Models of osteoarthritis: the good, the bad and the promising.

Osteoarthritis (OA) is a chronic degenerative disease of diarthrodial joints most commonly affecting people over the age of forty. The causes of OA are still unknown and there is much debate in the literature as to the exact sequence of events that trigger the onset of the heterogeneous disease we recognise as OA. There is currently no consensus model for OA that naturally reflects human disease. Existing ex-vivo models do not incorporate the important inter-tissue communication between joint components required for disease progression and differences in size, anatomy, histology and biomechanics between different animal models makes translation to the human model very difficult. This narrative review highlights the advantages and disadvantages of the current models used to study OA. It discusses the challenges of producing a more reliable OA-model and proposes a direction for the development of a consensus model that reflects the natural environment of human OA. We suggest that a human osteochondral plug-based model may overcome many of the fundamental limitations associated with animal and in-vitro models based on isolated cells. Such a model will also provide a platform for the development and testing of targeted treatment and validation of novel OA markers directly on human tissues. Crown

Introduction

Osteoarthritis (OA) is a chronic degenerative disease of diarthrodial joints, predominantly affecting the spine and peripheral joints of the body, particularly the hands, hips, knees and feet. OA most commonly affects people over the age of forty, with the risk of disease increasing with age. OA is a complex heterogeneous disease with different clinical and biochemical phenotypes.

The cause(s) of OA are unknown, and many studies have suggested that the pathobiology of OA is far more complex than a simple cartilaginous or bone disease. It is now acknowledged that OA affects many joint structures, including degeneration of cartilage, abnormal bone remodelling and synovial inflammation1, 2. Also, studies have shown that there is a complex interplay between the different joint components, making understanding of the degradative sequence of events involved in OA pathogenesis very difficult to dissect3, 4, 5.

The initial onset of OA disease is considered due to an imbalance between the cartilage degradation and repair process6, 7. The exact sequence of events that trigger the onset of the disease is however widely debated throughout the literature. One hypothesis, suggests that secretion of pro-inflammatory cytokines into the synovial joint induces matrix metalloproteinases which cause the fragmentation and degradation of cartilage extracellular matrix leading to bone remodelling and synovitis8, 9, 10. Contrary to this theory, some studies suggest that subchondral bone remodelling and synovitis precede articular degeneration in the early stages of OA8, 11, 12. While other studies suggest that meniscal degeneration evolving through fibrillation of tissue and a decrease in the levels of type I and II collagen within the meniscus, act as a predisposing or contributing factor to OA progression13, 14. In the later stages of OA, formation of subchondral cysts, subchondral sclerosis and osteophytes occur as a direct result of bone remodelling, cartilage degradation and synovitis15, 16, 17.

Treatment of OA is largely symptomatic due to insufficient understanding of aetiopathogenesis hindering the development of suitable disease-modifying drugs. This makes targeted treatment of OA a distinct challenge. Human OA tissue samples are usually collected for research once end stages of the disease have been reached, for example during joint replacement, by which time destructive changes in the joint are well established. This makes studying the early disease process very challenging. OA pathology, particularly early OA, is therefore very difficult to study, and so researchers turn to in-vivo and ex-vivo preclinical animal models to investigate early pathological changes in OA. These models offer unique advantages as well as limitations for studying human OA. This article will review the different models used for investigation of OA, discuss their advantages and disadvantages, and propose development of a gold standard model for OA that closely reflects natural human disease.

Current models used in OA research

OA research models can be categorised into either ex-vivo or in-vivo models. Depending on the research question, different models can be used to address different aspects of OA development and progression. Each model has its advantages, yet it has become clear that no single model provides the opportunity to study the disease as a whole. The different models currently used in OA research are discussed below.

Ex-vivo models

Ex-vivo models can be categorised into monolayer culture, co-culture, three-dimensional (3D) culture and explant-based culture. Each model has its advantages and disadvantages and so can be used to answer different questions in OA research.

Ex-vivo models such as monolayer culture and co-culture are easier and cheaper to produce than 3D cell cultures and explant-based models. Monolayer cultures are also easy to produce on a large scale and avoid the challenges associated with culturing different cell types at different conditions. However, monolayer and co-cultures are limited in their use due to the fact that they isolate only one or two tissue components at a time. Many studies have shown that there is a strong interplaying network of communication between different joint components that help regulate and maintain a healthy joint, and so isolation of specific joint components hinders this communication3, 19, 20. For example, healthy articular cartilage is dependent upon the release of soluble factors by subchondral bone, and interactions between chondrocytes and synovial fluid ensures the flow of growth factors, regulatory peptides and nutrients between them19, 20. When injured cartilage is co-cultured with synovium, a protective effect is produced on the synoviocytes. Similarly, culture of subchondral bone and cartilage separately results in increased chondrocyte death and cartilage degradation as well as decreased protein content in culture media compared to when cultured together3, 19, 22. Explant models and 3D cell cultures allow for this inter-tissue communication and so are arguably more useful models available to OA researchers to reproduce natural in-vivo environments. Despite this, these models are more difficult to produce in terms of tissue volume and maintaining cell viability over extended periods of time. Some of the advantages, disadvantages and applications of various ex-vivo models used in OA research are summarised in Table I.

Table I

A summary of the advantages and disadvantages of different ex-vivo models used in OA research

Ex-vivo Model	Advantages	Disadvantages	Example of the application of the model in OA research
Monolayer culture	- A large number of cells can be easily produced from a single sample23 - The configuration of cells cultured in a monolayer layout allows homogenous spread of nutrients and growth factor from the culture medium24	- Limited for certain tissue types such as cartilage, whose phenotype changes once in a monolayer culture environment, introducing inter-experimental variability25, 26 - Chondrocytes are very sensitive to their molecular environment and so need to remain in contact with the extracellular matrix to ensure that they reflect natural in-vivo samples23 - Cartilage has low cellularity, therefore, a large sample of cartilage is required to ensure sufficient numbers of cells are present to carry out a reliable experiment23 - Isolating a tissue in culture removes all systemic influences on that tissue, which does not reflect natural joint tissue - Cells in monoculture traditionally grow on a flat surface in glass or plastic flasks and so do not allow for growth in all directions, as seen in the natural 3D in-vivo environment24	- Monolayer cultures can be used to study the effects of cytokine stimulation and osmotic pressure23 - Synovial cell cultures useful to study the role of the synovium in OA
Co-culturing cells	- Co-culturing cells of different lineages is important to allow for changes in cell-specific physiology and cell–cell interactions that are important in regulating cell and tissue physiology23, 27	- Different conditions are required for culturing each cell type23 - Co-culturing cells can result in alterations of phenotype when cells are isolated23 - Co-cultures traditionally grow on a flat surface in glass or plastic flasks and so do not allow for growth in all directions, as seen in the natural 3D in-vivo environment24	- Co-culturing cells can be used to study the effects of cytokine stimulation and osmotic pressure23 - Osteoblast-chondrocyte co-culture useful in understanding bone-cartilage cross-talk28 - Co-culturing chondrocytes and osteoblasts results in greater cell growth, matrix production and deposition as well as reduced glycosaminoglycan deposition compared to culturing chondrocytes alone29, 30 - Co-culturing sclerotic osteoarthritic osteoblasts and chondrocytes from osteoarthritic articular cartilage results in an increased shift towards chondrocyte hypertrophy and release of matrix metalloproteinases and aggrecanases31, 32 - Culturing synovium and cartilage together produce very different results in terms of the break-down of proteoglycan and matrix structure compared to when cultured alone4 - Co-culturing synovium and injured cartilage produces a protective effect on synoviocytes21 - Synovium-cartilage cultures useful to study the role of the synovium in OA - Co-culture of bone components ensure balanced bone remodelling5
3D cell culture	- 3D cell culture allows for culture of different cell lines and important cell–cell interactions - 3D cell cultures grow as aggregates or spheroids in a matrix, allowing growth in all directions, similar to the natural in-vivo environment24 - The 3D structure provides structural strength to sensitive cells23	- The proliferation rate of cells tends to be slower in 3D cell cultures compared to 2D cultures33 - The structural strength provided to cultured cells depends on the scaffold used23	- 3D cell culture can be used to study the effects of cytokine stimulation and osmotic pressure, as well as the effects of physical injury and loading on tissue23 - A matrix structure of collagens and proteoglycans favours phenotypically normal cartilage28
Explant based models	- Simple, cheap and easy to produce23 - Explant models allow for the natural processes that occur within the extracellular matrix environment to be observed26	- Cell death often occurs at the explant edge - Only a limited number of cells can be extracted from a single source - Limited tissue availability and significant inter-experimental variability23	- Explant based models can be used to study the effects of cytokine stimulation and osmotic pressure, as well as the effects of physical injury and biomechanical loading on tissue23, 28 - Synovial tissue explants useful to study the role of the synovium in OA

A large number of cells can be easily produced from a single sample

The configuration of cells cultured in a monolayer layout allows homogenous spread of nutrients and growth factor from the culture medium

Chondrocytes are very sensitive to their molecular environment and so need to remain in contact with the extracellular matrix to ensure that they reflect natural in-vivo samples

Cartilage has low cellularity, therefore, a large sample of cartilage is required to ensure sufficient numbers of cells are present to carry out a reliable experiment

Cells in monoculture traditionally grow on a flat surface in glass or plastic flasks and so do not allow for growth in all directions, as seen in the natural 3D in-vivo environment

Monolayer cultures can be used to study the effects of cytokine stimulation and osmotic pressure

Different conditions are required for culturing each cell type

Co-culturing cells can result in alterations of phenotype when cells are isolated

Co-cultures traditionally grow on a flat surface in glass or plastic flasks and so do not allow for growth in all directions, as seen in the natural 3D in-vivo environment

Co-culturing cells can be used to study the effects of cytokine stimulation and osmotic pressure

Osteoblast-chondrocyte co-culture useful in understanding bone-cartilage cross-talk

Culturing synovium and cartilage together produce very different results in terms of the break-down of proteoglycan and matrix structure compared to when cultured alone

Co-culturing synovium and injured cartilage produces a protective effect on synoviocytes

Co-culture of bone components ensure balanced bone remodelling

3D cell cultures grow as aggregates or spheroids in a matrix, allowing growth in all directions, similar to the natural in-vivo environment

The 3D structure provides structural strength to sensitive cells

The proliferation rate of cells tends to be slower in 3D cell cultures compared to 2D cultures

The structural strength provided to cultured cells depends on the scaffold used

3D cell culture can be used to study the effects of cytokine stimulation and osmotic pressure, as well as the effects of physical injury and loading on tissue

A matrix structure of collagens and proteoglycans favours phenotypically normal cartilage

Simple, cheap and easy to produce

Explant models allow for the natural processes that occur within the extracellular matrix environment to be observed

Limited tissue availability and significant inter-experimental variability

In-vivo models

Many animal models in at least eighteen different species have been developed to study established pathological features of OA such as pain, synovitis, cartilage degeneration and bone remodelling. Animal models used in OA research (see Table II) can be categorised into either induced or spontaneous models. Induced models refer to models where OA disease (or OA like features) have been induced either chemically or surgically. On the other hand, spontaneous models are subcategorised into naturally occurring and genetically modified models that develop OA.

Table II

A summary of the different animal models used in OA research

Species/Model	Spontaneous	Surgically induced	Chemically Induced	Examples of the application of the model in OA research
Mouse	Naturally occurring OA1, 2, 34- Genetic models: PAR2−/−, CD4−/−, MMP17−/−, Tenascin C−/-, Ddr2−/−, SulPhatase−/- 1/2, Syndecan 4−/−, Fgf2−/−, Mmp13−/−,Hif2a+/-, GDF5+/−, Osteopontin, Ptges1, Tnfrsf11b+/-, Runx 2+/−, ADAMTS-5/4 −/−, Adamts5−/−, ADAMTS4−/−, MMP3−/−, ICE−/-, IL-1β−/-, iNOS −/−35- Transgenic models2, 36 Mutations in type II collagen gene1Brtl mouse37Mouse Del1: Short deletion in type II collagen18Col9a1 knockout36STR/ORT + C57/BL6 strains1, 2, 36, 38, 39, 40	- Anterior cruciate ligament transection (ACLT)18, 34, 41, 42, 43 - Articular groove model34 - Intra-articular tibial plateau fracture, cyclic articular cartilage tibial compression, anterior cruciate ligament, rupture via tibial compression overload2 - Ovariectomy34 - Partial discectomy44 - Medial partial meniscectomy34 - Destabilisation of medial meniscus, meniscectomy, tibial overload, fracture models34 - Meniscal destabilisation34, 38, 40, 43 - ACLT and removal of medial/lateral meniscus or transection of posterior/medial/lateral collateral ligament34	- Mono-iodoacetate (MIA) intra-articular injection34, 40 - Steroids, cytokines34, 40 - Papain34, 40 - Collagenase34, 40, 41	- Mouse models widely used for toxicology testing.1 - Mouse models used to study the molecular basis of OA.45 - Genetically modified mouse models used to investigate the genetic factors and specific genes involved in cartilage degeneration, bone remodelling and inflammation2, 18, 43, 45
Rat	- Naturally occurring OA uncommon18, 41	- ACLT34, 41, 46 - Medial meniscectomy (MMx)34, 40, 41, 46 - Articular groove model34 - Medial meniscal transection (MMT)1, 34, 41, 45 - Combination surgery40 - Ovariectomy34, 40, 41 - Partial medial meniscectomy45 - Immobilization40 - ACL injury47	- Intra-articular injection of steroids, cytokines40 - Collagenase40, 41 - Iodoacetate injection18, 34, 40, 41, 45, 48 - Papain34, 40, 41 - Immunotoxin41	- Rat model useful in toxicology testing of pharmaceutical compounds1. MMT, medial collateral ligament transection (MCLT) and iodoacetate induced models used to study pain40, 45 - Rat undergone partial medial meniscectomy are useful in cartilage restoration techniques45
Syrian hamster	- Naturally occurring OA1, 34 - Transgenic models1			- Syrian hamster OA models are naturally occurring, and transgenic models are used to study pathogenesis of OA1
Guinea pig	- Naturally occurring OA2, 34, 39, 40, 41, 45, 49 - Transgenic models1 - Naturally occurring OA in medial compartment of knee joint in Dunkin Hartley guinea pigs1, 18	- ACLT40, 41 - MCLT, osteotomy, patellectomy, sciatic neurectomy41 - Meniscal transection1, 34 - Ovariectomy34, 41 - MMx41, 50 - Combination surgery40	- Immunotoxin, papain, collagenase, copper II bisglycinate, lipopolysaccharide, chondromucoprotein41 - MIA34, 41 - Quinolone34	- Transgenic guinea pig models used to study pathogenesis of OA1 - Guinea pig models used to study age and BMI associated risk factors in OA49. Dunkin Hartley guinea pig used in therapeutic and pathogenic studies of knee OA51. Guinea pigs induced by medial meniscal tear and spontaneous OA models used to study slow and rapidly progressive OA34
Cat		- ACLT40		- Useful in pain studies2
Rabbit	- Naturally occurring OA2, 52	- ACLT, MMx, posterior cruciate ligament transection (PCLT), patellectomy34, 39, 40, 41 - ACL tear45 - Section of medial collateral and both cruciate ligaments, resection of medial meniscus53 - Immobilization40 - Combination surgery, impact loading, cartilage scarification40 - ACLT, ACLT and PCL/MCL34 - Ovariectomy34 - Articular groove34 - Partial and MMx1, 34, 39, 45 - Transarticular mechanical impact on patellofemoral joint, femoral condyle impact2	- Intra-articular injection of steroids and cytokines40 - Papain34, 40, 54, 55 - Allogeneic cartilage particles56 - Iodoacetate and collagenase40, 41 - Quinolone34 - Chymopapain, trypsin, IL-1β, chondroitinase, vitamin A, fibronectin fragments41	- Rabbit models useful in efficacy testing of various compounds such as hyaluronic acid45. - Partial meniscectomy models used in testing chondroprotective agents1
Canine	- Naturally occurring OA2, 34, 41	- Abrasion, valgus osteotomy, pelvic osteotomy, cartilage defect41 - Cranial cruciate ligament transection57 - Articular groove34, 41, 58 - ACLT1, 2, 41, 45, 58, 59 - Partial medial and MMx34, 41 - Immobilization1, 40 - Impact loading, cartilage scarification40 - ACLT60 - Groove model in femoral condyle - Transarticular impact to stifle34, 58	- MIA, papain, calcium pyrophosphate crystals34, 41 - Allogeneic cartilage particles61	- MMx model useful in toxicology testing and ACLT model used to study slow progression of OA and pathogenesis that mimics naturally occurring disease1 - Canine models that naturally develop OA have been used in therapeutic intervention preclinical trials2 - Transarticular impact models were used to identify whether osteoarthritic changes originate from cartilage or subchondral bone changes34 and to study early changes in OA in articular cartilage due to joint impact trauma34 - ACLT induced model have been used in identification of OA biomarkers45
Caprine	-Naturally occurring OA41	- MCLT, ACLT41 - MMx34, 39, 41, 45, 62 - Articular groove34 - Unilateral medial MMx, unilateral MCL, meniscal transection, cartilage scarification, unilateral ACLT63		- Goat models used to study cartilage repair45
Ovine	- Naturally occurring OA2	- Lateral meniscectomy, ACLT, MCLT41 - Articular groove model34 - MMx34, 39, 40, 41, 45, 62 - Bilateral and unilateral Mx, unilateral ACLT, medial MMx, unilateral MCLT, unilateral radial meniscal tear, unilateral caudal pole hemi-meniscectomy, unilateral medial meniscectomy63 - Ovariectomy34, 40		- Ovine models used to study early OA cartilage changes, meniscus changes and related treatment techniques45, 64
Equine	- Naturally occurring OA2, 45 - Post carpal fracture, exercise-induced41 - Trauma to medial femur and tibia65	- Metacarpophalangeal ligament transection41 - Osteochondral fragment66 - Articular groove model34	- Amphotericin, E. coli lipopolysaccharide, IL-1, carrageenan41 - Monosodium iodoacetate40, 41, 48 - Filipin34, 41, 67 - Lipopolysaccharide68 - Amphotericin69 - Polyvinyl alcohol foam particles41, 70 - Papain40 - Intra-articular injection of steroids, collagenase, cytokines40	- Equine models used to study articular cartilage repair, osteochondral defects and naturally occurring bone remodelling2
Zebrafish	- Genetic knockout e.g. COL10A171			- Zebrafish model useful in studying gene related pathology of OA71
Porcine	- Post-fracture72	- ACLT, ovariectomy, ACL reconstruction, articular groove model34 - Arthroscopy - Cartilage resurfacing Surgically in miniature pigs: - ACLT and ACL reconstruction41		- Porcine model used to study repair and regeneration of focal cartilage defects73
Bovine	- Naturally occurring OA in patella74, 75	- ACLT34
Non-human primates	- Naturally occurring OA1, 34, 76 - Naturally occurring OA in macaques34, 40, 77, 78 - Transgenic models1	- Meniscectomy34, 40 - Ovariectomy in macaques40, 77, 79 - Ovariectomy in cynomolgus monkeys34	- Collagenase induced in cynomolgus monkeys80	- Naturally occurring and transgenic models of non-human primates used to study general features of OA1

Articular groove model

Intra-articular tibial plateau fracture, cyclic articular cartilage tibial compression, anterior cruciate ligament, rupture via tibial compression overload

Ovariectomy

Partial discectomy

Medial partial meniscectomy

Destabilisation of medial meniscus, meniscectomy, tibial overload, fracture models

ACLT and removal of medial/lateral meniscus or transection of posterior/medial/lateral collateral ligament

Mouse models widely used for toxicology testing.

Mouse models used to study the molecular basis of OA.

Articular groove model

Combination surgery

Partial medial meniscectomy

Immobilization

ACL injury

Intra-articular injection of steroids, cytokines

Immunotoxin

Rat model useful in toxicology testing of pharmaceutical compounds. MMT, medial collateral ligament transection (MCLT) and iodoacetate induced models used to study pain40, 45

Rat undergone partial medial meniscectomy are useful in cartilage restoration techniques

Transgenic models

Syrian hamster OA models are naturally occurring, and transgenic models are used to study pathogenesis of OA

Transgenic models

MCLT, osteotomy, patellectomy, sciatic neurectomy

Combination surgery

Immunotoxin, papain, collagenase, copper II bisglycinate, lipopolysaccharide, chondromucoprotein

Quinolone

Transgenic guinea pig models used to study pathogenesis of OA

Guinea pig models used to study age and BMI associated risk factors in OA. Dunkin Hartley guinea pig used in therapeutic and pathogenic studies of knee OA. Guinea pigs induced by medial meniscal tear and spontaneous OA models used to study slow and rapidly progressive OA

ACLT

Useful in pain studies

ACL tear

Section of medial collateral and both cruciate ligaments, resection of medial meniscus

Immobilization

Combination surgery, impact loading, cartilage scarification

ACLT, ACLT and PCL/MCL

Ovariectomy

Articular groove

Transarticular mechanical impact on patellofemoral joint, femoral condyle impact

Intra-articular injection of steroids and cytokines

Allogeneic cartilage particles

Quinolone

Chymopapain, trypsin, IL-1β, chondroitinase, vitamin A, fibronectin fragments

Rabbit models useful in efficacy testing of various compounds such as hyaluronic acid.

Partial meniscectomy models used in testing chondroprotective agents

Abrasion, valgus osteotomy, pelvic osteotomy, cartilage defect

Cranial cruciate ligament transection

Impact loading, cartilage scarification

ACLT

Allogeneic cartilage particles

MMx model useful in toxicology testing and ACLT model used to study slow progression of OA and pathogenesis that mimics naturally occurring disease

Canine models that naturally develop OA have been used in therapeutic intervention preclinical trials

Transarticular impact models were used to identify whether osteoarthritic changes originate from cartilage or subchondral bone changes and to study early changes in OA in articular cartilage due to joint impact trauma

ACLT induced model have been used in identification of OA biomarkers

-Naturally occurring OA

MCLT, ACLT

Articular groove

Unilateral medial MMx, unilateral MCL, meniscal transection, cartilage scarification, unilateral ACLT

Goat models used to study cartilage repair

Naturally occurring OA

Lateral meniscectomy, ACLT, MCLT

Articular groove model

Bilateral and unilateral Mx, unilateral ACLT, medial MMx, unilateral MCLT, unilateral radial meniscal tear, unilateral caudal pole hemi-meniscectomy, unilateral medial meniscectomy

Post carpal fracture, exercise-induced

Trauma to medial femur and tibia

Metacarpophalangeal ligament transection

Osteochondral fragment

Articular groove model

Amphotericin, E. coli lipopolysaccharide, IL-1, carrageenan

Lipopolysaccharide

Amphotericin

Papain

Intra-articular injection of steroids, collagenase, cytokines

Equine models used to study articular cartilage repair, osteochondral defects and naturally occurring bone remodelling

Genetic knockout e.g. COL10A1

Zebrafish model useful in studying gene related pathology of OA

Post-fracture

ACLT, ovariectomy, ACL reconstruction, articular groove model

ACLT and ACL reconstruction

Porcine model used to study repair and regeneration of focal cartilage defects

ACLT

Transgenic models

Ovariectomy in cynomolgus monkeys

Collagenase induced in cynomolgus monkeys

Naturally occurring and transgenic models of non-human primates used to study general features of OA

Smaller animal models of OA such as mice, rats, rabbits and guinea pigs are much easier, quicker, cheaper and more readily available than larger animal models such as horses, pigs and dogs2, 40. Smaller animals can be handled and housed with greater ease than larger models, but due to their smaller size, tissue samples extracted are much smaller and therefore tend to differ to a greater extent in their anatomical and histological structure when compared to humans. Larger animal models therefore provide many advantages over the use of smaller animal models in terms of their greater anatomical similarity to the human model. A dog's articular cartilage for example is half the thickness of a humans, whereas that of a mouse is a minimum of 70 times thinner2, 18. Additionally, a wider range of tests can be performed on larger animals, such as repeated synovial fluid collection and imaging. They also have a longer life span allowing for slower disease progression and time to end stage OA, as seen in humans. Whilst slow progressive models most accurately reflect human OA, they are however more expensive and time consuming to conduct. There are also greater ethical considerations around the use of larger animal models such as non-human primates and canines. Based on this, some animal models are therefore better suited to OA research than others, such as the canine, caprine, bovine and porcine models. Some of the advantages and disadvantages of various animal species used in OA research are summarised in Table III.

Table III

A summary of the advantages and disadvantages of different animal models used in OA research

Animal Model	Advantages	Disadvantages
Mouse	- Mice have a short life span (generally one or 2 years) and so develop OA fairly rapidly, making mice an easy model to study the whole disease process18, 43 - Small animal size means the whole joint can be histologically sectioned81 - Mice are easily managed, with low maintenance cost, demonstrate rapid disease onset and their complete genome is available for study40 - Genetically modified mouse models are easy to produce and are useful to investigate the genetic factors involved in OA pathogenesis, specifically genes involved in cartilage degeneration, bone remodelling and inflammation18, 45 - Mouse models can be used in toxicology testing and to establish the molecular basis of OA1, 45	- Huge variation in results observed between different strains of mice18 - Disease severity varies with age, with older mice more representative of human disease82 - Difficult to ascertain skeletal maturity as growth plates often do not close completely82 - Mice are anatomically and histologically different to humans, for example, mice have a thicker layer of calcified cartilage, do not have three distinct chondrocyte layers and have a cartilage seventy times thinner than humans45 - Macroscopic lesions and degrees of damage are difficult to identify due to the small anatomical size of mice81 - The progression and process of disease is faster in mice than in humans (weeks rather than decades)36 - The small size of mice makes surgically inducing OA more challenging40 - Postoperative management of mice is difficult in surgically induced models40, 45
Rat	- Rat cartilage is thicker than that of mice, so it is possible to induce partial and full-thickness cartilage defects1, 45 - Rats rarely experience post-operative infection so are useful animal models to surgically induce OA1 - Rats are easily managed and require low maintenance costs40, 45 - It is easier to perform surgery in rats than in mice due to their larger size40 - The full rat genome is available for study40 - MMT, MCL transection and iodoacetate models useful to study pain40, 45 - Rat models useful in toxicology testing and studying cartilage restoration techniques1, 45	- Naturally occurring OA is uncommon in rats, variation in results is often observed between different strains of rat and disease severity varies with age, with older rats tending to present with more severe OA18 - It is difficult to ascertain the skeletal maturity of rats83 - Rats have greater volumes of highly vascularised adipose tissue and muscle in the medial knee region - Post-operative rats immediately resume load-bearing which accelerates joint degeneration1 - Genetically engineered rat models are not available and postoperative management of rats is challenging45
Guinea Pig	- The guinea pig model has similar OA histopathology to disease in humans84 - Guinea pigs are large enough that tissue samples can be easily collected for tests and the whole joint can be histologically sectioned49 - Guinea pigs are easy to manage40 - Naturally occurring guinea pig models are available and the disease pathogenesis is predictable and similar to that seen in humans1, 45 -Hartley guinea pigs can be used to study risk factors for OA such as BMI and age - Complete guinea pig genome available85	- The weight of each guinea pig and whether they are housed alone or in pairs influences the severity of their OA41, 49 - Unlike in humans, guinea pigs resume load bearing post-operatively which accelerates joint degeneration1 - The time to guinea pig skeletal maturity is fast45
Cat	- Cats are larger in size allowing for tissue and fluid collection40 - The full cat genome is available40	- Cats are difficult and costly to manage and there are ethical issues surrounding emotional attachment40 - Cats display genetic variability between individuals40
Rabbit	- Naturally occurring OA is very common in rabbits52 - Rabbit model useful in studying the efficacy of compounds45 - Complete rabbit genome available86	- Rabbits have a very different gait compared to humans and only rabbits over the age of eight or 9 months can be used to guarantee skeletal maturity52 - The cartilage of rabbits is ten times thinner compared to humans, with a higher chondrocyte density and cartilage zonal layers that varies highly within the same joint87, 88 - The rabbit meniscus is more cellular, has less vascular penetration and can heal faster than the human menisci89 - Rabbit cartilage can spontaneously heal and regenerate and there is no complete rabbit genome available for study40 - OA progression varies with the age of the rabbit after surgical OA induction, with faster progression seen in older rabbits41
Canine	- Canines have similar anatomy and disease progression to humans18, 90 - Canines display a widespread clinical incidence of OA18, 83 - Canines are easy to manage and train postoperatively40, 45 - Surgical lesions develop slowly in canines, similar to the human model1 - Canines have similar gastrointestinal physiology to humans45 - The canine model is widely used so comparison across different studies can be made, the larger size of canines allows for tissue and fluid collection and the full canine genome is available40 - Naturally occurring OA models are available for intervention preclinical trials2, 41	- Canines have different joint biomechanics and gait compared to humans, their skeletal maturity is not reached until 9 to 18 months of age and their cartilage is half the thickness of human cartilage64 - There are ethical issues surrounding emotional attachment of dogs and management is costly40, 45 - Canines display genetic variability between individuals40
Caprine	- Anatomically the caprine stifle joint is very similar to the human knee64 - The caprine stifle joint is closest in size to the human knee joint, the larger size of the animal allows for tissue and fluid collection and goat cartilage thickness is close to that of humans40 - Goats are cheap and easy to use in studies compared to most large animal models and they can be used to study cartilage repair45 - Complete goat genome available91	- Caprine cartilage thickness varies between individuals, the skeletal maturity of a goat is not reached until at least 2 years of age and cartilage healing capacity varies with a goat's age, with better capacity in younger animals87, 92 - Cartilage repair outcomes differ in the short and long term and so follow up is required to assess progress83 - Naturally occurring OA in goats is uncommon40, 45
Ovine	- Sheep are cheap and easy to use in studies compared to most large animal models. - The advantages of the sheep model are similar to the caprine model of OA	- The disadvantages of the sheep model are very similar to the caprine model of OA
Equine	- The large size of horses allows for easy tissue and fluid collection and a full genome is available40 - Anatomically and histologically the equine stifle joint is similar to the human knee, the articular cartilage is very similar in thickness and the cellular structure, biochemical makeup and properties of the cartilage are most comparable to humans2, 92, 93 - There are a wide range of imaging and clinical tests that can be performed on horses, including rehabilitation techniques94 - Naturally occurring OA models are available45	- Horses are difficult and expensive to house and manage due to their large size40
Zebrafish	- Zebrafish model is useful to study gene related pathology of OA and zebrafish genome available71	- Zebrafish do not have synovial joints71
Porcine	- The porcine stifle joint is anatomically similar to the human knee joint and pigs have similar immune systems and gastrointestinal physiology to humans41 - Pigs are most similar to humans in terms of their anatomy, neurobiology, cardiovasculature, gastrointestinal tract and genome73 - Genetically modified models are available and pigs are a useful model to study repair and regeneration of focal cartilage defects73 - Pigs have similar joint size, weight-bearing and cartilage thickness to humans73	- The porcine meniscus is wider, and the cruciate ligaments are longer than in humans64, 95 - Pig skeletal maturity is reached between 10 and 24 months of age41
Bovine	- Bovine cartilage thickness, cellularity and zonal cartilage layers of patella is similar to in human femoral condyles75, 88 - Bovine meniscus is biomechanically similar to the human meniscus95 - Complete bovine genome available96	- Bovine lateral tibial plateau cartilage is thinner, more cellular and varies in zonal cartilage thickness compared to the human75, 88
Non-human primates	- Non-human primates have similar anatomy, genetics, biology, behaviour and physiology to humans2 - The pathology of OA and the relationship between age and disease severity is very similar to in humans18, 76, 77 - The larger size of non-human primates allows for tissue and fluid collection and the full primate genome is available for some species40	- Non-human primates are expensive and ethically difficult to keep, for example chimpanzees display depression and post-traumatic stress disorder on a similar scale to that of humans97 - Non-human primates have a long-life span and so a long disease pathogenesis time scale which is both time consuming and costly2 - There are difficulties in obtaining adequate subject numbers for studies1 - Housing and management of non-human primates is challenging40

Small animal size means the whole joint can be histologically sectioned

Mice are easily managed, with low maintenance cost, demonstrate rapid disease onset and their complete genome is available for study

Huge variation in results observed between different strains of mice

Disease severity varies with age, with older mice more representative of human disease

Difficult to ascertain skeletal maturity as growth plates often do not close completely

Mice are anatomically and histologically different to humans, for example, mice have a thicker layer of calcified cartilage, do not have three distinct chondrocyte layers and have a cartilage seventy times thinner than humans

Macroscopic lesions and degrees of damage are difficult to identify due to the small anatomical size of mice

The progression and process of disease is faster in mice than in humans (weeks rather than decades)

The small size of mice makes surgically inducing OA more challenging

Rats rarely experience post-operative infection so are useful animal models to surgically induce OA

It is easier to perform surgery in rats than in mice due to their larger size

The full rat genome is available for study

Naturally occurring OA is uncommon in rats, variation in results is often observed between different strains of rat and disease severity varies with age, with older rats tending to present with more severe OA

It is difficult to ascertain the skeletal maturity of rats

Post-operative rats immediately resume load-bearing which accelerates joint degeneration

Genetically engineered rat models are not available and postoperative management of rats is challenging

The guinea pig model has similar OA histopathology to disease in humans

Guinea pigs are large enough that tissue samples can be easily collected for tests and the whole joint can be histologically sectioned

Guinea pigs are easy to manage

Complete guinea pig genome available

Unlike in humans, guinea pigs resume load bearing post-operatively which accelerates joint degeneration

The time to guinea pig skeletal maturity is fast

Cats are larger in size allowing for tissue and fluid collection

The full cat genome is available

Cats are difficult and costly to manage and there are ethical issues surrounding emotional attachment

Cats display genetic variability between individuals

Naturally occurring OA is very common in rabbits

Rabbit model useful in studying the efficacy of compounds

Complete rabbit genome available

Rabbits have a very different gait compared to humans and only rabbits over the age of eight or 9 months can be used to guarantee skeletal maturity

The rabbit meniscus is more cellular, has less vascular penetration and can heal faster than the human menisci

Rabbit cartilage can spontaneously heal and regenerate and there is no complete rabbit genome available for study

OA progression varies with the age of the rabbit after surgical OA induction, with faster progression seen in older rabbits

Surgical lesions develop slowly in canines, similar to the human model

Canines have similar gastrointestinal physiology to humans

The canine model is widely used so comparison across different studies can be made, the larger size of canines allows for tissue and fluid collection and the full canine genome is available

Canines have different joint biomechanics and gait compared to humans, their skeletal maturity is not reached until 9 to 18 months of age and their cartilage is half the thickness of human cartilage

Canines display genetic variability between individuals

Anatomically the caprine stifle joint is very similar to the human knee

The caprine stifle joint is closest in size to the human knee joint, the larger size of the animal allows for tissue and fluid collection and goat cartilage thickness is close to that of humans

Goats are cheap and easy to use in studies compared to most large animal models and they can be used to study cartilage repair

Complete goat genome available

Cartilage repair outcomes differ in the short and long term and so follow up is required to assess progress

The large size of horses allows for easy tissue and fluid collection and a full genome is available

There are a wide range of imaging and clinical tests that can be performed on horses, including rehabilitation techniques

Naturally occurring OA models are available

Horses are difficult and expensive to house and manage due to their large size

Zebrafish model is useful to study gene related pathology of OA and zebrafish genome available

Zebrafish do not have synovial joints

The porcine stifle joint is anatomically similar to the human knee joint and pigs have similar immune systems and gastrointestinal physiology to humans

Pigs are most similar to humans in terms of their anatomy, neurobiology, cardiovasculature, gastrointestinal tract and genome

Genetically modified models are available and pigs are a useful model to study repair and regeneration of focal cartilage defects

Pigs have similar joint size, weight-bearing and cartilage thickness to humans

Pig skeletal maturity is reached between 10 and 24 months of age

Bovine meniscus is biomechanically similar to the human meniscus

Complete bovine genome available

Non-human primates have similar anatomy, genetics, biology, behaviour and physiology to humans

The larger size of non-human primates allows for tissue and fluid collection and the full primate genome is available for some species

Non-human primates are expensive and ethically difficult to keep, for example chimpanzees display depression and post-traumatic stress disorder on a similar scale to that of humans

Non-human primates have a long-life span and so a long disease pathogenesis time scale which is both time consuming and costly

There are difficulties in obtaining adequate subject numbers for studies

Housing and management of non-human primates is challenging

Challenges presented by current OA models

At present, there is no gold standard animal model used in OA research. Differences in size, anatomy, histology (specifically cartilage thickness) biomechanics and physiology makes translatability between animal models and human disease very difficult52, 83. Challenges are posed by species-specific differences in disease pathology and progression, as well as normal joint homeostasis, specifically the repair processes that occur within the joint. Different OA induction methods used in certain animal species also sometimes results in differences in OA presentation. There is therefore a need to reduce experimental variability and increase the reliability of data interpretation. Furthermore, different animal models have been shown to represent different stages of the disease more effectively, making selection of an animal model that completely reflects natural human disease challenging. To add to this, it has become clear that there is much dispute as to what defines OA and which molecules are associated with the disease. This is in part hindered by the fact that current ex-vivo models do not allow for the important inter-tissue communication between different joint components required for natural OA disease processes to be studied. To gain a better understanding of the mechanisms of joint damage in OA, specifically the exact sequence of events and interactions between different joint components, it has been suggested that more focus should be placed on developing 3D cell cultures and explant-based models that allow for these important interactions, providing interesting opportunities for researchers to develop their understanding of OA.

In the ideal animal model, the disease must be induced reliably, with 100% penetrance, and within a suitable time frame, and yet still present with disease characteristics that are comparable to the human condition. Disease progression in the animal model should also allow for the examination of all stages of disease to ensure full detection of any therapeutic effects. The animal must be inexpensive, easy to house and manage but also be of large enough size to allow for a full range of analysis techniques to be performed. The animal must also be anatomically, biomechanically and histologically similar to humans. Some animal species, such as the canine, caprine, bovine and porcine models, are therefore better suited to OA research than others.

A promising model for OA

OA is a disease of the whole joint and therefore the gold standard model for human OA must allow for key communication between different tissues of the joint. An osteochondral (cartilage on bone) model may overcome many of the current challenges and limitations of the various models discussed above. The use of osteochondral plugs provides a model incorporating the key joint tissues affected in OA, maintaining the important interactions between these tissues as seen in human disease. There are a few studies that have used osteochondral plugs from bovine, equine and human samples as ex-vivo models of OA98, 99, 100. These osteochondral plug-based models appear very promising and so further studies should be encouraged as the basis for developing a gold standard model for OA. In the model design, cytokines such as interleukin-1 beta (IL-1β) and tumour necrosis factor alpha (TNF-α) may be used as the best method for inducing OA (cartilage damage) and tissue responses that closely replicate the natural disease, as these cytokines are known to contribute to the inflammatory effect of the synovium in the model. However, OA is a complex disease and many cytokines and chemokines have been shown to be expressed in OA synovium and detected in synovial fluid. Therefore, in the model design, investigators should consider using synovial tissue and/or synovial fluid from patients with active disease to induce OA in the osteochondral plugs. Osteochondral plugs can be harvested from joint surfaces, such as the femoral condyle, tibial plateau and patella. Methods of plug extraction available for use include different sizes of graft harvester, biopsy punch, mosaicplasty osteotome, diamond tipped cylindrical cutter or surgical trephine burr. Plugs can be cultured in serum-free culture medium such as Dulbecco's Modified Eagle Medium F-12 (DMEMF-12, Invitrogen, USA) or α-Minimum Essential Medium (α-MEM, 22,561 Gibco, The Netherlands) for up to 57 days with significant cell viability. Indeed, an early study reported that >99% chondrocyte viability can be maintained in the untraumatized areas at the centre of the osteochondral plugs.

The availability of an osteochondral plug-based model, particularly a human tissue-based one, would be invaluable in screening of new disease-modifying osteoarthritis drugs (DMOADs). Currently, there are many DMOADs under different stages of development. Early studies of these drugs were in animal models, but the availability of this new model will provide an opportunity to directly test these drugs on human tissues. An osteochondral plug system like this may also be used for discovery of novel markers for OA.

Conclusion

It is clear that we are limited in our understanding of OA because we do not have a suitable model that accurately reflects natural human OA. Whilst animal models provide crucial information about disease mechanisms, none of the current models used in OA research recreate the natural in-vivo environment and allow the whole disease process to be studied. Differences in anatomy and biomechanics also makes translatability to the human model a distinct challenge. It is also important to consider the cost and ease of management of using animal models in research. Whilst smaller animal models provide many benefits in terms of availability, handling and management, larger animal models such as canines and pigs are not only more comparable to humans physiologically but also in their progression to disease. Their larger size also allows for performance of a broader range of analysis techniques and investigations.

The models currently used in OA research each have their advantages and disadvantages; however, it has become clear that there are consistent problems with all of these models that hinders our ability to understand the pathogenesis of OA. The only way to achieve greater understanding of the pathological processes that underpin OA is to produce a ‘gold standard’ model for OA. Development of a consensus model will provide greater understanding of the specific stages and interactions involved in OA pathogenesis, as well as a model that can be used to compare data findings between different research groups, test pre-clinical drugs and identify and test possible biomarker targets directly on OA joint tissue. An osteochondral plug-based model could be a “promising” new model for OA, able to provide a reliable throughput model for proof of concept and mechanistic studies, aiding the discovery of targeted OA therapy. The model would also provide an opportunity to reduce the financial, ethical and time restraints associated with using animals in research, shifting OA research to embody the principle of the 3Rs; replacement, reduction and refinement of animal use in research.

Author's contributions

Conception and design: MS, KO, YL.

Drafting of the article: PC, MS, KO.

Provision of study materials: PC, YL.

Final approval of the article: MS.

Competing interest statement

The authors declare that they have no competing interests.

Funding

Not applicable.