy synovium or DRG culture alone, delineating the role of the synovium in OA pain as already indicated in vivo (Im et al., 2010; Li et al., 2011). However, as tissues were from rat and human origin, whether this is observed in human/human co-culture remains to be elucidated.

Co-culture systems have also been used for drug development. So far, various types of treatments, ranging from drugs to platelet-rich plasma and stem cells to biomechanics-changing additives were investigated in OA models (Table 4). Most of the co cultures systems used to this end have been based on the use of synovial and cartilage tissues or cells.

An insert-based co-culture system of IL-1β stimulated human chondrocytes and synoviocytes was used to look at the effect of the N-acetyl phenylalanine glucosamine derivative (NAPA) as a drug targeting the NFκB pathway involved in OA. The addition of synoviocytes to chondrocytes induced increased phosphorylation on serine 10 in histone 3, suggesting higher NFκB pathway activity and a shift to a more inflammatory state (Pagani et al., 2019). NAPA treatment was able to reduce the phosphorylation in the co-culture. Whether the co-culture model better mimics the in vivo drug response was not explained as no comparison of NAPA treatment on mono and co-culture was performed (Pagani et al., 2019). Even though chondrocytes and synoviocytes from both healthy and OA tissue were available, both cell sources were used interchangeably. However, most likely the strong stimulation with IL-1β would not have allowed for the detection of any differences. In combination with LPS stimulation, addition of sheep synovial capsule tissue to cartilage weakened the protective effect of S-(+)-ibuprofen on nitric oxide synthesis and aggrecan loss, possibly related to the higher levels of PGE₂ produced by the synovium (Bédouet et al., 2015). In contrast, the combination with synovial tissue eliminated the requirement for continuous presence of IL-1 receptor antagonist in order to prevent GAG and collagen loss and improve chondrocyte viability in IL-1α-stimulated cartilage explants (Mehta et al., 2019). The discrepancy in the role of the synovial tissue may lie in the type of stimulus, with LPS possibly inducing a more general and strong inflammatory response, also in the synovium via Toll Like receptors, compared to IL-1α. Principal component analysis of secretome data from the co-culture in comparison with respective monocultures revealed distinct clustering between the culture setups, which indicated crosstalk between cartilage and synovium (Mehta et al., 2019).

IL-1β stimulated co-cultures of cartilage and synovium from OA patients were also used to study the therapeutic effect of PRP and amniotic viscous fluid. Both significantly reduced ADAMTS-5 and TIMP-1 expression in both cartilage and synovium, back to levels seen in non-stimulated cartilage, and increased aggrecan expression in cartilage, although only PRP also reduced nitric oxide production (Osterman et al., 2015; O'Brien et al., 2019). Unfortunately, the added value of co-culture was not addressed in either study. Several recent clinical trials furthermore failed to show any effect on joint integrity and clinical outcomes of PRP treatment compared to placebo (Hohmann et al., 2020; Bennell et al., 2021). Also hyaluronic acid (HA), a key component of the synovial fluid, has been used for treating inflammation and pain in OA as so called visco-supplementation. Commercially available hyaluronic acid (Hyalgan or Synvisc) added to IL-1-stimulated canine synovium-cartilage co-cultures inhibited the loss of GAG content in cartilage, and reduced MMP-3 expression if Hyalgan was added (Greenberg et al., 2006). Yet, the in vitro effects of these biologicals are in contrast with the poor level of evidence of efficacy of HA treatment (Lin et al., 2019). This may further point towards the limited value of IL-1 stimulation in vitro models of OA, at the concentrations commonly used, which are between 100–1000 fold higher than those found in OA patients (Calich et al., 2010; Vangsness et al., 2011; Tsuchida et al., 2014). The IL-1 levels in the pg/ml range found in the synovial fluid of these patients, together with the increased concentrations of its natural inhibitor IL-RA may also explain why several clinical trials failing to show an effect of IL-1 inhibition in the treatment of OA (Kahle et al., 1992; Irie et al., 2003; Tsuchida et al., 2014). Another emerging OA treatment is the use of mesenchymal stem cells. Amniotic and adipose tissue derived MSCs were able to increase GAG content and chondrocyte viability and prevent an increase in OARSI score in cartilage, only in synovium cartilage co-culture from OA patients. (Topoluk et al., 2018). It was, however, not clear what processes in the synovial tissue were responsible for this modulating effect.

More recent approaches also investigated the use of an ADAMTS-5 targeting nanobody (Nanobody® M6495), which was able to dose-dependently inhibit the bovine synovium-induced GAG breakdown in bovine cartilage explants and in human OA cartilage monoculture (Siebuhr et al., 2020). This indicated potential as a treatment for OA, although for human OA cartilage proof of effectivity was obtained only in cartilage stimulated with high doses of proinflammatory cytokines.

Non-sclerotic and sclerotic osteoblasts were used in co-culture with chondrocytes to study the effect of carnosol, a polyphenol extracted from rosemary. Pretreatment of sclerotic osteoblasts with carnosol could prevent and even partially reverse the reduced aggrecan production by chondrocytes induced by untreated sclerotic osteoblasts. Carnosol pretreatment furthermore decreased MMP-3, ADAMTS-4 and ADAMTS-5 in chondrocytes compared to co-culture with non-pretreated osteoblasts. However, the reduced collagen II gene expression in chondrocytes could not be mitigated by pretreatment (Sanchez et al., 2015).