Currently 3D cell-based strategies include micromass systems, also referred as 3D pellet formation. 3D pellet culture systems are obtained through suspension of different cell densites in a solution (culture medium or a buffer) and followed by centrifugation to form a material-free 3D cell aggregates in order to mimic cell-cell interactions in high cell density cell agglomerates (29, 30). Watts et al. studied cell survival, growth and chondrogenic capacity of mesenchymal stem cells (MSCs) comparing 3D pellet systems and agarose, alginate, and fibrin-alginate hydrogels (30). They demonstrated that MSCs in 3D pellet form have higher chondrogenic differentiation capacity in contrast to 3D cell-laden hydrogels through expression of matrix proteins such as SOX9, Col-II and ACAN at 28 days’ time culture. Another group also showed that pellets formed, using both OA and non-OA chondrocytes, have higher matrix protein deposition comparing to chondrocytes cultured on a hydrogel (29). Despite the 3D pellet culture systems provide reliable results, they have limitations namely apoptosis of central cells, poor RNA quality in long-term culture, and only resemble cell-cell interaction neglecting cell-matrix interactions which are highly crucial for clinical applications to restore cartilage defects.

Nevertheless, this approach is suitable to address mechanisms related to chondrocyte differentiation and ECM production under healthy conditions, mainly by focusing on cell-cell interactions (21). 3D micromass systems also successfully mimic hypoxic conditions of native chondrocytes with higher chondrogenic phenotype (31). Mennan et al. demonstrated that hypoxic conditions (2% O₂, 3D pellets) upregulated SOX9, Col-II, and downregulated activin receptor-like kinase 1 (ALK1), which promoted vascular formation, when compared to normoxia (21% O₂, monolayer) (32). In other words, Öztürk et al. demonstrated that hypoxic conditions affected 3D-cultured chondrocytes by downregulating the release of β-catenin from the Wnt complex, and, consequently, promoting the chondrogenic phenotype (25). They indicated that chondrocytes cultured under hypoxia expressed higher Col-II, ACAN and GAGs while under expressed Col-I and RhoA which are markers for hypertrophic chondrocytes.

3D micromass models are obtained by using different cell sources, however the first patient-derived tissue grafts were developed based on autologous chondrocytes (33). Tew et al. demonstrated that transduced primary chondrocytes can be expanded to longer passages and successfully re-differentiated into chondrocytes as 3D pellets form (34). However, culture of primary chondrocytes presents issues related to the limited number of cells and high variability obtained from each donor (35). Grogan et al. showed that chondrogenic capacity of 3D pellet models formed using cells from different donors (n=21) can vary depending on the donors through checking the amount of GAG content and expression of CD44, CD151, and CD49c chondrogenic markers as well as the presence of catabolic genes (MMP-2, aggrecanase-2) in in vitro conditions (35). They showed that some of the donors expressed less amount of GAG, the above mentioned chondrogenic markers as well as higher number of catabolic genes comparing to other donors. To overcome these variabilities and optimize controlled culture conditions, stem cells, including mesenchymal stem cells (MSCs) mainly derived from bone marrow (BM) placenta, skeletal muscle synovium, synovial fluid and adipose tissue, human embryonic stem cells and recently induced-pluripotent stem cells (iPSCs) have been proposed (36, 37). Each cell type has different advantages and disadvantages, as showed in the Table 1 . To overcome the disadvantages from monoculture systems, Cooke et al. demonstrated that co-culture of MSCs with chondrocytes could be a promising alternative as chondrocyte hypertrophy is avoided, one of the characteristics of OA that results into cartilage mineralization (40). The chondrogenesis mechanism in co-culture systems is not yet well described, even though it is reported that co-culturing chondrocytes with MSCs led to higher expression of SOX9, aggrecan, and Col-II, and lower levels of MMP-13, Runx2 and collagen-I, when comparing to MSC pellets (40).

For better understanding of the joint tissue environment, other joint tissue models, both in healthy and inflammatory conditions, must be highlighted. Kiener et al. established a 3D pellet model by co-culturing synoviocytes and monocytes to mimic synovium using Matrigel^® (41). This study was followed by another work established by Broeren et al. They demonstrated that triggering the same micromass model through TNF-α cytokine creates proinflammatory environment and leads to fibrosis and fibroblastic phenotype similar to OA conditions (42). Later, more complex co-culture models have been developed to focus on the crosstalk of different joint tissues under inflammation. Peck et al. successfully established a scaffold-free model to mimic cartilage destruction by co-culturing FLS, macrophages and chondrocytes (43). In this 3D model, chondrocytes were activated through pro-inflammatory macrophages. The expression of ECM proteins (Col-II and Aggrecan) was observed to decrease while ECM degrading enzymes (MMP-1, 3, 13) and inflammatory genes (TNF-α, IL-1β, IL-6) expression increased. Furthermore, this model was also tested for celecoxib which is a highly adopted anti-inflammatory drug for OA patients. Applying celecoxib to the model resulted in decreasing of cartilage damage. Therapeutic effect of the drug was characterized through reduction of MMP-1, MMP-3 as well as increased amount of GAG and collagen. The combination of different cell types has opened new avenues regarding the investigation of the crosstalk between different tissues. However, it should not be neglected that the development of 3D OA models is highly dependent on the physiological relevance and biological accuracy. 3D models mentioned above have focused on cell-cell interactions, without considering cell-matrix interactions until cells produce their own ECM. This issue led to the establishment of scaffold-based 3D models for cartilage tissue engineering.

In tissue engineering approach, scaffolds support and induce not only 3D matrix synthesis and organization, but also promote cell adhesion and proliferation (44, 45). Cells may be seeded into the scaffold and then transplanted into the patients to contribute to tissue regeneration at the defected site (46). For this reason, a suitable scaffold should be biocompatible, mechanically stable to allow transplantation, non-immunogenic, permeable to serve as a carrier for growth factors, nutrients and cytokines, allow the promotion of chondrogenic phenotype and, in the end, support cartilage tissue formation (44). In addition to this, it should be biodegradable to assist tissue remodeling by allowing the replacement of damaged tissue with newly generated cartilage ECM (45). Depending on the defect, it might be preferred to be injectable to minimize surgical damage. Scaffold-based therapies have more advantages compared to cell-based therapies. They provide a 3D environment that diminishes chondrocyte dedifferentiation to encourage hyaline cartilage formation. They also provide support on the development of a mechanically more stable tissue that decreases postoperative recovery time (46).

Decellularized matrices and hydrogels are used for joint tissue engineering applications. For instance, cartilage and meniscus derived decellularized matrices have been studied to promote MSCs differentiation into chondrocytes and to stimulate cartilage ECM formation without causing hypertrophic differentiation (47, 48). Visser et al. demonstrated that the presence of decellularized matrices promotes cartilage formation due to biochemical cues, which includes Col-II (49). However, majority of scaffolds for joint tissue engineering applications are hydrogel-based scaffolds. Hence, hydrogels have been incorporated together to provide functional growth of articular cartilage both in in vitro and in vivo bioreactor systems due to their ability to form irregular shapes and porous structure (50, 51). There has already been certain number of biomaterials that are studied for joint tissue applications. These materials can be classified as natural [collagen, chitosan, alginate, hyaluronic acid (HA), fibrin] and synthetic [Polyethylene glycol (PEG), polylactic acid (PLA), polycaprolactone (PCL), poly(vinyl alcohol) (PVA)] (44). Natural biomaterials have weaker mechanical properties while synthetic materials might trigger toxicity (44). This issue led to the development of composite biomaterials, which aim to offer the best properties from natural and synthetic materials. To form native-like 3D joint constructs using biomaterials, different approaches have been applied, from cell encapsulation via simple pipetting to complex 3D bioprinting technologies (52).

Hydrogel-based materials can be combined with cells to obtain 3D systems that may be used as tissue grafts in joint regeneration (53). Encapsulation allows to culture cells in a scaffold that act not only as native-like ECM, but also mechanically supports cells, and promotes cell survival, proliferation, migration and differentiation (54). Hydrogels can be produced via two different crosslinking methods: chemical and physical crosslinking. Physical crosslinking occurs due to ionic interactions and hydrogen bonds. Due to the weaker bonds, these interactions are reversible. On the other hand, chemical crosslinking occurs through covalent interactions between polymeric solutions and additives to create irreversible and more stable hydrogels (54). One of the most common chemical crosslinking methods is photo-crosslinking through Ultraviolet (UV) and visible light. For instance, the scaffold might be exposed to UV light for a time period to obtain reticulation between the main polymer and photoinitiator (PI) (55). Photoinitiators are substances that produce reactive species, such as cations, anions and free radicals, when exposed to UV and visible light, that allow to obtain photocrosslinked hydrogels. Each PI type has different absorption spectrum wave length, being the most common Lithium phenyl-2,4,6 trimethyl benzoyl phosphinate (LAP), which is activated at 405nm and Irgacure 2959 (IC) is activated at 365nm. Camphorquinone, fluorescein and riboflavin are also widely used and are activated at 450nm, 490nm, and 444nm, respectively (56). These became popular since last decade and are able to finely tune the hydrogel properties in order to obtain the optimal conditions for cells. It has been reported that decreasing PI concentration and UV light intensity results in higher exposure time to form mechanically stable hydrogels (57). Furthermore, the mechanical properties of hydrogels and behavior of encapsulated cells can be controlled through changing certain parameters, such as precursor solution concentration, UV intensity and exposure time (55). Those parameters should be tailored to each cell type to avoid cell toxicity and to provide the necessary microenvironment for, in this case, chondrocyte proliferation, differentiation and chondrogenic phenotype maintenance (55). Otherwise, over crosslinking may result in poor nutrient, growth factor or cytokine diffusion and may lead to the generation of reactive oxygen species (ROS) that induces oxidative stress in cells. On the other hand, a low crosslinking degree may result in lower hydrogel viscosity, and, therefore, poor mechanical stability (58, 59). Several research groups investigated how the material viscosity affected chondrocyte and MSCs behavior through encapsulation of cells in different polymer-based hydrogels (60–64). They indicated that highly viscous hydrogels promoted the retaining of chondrogenic differentiation in both chondrocytes and MSCs, while lower viscosity ones are beneficial for proliferation of chondrocytes. Additionally, Pahoff et al. evaluated the effect of gelatine source [bovine (B) or porcine (P)] and PI type LAP and IC on the hydrogel mechanical properties and cellular behavior of chondrocytes that were encapsulated in GelMA/HAMA hydrogels after 28 days of cell culture (24). They demonstrated that hydrogels formed via B-IC displayed higher similarity on compressive strength of native articular cartilage, as well as higher expression of GAGs. Even though the gene expression analysis declared that chondrocytes encapsulated in B-IC showed an increased chondrogenic phenotype, all hydrogel constructs promoted the expression level of Col-II and aggrecan with no difference related to gelatine and PI type. To promote the chondrogenic phenotype additional polymer modifications may be needed. Salinas et al. and Hao et al. encapsulated MSCs in PEGDA hydrogels modified with RGD (Arginine-Glycine-Aspartate) peptide, an integrin binding peptide that is known as promoter of MSCs survival and inducer of chondrogenesis (65, 66). These research groups concluded that RGD peptide modification upregulated the expression of GAGs and Col-II protein via blocking the MMP-13 enzyme.

3D bioprinting is a computer-assisted technology that aims to precisely fabricate biological three-dimensional structures in an organized and optimized manner on a layer-by-layer basis, by depositing small droplets or filaments into specific substrates (67). These systems offer to print biological constructs in a desired shape through mixing cells with a solution of natural, synthetic or hybrid biomaterial (mainly hydrogels) and it is called bioink (39). After several iterations along the years, a number of influential researchers in the bioprinting field defined bioink as ‘a formulation of cells suitable for processing by an automated biofabrication technology that may also contain biologically active components and biomaterials’ (68). However, some researchers divide bio-inks into four categories depending on its function: support bioinks, fugitive bioinks, structural bioinks and functional bioinks (69). Support bioinks are the most common bioink type and act as artificial ECM to support cellular behaviour, while fugitive bioinks are temporary materials that can be easily removed by washing or temperature changes. The latter is generally applied when temporary supports are needed or to obtain channels inside the bioprinted structure. To provide mechanical integrity, structural bioinks are often used. These can also have fugitive properties on a long timescale. Finally, functional bioinks are adopted to increase the biomimicry level of the structure by providing biochemical, mechanical or electrical cues. Ideal bioinks should promote healthy cellular behaviour (viability, proliferation, differentiation) to assure functionality of printed tissue, as well as providing necessary mechanical, rheological properties to form a 3D construct (70).

3D bioprinting is divided on the following techniques: inkjet, laser-based, extrusion based bioprinting. Inkjet bioprinting was the first bioprinting technique, and is divided in thermal, electrostatic or piezoelectric inkjet bioprinting (71). Bioinks are ejected based on drop-on-demand technology (DOD) that provides high resolution printing (72, 73). This method offers low-cost production, digital control of highly resolution patterns and high printing speed. However, besides that, it is only suitable for low viscosity and low cell density bioinks (74). This limitation has been solved through laser-based bioprinting, that allows to print high resolution and homogenously distributed cells (39, 75). Laser based bioprinting is a nozzle-free process that works based on the pulsation of laser energy into a cell-containing layer that precisely ejects cell onto a substrate. Although this technique allows to print high-resolution patterns with high cell viability, it is an expensive and time-consuming bioprinting method (73). Extrusion-based bioprinting recently became popular due to the fact that it overcomes some of the limitations of the techniques previously described while allowing to print fiber sized constructs offering possibility to mix different bioinks (76). This technique works based on the deposition of bioink through a nozzle driven by pneumatic pressure, or a piston (72). However, this technique also has some disadvantages such as low resolution, deformation of construct and poor cell viability due to the increased shear stress that occurs during the extrusion process (77). Despite those advantages and disadvantages of all bioprinting techniques, they have been accepted as reliable systems in terms of cell viability and proliferation (67).

3D bioprinting technologies have been used for joint tissue applications and hydrogels are commonly used as bioinks for 3D bioprinting approaches due to their rheological behaviour, biocompatibility, and mechanical properties (78). Collagen, HA, gelatine, chitosan and alginate are the most adopted polymers on bioinks for printing cartilage tissue (79). Bioink viscosity must be suitable for each bioprinting technique to ensure printability and maintenance of the shape of the post-printing construct (80). Hence, the previously discussed parameters for hydrogel crosslinking are also critical for bioprinting technologies. This way, Markstedt et al. developed a nano fibril cellulose (NFC) and alginate based bioink to obtain optimal viscosity for chondrocyte culture (79). They concluded that their bioink (NFC/Alginate; 80:20) showed stable mechanical properties, when comparing to both before and after crosslinking, and further promoted cell viability after 7 days of culture. However, this study did not include any task regarding chondrogenic phenotypic characterization. Singh et al. performed a study on the development of silk-gelatine based bioink for cartilage tissue engineering applications (78). They successfully developed a crosslinker-free silk-gelatine based bioink with the required rheological and mechanical properties for printing. This material also promoted biocompatibility on both in vitro and in vivo after 14 days. Their bioink boosted the chondrogenic phenotype (ECM proteins: Col-II, SOX9, Aggrecan, GAGs) while decreasing cell’s hypertrophy. Lam et al. studied GelMA and hyaluronic acid methacrylate (HAMA) based hydrogel scaffolds regarding cell viability, morphology and chondrogenic phenotype for two different cell densities through stereolithography based bioprinting (81). They found that high cell density cell-laden GelMA hydrogels expressed more ECM proteins (Col-II and Aggrecan) at the end of 14 days of culture. It was also demonstrated that addition of HA to GelMA increased the polymer viscosity, provided more controlled, and homogenous bioprinting, and promoted expression of GAGs and Safranin (82). Duchi et al. introduced an apparatus called ‘Biopen’ that allowed to print viable adipose stem cells inside a hydrogel shell formed by GelMA and HAMA. This device is expected to be a promising tool for printing of 3D structures directly into chondral defects to promote in situ cartilaginous tissue formation (57, 83). Table 2 summarizes the most common materials for different joint tissue models.

Besides scaffold-based 3D in vitro models, tissue explants have also been investigated as they provide cells with their natural environment (3). Grenier et al. established an in vitro OA model through collagenase treatment of cartilage explants and by applying mechanical stress on the tissue (89). They demonstrated that collagenase treatment provided degradation of cartilage ECM in the superficial and intermediate zones of cartilage tissue, which was confirmed through histological analysis. To create more complex models, a cartilage–synovium explant co-culture model was developed by Beekhuizen et al. (90). Their results indicated that cartilage tissue co-cultured with synovium tissue from the same donor decreased GAGs expression when comparing to monoculture. Additionally, this group observed a similar pro-inflammatory cytokine expression in both conditions. Even though tissue explant models provide cells with similar physiological environments, availability of donor and/or intra-donor variability between samples limits the application of tissue explants in the field. These models are able to provide a proper environment similar to the in vivo situation, however they cannot incorporate the physiological stimulations of cartilage tissues, such as continuous nutrient supply and, more importantly, mechanical stimulations. In that view, more advanced cell culture platforms have been investigated, namely bioreactors and organ-on-chip systems.

Bioreactors are devices that enable to culture cells in a controlled environment (e.g., temperature, pH, nutrient supply, mechanical stimuli). Spinner flasks, rotating wall vessels, and perfusion bioreactors are examples of bioreactors that have been designed based on tissue and application. To mimic shear, compression, and hydrostatic pressure of healthy joint, Park et al. developed a bioreactor that applied compression and shear stress 1 hour per day on an engineered cartilage construct which was produced by encapsulating chondrocytes in a fibrin/hyaluronic acid hydrogel (91). This research group concluded that their engineered construct expressed more GAGs and Col-II which were horizontally parallel to the stimuli side when comparing to static group. To understand the interplay between bone and cartilage, Schwab et al. established an osteochondral culture platform that comprised two separated compartments to allow the seeding of bone and cartilage derived cells in one platform (92). This platform was able to provide tissue-specific growth factors to each cell type during long-term cultures (over 84 days) while maintaining the cartilage matrix content and its mechanical properties. Despite the fact that this osteochondral model provided new insights regarding the interaction between tissues, it is still limited to investigate the interplay between bone and cartilage for longer culture time.