Articular cartilage is the hyaline connective tissue that covers the ends of bones in the diarthrodial joint, and it serves to absorb shock from joint movement. Articular cartilage has an aneural, avascular, and alymphatic structure composed of 65–80% water, 20–40% extracellular matrix (ECM), and 1–5% chondrocytes. The major components of ECM are collagens (mainly collagen type II) and proteoglycans (Carballo et al., 2017; Chen et al., 2017). Articular cartilage presents zonal structures from the articular surface (superficial, middle, and deep zones) to the conjunction area of subchondral bone (calcified zone), and the cell density and morphology differ in each zone (Jiang et al., 2017).

Subchondral bone refers to the bone tissue underlying the calcified cartilage and tidemark (Figure 1), including both subchondral cortical plate and subchondral trabecular bone. Subchondral bone plate is a thin layer of cortical bone lying immediately beneath the calcified cartilage. The physiological properties of the cortical plate are similar to the cortical bone located at other skeletal sites, providing mechanical strength to support the overlying articular cartilage, although the cortical plate is less stiff in comparison to diaphyseal cortical bone (Elastic modulus: 1.372 vs. 14 GN/m²) (Brown and Vrahas, 1984; Burr and Gallant, 2012). Underneath the cortical bone plate lies the subchondral trabecular bone, which is more porous and metabolically active compared to the cortical plate. The trabeculae in the cancellous bone have a unique structural network that adjusts to local mechanical influences (Figure 1A).

The osteochondral unit in the joint, which transfers load during joint movement, consists of articular cartilage, calcified cartilage and subchondral cortical and trabecular bone (Lories and Luyten, 2011; Goldring and Goldring, 2016). The normal subchondral cortical plate has a dense structure to support articular cartilage, and there are canals (10–160 mm in diameter) in the cortical plate, allowing the exchange of nutrients and molecules between cartilage and bone (Duncan et al., 1987; Clark, 1990; Bian et al., 2016).

Articular cartilage stress distribution changes with subchondral bone expansion, and even a 1–2% incremental increase in the size of the subchondral bone could significantly increase stress to articular cartilage, as estimated by a recent human tibia data based computer-simulated model (Zhen et al., 2013). Upon injury or degeneration, the integrity of the osteochondral unit is breached, and the crosstalk between cartilage and subchondral bone is increased; the osteochondral unit is consequently altered and reshaped by the dynamic mechanical environment and biological microenvironment. Recently, Chen et al. (2020) reported a horizontal fissure at the osteochondral interface in obese OA patients, characterized by irregular cartilage erosion, fibro-granulation tissue infiltration, presence of free cartilage/bone debris and rupture of microcapillaries at the interface within the osteochondral unit. This represents a new type of OA pathological features, where neurovascular invasions were also identified in the degenerative osteochondral tissues (Li et al., 2013) (Figures 1B, 2A).

In OA joints, subchondral bone undergoes remarkable changes in both composition and structural organization, and has adverse effects on the overlying articular cartilage (Figure 1). With a bone scan (^99mTc pyrophosphate) and tetracycline labeling, Radin et al. (1984) provided early direct evidence that subchondral bone remodeling was significantly enhanced in response to mechanical loading in a rabbit model of hindlimb loading, and it preceded the fibrillation of the overlying articular cartilage (Radin et al., 1984). More histopathological features in osteoarthritic subchondral bone were then identified and discussed, including microdamage, bone marrow edema/lesions and bone cysts (Goldring and Goldring, 2016).

“Bone marrow edema” was termed and used about three decades ago, when Wilson et al. (1988) detected and localized regions of increased signal intensity in the subchondral bone of OA patients, with MRI based fluid-sensitive magnetic resonance sequences. However, histological examination of the subchondral bone in the anatomical edema sites revealed that local fat necrosis, marrow fibrosis and vascular changes were associated with microfractures of the trabecular bone, but not associated with edema (Leydet-Quilici et al., 2010). In current practice, people termed this pathological feature as bone marrow lesions (BMLs) in subchondral bone. Today, the different types and stages of BMLs can be identified by the radiological response to two MRI sequences, the fast spin-echo proton density (PDFS)-weighted and T1-weighted spin echo. Two major types of BMLs were designated based on the MRI readouts: BML1 represents an early-stage of the structural alteration in subchondral bone, which is detectable only by PDFS sequence, and BML2 is considered as a relatively more degenerative stage in the lesion sites, which is detectable by both PDFS and T1 sequence (Muratovic et al., 2016). Therefore, it appears that BMLs are correlated to the severity of OA progression (Taljanovic et al., 2008; Tanamas et al., 2010).

Subchondral bone cyst, also known as intra-osseous lesion, is another pathological feature often found in the load-bearing regions of the arthritic joints, but the exact pathogenesis of cyst formation is still unclear. Mechanical instability and overload may be the major cause. McErlain et al. (2012) created mechanically unstable OA models in rat knees by anterior cruciate ligament transection (ACLT) and partial medial meniscectomy (MMX); 75% of the knees developed subchondral bone cysts in 4 weeks, and all knees had subchondral bone cysts in 12 weeks. Moreover, enhanced bone turnover was also identified in human OA subchondral bone. Chen et al. (2015) examined cellular composition of human OA subchondral bone cysts and showed that the number of osteoclasts, osteoprogenitors (osterix⁺), osteoblasts (osteocalcin⁺, OCN⁺), and SOX9⁺ cells were higher in bone cyst sites than those in peri-cyst sites, suggesting enhanced bone turnover and abnormal bone remodeling.

Recently, the fine microarchitecture alterations of OA subchondral trabecular bone have been imaged and analyzed by advanced high-resolution micro-CT and MRI imaging systems. Liu et al. (2008) examined 71 human trabecular bone samples, and found that healthy trabecular bone consists of a network of rod-like and plate-like trabeculae microstructure (Figure 1A, right panel). Healthy subchondral trabecular bone displayed an even distribution of rod-like and plate-like trabeculae, and in human OA tibial plateau, a dramatic loss of rod-like trabeculae and thickening of plate-like trabeculae were identified using a novel microstructural analysis method, based on individual trabecula segmentation (Figures 1B,C, right panels) (Chen et al., 2018). Interestingly, in the same study, Chen et al. (2018) discovered that subchondral trabecular bone underneath the intact cartilage shared a similar phenotype with that underneath the severely damaged cartilage, which implied that the alterations of subchondral bone in early OA could precede cartilage degeneration.

Bone ECM is produced by osteoblasts, and compositionally consists of mineral, water, collagen and noncollagenous proteins and lipids; the proportion of each component varies depending on species, age and site (Boskey and Robey, 2013). The osteoid matrix located at the endosteum of the bone is the newly secreted unmineralized matrix which then undergoes mineralization during osteogenesis, and collagen type I serves as the scaffold for the deposition of hydroxyapatite crystals in osteoid mineralization. The mechanical properties of bone are a function of the hardness of the mineralized ECM and the flexibility of organic components of the ECM. Bone ECM is the body's reservoir of minerals, macromolecules, and bioactive factors. It is the primary source of calcium, phosphate, and magnesium ions for maintenance of general physiological functions, including the maintenance of bone homeostasis. Bioactive factors are released into the marrow microenvironment upon bone resorption, and subsequently contribute to the recruitment and differentiation of stem/progenitor cells, such as mesenchymal stem cells (MSCs), during bone formation (Crane and Cao, 2014). For instance, bone matrix-derived IGF1 could stimulate osteoblastic differentiation of MSCs by activating mammalian target of rapamycin (mTOR), thus maintaining proper bone microarchitecture and mass (Xian et al., 2012). TGFβs are also abundant in bone ECM (TGFβ1, 188 ng/g; TGFβ2, 14 ng/g) (Hering et al., 2001), and activation of the TGFβ signaling pathway has been confirmed in OA subchondral bone, resulting in recruitment of MSCs and enhancement of bone formation (Zhen et al., 2013). In this manner, the subchondral bone ECM also participates in the regulation of bone homeostasis during OA bone remodeling.

No current therapy is available to specifically target osteocytes.

No current therapy is available.

Apart from osteoblasts and osteoclasts, emerging evidence suggests that the matrix-embedded osteocytes also play essential roles in bone remodeling, particularly in arthritic subchondral bone. Tu et al. (2019) reported that osteocytes in subchondral bone expressed high level of cyclooxygenase-2 (COX-2) during the development of OA in a spontaneous OA model (STR/Ort mice), where there was no significant changes in the number of tartrate-resistant acid phosphatase (TRAP)-positive osteoclasts, supporting the importance of osteocytes in early OA subchondral bone remodeling. Specifically, in this STR/Ort mice OA model, the high level of COX-2 expression in osteocytes gave rise to abnormal subchondral bone formation, and accelerated cartilage degeneration at the onset of OA. Application of COX-2 inhibitors to animals at 12.5–25% of the clinical application dosage for 4 weeks resulted in normalization of the microstructure of subchondral bone and relief of cartilage degeneration in mice; a similar effect was also found in a rheumatoid arthritis mice model (TNF-α transgenic RA mice). This important piece of pre-clinical evidence suggested a treating window for bone remodeling during the development of arthritis, and the current clinical available COX-2 inhibitors could also manage the subchondral bone remodeling via regulating the osteocytes (Tu et al., 2019).

In addition, subchondral bone osteocyte-derived matrix metallopeptidase 13 (MMP13) was also identified as a regulator of articular cartilage health, by affecting PLR of osteocytes. Mazur et al. (2019) established a MMP13 knockout mouse model with a Cre recombinase driven dentin matrix acidic phosphoprotein 1 (DMP1) promoter to specifically ablate osteocyte MMP13, without affecting chondrocyte derived MMP13. The depletion of osteocyte MMP13 increased both cortical and subchondral bone mass, and suppressed PLR during OA development in animals. Results from this study provided insights on the important contribution of osteocyte derived MMP13 to PLR during OA development, and on the dynamic crosstalk between subchondral bone osteocytes and cartilage (Mazur et al., 2019).

The vasculature in bone is indispensable for skeletal development and homeostasis. Osteogenesis is coupled with angiogenesis during bone modeling and remodeling (Brandi and Collin-Osdoby, 2006; Kusumbe et al., 2014; Xie et al., 2014). The role of angiogenesis in subchondral bone has been studied recently with genetic mouse models. Su et al. (2020) used transgenic mice to prove that overexpression of platelet-derived growth factor-BB (PDGF-BB) in preosteoclast spontaneously recapitulated the OA-like phenotype in subchondral bone by enhancing both angiogenesis and subchondral bone remodeling; meanwhile, conditional PDGF-BB knockout could effectively prevent surgically-induced OA progression (Su et al., 2020). These results suggested that subchondral bone angiogenesis is one of the key processes in subchondral bone remodeling and may be considered as a potential target. Besides intrinsically inhibiting subchondral bone vascularization, exogenous drugs targeting subchondral bone angiogenesis have also been studied. For example, bevacizumab (a VEGF antibody), halofuginone (inhibiting phosphorylation of Smad2/3 and TGFβ) and defactinib (inhibitor of focal adhesion kinase, FAK) all showed promising beneficial effects against OA in animal models via inhibition of subchondral vascularization (Cui et al., 2016; Lu et al., 2018; Wu et al., 2020).

Nerves often grow along new blood vessels into subchondral bone structures that are normally not innervated, which is also a feature of OA pathogenesis (Figure 2A), and may be related to pain (Mapp and Walsh, 2012). Recently, in a study of the role of osteoclast-initiated subchondral bone remodeling in sensory innervation and OA pain, Zhu et al. (2019) showed that osteoclast-initiated aberrant subchondral bone remodeling induced sensory innervation via subchondral osteoclast-derived Netrin-1, resulting in enhanced subchondral bone sensory nerve growth. This sensory innervation and OA pain could be attenuated by either inhibiting subchondral osteoclasts (by means of gene knockout in DMP1-Rankl Flox mice or treatment with alendronate), or deleting Netrin1 in osteoclast-lineage cells (by means of knockout in Trap-Netrin1 flox mice) (Zhu et al., 2019). Our group has recently reported the role of NGF/TrkA signaling pathway in calcification of human healthy articular chondrocyte, representing another mechanism operating on the other side of the cartilage tidemark that regulates OA subchondral remodeling (Jiang and Tuan, 2019). Another osteoblast-secreted inflammatory mediator, prostaglandin E2 (PGE2), was also found to be increased in OA subchondral bone and sensory innervation. PGE2 activates an ion channel protein, Nav1.8, and inhibition of PEG2 production or Nav1.8 function with a selective Nav1.8 sodium channel blocker A-803467 could attenuate OA (Zhu et al., 2020). Taken together, these observations provide insights on the linkage between sensory innervation and OA subchondral bone remodeling (Zhu et al., 2020).

That angiogenesis/vascularization and nerve innervation appear to play functional roles in subchondral bone remodeling during OA progression suggests that they are potential therapeutic targets for the development of OA drugs (Figure 2F).

Regenerative therapies, including stem cell-based therapies are actively investigated for OA treatment. There are Phase I/II clinical trials of using bone marrow derived MSCs/BMSCs (Gupta et al., 2016; Shapiro et al., 2017), adipose tissue derived MSCs/ADSCs (Song et al., 2018; Lu et al., 2019), and other allogeneically derived MSCs (Matas et al., 2019) to treat OA knees (Table 2). For instance, two Phase II clinical trials showed that intra-articular injections of human ADSCs (5 × 10⁷ cells) exhibited the improvement in both pain relief and cartilage volume (Song et al., 2018; Lu et al., 2019). However, the favorable results reported are primarily focused on pain relief and functional improvement but few tissue regeneration and subchondral bone reestablishment, and the application is with the strict condition of administrative DOSES (Donor, Origin tissue, Separation method, Exhibited cell characteristics associated with behavior, and Site of delivery), i.e., cell derivation, cell number, harvesting method, sites of delivery (Murray et al., 2019) and frequency (Kim et al., 2019).

Regenerative therapy also represents an emerging strategy to enhance the subchondral bone health, particularly at the early stage of disease that could prevent OA progression. Subchondroplasty is a recently developed, clinically-used surgical treatment using bioactive materials, e.g., calcium phosphate, as bone substitute materials (BSMs), to fill the subchondral bone marrow lesions. The BSMs used for subchondroplasty are injectable, and are engineered products that have similar mechanical properties as the surrounding trabecular bone, thereby providing a bioactive scaffold for local subchondral bone remodeling, and preventing further damage to the osteochondral tissue (Farr and Cohen, 2013; Colon et al., 2015). Another approach is to use cell-based therapy to regenerate the osteochondral defects before it progresses to OA. Autologous chondrocytes (Brittberg et al., 1994; Peterson et al., 2010) and adult stem cells such as cartilage derived-progenitors (Jiang et al., 2016), bone marrow derived-MSCs (Wakitani et al., 2002, 2007; Kuroda et al., 2007) are the most frequently used cell types with satisfactory tissue repair results in clinical studies and applications (Jiang et al., 2017) (Table 2).

Tissue engineering constitutes a major strategy in regenerative therapy, i.e., to repair the tissue with exogenous cells, biomaterials, and/or bioactive factors. Another strategy is to enhance and push the limit of a tissue's self-healing capacity. Subchondral bone has a certain degree of self-repair ability in healthy joints. This healing capacity has been reported in different animal models, dependent on age and species (Pfeifer et al., 2017), size and site of the defect (Fisher et al., 2015; Orth et al., 2016), and is the foundation of the well-established surgical technique of microfracture. Microfracture has been used to treat small to medium sized cartilage/osteochondral defects by creating perpendicular channels into the subchondral bone plate and allowing the bone marrow to seep into the defect area (Pan et al., 2009). The cartilage repair/resurfacing resulting from this type of marrow stimulation techniques is thought to be partially due to the recruitment and local differentiation of bone marrow-derived MSCs, which act as a key player in subchondral bone remodeling and osteochondral homeostasis. We have previously established the application of a bioactive scaffold that achieved fast healing of subchondral bone in a rabbit osteochondral defect model. Interestingly, articular cartilage resurfacing was observed in the fast healing subchondral bone groups without any other additional treatments (Jiang et al., 2013). A subchondral remodeling step thus appears to take place in advance of cartilage repair in osteochondral healing, and the quality of cartilage generated from the marrow stimulation techniques is dependent on the healing speed and quality of subchondral bone remodeling (Chen et al., 2011; Jiang et al., 2013).

In this manner, a practical regenerative approach for OA is to enhance the performance of the subchondral bone by introducing additional bone marrow derived MSCs, to maintain osteochondral tissue homeostasis and prevent further cartilage loss. Some exciting results from a series of clinical trials have just been released by Hernigou et al. (2018, 2020a,b) (Table 3), with clinical data of up to 15 years of follow-ups. Patients with different OA conditions (moderate to advanced OA, 60 cases; advanced OA, 140 cases; OA secondary to knee osteonecrosis related to corticosteroids, 30 cases) were recruited. These patients have comparable clinical symptom in bilateral knees, and were randomized to accept subchondral bone MSC therapy or other surgical treatments. The subchondral bone MSC therapy was delivered to one knee, while the other knee was treated with either intraarticular MSC injection (n = 60) or total knee arthroplasty (TKA, n = 140) (Table 3, Study 1 and Study 2). These results showed that: (i) precise delivery of MSCs to subchondral bone could postpone TKA in OA patients for more than 10 years; (ii) subchondral bone MSC therapy achieved similar clinical scores as TKA knees in short and long-term follow ups; and (iii) subchondral bone MSC therapy could effectively reduce the size of BMLs in OA patients, and postpone or prevent the first TKA surgery in young patients with secondary osteonecrosis related to corticosteroids (Table 3, Study 3).

The exact injection site and optimal delivery technique for bone marrow concentrate/MSC are of high significance for these studies. Surgeons used fluoroscopy to direct the trocar to subchondral bone marrow lesions, reaching a distance of 5 mm distal from the tidemark to avoid perforating the cartilage and to keep the calcified cartilage intact; such an approach ensures precise delivery of MSCs to the injury sites of subchondral bone in clinical practice. Delivery sites (subchondral bone vs. intraarticular) of MSCs are also of importance: 20% of the knees (12 in 60) in the subchondral implantation group underwent TKA at a mean of 14 years (range, 9–16 years), while 70% of the knees (42 in 60) in the intra-articular implantation group underwent TKA at a mean of 7 years (range, 3–14 years) (Hernigou et al., 2020a).

Four potential OA subtype are proposed: (1) responders to bone-acting agents; (2) responders to NSAIDs (Bingham et al., 2006); (3) responders to MSC-subchondral bone/regenerative therapy; and (4) responders to mechanical stimuli.

Several new cellular and molecular targets of OA subchondral bone remodeling have been recently reported in pre-clinical animal studies, including TGFβ, PGE2/Cox2, vascularization, and mechanical sensor proteins (Figure 2). These mechanistic studies are investigated with small animals, and the targets and their possible pharmacological manipulation are summarized in Table 1. Among these, the known OA subchondral bone regulatory signaling pathways of estrogen, TGFβ, and COX2/PEG2 have been discussed in previous sections. Additional details on the regulators of the vascular and mechanical responses in OA subchondral bone are provided below.

Increased vascularization of subchondral bone and the invasion into osteochondral junction are important pathological characteristics of OA (Suri et al., 2007). The vascularization in subchondral bone is one of the potential targets. In the arthritic joint, VEGF signaling contributes to bone resorption and destruction via recruitment and activation osteoclasts (Matsumoto et al., 2002). Administration of bevacizumab, a neutralizing VEGF antibody, either intra-articularly (1 mL; 25 mg/mL; 3 weeks) or intravenously (5 mg/kg; 2 weeks), has been confirmed to attenuate OA in a rabbit model (Nagai et al., 2014). Su et al. (2020) discovered that preosteoclast-derived PDGF-BB enhanced the H type vessel numbers in OA subchondral bone (Kusumbe et al., 2014; Su et al., 2020). In addition, Hu et al. (2020) reported that H-type vessels and MSCs were coordinated in a positive feedback loop via focal adhesion kinase (FAK) signaling in MSCs, subsequently causing aberrant bone formation in ACLT-induced murine OA subchondral bone (Hu et al., 2020); furthermore, the FAK inhibitor defactinib effectively suppressed the positive loop between H type vessels and MSCs, and alleviated OA in mice (Hu et al., 2020) (Table 1).

As both cartilage and bone are mechanical sensitive joint tissues, the mechanosensitive pathway in subchondral bone represents another potential target. The expression of the mechanosensing molecule, Yes-associated protein (YAP), in human knee joint articular cartilage was positively correlated with cartilage ECM stiffness. In OA mouse cartilage, YAP expression was aberrantly increased, and conditional knockout of YAP protected against cartilage degeneration. Intra-articular administration of Verteporfin, a YAP-selective inhibitor, also lowered the expression of YAP and alleviated cartilage degeneration in DMM-induced murine OA model (Zhang et al., 2020). Wang et al. (2020) recently identified PIEZO1 as the major skeletal mechanosensor or mechanostat protein that tunes bone homeostasis. Mice deficient in Piezo1 gene expression in osteoblastic cells, while showing loss of bone mass and spontaneous fractures with increased bone resorption, are resistant to further bone loss and bone resorption due to mechanical unloading, suggesting that PIEZO1 in osteoblasts controls osteoblast-osteoclast crosstalk in response to mechanical forces. PIEZO1 regulates the YAP-dependent expression of collagen types II and IX, which in turn regulate osteoclast differentiation and bone remodeling. Additional support is provided by the findings of Zhou et al. (2020) who also reported that Piezo1/2 could regulate bone formation by mediating mechanotransduction via the NFAT/YAP1/ß-catenin axis in mice (Zhou et al., 2020). The manner and approach of regulating the mechano-modulatory factors and cellular activities in OA subchondral bone thus requires further exploration. Definitive classification of the nature of impairment of mechanoresponsive pathways in subchondral bone would be helpful to identify new subtypes of OA patients.