Human Osteochondral Explants: Reliable Biomimetic Models to Investigate Disease Mechanisms and Develop Personalized Treatments for Osteoarthritis

This article is published with digital features, including, e.g., a summary slide, to facilitate understanding of the article. To view digital features for this article go to https://​doi.​org/​10.​6084/​m9.​figshare.​13664675.

Osteoarthritis (OA) is the most prevalent chronic age-related joint disease, causing pain and disability [1]. Likely due to ignored heterogeneity in disease pathophysiology, osteoarthritis has become the most common disabling joint disease, without effective disease-modifying OA drugs (DMOADs) causing great social and economic burdens. As a result, OA significantly decreases quality of life while increasing absenteeism from work and healthcare costs [2].

The OA disease process itself is characterized by unfavourable dynamic regulation of chondrocytes upon environmental perturbations such as age or mechanical stress, likely in interaction with genetic variants that cause subtle changes in expression of OA risk genes. The OA pathophysiological process itself has been linked to enhanced metabolic activity of articular chondrocytes, resembling growth plate chondrocytes undergoing endochondral ossification [3]. OA chondrocytes enter a cascade of proliferation and hypertrophic differentiation, accompanied by expression of genes such as alkaline phosphatase (ALPL), collagen X (COL10A1) and matrix metallopeptidase 13 (MMP13), resulting in apoptotic death and mineralization of cartilage [4‐8]. Other hallmarks of the OA disease pathophysiology include new bone formation at the joint margins, limited inflammation and changes in subchondral bone structure. Together, these OA risk factors impose a persistent, yet variable, negative influence on joint tissue homeostasis throughout life, inevitably leading to progressive joint tissue destruction with age [9].

To address shortcomings of translational research and the challenges of translating data from in vitro models and a preclinical animal model to humans and increase efficiency of effective and safe drug development, while being compliant with the guiding principles of reduction, refinement and replacement of animal experiments, validated human models mimicking the different aspects of OA pathophysiology are required. Nonetheless, preclinical models thus far are limited to post-traumatic animal models or analyses of cell signalling in 2D and 3D in vitro cultures of neo-cartilage derived from human articular chondrocytes or stem cells. However, none of these models reliably recapitulate the osteochondral compartment, let alone faithfully representing age-related joint tissues prone to enter the OA process upon disease-initiating cues.

Osteochondral explant-based models allow investigation of both bone and cartilage compartments at the same time. The major advantage of such a model is that the cell response can be determined in their natural environment and they are relatively simple and easy to produce. Most commonly used explant-based models thus far were of bovine origin and applied a super-physiological perturbing factor of either a fierce inflammatory cytokines treatment [10‐12] or cartilage loading [13‐15]. Next to inflammation and mechanical loading, recapitulation of endochondral ossification and thereby hypertrophy is also thought to be one of the major mechanisms driving the processes in OA [16].

The aim of the current study is to explore and compare responses of aged human osteochondral explants triggered by three different physiological perturbing cues: inflammation (IL-1β) [17, 18], hypertrophy [triiodothyronine (T3)] [19, 20] and mechanical stress (65% strain) [21]. We determined different output measures related to catabolic, anabolic and hypertrophic chondrocyte signalling, sGAG released into the media, cartilage structure by histology and changes in mechanical properties. The presented models enable in-depth studies on how such cues interfering with homeostasis of aged cartilage contribute to human OA onset. They also allow for personalized testing of new treatment regimes in a validated human model including interaction of joint tissues and essential environmental cues.

We present human ex vivo osteochondral explants as a model system to study OA-related changes after three known pathophysiological perturbations. We applied IL-1β, T3 and 65% MS as relevant perturbing factors and studied a variety of output measures including chondrocyte signalling, cartilage structure and breakdown, and mechanical properties. Our data provide a relevant personalized human model for research on OA, which can be used for target identification and/or drug efficacy testing. The biomimetic model also complies with the guiding principles of reduction, refinement and replacement of animal experiments.

An increased catabolic response was measured after perturbation in all three models. The highest increase in MMP13 gene expression was measured in response to IL-1β (FC = 12.7), followed by mechanical stress (FC = 10.3), while the lowest increase was observed after T3 treatment (FC = 3.7). Strikingly, none of the treatments induced a significant increase in ADAMTS5 gene expression. Moreover, we measured a greatly significant increase of EPAS1 in all three OA models, indicating its sensitivity to a perturbed cartilage homeostasis. The EPAS1 gene encodes HIF-2α and its role in the onset of OA in humans is unclear as both increased [27, 28] and decreased [29] expression has been reported in human OA cartilage. Functionally, HIF-2α has been shown to regulate endochondral ossification in mouse studies by inducing expression of genes mediating chondrocyte hypertrophy (Col10a1), matrix degradation (Mmp13) and vascular invasion (Vegfa) [27].

We observed that the three different perturbations were diverse in the other outcome measures. The most severe cartilage breakdown was observed after treatment with the pro-inflammatory IL-1β and this breakdown was also characterized by an increased chondrocyte cell signalling of catabolism (MMP13 and EPAS1, Fig. 1a) and abolishment of anabolic cell signalling (ACAN and COL2A1, Fig. 2a). Gene expression of COL2A1 and ACAN was downregulated by IL-1β 100 and 33 times, respectively, suggesting a very low expression of these normally highly expressed cartilage genes. This shift in chondrocyte signalling towards catabolism is confirmed by cartilage breakdown, as measured by a stark 99% increased release of sGAG from cartilage (Fig. 4a) and by a 1.95 times increased Makin score (Fig. 5a). Upon investigation of the different subcategories of the Mankin scoring, we observed that IL-1β greatly reduced cartilage quality as measured by a 3.2 times reduction of staining for sGAG (Fig. S4C) and 1.7 times increased cartilage surface damage (Fig. S4A). These results of high cartilage breakdown in response to IL-1β are in line with many previous studies, which often observed an increased release of matrix metalloproteinases (MMPs) and other degradative enzymes, production of nitric oxide and inhibition of the synthesis of matrix proteins [17, 30]. This model might be most suitable to study interventions aimed at a subgroup of OA patients that have more inflammatory characteristics and might even suffer from rheumatic arthritis.

The perturbation with 65% MS can be considered a posttraumatic model, triggering modest OA-related changes particularly via catabolism, as reflected by the consistent yet particular effect on MMP13 and EPAS1 (FC = 1.8; P = 1.8 × 10^–20 and FC = 10.3; P = 1.4 × 10^–2, respectively). In addition, we showed a slight decrease in cartilage anabolism as measured by reduced COL2A1 gene expression (FC = 0.9; P = 8.7 × 10^–2). At the protein level we measured a 30% increase of sGAG released from cartilage (Fig. 4c) after 65% MS, corresponding to the measured slightly higher scoring for sGAG loss in toluidine blue staining (Fig. S4C). Macroscopically, we observed more macrocracks on the cartilage surface (Fig. S5) of explants receiving 65% MS and this damage was reflected in a substantial unbeneficial change of mechanical properties of the cartilage (Fig. 6). Compared to controls, explants receiving 65% MS had a 48% reduced tensile stiffness (Fig. 6a), 55% reduced Young’s modulus (Fig. 6b) and 300% increased hydraulic permeability (Fig. 6c). These results suggest that the cartilage extracellular matrix is damaged after 65% MS has been applied as it no longer appears to have the normal elastic properties and water-retaining capabilities that allow cartilage to withstand high loads. We hypothesize that this mechanism of function could be similar to exceeding the injury threshold of mechanical loading during one’s life [31]. Exceeding this threshold could occur when the mechanical load is suddenly increased or when the joint has lost its natural mechanoprotective properties.

It is generally accepted that biomechanical loading is necessary for the maintenance of cartilage homeostasis, as evidenced by the rapid loss of proteoglycans in joints that are immobilised or in disuse [32]. However, abnormal, altered or injurious loading is associated with inflammatory and metabolic imbalances that may eventually lead to OA-like damage [13, 15, 33‐35]. Moreover, ex vivo cartilage explants subjected to these magnitudes of stress exhibit a significant suppression of metabolic activity, and particularly biosynthesis of aggrecan and collagen is affected [13, 15, 33, 34, 36] similar to the in vivo situation [37]. Consistent upregulation of catabolic genes such as RUNX2, MMP1, MMP3, MMP13 and ADAMTS5 has been found in several mechanical injury models using either chondrocytes or cartilage explants [14, 15, 38‐40]. The literature has shown that levels of measured genes can vary greatly, depending on the magnitude of force, speed, age of cartilage and at which time point gene expression is measured [41‐43]. In our model, we measured targeted genes and in follow-up studies it would be interesting to measure the whole genomic transcript using RNA sequencing to identify different pathways modulating the lasting response to mechanical stress. Our model applying 65% MS might be most suitable to study interventions aimed at post-traumatic OA patients who would benefit most from a reduction of the (early) response of cartilage to mechanical stress.

In our third model we showed that in response to T3, chondrocyte signalling increased expression of the early hypertrophic markers COL10A1 and MMP13 (FC = 5.0; P = 6.1 × 10^–3 and FC = 3.7; P = 3.0 × 10^–3, respectively), while also greatly increasing the mineralization markers COL1A1 and ALPL (FC = 144.7; P = 3.0 × 10^–3 and FC = 665.8; P = 7.4 × 10^–9, respectively). Together, these results suggest that T3 induces terminal differentiation towards bone in chondrocytes. Treatment with T3 also induced a greatly consistent increased gene expression of COL2A1 (FC = 3.5; P = 2.4 × 10^–10), but did not affect ACAN expression. Nonetheless, upregulation of COL2A1 does not necessarily mean that T3 induces a beneficial response of chondrocytes, as COL2A1 is also upregulated in response to damage. In addition, a microarray study has shown that COL2A1 gene expression is upregulated in preserved compared to healthy cartilage, suggesting that there might be an early role for COL2A1 in the OA process when the cartilage is still trying to repair matrix damage [44]. To understand downstream transcriptional effects of T3, we measured RUNX2 and EPAS1, two critical transcription factors hallmarking OA and acting downstream of T3. We measured an upregulation of both EPAS1 and RUNX2 (FC = 1.8; P = 1.0 × 10^–3 and FC = 1.2; P = 7.1 × 10^–2, respectively), suggesting a possible role for both transcription factors as downstream targets of T3. The changes in chondrocyte signalling after T3 perturbation did not lead to significant changes of cartilage matrix integrity. Our results indicate that hypertrophy was induced by T3 in our explant model and that this was not necessarily detrimental to the cartilage matrix. T3 can induce changes in chondrocyte signalling directly by binding to specific thyroid responsive elements (TREs) on the DNA whereby it regulates transcription or more indirectly by activating the transcription of another transcriptional regulator such as RUNX2. However, which genes are transcriptionally regulated by T3 needs to be elucidated and regulation has been shown to be very tissue-specific because of the different levels and isotypes of thyroid hormone receptors present in different cell types. It is possible that T3 is able to induce multiple genes such as MMP13 in cartilage via binding to TREs. For example, in trβ crispant tadpoles, T3 did not induce MMP13 gene expression, suggesting that T3 acts via Trβ on inducing transcription [45].

Other researchers have seen similar effects using T3 and T4, with T3 being a more potent inducer of collagen production [46, 47]. However, these two studies did not observe an increase in hypertrophic markers such as COL10 and COL1, and this could be due to the cell type and concentration used in their experiments. On the contrary, in an in vitro chondrogenesis model using human bone marrow-derived stem cells (hBMSCs), perturbation with T3 increased chondrocyte cell signalling of terminal maturation markers (ALPL, COL1A1) [37]. Overexpression of DIO2, encoding for the D2 enzyme which converts T4 into T3, in the same model had even more detrimental effects. This explant model perturbated by T3 might be most suitable to studying interventions aimed at investigating mild types of OA that are more characterized by occurrence of hypertrophy and mineralization of cartilage.

The observation that we did not measure a response of ADAMTS5 in our three models was unexpected. Possible explanations could be that in general expression levels of ADAMTS5 were too low to be accurately assessed or that ADAMTS5 expression was too heterogeneous between patients to lead to concluding results. A more biological explanation could be the temporal and tight regulation of ADAMTS5 gene expression, peaking 10 h after injury and declining thereafter [42].

A major strength of our models is that they consider aged, yet preserved, human osteochondral explants of a heterogeneous OA patient population. As a result cartilage explants may reflect a reliable biomimetic model, prone to OA onset. Moreover, despite the heterogeneous patient population we present consistent output specific for three different relevant triggers of OA, allowing for development of different treatment modalities. Some weaknesses of the models concern the scalability and dependency of patients undergoing joint replacement surgery. In addition, we only measured changes of the overall cartilage matrix and not changes of specific proteins that make up the articular cartilage, such as collagen type II. Nonetheless, we advocate that focusing clinical development on directly counteracting these specific unbeneficial responses of chondrocytes upon these OA triggers will facilitate further personalized development and testing of desperately needed disease-modifying OA drugs. Our data provide a reference for development of advanced 3D in vitro model systems of cartilage, bone or osteochondral models aiming towards a joint on a chip using the sensitive changes in gene expression. Moreover, our model offers a next step opportunity for in depth molecular exploration with and without perturbations, e.g., by RNA sequencing in bulk or at the single-cell level.

Our study demonstrates that it is possible to set up personalized human OA disease models reflecting different relevant aspects (inflammation, hypertrophy and mechanical stress) of OA pathophysiology. The different perturbing factors and their variety in downstream effects could facilitate the development of novel targeted treatment modalities reflecting different aspects of the OA pathophysiology. Applying the here presented aged human explant model could result in a paradigm shift for biomedical research and the pharmaceutical industry leading to new ways to identify desperately needed effective drugs for OA.