Combined enzymatic degradation of proteoglycans and collagen significantly alters intratissue strains in articular cartilage during cyclic compression

Abstract

As degenerative joint diseases such as osteoarthritis (OA) progress, the matrix constituents, particularly collagen fibrils and proteoglycans, become damaged, therefore deteriorating the tissue’s mechanical properties. This study aims to further the understanding of the effect of degradation of the different cartilage constituents on the mechanical loading environment in early stage OA. To this end, intact, collagen- and proteoglycan-depleted cartilage plugs were cyclically loaded in axial compression using an experimental model simulating in vivo cartilage-on-cartilage contact conditions in a micro-MRI scanner. Depletion of collagen and proteoglycans was achieved through enzymatic degradation with collagenase and chondroitinase ABC, respectively. Using a displacement-encoded imaging sequence (DENSE), strains were computed and compared in intact and degraded samples. The results revealed that, while degradation with one or the other enzyme had little effect on the contact strains, degradation with a combination of both enzymes caused an increase in the means and variance of the transverse, axial and shear strains, particularly in the superficial zone of the cartilage. This effect indicates that the balance between cartilage matrix constituents plays an essential role in maintaining the mechanical properties of the tissue, and a disturbance in this balance leads to decrease of the load bearing capacity associated with degenerative joint diseases such as OA.

1. INTRODUCTION

Articular cartilage is a connective tissue whose main function is to absorb impact, minimize peak loads in the underlying subchondral bone, and ensure low friction between the joint surfaces. It is composed of an extracellular matrix (ECM) consisting of 70-80% water, 10-25% type II collagen and 5-15% proteoglycans (PGs), in which cartilage cells – chondrocytes – are embedded [1], [2]. Due to its avascular nature, low tissue turnover and low cell density, the regenerative properties of articular cartilage are very limited [3], and cartilage defects, overload or natural wear and tear are often a first step in developing osteoarthritis (OA) [4]. This disease is estimated to affect more than 40 million people across Europe and is predicted to become the fourth leading cause of disability worldwide by 2020 [5], [6].

The pathophysiology of knee OA involves a degenerative cascade in the cartilage, starting with inflammation and the release of cytokines and other signaling molecules that might increase catabolic chondrocyte activity, apoptosis, and production of catabolic agents such as cartilage-degrading proteinases [7]. These alterations lead to cartilage loss, with the earliest visible changes in the structure of the articular surface being fibrillation or disruption of the superficial zone, which later develops to more complex fissures that might extend to the deeper zones. The extent of knee OA can be estimated using the OARSI score, which grades the severity of the disease from 0 to 6, according to depth progression into cartilage [8], with early stage OA defined histologically as having a grade of 1–3 [9] and affecting only the superficial and middle zones of the cartilage.

Biochemical changes in the extracellular matrix typical of OA initiation and progression have profound effects on the mechanical properties of the tissue, diminishing its function and altering the mechanical loading environment in the joint. The network of collagen type II fibers give cartilage most of its tensile strength, which is highest in the superficial zone, where tightly packed collagen fibrils are aligned parallel to the articular surface. Damage to this network in osteoarthritic cartilage is correlated with a decrease in the tensile stiffness [10]. PGs, and, in particular, their interaction with water, provide cartilage with resistance to compression through negative electrostatic repulsion forces. The negative or fixed charges of the glycosaminoglycan (GAG) sidechains give rise to an osmotic pressure difference, which causes swelling of the tissue, balanced by the elastic pre-stress developed in the collagen network [2], [11]. The compressive stiffness of cartilage increases as a function of the total GAG content [12], and decrease of PG concentration as it occurs in early OA was shown to predispose cartilage tissue to microdamage from mechanical loading, weakening and disruption of matrix structural integrity [13], [14]. Water content in cartilage is highly correlated with permeability and compressive stiffness, as measured by the aggregate modulus: As the water content increases, the tissue becomes less stiff and more permeable most likely due to a decrease in solid matrix [13]. OA is associated with an increase in water content due to the extra space created in the tissue by the decrease in PG content [15].

To study the biochemical changes occurring in early stage OA, assessment of enzymatically-degraded cartilage samples with quantitative magnetic resonance imaging (MRI) relaxometry has proven a valuable tool. The T₂ relaxation time has been associated with variations in the collagen fibril angle [16], [17] and concentration [18], while T_1ρ is strongly correlated with PG concentration [18], and T₁ is directly proportional to the water content of the tissue [19]. Extensive research has been performed on linking concentration and structure of cartilage components with degradation- and loading-induced changes in the T₂, T_1ρ, and T₁ relaxation times [20]–[22]. Nevertheless, little attention has been paid to the combined effects of enzymatic degradation and mechanical loading, i.e., to the investigation of loading-induced changes in relaxation times in enzymatically-degraded cartilage. These changes could provide information on the biochemical changes taking place in early OA due to loading, appearing already before visible tissue damage.

Furthermore, to understand the effect of these biochemical changes on the mechanical loading environment of the joint cartilage, various studies have investigated in vivo how knee joint loading is affected in patients with OA by studying contact forces or contact pressures during various tasks [23]–[25]. Nevertheless, these studies either consider the mechanical environment at the level of the whole joint, or mostly relate to more advanced OA stages, and thus may not detect significant changes in cartilage loading with early stages of the disease [26]. To investigate the direct relationship between matrix constituent damage in early OA and alterations in the mechanical environment in terms of deformations/strains or loads, in vitro experiments have been used which mimic early stage OA by using enzymatic degradation [27]–[30]. While allowing for controlled induction of OA by depletion of PGs or collagen and for determination of altered mechanical environment at the microscopic scale, such studies are mostly conducted using loading platens that are much stiffer than cartilage, therefore creating conditions that are significantly different than those in vivo.

The main objective of this study was to investigate the influence of the selective and combined degradation of cartilage matrix components on the mechanical environment in the tissue. We computed and compared strains under compressive loading in intact and enzymatically degraded cartilage plugs, depleted of either collagen, PGs, or both, thereby mimicking early stage OA. To this end, we used an experimental setting simulating in vivo cartilage-on-cartilage contact conditions in a micro-MRI environment, and investigated the depth-dependent strain magnitudes using descriptive statistics. Additionally, we analyzed the loading-induced degradation of cartilage matrix constituents and the associated biochemical changes as reflected by the T₂, T_1ρ, and T₁ relaxation times.

2. MATERIALS AND METHODS

2.1. Sample preparation and staining

Bovine knee joints from 9-month-old calves were obtained from a local abattoir within one day of slaughtering. Cylindrical osteochondral plug pairs with a diameter of 30 mm were harvested from the load-bearing area of the knees, one from the medial and one from the lateral femoral condyle. After preparation, the subchondral bone of the plugs was embedded in bone cement (VersoCit-2, Struers, Denmark) for later fixation in the sample holder.

Osteochondral plug pairs were then randomly assigned to one of the following groups: one group (n = 4) was treated with 6 U/ml collagenase type VII (Sigma Aldrich, UK) to induce collagen degradation [31]; another group (n = 4) was treated with 0.1 U/ml chondroitinase ABC (Sigma Aldrich, UK) to induce PG degradation [32], [33]; a third (n = 4) was treated with a combination of the two enzymes to induce both PG and collagen degradation; the last group (n = 3) was not treated and served as controls. The enzyme concentrations were based on previous studies showing that such enzymatic protocols had an effect on the biochemical and biomechanical properties of articular cartilage, as demonstrated by histology [32], digital densitometry [33] and stress-relaxation tests [31]–[33]. Both the control and the enzymatically degraded sample pairs were immersed in DMEM:F12 medium and incubated for 44h at 37°C. After the incubation period, enzymatic digestion was stopped, the medium was discarded and replaced with fresh phosphate buffered saline (PBS) with enzyme inhibitors (5mM EDTA and 5mM benzamidine HCL, Sigma Aldrich, UK). Explants were then snap frozen with liquid nitrogen and stored at −80° C until testing day. Before testing, the samples were incubated 30 minutes at 37°C in PBS. Previous studies on cartilage cryopreservation have shown that sample freezing at −80°C preserves tissue stiffness, especially in the case of only one freeze-thaw cycle [34], [35], while rapid thawing as performed in our study further ensures preservation of tissue mechanical properties by inhibiting ice lens formation [35]. To visualize the enzymatic activity, 5 μm sections across the full cartilage thickness of representative samples were obtained and Safranin-O/hematoxylin and Picrosirius Red staining were performed to visualize PGs and collagen, respectively. Images were taken using a Leica DMR microscope.

2.2. Cyclic compression

The paired plugs were mounted in a pneumatically-controlled, custom-built loading device [36], with the lateral plug fixed in the moving part while the medial was static. Thus, the more rounded, lateral plug was used to indent the flatter, medial plug, thereby approximating the loading between femoral and tibial joint surfaces. Cyclical load in axial compression was performed with device inside a 9.4T MRI scanner and a quadrature transmit/receive volume coil (72 mm diameter, Biospec 94/20 USR, Bruker Biospin, Ettlingen, Germany). During loading, the samples were submerged in PBS to avoid dehydration. Each cycle consisted of loading up to 350 N, 2 s holding time, and unloading in 1.4 s (Figure 1). The loading profile was gated with the DENSE imaging protocol (see next section), resulting in approximately 1200 loading cycles during imaging alone. Before image acquisition, the explants were pre-loaded for 500 cycles to reach quasi-steady state load-deformation behaviour [36].

Figure 1:

The protocol for cyclic compressive loading was gaited with the DENSE imaging protocol.

2.3. Micro-MR image acquisition

Before and after loading, high-resolution 2D T₂-weighted ‘rapid acquisition with relaxation enhancement’ (RARE) anatomical reference scans were acquired (TEeff/TR: 16.9/5692 ms, RARE factor: 4, FOV: 60×60 mm², matrix: 392×392 pixel conferring an in-plane resolution of 0.153 mm², 9 contiguous sagittal slices of 3 mm thickness acquired in an interleaved scheme, acquisition time: 9 min 17 s).

Additionally, to assess changes in the biochemical composition, quantitative T₁, T₂, and T_1ρ relaxation time measurements were conducted for a single image slice (3 mm thickness) through the center of the cartilage explants before and after loading. Data acquisition for T₁ mapping was performed using variable repetition times of 220, 350, 500, 1000, 2000, and 4000 ms with a RARE readout (TEeff: 9.26 ms, RARE factor: 2, acquisition time: 17 min 37 s). T₂ was assessed after acquisition of a multi-slice, multi-echo sequence with echo times (TE) of 11.5, 23, 34.5, 46, 57.5, 69, 80.5, and 92 ms (TR: 1000 ms, acquisition time: 4 min 22s). For T_1ρ, a RARE sequence (TEeff/TR = 50.27/2000 ms, RARE factor: 8) with a spin lock pulse of 851 Hz and spin-lock durations (TSL) of 10, 20, 40, 60, 80, 100, 120, 140, 160, and 180 ms was acquired (acquisition time: 21min 20 s). The matrix size for the relaxation scans was 392×392 pixel with an in-plane resolution of 153 μm².

To determine the axial and transverse displacements of the cartilage plugs during compressive loading, DENSE sequences [36]–[39] were used with an encoding strength of 2.55 π/mm or 60 mT/m in the x direction and 1.91 π/mm or 45 mT/m in the y direction. For the x- and y-encoded scans, the displacement encoding gradient was applied during loading, and the readout occurred during the loading plateau (Figure 1). Reference scans without a displacement encoding gradient were also acquired to eliminate displacement-independent phase contributions. Imaging was performed with the following parameters: acquisition time approximately 16 min; echo time = 1.32 ms; bandwidth = 40 kHz; number of averages to improve the signal-to-noise ratio = 20; number of repetitions for cosine and sine modulation (CANSEL) to eliminate anti-echo and T₁ decay artifacts = 4; field of view = 60×60 mm²; slice thickness = 3 mm; image size = 512×512 pixel²; resolution: 0.117 mm.

2.4. Data processing

The medial and lateral cartilage layer in each plug pair were segmented manually using ImageJ and converted into binary masks. Separate masks were segmented from the anatomical and relaxation scans before and after loading, when the plugs were not in contact; and from the loaded DENSE scans, when the plugs were in contact. All subsequent data processing was performed using MATLAB (2016b, The Mathworks, Natick, MA). Relaxation time and strain distribution comparisons between intact and degraded sample pairs were performed in the contact area between the cartilage layers for the superficial (SZ), the middle (MZ), and the deep zone (DZ), shown in Figure 2. The three zones were determined by dividing the cartilage thickness by 3, with the non-zero rest being added to the middle zone (MZ), as this is known to be thickest [40].

Figure 2:

The cartilage layers (B) were segmented from MRI DENSE reference scans, with the paired cartilage plugs in contact (A). Segmented cartilage layers, zones, contact area (between the black lines), and displacement encoding directions are shown in B.

2.4.1. Thickness analysis

To determine the thickness of the cartilage plugs before and after loading, masks were segmented from the central slice of the RARE anatomical scans taken before and immediately after loading the samples (cartilage plugs not in contact). The thickness of each plug was calculated as the average thickness of each pixel column across the contact area. The reported thickness value for a plug therefore represents a contact area-specific average thickness.

2.4.2. Relaxation time analysis

The masks used for analysis of the T₁, T₂, and T_1ρ relaxation times before and after loading were obtained by segmentation from the relaxation scans (cartilage plugs not in contact). Pixel-wise determination of the relaxation times was obtained with a Levenberg-Marquardt mono-exponential curve-fitting algorithm [41], [42] using the Anaconda distribution of Python 2.7.13 (https://www.anaconda.com/). To generate relaxation profiles as a function of the normalized cartilage thickness in the contact area, relaxation times were averaged per layer of equally distanced pixels, resulting in 15 values along the cartilage depth.

2.4.3. Strain analysis

The masks used for strain analysis during loading were obtained by segmentation from the DENSE scans (cartilage plugs in contact). From the raw DENSE data, phase differences between reference and displacement-encoded scans were computed and converted into transverse (dx) and axial displacements (dy), which were then smoothed using a two-dimensional locally weighted linear regression method, with matrix size of 256. From the smoothed displacements, Green-Lagrange transverse (E_xx), axial (E_yy), and shear strain components (E_xy) were calculated pixel-wise using a maximum likelihood estimation of the deformation gradient tensor [43].

2.5. Statistical analysis

To investigate loading-induced changes in thickness, a paired t-test was performed between the average thickness of each plug pre- and post-loading (p = 0.05). To investigate loading-induced changes in the relaxation times, paired t-tests between the average relaxation times of each zone for all samples pre- and post-loading were performed (p = 0.05). Due to the low number of samples in each group (n = 3 for controls and n = 4 for degraded samples), no statistical tests were performed on the relaxation and strain results between sample groups. Instead, a descriptive statistical analysis was employed and boxplots, means, standard deviations, and the 5% and 95% percentiles (to account for outliers) are reported.

3. RESULTS

3.1. Staining

The reduction of the histochemical staining in the degraded samples confirmed the loss of PG (Figure 3A–D) and collagen (Figure 3E–H) content compared to controls, thus suggesting effective enzymatic degradation of the cartilage. Interestingly, the collagenase degradation resulted in a lower PG concentration throughout the cartilage thickness (Figure 3B), as opposed to chondroitinase ABC degradation (Figure 3C), which strongly depleted the PGs at the cartilage surface. Degradation with a combination of both enzymes resulted in a stronger depletion of PGs (Figure 3D).

Figure 3:

Change in intensity and consistency of the histochemical staining due to enzymatic activity compared to controls (A, E) in representative samples degraded with collagenase (B, F), chondroitinase ABC (C, G), and a combination of collagenase and chondroitinase ABC (D, H). Staining was performed with Safranin-O/hematoxylin to visualize PGs (A-D) and with Picrosirius Red to visualize collagen (E-H).

3.2. Thickness

The average length of the contact area between the cartilage plugs was 14.04±1.71 mm. The mean thickness of the cartilage plugs in the contact area decreased significantly after loading, from 2.05±0.31 mm to 1.92±0.31 mm (p = 1.157e-4) in the medial plug, and from 1.92±0.28 mm to 1.78±0.23 mm (p = 2.77e-4) in the lateral plug (Figure 3). There was no significant difference in the mean thickness of the medial and lateral plugs either before loading (p = 0.0572) or after loading (p = 0.0511). No differences in thickness were observed between control and degraded samples, however, the largest decrease in average thickness due to loading (~10.06%) was observed in the medial plugs of the combination-degraded samples.

3.3. Relaxation times

The magnitude of the T₁, T₂, and T_1ρ relaxation times showed a depth-dependence, with the average relaxation times in the contact area decreasing from the superficial (SZ) to the middle (MZ) and deep zones (DZ), both before and after loading (Figure 5A and 5B). While the SZ had clearly higher values for T₁, T₂, and T_1ρ, the differences between the MZ and SZ were less pronounced.

Figure 5:

T₁, T₂, and T_1ρ relaxation times in the cartilage plug pairs exhibited a depth-dependent behavior: A. Relaxation maps before and after loading (example of a collagenase-degraded sample); B. Sample-specific relaxation times profiles along the normalized depth of the cartilage in the medial and lateral plugs before and after loading^†. The x axis represents the normalized cartilage thickness, where 0 = cartilage surface; 1.0 = cartilage-bone interface; the superficial (SZ), middle (MZ) and deep zones (DZ) are marked at 0.33, 0.66 and 1.0, respectively.

The average T₂ and T_1ρ relaxation times in the contact area decreased significantly immediately after loading in all three cartilage zones, as shown in Table S1. Conversely, the average T₁ relaxation time increased significantly after loading. The largest changes in the relaxation times post-loading occurred in the SZ, with smaller changes in the MZ and DZ.

Relaxation time profiles throughout the sample thickness are depicted in Figure 5B. Table 1 summarizes the average relaxation times per sample group at the articular surface (corresponding with normalized cartilage thickness equals 0 in Figure 5B). Although the average T₂ and T_1ρ were generally increased at the articular surface in the degraded samples compared to controls, no clear trend was visible for degradation only with collagenase. Conversely, degradation with a combination of both enzymes always resulted in increased T₂ and T_1ρ at the articular surface compared to controls, for both medial and lateral cartilage layers, before and after loading. No clear conclusion could be drawn regarding degradation-related changes in T₁ versus controls.

Table 1:

Average relaxation times [ms] and difference in the relaxation times before and after loading per sample group at the articular surface of the medial and lateral cartilage plugs, corresponding with normalized cartilage thickness equals 0 in Figure 5B.

	Medial plug articular surface									Lateral plug articular surface
	before loading			after loading			ΔT			before loading			after loading			ΔT
	T11	T2	T1ρ	T11	T2	T1ρ	T11	T2	T1ρ	T11	T2	T1ρ	T11	T2	T1ρ	T11	T2	T1ρ
Control	3051.1±307.8	193.3±13.1	289.6±26.2	5131.7±747.5	172.7±8.2	269.3±18.4	−2080.6±1055.2	20.5±8.8	20.3±8.8	3281.9±391.4	187.8±9.2	260.1±25.0	4366.3±752.9	172.6±20.6	239.4±24.1	−1084.4±361.4	15.2±13.1	20.8±2.24
Degraded collagen	3336.9±215.9	232.4±55.2	334.8±56.5	5108.7±1116.4	189.3±49.1	206.9±15.0	−1771.8±1235.6	43.1±20.6	127.8±47.2	2384.3±342.6	174.7±17.1	282.0±23.5	3966.9±135.0	135.0±16.8	233.9±23.1	−1582.6±217.2	39.8±20.2	48.1±43.6
Degraded PGs	3118.1±342.1	243.5±45.6	360.2±93.6	5240.6±796.3	188.6±28.8	248.5±56.2	−2122.5±583.9	55.0±20.8	111.7±58.1	2825.6±325.3	249.4±19.7	371.8±47.9	4070.6±1180.8	160.8±22.4	266.4±85.1	−1245.0±1201.6	88.6±5.3	105.3±114.9
Degraded collagen + PGs	3326.9±60.0	289.8±12.2	437.8±28.8	5968.2±235.2	235.2±21.1	319.9±34.1	−2641.3±209.2	54.6±18.7	117.9±40.2	3303.6±133.8	281.2±20.1	440.7±31.5	5091.4±301.6	222.8±17.5	339.0±17.3	−1787.9±317.5	58.4±14.7	101.7±27.5

The change in relaxation times due to loading proved to be least pronounced in the control samples and highest in those degraded with chondroitinase or a combination of chondroitinase and collagenase (see grey columns in Table 1).

3.4. Strains

The transverse (E_xx), axial (E_yy), and shear strains (E_xy) were calculated pixel-wise in the contact area (see Figure 2), separately for the three zones (SZ, MZ, DZ) in each cartilage plug (medial, lateral). The sample-specific results in the form of boxplots are summarized as distributions in Figure S1.

The average axial strains showed a depth-dependent behavior, with maximum values in the superficial zone (SZ) and decreasing towards the subchondral bone interface, i.e. in the deep zone (DZ) (Figure 6, E_yy). The same trend was observed for the average shear strains (Figure 6, E_xy), although the difference in magnitude between SZ and DZ faded in the lateral plug. This behavior was less pronounced in the average transverse strains, although values were still higher in the SZ than the DZ (Figure 6, E_xx).

Figure 6:

Sample-specific average transverse (E_xx), axial (E_yy), and shear strains (E_xy) in the medial and lateral plug, in each of the three zones: superficial (SZ), middle (MZ), and deep (DZ). The average axial strains exhibited depth-dependent behaviour, which was less visible for the transverse and shear strains.

In the medial (static) plug, the largest average transverse and axial strains occurred in the SZ and MZ of the combination-degraded samples (Table S3). In the DZ, these samples only showed larger values than the other groups in the axial direction (of loading). In the lateral (moving) plug, these differences were less pronounced, although still visible in the transverse strains in the SZ and MZ. Nevertheless, when looking at the absolute maximum strain values (considered here as the 5% and 95% percentiles) these differences also become apparent in the lateral plug (Table S4). The maximum absolute shear strains were higher in the combination-degraded samples for both plugs in all three zones. In all samples, degradation with one of the two enzymes, collagenase or chondroitinase, produced a limited increase in the average and maximum contact area strains. Conversely, degradation with the combination of the enzymes resulted in much higher average and maximum strains compared to controls, with the most dramatic differences between control and combination-degraded samples noticeable in the SZ of the medial (static) plug, in the axial direction (Figure S1 and Figure 7).

Figure 7:

The most pronounced difference in strain magnitude in the contact area was found between control (A) and combination-degraded (B) samples. The strain maps in the two selected samples also illustrates the strain gradient from the articular surface towards the subchondral bone.

3.5. Case study: broken collagenase-degraded sample

Beside the 15 samples for which we reported pre- and post-loading relaxation times and strain values, we tested a number of other samples, which proved to be damaged, or became damaged during testing due to pre-existing flaws. While we removed these samples from the analysis of intact samples due to unreliable data, we here report these values separately for one particularly interesting sample, as a case study of cartilage undergoing both enzymatic degradation and structural damage.

One sample from the collagenase-degraded group had a fissure throughout the depth of the cartilage in the lateral plug, visible only in the anatomical scans (Figure 8A), which opened to a flap during loading (Figure 8B and C). At the articular surface, all three relaxation times, both before and after loading, were higher in the broken plug than in the intact plugs. The T₂ and T_1ρ relaxation times stayed higher in the broken plug as far in depth as the SZ before loading (Figure 9A, left); and after loading, T₁, T₂ and T_1ρ were all higher in the broken plug compared to the intact plugs as far as the SZ (T₁ and T₂), or SZ and MZ (T_1ρ) (Figure 9A, right). The transverse (tensile) strains were much higher in the broken (collagenase-degraded) plug compared to the intact (collagenase-degraded) plug in all three zones of the cartilage, suggesting decreased tensile stiffness (Figure 9B).

Figure 8:

One of the collagenase-degraded samples had a fissure extending throughout the cartilage depth of the lateral plug, shown in the anatomical scan (A), which opened to a flap in the loaded scans (B), and was also visible macroscopically after removal from the loading device (C).

Figure 9:

A. Relaxation times depth profiles (A) and transverse strains (B) were altered in the broken collagenase-degraded sample compared to the intact collagenase-degraded samples (lateral plug) before and after loading³.

4. DISCUSSION

This study aimed to evaluate the altered mechanical environment during compressive cartilage on cartilage loading in enzymatically degraded cartilage plugs mimicking early-stage OA. We used an MRI-compatible loading device, allowing for non-destructive determination of strains during cartilage-on-cartilage dynamic loading, as well of loading-induced changes in thickness and relaxation times. We previously showed that the chosen loading protocol results in average contact stress of 2.78±1.47 MPa applied for 2 s during the holding phase [44], representative for moderate daily physical activities in humans [45], and on the order of magnitude of the average contact pressures determined in human knee condyles during the stance phase of gait [46]. On the longer term, this method could allow for non-invasive assessment of early OA cartilage mechanical properties, thus contributing to the design of mechanically tailored scaffolds for regenerative procedures.

Staining and relevance for early stage OA

Safranin-O and Picrosirius Red staining in representative samples was used to confirm the enzymatic degradation of cartilage. Interestingly, PG depletion in the samples degraded with collagenase seems to have advanced further in the tissue depth than in the samples degraded with chondroitinase ABC, which directly targets PGs. This may be explained by the release of PGs from their positioning in the extracellular matrix once the collagen network is disrupted. On the other hand, degradation with chodroitinase ABC caused a more dramatic depletion of the PGs directly at the cartilage surface. Although we did not quantify the extent of the enzyme penetration for each sample used in this study, Nieminen et al. [33] showed that with the same protocol used here, chondroitinase ABC removed PGs to a typical depth of 20% to 30% of cartilage thickness in osteochondral plugs with 13 mm diameter. Therefore, despite analyzing all three zones of the cartilage layers for degradation-related differences in relaxation times and strains, we expected that most of these changes would be visible in the SZ, defined here as 1/3 of the cartilage thickness. These arguments, supported by our Safranin-O and Picrosirius Red staining images, lead us to conclude that our experimental setup is clinically relevant for early stage OA, more specifically, grade 1-2 (as defined in [8]). OA grade 1 is characterized by disruption or alteration of the cartilage matrix, cartilage swelling, and mild fibrillation in the superficial zone, as well as swelling of the aggrecan molecules, increase in the tissue permeability, a decrease of the matrix stiffness, and formation of surface irregularities [47], [48]. In grade 2, production of nitric oxide [49] may eventually lead to increased release of proteases that progressively degrade collagen fibrils and PGs, particularly aggrecan, in the superficial zone. Grade 2 OA is characterized by matrix depletion proven by staining (with Safranin O or Toluidine Blue) in the upper 1/3 of cartilage, as well as increased chondrocyte death, proliferation and clustering.

Thickness

There was no difference between the thickness of control and enzymatically degraded samples, as found previously both with collagenase and with chondroitinase ABC degradation [30]. As expected, cartilage thickness decreased immediately after loading. Nevertheless, no difference was observed between the thickness decrease in control and enzymatically degraded samples. The highest average decrease in thickness was observed in the medial cartilage layer for the combination-degraded samples, which coincides with the condition presenting the maximum compressive strains.

Relaxation times – before loading

As the Safranin O/hematoxylin staining suggested a limited penetration of the enzymes in the tissue depth, we initially analyzed the sample group-specific average relaxation times only in the SZ. Nevertheless, the influence of the different enzymes on the average relaxation times in the SZ before and after loading did not show a clear trend (Table S2). This effect may be lost due to averaging the values in the SZ, representing one third of the sample depth, when in reality the degradation might affect relaxation times more superficially. We therefore refined the analysis and determined the relaxation throughout the whole sample thickness, thus obtaining the profiles depicted in Figure 5B.

Our results clearly showed increased relaxation times in the samples degraded with a combination of collagenase and chondroitinase ABC compared to controls, both before and after loading, particularly at the articular surface and up to at least 10% cartilage depth (Figure 5B, Table 1). As previously shown in intact cartilage samples [44], we found that the relaxation times were depth dependent for all sample groups, with maximum values at the cartilage surface and decreasing towards the subchondral bone. This behavior reflects the depth dependent distribution of cartilage components. The orientation of the collagen fibrils changes from parallel to the cartilage surface in the SZ, to oblique in the MZ and perpendicular to the subchondral bone in the DZ, and the depth-wise variation of T₂ observed here has been associated with variations in the collagen fibril angle [16], [17] and concentration [18]. Increased values of T_1ρ have been associated with a decreased concentration of PGs [18]. This, corroborated with the fact that the DZ has the highest PG content [40], results in the herein observed decrease in T_1ρ from the articular surface towards the subchondral bone. Finally, the T₁ relaxation time, directly proportional to the water content, was found to decrease from the articular surface towards the subchondral bone, as previously shown in literature [19]. This variation reflects the distribution of water in cartilage, which has the highest concentration at the surface, due to the high tissue permeability, and the lowest concentration in the DZ.

Nevertheless, results obtained for samples degraded only with one or the other enzyme were less conclusive, particularly in the case of collagenase. In order to compare the outcomes obtained here with other findings, we will first focus on the effect of degradation on the relaxation times before loading, to remove a possible bias due to compression of the tissue. Increases in T₂ and T_1ρ relaxation times with enzymatic degradation have been previously reported in literature. Nissi et al. [32] and Nieminen et al. [33] showed significantly increased values of T₂ in the superficial zone (up to ~20% of tissue depth) of collagenase-degraded bovine patellar cartilage. Conversely, degradation with chondroitinase ABC did not affect the superficial T₂. Our results showed increased T₂ at the articular surface with collagenase degradation, but not consistently, e.g., the average T₂ for the control group in the lateral plug before loading was larger than the average T₂ for the collagenase-degraded group. This inconsistency might be related to experimental parameters: although we used a similar degradation protocol in terms of degradation time, Nissi et al. [32] and Nieminen et al. [33] used a concentration of 30 U/ml collagenase, i.e. five-fold higher than ours. The mentioned studies also performed degradation on smaller samples, with a diameter of 13 mm [33], or resulting from 25-mm-diameter plugs being cut into 3 sectors [32]. These factors could contribute to a much stronger degradation of the collagen network. Moreover, the resolution used in our study for T₂ imaging was 153 μm, while the other studies used 62.5 μm [32] and 78 μm [33], most likely resulting in higher sensitivity to degradation-induced changes. Regarding T_1ρ, our results showed increased values in chondroitinase ABC degraded samples compared to controls before loading at the articular surface, as well as up to ~10% of the tissue depth (Figure 5B). This is consistent with studies using the same concentration of enzyme (0.1 U/ml) [32].

Strains

Interestingly, the spread of the strain distributions in the three zones of the cartilage plugs was larger in the degraded than in the control samples, indicating that enzymatic degradation induced a higher material heterogeneity. In the descriptive analysis performed, this effect is illustrated by the larger interquartile range of the strain distributions of the combination-degraded samples (Figure S1), as well as in the larger standard deviations of these strains (Table S3). The effect of degradation on the contact strains was highest in the SZ, most likely due to the limited penetration of the enzymes, expected to reach approximately 1/3 of the cartilage depth or less considering the size of the plugs used here. Nevertheless, while degradation-related changes in the relaxation times were visible up to a cartilage depth of ~10%, changes in strains reached in some cases to the MZ, representing 2/3 of the depth. We therefore conclude that our method is more sensitive to detecting biomechanical changes than biochemical changes, partially also due to the different imaging resolutions (117 μm for DENSE vs 153 μm for the relaxations scans).

As we previously showed in intact cartilage samples [44], axial strains E_yy were compressive and had a depth-dependent behavior, decreasing from the articular surface towards the subchondral bone, findings consistent with studies in both adult [50]–[52] and immature [51] bovine cartilage. This reflects the PG concentration in the tissue, which is highest at the DZ and decreases towards the SZ, resulting in the opposite depth-dependence for the cartilage compressive stiffness [53]. A similar depth-dependent pattern was partially also seen in the shear strains E_xy (more prominently in the medial/static plug), consistent with our previous study [44] on the same type of bovine cartilage samples (9 month old). This depth-dependence has been reported in literature and associated with an increase of the shear modulus from the articular surface towards the subchondral bone in immature bovine cartilage of a similar age to ours (6 months old). Despite a less clear depth-dependence of the transverse (tensile) strains, E_xx was higher in the SZ than the DZ for most of the samples, as has been shown before in both adult and young bovine cartilage [51], [52]. Considering the orientation and density of collagen fibrils, it would be expected that cartilage has the highest tensile stiffness at the SZ, where fibrils are densely packed and parallel to the articular surface, and that this mechanical property decreases with increasing depth, resulting in lower tensile strains at the articular surface. While this is true for static loading, with loading rates low enough to allow superficial collagen fibrils to stretch and provide tensile resistance, the protocol used in our study resulted in faster strain rates, at which PGs limit stretching of the collagen fibers [27].

An interesting outcome is that the individual enzymes, collagenase or chondroitinase-ABC, produced less significant changes than their combination in the contact region-specific strains. In fact, samples degraded with collagenase showed very similar or even lower average strains compared to controls, particularly for E_xx, which would be expected to be increased at least in the SZ due to degradation of the collagen network. Higher strain values in the collagenase-degraded samples compared to controls only become apparent when looking at the maximum (p95) values of E_xx in the SZ of the cartilage layers (Table S4). Various studies have confirmed a decrease in cartilage mechanical properties with collagenase degradation: Töyräs et al. [30] reported a significantly lower Young’s modulus from indentation tests in collagenase-degraded full-thickness cartilage versus controls; Laasanen et al. [53] showed that the dynamic modulus of full-thickness cartilage samples (without subchondral bone) decreased by an average of 45% and the equilibrium modulus by an average of 70% after collagenase treatment; and Griffin et al. [54] showed a decrease in shear modulus advancing deeper in the tissue depth with degradation time. Nevertheless, all these studies used either smaller samples or a higher enzyme concentration than ours, cartilage samples without subchondral bone, or a combination of these factors. We therefore attribute the fact that our results do not reflect the biomechanical changes expected with collagenase degradation to the large sample diameter and low enzyme concentration.

On the other hand, depletion of PGs due to chondroitinase ABC degradation resulted in an increase of E_yy in the SZ of the cartilage plugs. The average superficial axial strains in chondroitinase degraded samples were 25% and 39% higher than in controls in the medial and lateral plug, respectively. These results are consistent with others showing that chondroitinase ABC degradation results in a significantly lower Young’s modulus of cartilage from indentation tests [30] and lower dynamic and equilibrium compressive modulus [53], [55]. In our study, depletion of PGs also led to increased maximum (p95) shear strains in the SZ by 68% and 60% in the medial and lateral plug, respectively, suggesting a decreased shear modulus, which has been confirmed in samples treated with chondroitinase ABC [55].

Finally, enzymatic degradation with both collagenase and chondroitinase ABC resulted in the most dramatic changes compared to controls in the axial, transverse, and shear strains in the cartilage contact region, with the highest increase in the SZ and decreasing towards the DZ. This effect was more visible in the medial plug when comparing the average strain values, but also became apparent in the lateral counterpart when considering the extreme values (5% and 95% percentile). Indeed, it has been previously shown that cartilage covering the medial femoral condyle has inferior mechanical properties to the lateral and is more prone to developing OA, as it carries most of the joint load [56]. Another source of differences between the two plugs may be of experimental nature, as the medial explant is static while the lateral is moving through PBS, which might influence the image acquisition during contact. The largest difference in the average E_yy (158%) between combination degraded and control samples was in the SZ of the medial plug, followed by the MZ (116%) of the same plug. Interestingly, Eyy was also increased in the DZ (70%), suggesting that degradation with both enzymes depleted the matrix constituents up to larger depths. The average E_xx was increased due to combination degradation by 75% in the medial plug SZ, and by 60% in the lateral plug SZ. The maximum (p95) value of E_xy for the combination-degraded samples was 84% and 76% higher than controls in the medial plug SZ and lateral plug SZ, respectively. The combined effect of both collagen and PG degradation was illustrated by previous studies using trypsin to degrade both components, resulting in a lower equilibrium compressive modulus of cartilage with this enzyme than with degradation of one or the other component using only collagenase or only chondroitinase ABC [53].

Relaxation times – effect of loading

Determination of the relaxation times before and after loading was performed to evaluate the effect of dynamic compression on the biochemical composition of cartilage. Similar to our previous study [44], we found that the T₂ and T_1ρ relaxation times increased after dynamic loading in all three cartilage zones; while T_l decreased after loading. The decrease of T₂ due to compressive loading has been previously reported in literature, both in vivo and in vitro, and linked with the deformation and orientation changes of the collagen fibrils, as well as with the increase of collagen concentration due to matrix compaction and extrusion of water [21], [22], [57]. The decrease in T_1ρ observed here immediately after dynamic loading is consistent with other studies, and may be related to an overall increase in PG concentration due to tissue compression [22]. The increase in T₁ relaxation time after loading, although reported previously by our group [44] is, nevertheless, not consistent with results reported in literature [20]. The latter show a decrease in T₁ with loading in vivo, most likely caused by a decrease in the mobile water due to compression of cartilage. The increase in T₁ observed here would suggest, by contrast, an increase in the water concentration. This could be, on the one hand, due to the release of “trapped” water from the extracellular space, which had been previously bound to proteins and was freed due to tissue deformations. On the other hand, as previously hypothesized [44], this increase could also be related to temperature effects due to the heat transfer associated with magnetization.

Looking at the effect of the enzymes on the loading-induced changes in relaxation times in the different sample groups, we conclude that degradation with chondroitinase ABC or a combination of chondroitinase and collagenase results in the largest changes at the articular surface (Table 1). These changes were largest in the combination-degraded samples in the medial and lateral plug for T₁ in the chondroitinase-degraded samples in both plugs for T₂, and in the lateral plug for T_1ρ. Interestingly, T_1ρ in the medial plug had the largest loading-induced changes in the collagenase-degraded group. As this result is not consistent with other outcomes in this sample group, we attribute it to the limitations in the method used to determine relaxation times at the articular surface (see section Limitations). Taken together, these outcomes show that alteration of the biochemical composition of cartilage due to mechanical loading is more pronounced when the tissue components are depleted and the matrix structure is already altered before loading.

Case study – relevance to later stage OA

We presented herein a case study of a collagenase-degraded sample with a fissure throughout the depth of the cartilage in the lateral plug, which opened to a flap during loading. This plug exhibited higher T₂ and T_1ρ relaxation times than matched intact collagenase-degraded plugs both before and after loading, indicative of the altered structure and concentration of the PGs and collagen fibers. Most interestingly, it exhibited significantly higher transverse (tensile) strains compared to the intact plugs in all three cartilage zones, while the axial (compressive) strains were not significantly increased. This suggests that the discontinuity induced in the cartilage structure by the flap mostly disrupts the collagen network, while PGs are less affected due to tissue compaction during loading. The outcomes obtained for this sample could be clinically representative for later stage OA, i.e. grade 3-4, where vertical fissures extend into the middle zone and delamination of the superficial zone takes place [8].

Limitations

One of the clear limitations of the current study is the low concentration of collagenase used to degrade collagen in the cartilage samples, which did not allow for proper detection of biochemical and biomechanical changes induced by damage to the collagen network. In the future, increasing the enzyme concentrations in the separate treatments could provide more clear information on how degrading only one or the other matrix component affects the contact strains in cartilage-on-cartilage compressive loading. Moreover, the ability to quantify the degree of enzyme penetration into the cartilage thickness could offer a quantitative relationship between degradation and the related alterations in the tissue’s mechanical properties.

Another limitation we acknowledge is related to the method used to determine relaxation times and thickness in the plugs, which is based on manual segmentation of the cartilage layers. This step in the relaxation time analysis is prone to errors due to potential inclusion in the cartilage masks of pixels which lay outside the cartilage layers. This error may lead to partial volume effects and inflated values of the relaxation times, particularly at the articular surface. In the thickness measurements, this error might explain the inconsistency in thickness decrease between individual plugs, seen in Figure 4, as well as the fact that we found no significant differences between thickness decrease in control and degraded samples. Nevertheless, the masks were segmented by the same author in the same manner for all samples, thus not affecting the validity of the conclusions drawn herein.

Figure 4:

The average contact area-specific thickness decreased after loading for the 15 samples.

5. CONCLUSION

This study showed that degradation of the main components of cartilage, collagen and proteoglycans, changes the biomechanical response of the tissue during cyclic compressive loading. Nevertheless, while degradation of one or the other matrix component induced a low increase in strains, samples depleted of both collagen and proteoglycans showed significantly higher maximum axial, transverse and shear strain components than controls. This indicates that the interplay between cartilage matrix constituents plays an essential role in the decrease of the load bearing capacity and the altered mechanical environment characteristic of degenerative joint diseases such as OA. This effect seems to be even more important than that of an isolated full thickness defect, investigated previously by our group [58], therefore confirming the existence of altered mechanical environment already in early OA, even before macroscopic disruption of the cartilage structure.

The MRI DENSE method presented here allows to non-invasively measure intratissue strains during cartilage-on-cartilage cyclic compression using physiologically relevant loading. Importantly, this method enables the measurement of displacements or strains in individual pixels throughout the tissue, thereby providing detailed deformation fields and high resolution information on local deformations. Such information is essential for the determination of localized biomechanical stimuli e.g. those affecting individual chondrocytes, which may provide input for highly accurate design of tissue engineering solutions for articular cartilage repair. Furthermore, this setup allowed us to enzymatically induce early stage OA in the plugs and study the mechanical behavior of cartilage under these conditions. We could, therefore, show altered mechanics of the tissue due to altered biochemical contents with an OA stage that is most likely not detectable in clinical applications. In the future, this method could be used to gradually induce OA in osteochondral plugs to study the altered mechanics of the tissue with disease progression. Because such experiments are performed in vitro and the magnitude and duration of the applied loading is tunable, extreme mechanical conditions can be also be simulated, such as overloading or fatigue. Finally, this method could contribute to the development of mechanically tailored scaffolds for cell-based joint replacement techniques in early OA cartilage defects.

Displacement-encoded MRI as presented in this manuscript has been recently applied for the first time to also measure articular cartilage deformation in vivo [59]. The results showed that this method could be established as a new tool for the non-invasive, biomechanical assessment of articular cartilage at high resolutions in vivo. In the future, this could allow for non-invasive determination of the mechanical properties of early stage OA cartilage, potentially as part of in vivo clinical screening.