Research ArticleOsteoarthritis

Topographic modeling of early human osteoarthritis in sheep

See allHide authors and affiliations

Science Translational Medicine  04 Sep 2019:
Vol. 11, Issue 508, eaax6775
DOI: 10.1126/scitranslmed.aax6775

Deciphering degeneration

Aging and injury are known to contribute to osteoarthritis, a disease characterized by the breakdown of cartilage. Here, Oláh et al. induced meniscal injury in sheep and created detailed maps of disease progression. They found that cartilage and bone breakdown occurred first in the injured region before spreading throughout the joint. Patterns of degeneration observed in the sheep 6 months after injury mirrored pathology observed in human osteoarthritis. This large animal model of osteoarthritis could be useful for furthering our understanding of disease progression and for testing potential therapies.


Articular cartilage damage occurring during early osteoarthritis (OA) is a key event marking the development of the disease. Here, we modeled early human OA by gathering detailed spatiotemporal data from surgically induced knee OA development in sheep. We identified a specific topographical pattern of osteochondral changes instructed by a defined meniscal injury, showing that both cartilage and subchondral bone degeneration are initiated from the region adjacent to the damage. Alterations of the subarticular spongiosa arising locally and progressing globally disturbed the correlations of cartilage with subchondral bone seen at homeostasis and were indicative of disease progression. We validated our quantitative findings against human OA, showing a similar pattern of early OA correlating with regions of meniscal loss and an analogous late critical disturbance within the entire osteochondral unit. This translational model system can be used to elucidate mechanisms of OA development and provides a roadmap for investigating regenerative therapies.


Osteoarthritis (OA), the major cause of chronic disability in the United States (1) that affects more than 70% of its population above 55 years of age (>67 million Americans by 2030) (2, 3) with total costs exceeding $3 billion annually (4), confronts our society with enormous problems. Knee OA accounts for ~83% of the total OA burden (5) and is increasingly recognized as a complex, highly heterogeneous disorder (6). Emerging research suggests that various factors may initiate the breakdown of the articular cartilage, the hallmark of the disease. Lesions of the menisci are among the most important known causes leading to knee OA (7), and the strong association of meniscal damage with the onset and progression of OA in the tibiofemoral compartment is clinically well established (8, 9).

The early phase of OA, when the articular cartilage damage commences, is a key period during the development of the disease. However, we currently lack a comprehensive understanding of the underlying structural patterns that pathologically alter the spatiotemporal osteochondral trajectories, complicating an accurate numerical assessment of the disease, which is an essential factor to quantitatively evaluate progression and novel therapies. Models of induced meniscal instability in small animals enable only limited insights into topographical changes that result from a defined OA induction because of their size.

Our aim was to investigate and quantify the early spatiotemporal profile of OA in a preclinical large animal model and to compare it with the clinical situation, addressing the fundamental problem in translational medicine of how experimental models reflect clinical disease processes. We hypothesized that loss of medial meniscus stability in sheep induces a topographical pattern of osteochondral changes reflecting characteristics of early human knee OA.


Mapping the early topographic OA pattern

In an adult sheep model of knee OA induced by partial medial anterior meniscectomy (pMMx) (Fig. 1A) (10, 11), Kellgren-Lawrence (KL) grading of standard radiographs revealed mild early OA 6 weeks after pMMx, which was worse but insignificant compared to normal controls (P > 0.05) (Fig. 1, B and C). KL grading of OA significantly increased by 6 months, resembling mild established OA (P < 0.01) (Fig. 1, B and C). The total cartilage area affected by OA as assessed at the macroscopic level was always significantly larger after pMMx (P < 0.01) (Fig. 1D). The anterior central and anterior peripheral regions, both directly below the induced meniscal lesion (fig. S1), were particularly affected by OA at the macroscopic level, with the anterior central region showing the largest affected area (>80%) at both time points (P < 0.01) (Fig. 1, E to I). Thus, OA cartilage changes originate from the region directly below the affected meniscus.

Fig. 1 Ovine OA changes.

(A) Schematic of surgical approach. OA was induced by removing the anterior third of the medial meniscus (pMMx). (B) Representative x-ray images (arrowhead, osteophytes; arrows, suspected joint space narrowing) and (C) box plot of the KL scores (dots, individual data points; +, mean values; whiskers, minimum and maximum values of the datasets; upper and lower borders of the boxes, the 75th and 25th percentiles) of the ovine knees. Coverage of OA lesions (D) in the total area of the ROIs and (E to I) within the individual subregions. (J) Area of osteophytes and (K) schematic maps of their localization in the medial tibial plateau (larger images; insets show the entire tibial plateau). (L) Representative 2D micro-CT images showing osteophyte development (arrowheads) and the original margin of the tibial plateau (dashed lines). A schematic diagram of the subregions evaluated in the medial tibial plateau is shown in fig. S1. a, anterior; d, distal; l, lateral; m, medial; p, posterior; pr, proximal. n = 7 to 8 per group. *P < 0.05, **P < 0.01, normal versus pMMx with paired t test or Wilcoxon test.

Six weeks after pMMx, small osteophytes had developed exclusively at the periphery of the anterior and intermediate regions of the medial tibial plateaus. By 6 months, the size of the osteophytes had significantly increased (P < 0.05), although their location remained unchanged (Fig. 1, J to L). There was no correlation between the regional distribution of osteophytes and the region affected by OA as detected macroscopically at 6 weeks (P > 0.05), whereas there was a significant moderate correlation at 6 months (P = 0.00012, r = 0.571) (fig. S2), indicating their concomitant presence in the medial plateau.

Similar to the previous findings, the most severe OA consistently developed in the anterior central subregion as detected at both time points after pMMx (P < 0.01) (Fig. 2 and fig. S3). Fissures, fibrillations, and erosions of cartilage were significantly worse in the anterior peripheral and anterior central subregions (subscore for structure; P < 0.01 and P < 0.05) at both time points after pMMx (Fig. 2, A and C). Likewise, chondrocyte density decreased within these two anterior subregions (subscore for chondrocyte density; P < 0.01 and P < 0.05) at both time points after pMMx (Fig. 2, B and D). Polarized light microscopy revealed a disturbed collagenous architecture, particularly in the superficial cartilage layer of the anterior region at both time points (Fig. 2, E and F). OA changes after pMMx present in the anterior peripheral and anterior central regions were significantly more severe as assessed histologically at both time points (P < 0.01 and P < 0.05), as were changes within the intermediate central region (P < 0.01) at 6 months (Fig. 2, G to K), indicating that the topographical histopathological pattern of OA changes originates from the affected region below the meniscus and subsequently extends further to the major part of the medial tibial plateau.

Fig. 2 OA changes of the ovine articular cartilage.

(A, B, and E) Six weeks and (C, D, and F) 6 months after pMMx. Box plots of histological scoring of (A and C) surface structure and (B and D) chondrocyte density according to subregion. Higher scores indicate more deviation from normal; the color code is identical for all box plots. (E and F) Representative safranin O–stained histological sections and polarized light microscopy images of Masson-Goldner trichrome–stained histological sections (in insets; all panels with identical magnification) from the subregions of normal and pMMx samples, showing the severity of OA (E) 6 weeks and (F) 6 months after pMMx. Dashed lines indicate the alignment of the images according to the cement line. (G to K) Total OARSI histopathological scores (44) determined in the individual subregions, where higher values indicate more severe OA. n = 8 per group. *P < 0.05, **P < 0.01, normal versus pMMx with Wilcoxon test.

Differential alterations within the subchondral bone

Six months after pMMx, cartilage in the entire anterior region was significantly (P < 0.05) thicker than normal (fig. S4, A and B). Determining the ratio of the calcified cartilage to the entire cartilage thickness revealed that this increase originated from within the uncalcified articular cartilage, which is separated from the calcified cartilage by the tidemark (fig. S4, C to F). Our analyses further identified a specific spatiotemporal pattern of OA changes of the subchondral bone plate thickness (Fig. 3A and fig. S4G). Particularly, in the anterior peripheral and central subregions, a pronounced thickening occurred by 6 months after pMMx (P < 0.01 or P < 0.05), similar to the changes of the articular cartilage. A distinct topographical pattern of articular cartilage thickening also existed in all groups (fig. S4A). Significant (P < 0.01) and strong overall correlations between the cartilage and subchondral bone plate thickness were observed in all subregions of normal joints; after pMMx, these correlations were only moderate at 6 weeks (r = 0.478, P = 1.7 × 10−7), and overall correlation deteriorated by 6 months (r = 0.283, P = 0.003) (Fig. 3, B and C). When we analyzed the anterior, intermediate, and posterior regions separately, significant (P < 0.01) positive correlations between cartilage and subchondral bone plate thickness were found in each normal subregion. In contrast, 6 months after pMMx, the anterior region exhibited an inverse correlation (Fig. 3, B and C, and table S1).

Fig. 3 Ovine normal and OA osteochondral correlations.

(A) Schematic representation of differences in subchondral bone plate thickness expressed as the percentage of change in pMMx compared to normal. Scatter plot and linear regression showing the correlation between cartilage thickness and subchondral bone plate thickness in the normal and pMMx samples. (B) Six weeks and (C) 6 months after OA induction. Correlation coefficients of the individual regions are shown in table S1. a, anterior; d, distal; l, lateral; m, medial; p, posterior; pr, proximal. n = 18 to 48 locations per group. *P < 0.05, **P < 0.01, normal versus pMMx with paired t test or Wilcoxon test.

Next, we studied possible microstructural alterations within the subchondral bone plate by mapping individual regions at high spatial resolution (Fig. 4, A to E, and fig. S5). The bone surface density (BS/TV) significantly decreased in both anterior regions 6 months after pMMx (P < 0.05; Fig. 4D). The porosity of the anterior peripheral region significantly increased at 6 weeks (P < 0.05) but significantly decreased at 6 months (P < 0.01), reflecting the decrease of the structural complexity of the subchondral bone plate as seen in the three-dimensional (3D) models of porosity (Fig. 4, B, C, and E). More detailed analysis revealed that both the open and closed porosity showed similar patterns of spatiotemporal changes as the total porosity (fig. S5, F, G, K, and L). Assessment of other microstructural parameters of the subchondral bone plate revealed similar changes, with minimal alteration affecting only the anterior peripheral region at 6 weeks, whereas the anterior region and occasionally other subregions were affected by 6 months after pMMx (fig. S5).

Fig. 4 OA structural changes in the ovine subchondral bone.

Regional analysis of the (A and D) bone surface density (BS/TV) and (B and E) total porosity of the subchondral bone plate (A and B) 6 weeks and (D and E) 6 months after pMMx. (C) 3D reconstructed micro-CT models of the subchondral bone plate porosity showing the spatial distribution of open and closed pores in the anterior central region, with lower magnification view in insets. Progressive trabecular degradation in the subarticular spongiosa (F to M) 6 weeks and (N to U) 6 months after pMMx. (F and N) Representative Masson-Goldner trichrome–stained histological sections, 3D reconstructed micro-CT models, color-coded 2D micro-CT images, and 3D models of trabecular separation of the subarticular spongiosa of the intermediate peripheral region of normal and pMMx (F) 6 weeks and (N) 6 months after surgery. Regional analysis of (G and O) BS/TV, (H and P) bone volume fraction (BV/TV), (I and Q) bone mineral density (BMD), (J and R) trabecular number (Tb.N), (K and S) connectivity density (Conn.Dn), (L and T) trabecular separation (Tb.Sp), and (M and U) trabecular thickness (Tb.Th). Ant, anterior region; Interm, intermediate region; Post, posterior region; centr, central; periph, peripheral. n = 7 to 8 per group. *P < 0.05, **P < 0.01, normal versus pMMx with paired t test or Wilcoxon test.

The 3D structure of the subarticular spongiosa was subsequently analyzed to evaluate bone quality (12). After pMMx, the relative bone volume (BV/TV) and trabecular thickness (Tb.Th) decreased significantly by 6 weeks (P < 0.01 or P < 0.05) in most regions, whereas other parameters remained largely unchanged (Fig. 4, F to M) or varied only in the anterior subregion (fig. S6). By 6 months, the significant (P < 0.01 or P < 0.05) global losses of bone volume and trabecular structural complexity after pMMx affected all subregions of the medial tibial plateau (Fig. 4, N to U, and fig. S6, A and G to K). The BS/TV, BV/TV, bone mineral density (BMD), trabecular number (Tb.N), and connectivity density (Conn.Dn) all decreased (P < 0.01 or P < 0.05), whereas the trabecular separation (Tb.Sp) increased in the entire medial tibial plateau (P < 0.01 or P < 0.05; Fig. 4, N to U).

To begin deciphering the contributions of each component of the osteochondral unit during the disease progression, we performed a multivariate analysis. At 6 weeks postoperatively, a fair separation between groups was achieved with principal components analysis (PCA) solely when analyzing the major articular cartilage structural parameters in the anterior region, suggesting that at this time point the cartilage parameters of this specific region are the only usable indicators of OA (Fig. 5, A and B, and fig. S7). By 6 months, an adequate separation of articular cartilage parameters was observed in both the anterior and posterior regions by multivariate analysis (Fig. 5C). Unexpectedly, a considerably more distinct separation of the two treatment groups was observed in each of the subregions when the subchondral bone microstructure–related parameters were analyzed (Fig. 5D and fig. S8). These data suggest that at mid-term, subchondral bone parameters may serve as robust structural indicators of OA degeneration. PCA and cluster analysis identified BMD, DA (degree of anisotropy), BS/BV, and Tb.Sp, all of the subarticular spongiosa, to be sufficient and necessary for clear separation between the normal and pMMx group for all of the regions (fig. S9).

Fig. 5 PCA of ovine OA.

(A and C) Articular cartilage and (B and D) subchondral bone, (A and B) 6 weeks and (C and D) 6 months postoperatively. All determined cartilage and bone parameters were used as input data to visualize the separation of the samples into groups based on the individual subregions. Data points represent individual normal and pMMx samples. Information content (% of variance) of the axes for principal components 1 (PC1) and 2 (PC2) are shown above the graphs.

Topographic characterization of human OA

To assess whether the spatiotemporal patterns observed within our ovine model recapitulated human OA progression, we analyzed the knees of human subjects without or with early or late OA. Radiographic evaluation and KL grading confirmed early OA (Fig. 6A) and subsequent macroscopic examination by direct arthroscopic inspection revealed the highest extent of meniscal loss in the posterior region (Fig. 6B and fig. S10A). Correspondingly, articular cartilage lesions were quantified (13) as severe in the posterior and intermediate central subregions (Fig. 6B and fig. S10B). The localization of the meniscal defects and the cartilage lesions showed a significant, strong, positive correlation (r = 0.704, P = 1.9 × 10−7) (fig. S10C).

Fig. 6 Evaluation of human OA.

(A) Representative x-rays (arrowheads, osteophytes; arrows, joint space narrowing) and box plot of KL scores of human knees. (B) Schematic topographical comparison of ovine subregions (6 weeks, 6 months) and human early and late OA. (C) Localization and area of osteophytes in late human OA (large image, medial tibial plateau; inset, entire tibial plateau). (D) Representative India ink–stained macroscopic images and 3D reconstructed micro-CT models of a normal and late OA tibial plateau: regional analysis of coverage by OA. (E) Structure, (F) chondrocytes, (G) matrix staining subscores, and (H) total Mankin (51) and (I) OARSI scores (52) of human late OA. (J) Schematic map of mean human cartilage thickness. (K) Structural overview and color-coded 2D and 3D micro-CT models of osteochondral thickness of representative human normal and OA samples. (L) Schematic map of mean human subchondral bone plate thickness. (M) Scatter plot and linear regression showing the correlation between cartilage thickness and subchondral bone plate thickness (normal, OA samples). (N) Schematic map of percentage of change of cartilage and subchondral bone plate thicknesses in late OA compared to normal. Ant, anterior; Interm, intermediate; Post, posterior; centr, central; periph, peripheral; a, anterior; l, lateral; m, medial; p, posterior. n = 4 to 13 per group. *P < 0.05, **P < 0.01 with unpaired t test or Mann-Whitney test.

Osteophytes were found adjacent to all regions in the knees of patients with advanced OA (P < 0.05) (Fig. 6C), with evidence of joint space narrowing and sometimes bony deformity, reflected in a high KL score (P < 0.01 compared to normal controls; Fig. 6A). Articular cartilage damage (Fig. 6D) as well as individual parameters and overall scores (Fig. 6, E to I) were severe and evenly distributed across regions. The topographical pattern of global OA cartilage changes in the affected regions was similar to sheep at 6 months, with stronger changes in humans, reflecting the late disease stage (Fig. 6B). Mapping of normal human articular cartilage and subchondral bone plate thicknesses revealed that both were thinner in the peripheral subregions covered by the meniscus (Fig. 6, J to L). These parameters showed significant (P < 0.01), strong overall and region-specific correlations between cartilage and subchondral bone thickness (Fig. 6M and table S1). In contrast to these findings, in OA, the cartilage thickness was decreased in the anterior peripheral and intermediate peripheral subregions (Fig. 6, J, K, and N). The topographical distribution of OA changes in the subchondral bone plate thickness (Fig. 6N) revealed increases primarily in the submeniscal and posterior parts of the medial tibial plateau. The normal strong correlation of the articular cartilage and subchondral bone plate thicknesses was severely disturbed in human OA, as previously seen in the ovine model (Fig. 6M and table S1).

The subarticular spongiosa was affected in most regions (Fig. 7, A to G). In particular, the BS/TV was significantly higher (P < 0.01 or P < 0.05) compared to normal in all subregions (Fig. 7A). Numerical changes of other parameters were mostly unidirectional with a random pattern of significances (Fig. 7, B to D, and fig. S11). As the trabeculae itself expanded in late OA (BV/TV, Tb.N; Fig. 7, E and F), the diameter of the spaces between the trabeculae decreased (Tb.Sp; Fig. 7, G and H). Because in humans the posterior region is most affected by meniscal damage (14), we compared the structural parameters of the sheep osteochondral unit to comprehend the relevance of these changes. A structural similarity exists between the normal ovine anterior and posterior regions (Fig. 7, I to K) based on multivariate analysis of merged data from both time points (Fig. 7J). Because most (17 of 22) parameters were not significantly different (P > 0.05) (Fig. 7K), the observed pattern after anterior meniscal damage in sheep reproduces the posterior meniscal damage in humans.

Fig. 7 Evaluation of human OA in the subarticular spongiosa.

Regional analysis of (A) BS/TV, (B) Conn.Dn, (C) Tb.Th, (D) fractal dimension (FD), (E) BV/TV, (F) Tb.N, and (G) Tb.Sp. n = 5 to 8 per group. (H) Representative 3D reconstructed models, color-coded 2D micro-CT images, and 3D models of trabecular separation (intermediate central region, human normal and late OA knees). (I) Representative Masson-Goldner trichrome–stained histological sections and 3D reconstructed micro-CT models (anterior and posterior subregions) in normal sheep knees. Evaluation of the similarity of the anterior and posterior subregions in all normal ovine samples (n = 16) by (J) PCA and (K) plotting the distribution of nonsignificant (n = 17) and significantly different (P < 0.05; n = 5) parameters within all of the evaluated micro-CT parameters (n = 22). (L) PCA and (M) cluster analysis of human normal and late OA subarticular spongiosa (considering all determined micro-CT parameters, all subregions). (N) PCA and (O) cluster analysis with heat-map comparison of human normal and late OA articular cartilage based on Mankin and OARSI scores. Ant, anterior; Interm, intermediate; Post, posterior; centr, central; periph, peripheral. *P < 0.05, **P < 0.01 with unpaired t test or Mann-Whitney test.

Reconstructing pathological OA trajectories

A distinct separation between human normal and late OA exists on the microstructural level of both the subchondral bone plate and subarticular spongiosa (Fig. 7, L and M). Separation between normal and late OA was also revealed using the parameters of the articular cartilage (Fig. 7, N and O). Thus, the spatiotemporal sequence of OA changes is initiated in the region directly below the meniscal injury, progressively affecting the entire tibial plateau over time, especially at the level of the subarticular spongiosa in otherwise stable knees, in both sheep and humans (fig. S12).


By analyzing the spatiotemporal changes of OA development of the articular cartilage and subchondral bone microarchitecture, we show here that the severe topographical pattern of OA changes originates from the affected region below the meniscus, and that the normal strong correlation of cartilage and subchondral bone is disturbed during OA progression. Our data also reveal that OA changes in the subarticular spongiosa develop in the same location as the meniscal injury and then progress globally, and that these subchondral bone alterations may serve as precise indicators of OA. Last, by validating these findings against clinical OA changes in patients with medial meniscal lesions in otherwise stable knees, we identified a pattern of early human cartilage damage correlating with regions of meniscal tissue loss, similar to the changes in the preclinical ovine model, and an analogous late disturbance of the previously normal configurations within the osteochondral unit.

Our investigation of the spatiotemporal profile of OA in a preclinical large animal model provides a previously unexplained link between clinical meniscal damage and OA development. The pattern of early OA changes started in the area affected by the partial meniscus loss in sheep. In patients, a strong correlation between meniscal lesions and articular cartilage damage exists, supporting the view that meniscal damage instructs the development of OA cartilage lesions in its close proximity. In humans, the posterior meniscal horn is commonly affected, making it clinically the most relevant portion (14), because the human tibial plateau slides back on the femoral condyles during knee flexion, transmitting higher compressive loads (15). As an intact medial meniscus transmits ~50% of the load in its compartment (1517), its partial removal significantly increases the tibiofemoral load (15, 1820), which may initiate cartilage degeneration (15, 21, 22). The anterior root tear-induced destabilization of the remaining part of the meniscus causes a ~25% increase in peak contact pressure, comparable to a total meniscectomy (23), progressively resulting in direct contact in nonphysiological locations of the joint surfaces, increasing the friction within the entire tibiofemoral compartment. Thus, major meniscal functions are most likely considerably disrupted in the pMMx ovine model, resulting in increased loads that possibly lead to the erosion of the cartilage surface over time. Our data suggest that the pattern of early cartilage degeneration is similar between sheep and humans.

The trabecular deterioration of the subarticular spongiosa of the entire medial tibial plateau is a hallmark of subchondral bone changes in mid-term OA, one of the major findings of this study. Alterations of the subarticular spongiosa developed at the same relative location as the damage of the articular cartilage and then progressed globally, suggesting a continuous weakening through extensive remodeling. From a biomechanical standpoint, the underlying subchondral bone may be even more sensitive to local altered forces, as it absorbs ~10× more force than the cartilage (24). Similar clinical changes (25) in BMD at 6 months, the marked loss of complexity (DA, FD) and connectivity (Conn.Dn, Tb.Pf) of the trabecular structure, and the decreased number (Tb.N) and increased separation (Tb.Sp) of the trabeculae in all subregions indicate the spread of OA into the deeper subchondral regions (2631). This may reflect the higher bone turnover rate in the subarticular spongiosa than in the subchondral bone plate (32) and a pathological discrepancy of the much higher force normally absorbed in the subchondral bone compared to cartilage (24), likely causing an expedited dissemination of the damage in the subarticular spongiosa. Such OA changes in the subarticular spongiosa may possibly serve as precise indicators of OA. Likewise, correlations of cartilage and subchondral bone plate thickness were severely disturbed in human OA, as in the ovine model.

Limitations of the study include the lack of sham-operated controls, considering historical nonsignificant differences between normal and sham-operated controls (33, 34), and the clinical observation that medial meniscal lesions are frequently in posterior regions (14). Surgical OA induction in the posterior region of the knee in sheep is complicated because it potentially destabilizes the medial collateral ligament (MCL), which may alter the course of the experimental OA. However, our results confirmed the high structural similarity between the ovine anterior and posterior regions. Considering human ex vivo data (15, 23) and the lack of tools for measuring loading force in vivo, no biomechanical evidence is provided to support the claim that pMMx alters intra-articular load distribution in this model. Last, a direct structural comparison between early OA in human and sheep was not feasible because of the well-known difficulties of obtaining human samples of early OA. Before direct application to human OA, clinical high-resolution imaging modalities capable of reliably detecting subtle changes within the subchondral bone structure require further development to apply new radiological markers of the early and mid-stage of the disease. Further validation of this model with long-term (>5 years) ovine studies is desirable to simulate late clinical OA. Moreover, such topographical analyses could be translated to study other important causes of OA, such as anterior cruciate ligament (ACL) injuries.

This investigation has at least three important clinical implications: First, the specific microarchitectural changes of the subchondral bone discovered here might find application as new radiological markers to detect early OA, which is yet difficult due to its largely asymptomatic nature. Fractal signature analysis of trabecular morphometry using standard (2D) radiographs has been proposed as a prognostic tool to predict knee OA progression (35, 36), and high-resolution peripheral quantitative computed tomography (CT) may provide even more informative 3D analyses (37). Second, the initiation of focal OA cartilage changes below the area of the meniscal injury is in good agreement with the recent clinical findings of a highly spatially heterogeneous cartilage loss within a joint compartment (38), challenging the clinical dogma that radiographic changes based on the established KL score (21) are sufficient clinical endpoints when investigating OA and supporting the value of 3D image–based quantitative cartilage and subchondral bone analyses. Third, these findings suggest a potential benefit of surgical repair of traumatic, nondegenerative meniscal injuries and root tears to decelerate OA progression (39, 40), although more research is needed to assess if and how these findings are affected by meniscal repair.

In summary, this work reduces the gaps in our fundamental understanding of early OA by modeling the complex 3D structural variations of the osteochondral unit. We anticipate that such profiling will be transformative for studying fundamental mechanisms of OA and for investigating novel therapies.


Study design

We developed a model of early human OA by gathering detailed spatiotemporal data on OA development in sheep. Although medial meniscal lesions are more common in a posterior location in humans, we selected a partial anterior meniscectomy including a root tear (pMMx) to avoid a possible postoperative instability of the MCL that would have to be released during the surgical approach to the posterior meniscus and that may potentially alter the course of the experimental OA in the right hind legs of sheep (n = 16; Fig. 1A). Contralateral knees served as unoperated controls. On the basis of previous data (41, 42), sample size requirements were calculated to detect a mean difference of 5 in total points and an SD of 3 points of the semiquantitative Osteoarthritis Research Society International (OARSI) histopathological scoring system (43, 44). To accommodate the possible loss of one animal accounting for complications, a sample size of n = 8 animals per treatment group was chosen. Endpoints of the animal experiments were 6 weeks and 6 months after pMMx to study early and mid-term time points of OA development, respectively. Criteria for terminating the study before completion included persistent lameness or infection after surgical recovery. None of the animals met these criteria. At the times indicated, animals were euthanized and OA changes of the medial tibial plateau were quantified by macroscopy, radiography, histomorphometry, semiquantitative histopathological scoring, and micro-CT. Correlations and clustering of the samples based on the subchondral bone and articular cartilage were evaluated by multivariate analyses. Correlation between the location of medial meniscal damage and cartilage lesions of the tibial plateaus were based on arthroscopic examination of human knees because of symptomatic meniscal lesions. Human normal and late OA tibial plateaus were submitted to analyses of articular cartilage histomorphometry and subchondral trabecular bone characteristics. Reproducibility of micro-CT measurements was confirmed by two repeated measurements of a human tibial plateau sample, resulting in similar values of all tested parameters. KL grading of x-ray images was repeated by two observers, resulting in similar scores. During histological analysis and semiquantitative scoring, three technical repeats (sections) per sample were examined, giving similar results for the subscores and total scores. No outliers or other observations were removed on account of their deviation from the mean in any of the experimentation shown. Samples with artifacts compromising the reliability of the measurements were excluded when necessary. Data were collected blindly and decoded only after analysis.

Animal experiments

Sixteen healthy female Merino ewes (mean body weight, 70 ± 20 kg; mean age, 18 months) were randomly divided into two treatment groups (6 weeks, 6 months; n = 8 each), all of them receiving a standardized mini-open medial partial anterior meniscectomy (Fig. 1A) (45) on their right knees. Left knees served as unoperated controls. All animal experiments were conducted in accordance with the German legislation on protection of animals and were approved by the Saarland University Animal Committee according to German guidelines (Versuchsvorhaben Nr. 43/2015).

Before surgery, preoperative radiographs were taken to exclude OA. Animals were monitored at all times by a veterinary surgeon as previously described (46). After a 12-hour fast, animals were sedated with 2% Rompun (0.05 mg/kg body weight; Bayer) and received general anesthesia with intravenous application of 20 ml of 2% propofol (AstraZeneca) and carprofen (1.4 mg/kg body weight; Pfizer). After endotracheal intubation, anesthesia was maintained with inhalation of 1.5% isoflurane (Baxter) and intravenous application of propofol (6 to 12 mg/kg body weight per hour).

The right stifle joint was entered using a medial parapatellar mini-approach without dislocation of the patella. The anterior horn of the medial meniscus was resected directly anterior to the MCL until the anterior meniscal root. The joint was rinsed thoroughly, incisions were closed in layers, and a spray bandage was applied. Postoperatively, 3 ml of 0.25% fenpipramide/levomethadone (MSD) and amoxicillin clavulanate (30 mg/kg body weight; Pfizer) were administered. Animals were allowed full weight bearing immediately after surgery. After 6 weeks and 6 months, animals were euthanized in general anesthesia and the stifle joints were subjected to the subsequent analyses.

Human samples

All human OA samples were from patients with otherwise stable knees undergoing total knee replacement. Late OA tibial plateau samples (n = 8; n = 4 females, n = 4 males; mean age, 68 ± 9 years) were obtained as surgical discards from the Department of Orthopaedic Surgery, Saarland University Medical Center. Patients were selected exclusively with known medial meniscal involvement, excluding other causes such as ACL ruptures. Normal human tibial plateaus (n = 5; n = 2 females, n = 3 males; mean age, 67 ± 21 years) were obtained from the Department of Pathology, Saarland University Medical Center, Homburg. Early OA patients (n = 14; n = 9 females, n = 5 males; mean age, 70 ± 10 years) of the Department of Orthopaedic Surgery (Saarland University Medical Center) following a meniscal injury with otherwise stable knees were examined arthroscopically to evaluate their medial tibial plateaus and menisci. Informed consent was obtained from all participants and from the families of the cadaver donors providing normal control samples. The study was approved by the Ethics Committee of the Saarland Physicians Council (Ärztekammer des Saarlandes, Ethik-Kommission, No. 267/17).

Macroscopic grading of OA and definition of subregions

The area of OA and normal cartilage was visualized by India ink staining (41) (n = 16 normal, n = 16 pMMx sheep, n = 5 normal, n = 8 late OA human samples). For sophisticated analyses, the medial tibial plateau was divided into five zones: (i) anterior peripheral and (ii) central zone (divided at 50% of the anterior tibial plateau width), (iii) intermediate meniscal covered (peripheral) and (iv) not covered (central) area (40 and 60% of the intermediate tibial plateau width, respectively), and a (v) posterior zone (fig. S1) (47). To avoid overlapping of the regions, they were separated from each other with 6-mm (between the thirds) or 2-mm (between the peripheral and central regions) distance. The area with OA within the regions of interests (ROIs) was calculated on macroscopic photographs using a Canon PowerShot A480 camera (Canon) with 10 megapixels and a specific macroscopic lens under standardized conditions including illumination by two blinded observers [T.O. (postdoctoral biomedical scientist with special training in macroscopic, histopathologic, and microcomputed analysis of OA; for KL grading, he received extensive training from L.K.H.G. and H.M.) and J.R. (sixth-year medical student with special training in macroscopic, histopathologic, and microcomputed analysis of OA)]. The cellSens Standard software version 1.12 (Olympus) was used for image evaluation.

Radiographic analysis

After euthanasia, long leg anteroposterior radiographs of each knee joint were acquired using a Siemens Arcadis Varic image intensifier (Siemens Healthcare). X-ray images of sheep (n = 16 normal, n = 15 pMMx; one pMMx sample at 6 months was omitted because of low image quality) and human (n = 5 normal, n = 13 early OA, n = 8 late OA) samples were graded according to the semiquantitative KL grading system (48, 49), with one patient (early OA) omitted due to missing x-ray images. Evaluations were performed by two independent, blinded observers [T.O. and L.G. (registrar for orthopedic surgery with special training in OA and KL grading)]. Disagreement was resolved by consultation with a third observer [H.M., senior consultant for orthopedic surgery and specialist for macroscopic, histopathologic, radiologic, (including KL), and microcomputed analysis of OA].

Arthroscopic mapping of meniscal and cartilage damage

Medial tibial plateaus and menisci of patients with symptomatic meniscal lesions (n = 14) were examined arthroscopically and scored on the basis of the extent of early OA using the International Cartilage Repair Society (ICRS) grading system (13) and meniscal damage (0, no damage; 1, less than one-third of the ROI area affected; 2, one-third to two-thirds of the ROI area affected; 3, more than two-thirds of the ROI area affected) by H.M. Because of the strict indication criteria for arthroscopy of the patients with symptomatic medial meniscal lesions and early OA, blinding was not possible, as this would have needed to include an asymptomatic group of patients undergoing similar arthroscopy.

Semiquantitative histological scoring

Paraffin-embedded samples were sectioned (5 μm) at constant positions within the medial tibial plateaus corresponding to the previously described regions. A total of 720 ovine and 180 human sections were analyzed.

Safranin O/fast green (safranin O) (50)–stained ovine sections (n = 480; three technical repeats for five subregions of n = 8 normal and n = 8 pMMx samples of the 6-week and 6-month time points) were evaluated with the OARSI semiquantitative histopathological scoring system according to Little et al. (44) (inverse score, 0 to 25 points with higher points for more severe OA) and the Mankin score (51) (inverse score, 0 to 14 points with higher points for more severe OA) by two blinded observers (T.O. and J.R.). Human safranin O–stained histological sections (n = 180; one normal sample was omitted because of sample processing artifacts; three technical repeats for five subregions of n = 4 normal and n = 8 late OA samples) were evaluated with the Mankin and OARSI (52) scores (inverse score, 0 to 6 grades with higher grades for more severe OA) by two blinded observers [T.O. and S.H. (medical student with special training in histopathologic analysis of OA)]. Mean values of the analyzed sections were calculated for further analyses. Masson-Goldner trichrome–stained histological sections (n = 160; five subregions of n = 8 normal and n = 8 pMMx samples of the 6-week and 6-month time points) were evaluated with polarized light microscopy (53). To determine the ratio of calcified cartilage to total cartilage thickness, hematoxylin and eosin–stained sections were used (n = 80; five subregions of n = 8 normal and n = 8 pMMx samples of the 6-month time point) (41). Photos were taken with an Olympus BX45 microscope (Olympus), and images were stitched in Photoshop CS5 (Adobe).

Micro-CT imaging

Thirty-two ovine tibial plateau specimens (8 from left and 8 from right knees of the 6-week and 6-month groups) were scanned in a micro-CT scanner (SkyScan 1176, Bruker microCT; tube voltage, 90 kV; current, 278 μA; resolution, 18 μm; combined 0.5-mm aluminum/copper filter; 0.4° intervals; 270-ms exposure time, averaging three frames) as described previously (47). Because of their large size, n = 8 OA and n = 5 normal human samples were scanned with a spatial resolution of 35 μm. Images were reconstructed by a modified Feldkamp cone-beam algorithm (54) with NRecon software (v., Bruker microCT). To obtain comparable image sets, the reconstructed datasets were rotated uniformly (DataViewer software v., Bruker microCT) before saving the coronal section images for further evaluations (47). Evaluations were performed by two blinded observers (T.O. and J.R.).

Volumes of interests (VOIs) selected within the five main regions of the medial tibial plateaus were evaluated separately. The following 3D structural parameters were determined in all VOIs using the software provided by the manufacturer (CTAnalyzer v., Bruker microCT) (46): BMD, BV/TV, bone surface–to–volume ratio (BS/BV), and BS/TV. Percentage of total porosity as well as open and closed pores were calculated only in the subchondral bone plate. Tb.Th, Tb.Sp, Tb.N, trabecular pattern factor (Tb.Pf), structure model index (SMI), DA, fractal dimension (FD), and Conn.Dn were assessed only in the subarticular spongiosa. Because of the limited scanning resolution of the human samples, structural parameters of the subchondral bone plate were evaluated only in sheep. For more detailed evaluation of the subchondral bone plate thickness [or cortical thickness (Ct.Th)] and cartilage thickness (Ca.Th), these parameters were measured in three thirds (outer, middle, and inner) of the subregions. Locations where either the cartilage thickness or the subchondral bone plate thickness could not be accurately determined were excluded from the analysis.

For 3D reconstruction of the micro-CT image sets and modeling of trabecular separation, cartilage, and subchondral bone plate thickness, the CTVox v. 3.2.0 (Bruker microCT) program was used, using shadows and surface lighting to enhance the visibility of the surface structures of the samples or using color coding to visualize structural thickness. To generate a 3D model of the porosity of the subchondral bone plate, the CTVol v. (Bruker microCT) program was used. Area of osteophytes was measured on 3D reconstructed micro-CT images using the cellSens Standard software.

Statistical analysis

Normal distribution and equal variance of the data were tested with the Shapiro-Wilk normality test and f test, respectively. When matching sheep normal control and pMMx data were compared, depending on a normal distribution, statistical significance between the corresponding normal and pMMx samples was tested by Student’s paired t test (two-tailed) or Wilcoxon signed-rank test. When human normal control and OA data were compared, depending on a normal distribution, statistical significance was tested by Student’s unpaired t test (two-tailed) or Mann-Whitney rank sum test. The Pearson correlation coefficient was calculated to test correlation between the thicknesses of the articular cartilage and the subchondral bone plate.

Multivariate analyses were performed for all available bone (BMD, BV/TV, BS/BV, BS/TV, total, open and closed porosity, and thickness of the subchondral bone plate; BMD, BV/TV, BS/BV, BS/TV, Tb.Th, Tb.Sp, Tb.N, Tb.Pf, SMI, DA, FD, and Conn.Dn of the subarticular spongiosa) and cartilage parameters (semiquantitative histological scores and sub-scores, cartilage thickness, and coverage of OA lesions) of the relevant individual subregions of the samples. For hierarchical cluster analysis, the unweighted pair-group average (UPGMA) algorithm was used to produce a dendrogram. Clusters were joined on the basis of the average distance between all members in the two groups. The distance matrix was computed using the Gower index. PCA with a correlation matrix routine was used for reduction of the multidimensional dataset to only two variables, which account for most of the variance in the original dataset (55). Missing data were first replaced with their column average, and then their regression values were computed with an initial PCA run, iterated until convergence (55).

All calculations were performed with SigmaPlot version 12.0 (Systat Software Inc.), Prism version 6.01 (GraphPad Software Inc.), or Past version 3.16 (55); P < 0.05 was considered statistically significant. Data were expressed as mean ± SD. Box plot diagrams always show the individual data points (dots), mean values (+), the minimum and maximum values of the datasets (whiskers), and the 75th and 25th percentiles (upper and lower borders of the boxes). Raw data of the figures are included in data file S1.


Fig. S1. Subregions evaluated in the ovine medial tibial plateau.

Fig. S2. Correlation between the location of osteophytes and OA lesions in sheep.

Fig. S3. Histopathological scores of the ovine articular cartilage.

Fig. S4. OA patterns of ovine articular cartilage and subchondral bone plate thickness.

Fig. S5. Topographical patterns of the OA changes of the ovine subchondral bone plate.

Fig. S6. Topographical patterns of the OA changes of the ovine subarticular spongiosa.

Fig. S7. Cluster analysis of the ovine articular cartilage and the subchondral bone parameters at 6 weeks after pMMx.

Fig. S8. Cluster analysis of the ovine articular cartilage and the subchondral bone parameters at 6 months after pMMx.

Fig. S9. Minimally required ovine subchondral bone parameters to clearly separate normal from pMMx with multivariate analysis at 6 months.

Fig. S10. Human early OA meniscus and cartilage scores.

Fig. S11. Additional structural parameters of subarticular spongiosa in human late OA.

Fig. S12. Proposed schematic of the initiation and spatiotemporal progression of OA by a defined meniscal tissue loss.

Table S1. Pearson correlation analysis between the subchondral bone plate thickness and the articular cartilage thickness within the three main regions.

Data file S1. Raw data of the figures.


Acknowledgments: Dedicated to the memory of H. J. Mankin. We thank R. M. Bohle, D. Kohn, M. Laschke, C. B. Little, M. D. Menger, and P. Orth for helpful discussions and G. Schmitt and S. Speicher-Mentges for excellent technical assistance. Funding: This project was supported by the German Federal Ministry of Education and Research (BMBF; OVERLOAD-PrevOP: funding number: 01EC1408C). Author contributions: H.M. conceptualized the study and performed animal surgery. T.O., L.G., J.R., S.H., and H.M. acquired and analyzed data. T.O., M.C., L.K.H.G., and H.M. interpreted data. T.O. and H.M. wrote the initial draft. All authors contributed to editing and revising the manuscript and have approved the submitted version of the manuscript. Competing interests: H.M. has consulted for Bone Therapeutics and received a speaker honorarium from Novartis. All other authors declare that they have no competing interests. Data and materials availability: All data associated with this study are available in the main text or the Supplementary Materials.

Stay Connected to Science Translational Medicine

Navigate This Article