Osteoarthritis (OA) is a degenerative joint disease, which causes pain and disability and is characterized by progressive damage of articular cartilage, changes in the underlying (subchondral) bone, and occasional mild synovial inflammation.

Increasing evidence suggests that subchondral bone plays an important role in the etiology of OA [1, 2], but studies thus far do not provide a consistent view on this subject. Subchondral bone changes have been studied in both humans with OA and in animal models of OA. In human studies, an increase in trabecular bone volume fraction and trabecular thickness was found [3, 4], as well as an increase in cortical subchondral plate thickness [3]. However, other studies found a lower bone volume fraction and trabecular thickness in patients with OA [5, 6] or a decrease in stiffness [7, 8]. Even within one patient, areas with high and low bone volume fraction have been reported, depending on the condition of the overlying cartilage [9]. A problem of the human studies is that mostly established (severe) OA is studied, and longitudinal data showing the changes from onset until full clinical osteoarthritic signs do not exist. A problem is that there are no objective criteria that indicate early OA with mild pre-clinical signs and therefore the design of longitudinal studies is difficult.

Several animal models have been developed to study osteoarthritis and changes in the subchondral bone. Some animal studies reported a decrease in bone volume fraction and trabecular thickness [10–13], whereas in other studies these parameters increased [14, 15]. These differences may be explained by the type of model used and the time at which the measurements were performed. Some bone parameters may occur in two phases: an initial decrease followed by an increase [16].

A frequently used animal model of OA is anterior cruciate ligament transection (ACLT) in dogs. ACLT results in permanent instability of the knee joint, which is followed by osteoarthritic features [17]. The ACLT model has been used for in-vivo evaluation of several treatment strategies [18–21]. However, the instability remains present, and may counteract the possible beneficial effects of treatment.

For this reason, the canine groove model has been developed. In this canine model, surgically applied damage to the articular cartilage of the weight-bearing areas of the femoral condyles, not damaging the subchondral bone, is the trigger for development of OA features [22]. The model is distinctive in that the osteoarthritic trigger is not permanent and the degenerative changes are progressive and not just the expression of surgically applied chondral damage, while synovial inflammation diminishes over time [22–24].

In the current study, we report changes in the subchondral bone of the canine groove model and compare these with changes in the ACLT model. Because the cartilage damage induced in the groove model appeared less drastic than in the ACLT model, the groove model could be very useful to investigate the subtle relationship between bone and cartilage during the development of OA. Therefore, we studied the groove model also at a very early time point. Specifically we used micro-CT analyses to quantify subchondral trabecular bone volume and architecture, the subchondral plate thickness and porosity, and osteophytosis and related this to the changes in cartilage integrity.

OA was induced according to the ACLT model [25] or the groove model [22]. For the ACLT model, knee joints were available from 10 and 20 weeks post-surgery (both n = 5). For the groove model, knees were available from 3, 10 and 20 weeks post-surgery (all n = 4). In short, the following procedures were followed:

Groove model

In 12 animals, the cartilage of the lateral and medial femoral condyles was damaged with a Kirschner-wire (1.5 mm diameter) that was bent 90° at 0.5 mm from the tip as described before [22–24]. In this way the depth of the grooves was restricted to 0.5 mm. In utmost flexion, ten longitudinal and diagonal grooves were made on the weight-bearing parts of femoral condyles without damaging the subchondral bone (Fig. 1A). The latter was checked by histology at the end of the experiment. There was no absolute visual control over the procedure, but macroscopic evaluation after termination of the animals showed similar patterns in all knees treated. Two days after surgery, the dogs were forced to load the joint with the mechanically damaged cartilage by fixing the contra-lateral left limb to the trunk 3 days per week for approximately 4 h per day until the end of the experiment. The cartilage integrity and bone changes were evaluated 3, 10, and 20 weeks post-surgery (n = 4 in each group).

Changes in cartilage and in bone were similar for femoral condyles and tibial plateau. But for reasons of clarity the tibial plateau is shown as representative for both cartilage and bone parameters, since this surface was not surgically damaged when osteoarthritis was induced in the groove model, making comparison with the ACLT model the most sound.

Groove vs. ACLT at 10 and 20 weeks post-surgery

Cartilage integrity

Histological cartilage damage was increased in the experimental tibias of all animals in both models. (Table 1 and Fig 2A). This cartilage degradation was supported by biochemical analysis. A decrease in GAG content, representing impaired cartilage integrity, was found in the tibial plateau cartilage of both models. The GAG content was decreased with 10–25% in the experimental knee compared to the control knee (Table 1 and Fig 2B).

	Cartilage				Epiphyseal trabecular bone								Metaphysis		Subchondral plate
	GAG		Mankin		BV/TV		TbTh		SMI		CD		BV		PlTh		PlPor
	δ(%)	p	δ(-)	p	δ(%)	p	δ(%)	p	δ(-)	p	δ(%)	p	δ(%)	p	δ(%)	p	δ(%)	p
Groove
3w	-4.5	0.137	+0.17	0.055	+3.9	0.465	0.0	1.000	-0.05	0.465	+3.8	0.068	-0.9	0.715	-40.7	0.068	+84.8	0.068
10w	-11.1	0.233	+2.15	0.034	+6.0	0.068	+4.2	0.144	-0.30	0.068	+0.3	1.00	-3.1	0.068	-28.6	0.068	+47.7	0.068
20w	-20.9	0.034	+1.56	0.034	-3.5	0.144	-4.2	0.144	+0.28	0.068	+15.3	0.068	-12.5	0.144	-35.7	0.068	+72.2	0.068
ACLT
10w	-22.3	0.022	+1.95	0.021	-16.6	0.043	-12.2	0.043	+0.67	0.043	+20.9	0.225	-28.1	0.042	-28.7	0.043	+37.5	0.043
20w	-16.5	0.022	+1.45	0.021	-17.2	0.043	-13.6	0.043	+0.77	0.043	+19.5	0.043	-16.0	0.043	-30.9	0.043	+26.2	0.043

Osteophytes

In the groove model at 10 weeks post-surgery no osteophytes were found whereas at 20 weeks post-surgery they were clearly seen at the micro-CT images of the experimental tibial plateau in all four animals (Fig 3). In the ACLT model already at 10 weeks and also at 20 weeks post-surgery osteophytes were found at the experimental joint in all animals. For both models, the osteophytes were located predominantly at the medial site, below the rim of the tibia plateau. The osteophytes in the groove model were smaller than in the ACLT model. In none of the control joints osteophytes were observed.

Bone changes

Subchondral trabecular bone

Overall, the trabecular bone changes in the epiphysis of the experimental groove knee compared to its contralateral control were small. At 10 weeks there was a small increase in the trabecular bone volume fraction in the groove model. Also trabecular thickness was slightly elevated at 10 weeks. At 20 weeks the bone volume fraction and the trabecular thickness were slightly decreased (Table 1 and Fig. 4A, B).

The subchondral trabecular bone in the ACLT model showed a decrease in bone volume fraction (BV/TV) and trabecular bone thickness (TbTh) in all animals, at 10 and at 20 weeks. This was also reflected in the increase of the Structure Model Index (SMI) and Connectivity Density (CD) that indicate a more rod-like structure by the generation of more pores in the original structure, see Table 1.

Metaphyseal trabecular bone

In the metaphyseal region (region 3 in Fig 1B), which contained only trabecular bone, the differences in bone volume between control and experimental knee in the groove model were small. In the experimental ACLT knee, the bone volume was decreased up to 28% at 10 weeks (Table 1).

Subchondral plate

In contrast to the trabecular bone parameters, the changes in the subchondral plate were similar in both models. The thickness of the subchondral plate in the cylinders decreased in all animals with about 25 to 40% in both the groove and ACLT model at both time points. The porosity of the subchondral plate increased severely in both ACLT and groove model, at all time points in all animals (Table 1 and Fig 4C, D).

Diaphyseal cortical bone

In the diaphyseal part of the tibias, more distal from the joint, there were no differences in bone area and moment of inertia between the control knee and the experimental knee (data not shown) for both models.

Groove model at 3 weeks post-surgery

The development of OA appeared less advanced in the groove model than in the ACLT model. Therefore we used the groove model to gain further insight in the subtle relationship between cartilage and bone in the process of OA development. Thus, an additional time point was studied, at 3 weeks post-surgery.

Cartilage integrity

The histological cartilage damage in the experimental tibia was minimal at 3 weeks, while at 10 and 20 weeks, more cartilage damage was present and in all animals. The GAG content was minimally reduced at 3 weeks and gradually decreased over time (Table 1 and Fig 5A).

Bone changes and osteophytes

Subchondral trabecular bone

At 3 weeks, there were no consistent changes in trabecular bone. Also in the metaphyseal area, no changes in trabecular bone were observed between experimental and control tibia.

Subchondral plate

In contrast to the trabecular bone, there were already clear changes in the subchondral plate at 3 weeks in the groove model. In all animals the subchondral plate thickness was decreased, on average with 40%. The plate porosity was increased in all animals, with on average 85%, which is even larger than at the later time points (Table 1 and Fig 5B).

No diaphyseal cortical bone changes or any osteophytes were found at 3 weeks post-surgery in the groove model.

The thickness of the subchondral plate decreased very consistently in two different canine models of osteoarthritis: the groove model and the ACLT model. In contrast, the changes in the trabecular bone at the tibial epiphysis in the groove model were relatively small and not consistent over time whereas these changes in the ACLT model were larger, with up to 20% loss in bone volume fraction with significant changes in the corresponding architectural parameters. Due to the low number of animals in the groove model, the bone parameters could not reach statistical significance in this model. Although the trabecular parameters were not consistent, the changes in the subchondral plate were very consistent in the groove model, with a clear and early reduction of the plate thickness and an increase in plate porosity.

Although the grooves in the groove model were made in the femur only, the changes in subchondral bone were found in both the femur and in the tibia. This is in concurrence with the cartilage changes found in the groove model which also showed changes in both femur and tibia [22]. Since the subchondral bone changes in the tibia cannot be caused directly by the grooves, we believe that these changes are part of the osteoarthritic process. This suggests an interaction between the bone and the cartilage through diffusive molecules that originate from the degenerated cartilage or the synovial fluid.

The cartilage changes in both models were similar to the changes previously described for larger groups of animals [22, 24] and thus the data concerns a representative set of these earlier studies. The groove model showed only very mild changes in cartilage integrity at 3 weeks, which progressed at 10 and 20 weeks. In the ACLT model the changes were comparable to those in the groove model, but slightly more progressive.

Osteophytosis, visible on the CT-images, occurred in all the experimental ACLT knees at 10 and 20 weeks. This contrasts the groove model in which osteophytes only were detected at 20 weeks and not at 3 and 10 weeks. This corroborates the less progressive development of OA in the groove model compared to the ACLT model. However, a cartilaginous pre-form of the osteophytes may develop earlier, but is not detectable on the micro-CT images. In both models the osteophytes start below the rim of the medial tibia plateau and extend to more distant regions. This location is in line with osteophyte location in a rabbit ACLT model [16]. In human osteoarthritis, osteophytes are found close to the joint surface; it has been suggested that the load bearing area increases as to compensate for instability [33]. However, in our study, the osteophytes were also found in the groove model, in which the joint does not become unstable arguing against their role in joint stabilization. An explanation for the different location in comparison to humans may be that, in dogs, the ligaments are attached to the bone at a different location than in humans, thereby causing high stresses on the bone in a different location. In addition to this, cytokines such as TGFβ, which is elevated after OA induction [34, 35], stimulate osteophyte formation [36]. Since the synovial capsule in dogs extends more to the proximal and distal part of the joint than in humans, the interface between synovial capsule and bone is more distant from the joint space. Assuming synovial tissue derived cytokines to play an important role in osteophyte formation [37], this may explain their location in dogs compared to humans.

The changes in the trabecular bone were not very pronounced in the groove model. However, in the ACLT model, the bone volume fraction and trabecular thickness were clearly reduced. This corroborates the difference in rate of development of cartilage changes in both models. The changes in the ACLT model fit with previous studies in this model in dogs as well as cats [10–13]. The fact that other studies find an increase in bone volume fraction and trabecular thickness [14, 15] may be explained by the use of a different type of model, evaluated at a longer time period. Irrespective of the different changes in trabecular bone, similar changes in cartilage and subchondral plate were found in both models. Thus, it seems that the trabecular bone changes are not directly related to the changes in subchondral plate and cartilage. Since the subchondral plate changes consistently follow the cartilage changes, and the trabecular bone changes do not, the subchondral plate may play a more important role in the OA process than the trabecular bone changes.

The subchondral plate thickness decreased in both models at all time points in all experimental knees. This is in line with findings from previous studies concerning various animal models for OA, where subchondral plate thinning was documented in the early stage of the disease [10, 11, 38, 39]. In some of these studies, this early phase of thinning was followed by a later phase of plate thickening [11, 38]. This also explains the discrepancy with the sclerosis seen in most human studies [3, 4, 9], since such studies often concern patients with late osteoarthritis, whereas our present study examined only relatively early time points.

In order to justify the use of the contralateral knee as control, we calculated bone parameters in the diaphyseal and metaphyseal tibia, distal from the joint, containing cortical and trabecular bone, respectively. The bone volume of the metaphyseal tibia was significantly decreased in the experimental ACLT tibias, indicating disuse of the experimental ACLT knee. Thus, the trabecular bone loss in the epiphysis in the ACLT model may be explained by disuse. However, the tibias of the groove model showed hardly any changes in the diaphyseal and metaphyseal bone parameters. Hence, we have no signs of disuse in this model. Both the ACLT and groove model show similar subchondral plate thinning and increased porosity. Since the diaphyseal cortical bone showed no differences between control and experimental knee, we assume that in both models these subchondral plate changes are not caused by disuse of the treated leg.

The consistent decrease in subchondral plate thickness occurred already at 3 weeks post-surgery in the groove model, whereas the cartilage changes were only very mild at this early time point (Fig 5, table 1). This suggests that the subchondral plate changes occur fast. Taken together with the fact that this cannot be explained by disuse, this indicates (at least in the groove model) an interaction between cartilage and subchondral plate that induces bone resorption as a consequence of initiation of cartilage damage induced by the grooves. The thinning and drastically increased porosity of the subchondral plate may facilitate vascular invasion of the cartilage and diffusion of molecules from the damaged cartilage through the subchondral plate and vice versa, thereby enhancing the biochemical communication between bone and cartilage [40]. It is not clear if this bone cartilage communication interacts with an intrinsic repair activity of cartilage [41] or plays a role in the progression of the disease process [42].

We see differences in subchondral trabecular bone changes and osteophyte formation between the groove model and the ACLT model, with the groove model clearly showing a slower development of these changes. However, the severe loss of thickness and increased porosity in the subchondral plate are the same in both models. This quick and extensive loss of the subchondral plate thickness and increase in plate porosity cannot be explained by unloading and strongly suggests that cartilage-bone interplay is part of the etiology of osteoarthritis.